godotengine
/
godot
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
godot/drivers/vulkan/rendering_device_vulkan.cpp

9520 lines
386 KiB
C++
Raw Normal View History

Add static Vulkan loader. Initial Vulkan support for Windows. Initial Vulkan support for macOS.
2019-06-22 16:34:26 +00:00
/*************************************************************************/
/* rendering_device_vulkan.cpp */
/*************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* https://godotengine.org */
/*************************************************************************/
Update copyright statements to 2022 Happy new year to the wonderful Godot community!
2022-01-03 20:27:34 +00:00
/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */
/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */
Add static Vulkan loader. Initial Vulkan support for Windows. Initial Vulkan support for macOS.
2019-06-22 16:34:26 +00:00
/* */
/* Permission is hereby granted, free of charge, to any person obtaining */
/* a copy of this software and associated documentation files (the */
/* "Software"), to deal in the Software without restriction, including */
/* without limitation the rights to use, copy, modify, merge, publish, */
/* distribute, sublicense, and/or sell copies of the Software, and to */
/* permit persons to whom the Software is furnished to do so, subject to */
/* the following conditions: */
/* */
/* The above copyright notice and this permission notice shall be */
/* included in all copies or substantial portions of the Software. */
/* */
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
/*************************************************************************/
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
#include "rendering_device_vulkan.h"
Vulkan: Use VK_NULL_HANDLE for handles, fixes 32-bit builds Some Vulkan types are defined as "non dispatchable handles" and use a different typedef on 32-bit and 64-bit systems (struct pointer on 64-bit, `uint64_t` otherwise). https://github.com/KhronosGroup/Vulkan-Headers/blob/0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5c/include/vulkan/vulkan_core.h#L59-L65 Contrarily to `NULL`, `nullptr` can't be converted to `uint64_t` so build was now failing on 32-bit after converting the codebase from using `NULL` to `nullptr`. Fixes #37620.
2020-04-06 08:17:42 +00:00
Reorganized core/ directory, it was too fatty already -Removed FuncRef, since Callable makes it obsolete -Removed int_types.h as its obsolete in c++11+ -Changed color names code
2020-11-07 22:33:38 +00:00
#include "core/config/project_settings.h"
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
#include "core/io/compression.h"
Core: Move DirAccess and FileAccess to `core/io` File handling APIs are typically considered part of I/O, and we did have most `FileAccess` implementations in `core/io` already.
2021-06-11 12:51:48 +00:00
#include "core/io/file_access.h"
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
#include "core/io/marshalls.h"
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
#include "core/os/os.h"
Reorganized core/ directory, it was too fatty already -Removed FuncRef, since Callable makes it obsolete -Removed int_types.h as its obsolete in c++11+ -Changed color names code
2020-11-07 22:33:38 +00:00
#include "core/templates/hashfuncs.h"
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
#include "drivers/vulkan/vulkan_context.h"
Vulkan: Initialize struct members on declaration Supersedes and closes #38945.
2020-06-03 08:45:44 +00:00
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
#include "thirdparty/misc/smolv.h"
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
#include "thirdparty/spirv-reflect/spirv_reflect.h"
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
//#define FORCE_FULL_BARRIER
Add a separate pool for small allocations in Vulkan RD
2022-02-12 11:47:08 +00:00
static const uint32_t SMALL_ALLOCATION_MAX_SIZE = 4096;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
// Get the Vulkan object information and possible stage access types (bitwise OR'd with incoming values)
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RenderingDeviceVulkan::Buffer *RenderingDeviceVulkan::_get_buffer_from_owner(RID p_buffer, VkPipelineStageFlags &r_stage_mask, VkAccessFlags &r_access_mask, uint32_t p_post_barrier) {
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
Buffer *buffer = nullptr;
if (vertex_buffer_owner.owns(p_buffer)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = vertex_buffer_owner.get_or_null(p_buffer);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT;
r_access_mask |= VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT;
if (buffer->usage & VK_BUFFER_USAGE_STORAGE_BUFFER_BIT) {
if (p_post_barrier & BARRIER_MASK_RASTER) {
r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
}
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
}
}
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
} else if (index_buffer_owner.owns(p_buffer)) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT;
r_access_mask |= VK_ACCESS_INDEX_READ_BIT;
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = index_buffer_owner.get_or_null(p_buffer);
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
} else if (uniform_buffer_owner.owns(p_buffer)) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
if (p_post_barrier & BARRIER_MASK_RASTER) {
r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
}
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
}
r_access_mask |= VK_ACCESS_UNIFORM_READ_BIT;
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = uniform_buffer_owner.get_or_null(p_buffer);
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
} else if (texture_buffer_owner.owns(p_buffer)) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
if (p_post_barrier & BARRIER_MASK_RASTER) {
r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
r_access_mask |= VK_ACCESS_SHADER_READ_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
}
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
r_access_mask |= VK_ACCESS_SHADER_READ_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
}
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = &texture_buffer_owner.get_or_null(p_buffer)->buffer;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
} else if (storage_buffer_owner.owns(p_buffer)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = storage_buffer_owner.get_or_null(p_buffer);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
if (p_post_barrier & BARRIER_MASK_RASTER) {
r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (buffer->usage & VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT) {
r_stage_mask |= VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT;
r_access_mask |= VK_ACCESS_INDIRECT_COMMAND_READ_BIT;
}
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
}
return buffer;
}
static void update_external_dependency_for_store(VkSubpassDependency &dependency, bool is_sampled, bool is_storage, bool is_depth) {
// Transitioning from write to read, protect the shaders that may use this next
// Allow for copies/image layout transitions
dependency.dstStageMask |= VK_PIPELINE_STAGE_TRANSFER_BIT;
dependency.dstAccessMask |= VK_ACCESS_TRANSFER_READ_BIT;
if (is_sampled) {
dependency.dstStageMask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
dependency.dstAccessMask |= VK_ACCESS_SHADER_READ_BIT;
} else if (is_storage) {
dependency.dstStageMask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
dependency.dstAccessMask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
} else {
dependency.dstStageMask |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
dependency.dstAccessMask |= VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
}
if (is_depth) {
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
// Depth resources have additional stages that may be interested in them
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
dependency.dstStageMask |= VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
dependency.dstAccessMask |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
}
}
Additional synchronization fixes w/o FORCE_FULL_BARRIER Additional synchronization fixes from hazards arising from disabling FORCE_FULL_BARRIER.
2020-07-29 19:12:21 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::_add_dependency(RID p_id, RID p_depends_on) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (!dependency_map.has(p_depends_on)) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
dependency_map[p_depends_on] = Set<RID>();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
dependency_map[p_depends_on].insert(p_id);
if (!reverse_dependency_map.has(p_id)) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
reverse_dependency_map[p_id] = Set<RID>();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
reverse_dependency_map[p_id].insert(p_depends_on);
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::_free_dependencies(RID p_id) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//direct dependencies must be freed
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
Style: Set clang-format Standard to Cpp11 For us, it practically only changes the fact that `A<A<int>>` is now used instead of the C++03 compatible `A<A<int> >`. Note: clang-format 10+ changed the `Standard` arguments to fully specified `c++11`, `c++14`, etc. versions, but we can't use `c++17` now if we want to preserve compatibility with clang-format 8 and 9. `Cpp11` is still supported as deprecated alias for `Latest`.
2020-03-17 06:33:00 +00:00
Map<RID, Set<RID>>::Element *E = dependency_map.find(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (E) {
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
while (E->get().size()) {
free(E->get().front()->get());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
dependency_map.erase(E);
}
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
//reverse dependencies must be unreferenced
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
E = reverse_dependency_map.find(p_id);
if (E) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
for (Set<RID>::Element *F = E->get().front(); F; F = F->next()) {
Style: Set clang-format Standard to Cpp11 For us, it practically only changes the fact that `A<A<int>>` is now used instead of the C++03 compatible `A<A<int> >`. Note: clang-format 10+ changed the `Standard` arguments to fully specified `c++11`, `c++14`, etc. versions, but we can't use `c++17` now if we want to preserve compatibility with clang-format 8 and 9. `Cpp11` is still supported as deprecated alias for `Latest`.
2020-03-17 06:33:00 +00:00
Map<RID, Set<RID>>::Element *G = dependency_map.find(F->get());
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
ERR_CONTINUE(!G);
ERR_CONTINUE(!G->get().has(p_id));
G->get().erase(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
reverse_dependency_map.erase(E);
}
}
const VkFormat RenderingDeviceVulkan::vulkan_formats[RenderingDevice::DATA_FORMAT_MAX] = {
VK_FORMAT_R4G4_UNORM_PACK8,
VK_FORMAT_R4G4B4A4_UNORM_PACK16,
VK_FORMAT_B4G4R4A4_UNORM_PACK16,
VK_FORMAT_R5G6B5_UNORM_PACK16,
VK_FORMAT_B5G6R5_UNORM_PACK16,
VK_FORMAT_R5G5B5A1_UNORM_PACK16,
VK_FORMAT_B5G5R5A1_UNORM_PACK16,
VK_FORMAT_A1R5G5B5_UNORM_PACK16,
VK_FORMAT_R8_UNORM,
VK_FORMAT_R8_SNORM,
VK_FORMAT_R8_USCALED,
VK_FORMAT_R8_SSCALED,
VK_FORMAT_R8_UINT,
VK_FORMAT_R8_SINT,
VK_FORMAT_R8_SRGB,
VK_FORMAT_R8G8_UNORM,
VK_FORMAT_R8G8_SNORM,
VK_FORMAT_R8G8_USCALED,
VK_FORMAT_R8G8_SSCALED,
VK_FORMAT_R8G8_UINT,
VK_FORMAT_R8G8_SINT,
VK_FORMAT_R8G8_SRGB,
VK_FORMAT_R8G8B8_UNORM,
VK_FORMAT_R8G8B8_SNORM,
VK_FORMAT_R8G8B8_USCALED,
VK_FORMAT_R8G8B8_SSCALED,
VK_FORMAT_R8G8B8_UINT,
VK_FORMAT_R8G8B8_SINT,
VK_FORMAT_R8G8B8_SRGB,
VK_FORMAT_B8G8R8_UNORM,
VK_FORMAT_B8G8R8_SNORM,
VK_FORMAT_B8G8R8_USCALED,
VK_FORMAT_B8G8R8_SSCALED,
VK_FORMAT_B8G8R8_UINT,
VK_FORMAT_B8G8R8_SINT,
VK_FORMAT_B8G8R8_SRGB,
VK_FORMAT_R8G8B8A8_UNORM,
VK_FORMAT_R8G8B8A8_SNORM,
VK_FORMAT_R8G8B8A8_USCALED,
VK_FORMAT_R8G8B8A8_SSCALED,
VK_FORMAT_R8G8B8A8_UINT,
VK_FORMAT_R8G8B8A8_SINT,
VK_FORMAT_R8G8B8A8_SRGB,
VK_FORMAT_B8G8R8A8_UNORM,
VK_FORMAT_B8G8R8A8_SNORM,
VK_FORMAT_B8G8R8A8_USCALED,
VK_FORMAT_B8G8R8A8_SSCALED,
VK_FORMAT_B8G8R8A8_UINT,
VK_FORMAT_B8G8R8A8_SINT,
VK_FORMAT_B8G8R8A8_SRGB,
VK_FORMAT_A8B8G8R8_UNORM_PACK32,
VK_FORMAT_A8B8G8R8_SNORM_PACK32,
VK_FORMAT_A8B8G8R8_USCALED_PACK32,
VK_FORMAT_A8B8G8R8_SSCALED_PACK32,
VK_FORMAT_A8B8G8R8_UINT_PACK32,
VK_FORMAT_A8B8G8R8_SINT_PACK32,
VK_FORMAT_A8B8G8R8_SRGB_PACK32,
VK_FORMAT_A2R10G10B10_UNORM_PACK32,
VK_FORMAT_A2R10G10B10_SNORM_PACK32,
VK_FORMAT_A2R10G10B10_USCALED_PACK32,
VK_FORMAT_A2R10G10B10_SSCALED_PACK32,
VK_FORMAT_A2R10G10B10_UINT_PACK32,
VK_FORMAT_A2R10G10B10_SINT_PACK32,
VK_FORMAT_A2B10G10R10_UNORM_PACK32,
VK_FORMAT_A2B10G10R10_SNORM_PACK32,
VK_FORMAT_A2B10G10R10_USCALED_PACK32,
VK_FORMAT_A2B10G10R10_SSCALED_PACK32,
VK_FORMAT_A2B10G10R10_UINT_PACK32,
VK_FORMAT_A2B10G10R10_SINT_PACK32,
VK_FORMAT_R16_UNORM,
VK_FORMAT_R16_SNORM,
VK_FORMAT_R16_USCALED,
VK_FORMAT_R16_SSCALED,
VK_FORMAT_R16_UINT,
VK_FORMAT_R16_SINT,
VK_FORMAT_R16_SFLOAT,
VK_FORMAT_R16G16_UNORM,
VK_FORMAT_R16G16_SNORM,
VK_FORMAT_R16G16_USCALED,
VK_FORMAT_R16G16_SSCALED,
VK_FORMAT_R16G16_UINT,
VK_FORMAT_R16G16_SINT,
VK_FORMAT_R16G16_SFLOAT,
VK_FORMAT_R16G16B16_UNORM,
VK_FORMAT_R16G16B16_SNORM,
VK_FORMAT_R16G16B16_USCALED,
VK_FORMAT_R16G16B16_SSCALED,
VK_FORMAT_R16G16B16_UINT,
VK_FORMAT_R16G16B16_SINT,
VK_FORMAT_R16G16B16_SFLOAT,
VK_FORMAT_R16G16B16A16_UNORM,
VK_FORMAT_R16G16B16A16_SNORM,
VK_FORMAT_R16G16B16A16_USCALED,
VK_FORMAT_R16G16B16A16_SSCALED,
VK_FORMAT_R16G16B16A16_UINT,
VK_FORMAT_R16G16B16A16_SINT,
VK_FORMAT_R16G16B16A16_SFLOAT,
VK_FORMAT_R32_UINT,
VK_FORMAT_R32_SINT,
VK_FORMAT_R32_SFLOAT,
VK_FORMAT_R32G32_UINT,
VK_FORMAT_R32G32_SINT,
VK_FORMAT_R32G32_SFLOAT,
VK_FORMAT_R32G32B32_UINT,
VK_FORMAT_R32G32B32_SINT,
VK_FORMAT_R32G32B32_SFLOAT,
VK_FORMAT_R32G32B32A32_UINT,
VK_FORMAT_R32G32B32A32_SINT,
VK_FORMAT_R32G32B32A32_SFLOAT,
VK_FORMAT_R64_UINT,
VK_FORMAT_R64_SINT,
VK_FORMAT_R64_SFLOAT,
VK_FORMAT_R64G64_UINT,
VK_FORMAT_R64G64_SINT,
VK_FORMAT_R64G64_SFLOAT,
VK_FORMAT_R64G64B64_UINT,
VK_FORMAT_R64G64B64_SINT,
VK_FORMAT_R64G64B64_SFLOAT,
VK_FORMAT_R64G64B64A64_UINT,
VK_FORMAT_R64G64B64A64_SINT,
VK_FORMAT_R64G64B64A64_SFLOAT,
VK_FORMAT_B10G11R11_UFLOAT_PACK32,
VK_FORMAT_E5B9G9R9_UFLOAT_PACK32,
VK_FORMAT_D16_UNORM,
VK_FORMAT_X8_D24_UNORM_PACK32,
VK_FORMAT_D32_SFLOAT,
VK_FORMAT_S8_UINT,
VK_FORMAT_D16_UNORM_S8_UINT,
VK_FORMAT_D24_UNORM_S8_UINT,
VK_FORMAT_D32_SFLOAT_S8_UINT,
VK_FORMAT_BC1_RGB_UNORM_BLOCK,
VK_FORMAT_BC1_RGB_SRGB_BLOCK,
VK_FORMAT_BC1_RGBA_UNORM_BLOCK,
VK_FORMAT_BC1_RGBA_SRGB_BLOCK,
VK_FORMAT_BC2_UNORM_BLOCK,
VK_FORMAT_BC2_SRGB_BLOCK,
VK_FORMAT_BC3_UNORM_BLOCK,
VK_FORMAT_BC3_SRGB_BLOCK,
VK_FORMAT_BC4_UNORM_BLOCK,
VK_FORMAT_BC4_SNORM_BLOCK,
VK_FORMAT_BC5_UNORM_BLOCK,
VK_FORMAT_BC5_SNORM_BLOCK,
VK_FORMAT_BC6H_UFLOAT_BLOCK,
VK_FORMAT_BC6H_SFLOAT_BLOCK,
VK_FORMAT_BC7_UNORM_BLOCK,
VK_FORMAT_BC7_SRGB_BLOCK,
VK_FORMAT_ETC2_R8G8B8_UNORM_BLOCK,
VK_FORMAT_ETC2_R8G8B8_SRGB_BLOCK,
VK_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK,
VK_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK,
VK_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK,
VK_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK,
VK_FORMAT_EAC_R11_UNORM_BLOCK,
VK_FORMAT_EAC_R11_SNORM_BLOCK,
VK_FORMAT_EAC_R11G11_UNORM_BLOCK,
VK_FORMAT_EAC_R11G11_SNORM_BLOCK,
VK_FORMAT_ASTC_4x4_UNORM_BLOCK,
VK_FORMAT_ASTC_4x4_SRGB_BLOCK,
VK_FORMAT_ASTC_5x4_UNORM_BLOCK,
VK_FORMAT_ASTC_5x4_SRGB_BLOCK,
VK_FORMAT_ASTC_5x5_UNORM_BLOCK,
VK_FORMAT_ASTC_5x5_SRGB_BLOCK,
VK_FORMAT_ASTC_6x5_UNORM_BLOCK,
VK_FORMAT_ASTC_6x5_SRGB_BLOCK,
VK_FORMAT_ASTC_6x6_UNORM_BLOCK,
VK_FORMAT_ASTC_6x6_SRGB_BLOCK,
VK_FORMAT_ASTC_8x5_UNORM_BLOCK,
VK_FORMAT_ASTC_8x5_SRGB_BLOCK,
VK_FORMAT_ASTC_8x6_UNORM_BLOCK,
VK_FORMAT_ASTC_8x6_SRGB_BLOCK,
VK_FORMAT_ASTC_8x8_UNORM_BLOCK,
VK_FORMAT_ASTC_8x8_SRGB_BLOCK,
VK_FORMAT_ASTC_10x5_UNORM_BLOCK,
VK_FORMAT_ASTC_10x5_SRGB_BLOCK,
VK_FORMAT_ASTC_10x6_UNORM_BLOCK,
VK_FORMAT_ASTC_10x6_SRGB_BLOCK,
VK_FORMAT_ASTC_10x8_UNORM_BLOCK,
VK_FORMAT_ASTC_10x8_SRGB_BLOCK,
VK_FORMAT_ASTC_10x10_UNORM_BLOCK,
VK_FORMAT_ASTC_10x10_SRGB_BLOCK,
VK_FORMAT_ASTC_12x10_UNORM_BLOCK,
VK_FORMAT_ASTC_12x10_SRGB_BLOCK,
VK_FORMAT_ASTC_12x12_UNORM_BLOCK,
VK_FORMAT_ASTC_12x12_SRGB_BLOCK,
VK_FORMAT_G8B8G8R8_422_UNORM,
VK_FORMAT_B8G8R8G8_422_UNORM,
VK_FORMAT_G8_B8_R8_3PLANE_420_UNORM,
VK_FORMAT_G8_B8R8_2PLANE_420_UNORM,
VK_FORMAT_G8_B8_R8_3PLANE_422_UNORM,
VK_FORMAT_G8_B8R8_2PLANE_422_UNORM,
VK_FORMAT_G8_B8_R8_3PLANE_444_UNORM,
VK_FORMAT_R10X6_UNORM_PACK16,
VK_FORMAT_R10X6G10X6_UNORM_2PACK16,
VK_FORMAT_R10X6G10X6B10X6A10X6_UNORM_4PACK16,
VK_FORMAT_G10X6B10X6G10X6R10X6_422_UNORM_4PACK16,
VK_FORMAT_B10X6G10X6R10X6G10X6_422_UNORM_4PACK16,
VK_FORMAT_G10X6_B10X6_R10X6_3PLANE_420_UNORM_3PACK16,
VK_FORMAT_G10X6_B10X6R10X6_2PLANE_420_UNORM_3PACK16,
VK_FORMAT_G10X6_B10X6_R10X6_3PLANE_422_UNORM_3PACK16,
VK_FORMAT_G10X6_B10X6R10X6_2PLANE_422_UNORM_3PACK16,
VK_FORMAT_G10X6_B10X6_R10X6_3PLANE_444_UNORM_3PACK16,
VK_FORMAT_R12X4_UNORM_PACK16,
VK_FORMAT_R12X4G12X4_UNORM_2PACK16,
VK_FORMAT_R12X4G12X4B12X4A12X4_UNORM_4PACK16,
VK_FORMAT_G12X4B12X4G12X4R12X4_422_UNORM_4PACK16,
VK_FORMAT_B12X4G12X4R12X4G12X4_422_UNORM_4PACK16,
VK_FORMAT_G12X4_B12X4_R12X4_3PLANE_420_UNORM_3PACK16,
VK_FORMAT_G12X4_B12X4R12X4_2PLANE_420_UNORM_3PACK16,
VK_FORMAT_G12X4_B12X4_R12X4_3PLANE_422_UNORM_3PACK16,
VK_FORMAT_G12X4_B12X4R12X4_2PLANE_422_UNORM_3PACK16,
VK_FORMAT_G12X4_B12X4_R12X4_3PLANE_444_UNORM_3PACK16,
VK_FORMAT_G16B16G16R16_422_UNORM,
VK_FORMAT_B16G16R16G16_422_UNORM,
VK_FORMAT_G16_B16_R16_3PLANE_420_UNORM,
VK_FORMAT_G16_B16R16_2PLANE_420_UNORM,
VK_FORMAT_G16_B16_R16_3PLANE_422_UNORM,
VK_FORMAT_G16_B16R16_2PLANE_422_UNORM,
VK_FORMAT_G16_B16_R16_3PLANE_444_UNORM,
};
const char *RenderingDeviceVulkan::named_formats[RenderingDevice::DATA_FORMAT_MAX] = {
"R4G4_Unorm_Pack8",
"R4G4B4A4_Unorm_Pack16",
"B4G4R4A4_Unorm_Pack16",
"R5G6B5_Unorm_Pack16",
"B5G6R5_Unorm_Pack16",
"R5G5B5A1_Unorm_Pack16",
"B5G5R5A1_Unorm_Pack16",
"A1R5G5B5_Unorm_Pack16",
"R8_Unorm",
"R8_Snorm",
"R8_Uscaled",
"R8_Sscaled",
"R8_Uint",
"R8_Sint",
"R8_Srgb",
"R8G8_Unorm",
"R8G8_Snorm",
"R8G8_Uscaled",
"R8G8_Sscaled",
"R8G8_Uint",
"R8G8_Sint",
"R8G8_Srgb",
"R8G8B8_Unorm",
"R8G8B8_Snorm",
"R8G8B8_Uscaled",
"R8G8B8_Sscaled",
"R8G8B8_Uint",
"R8G8B8_Sint",
"R8G8B8_Srgb",
"B8G8R8_Unorm",
"B8G8R8_Snorm",
"B8G8R8_Uscaled",
"B8G8R8_Sscaled",
"B8G8R8_Uint",
"B8G8R8_Sint",
"B8G8R8_Srgb",
"R8G8B8A8_Unorm",
"R8G8B8A8_Snorm",
"R8G8B8A8_Uscaled",
"R8G8B8A8_Sscaled",
"R8G8B8A8_Uint",
"R8G8B8A8_Sint",
"R8G8B8A8_Srgb",
"B8G8R8A8_Unorm",
"B8G8R8A8_Snorm",
"B8G8R8A8_Uscaled",
"B8G8R8A8_Sscaled",
"B8G8R8A8_Uint",
"B8G8R8A8_Sint",
"B8G8R8A8_Srgb",
"A8B8G8R8_Unorm_Pack32",
"A8B8G8R8_Snorm_Pack32",
"A8B8G8R8_Uscaled_Pack32",
"A8B8G8R8_Sscaled_Pack32",
"A8B8G8R8_Uint_Pack32",
"A8B8G8R8_Sint_Pack32",
"A8B8G8R8_Srgb_Pack32",
"A2R10G10B10_Unorm_Pack32",
"A2R10G10B10_Snorm_Pack32",
"A2R10G10B10_Uscaled_Pack32",
"A2R10G10B10_Sscaled_Pack32",
"A2R10G10B10_Uint_Pack32",
"A2R10G10B10_Sint_Pack32",
"A2B10G10R10_Unorm_Pack32",
"A2B10G10R10_Snorm_Pack32",
"A2B10G10R10_Uscaled_Pack32",
"A2B10G10R10_Sscaled_Pack32",
"A2B10G10R10_Uint_Pack32",
"A2B10G10R10_Sint_Pack32",
"R16_Unorm",
"R16_Snorm",
"R16_Uscaled",
"R16_Sscaled",
"R16_Uint",
"R16_Sint",
"R16_Sfloat",
"R16G16_Unorm",
"R16G16_Snorm",
"R16G16_Uscaled",
"R16G16_Sscaled",
"R16G16_Uint",
"R16G16_Sint",
"R16G16_Sfloat",
"R16G16B16_Unorm",
"R16G16B16_Snorm",
"R16G16B16_Uscaled",
"R16G16B16_Sscaled",
"R16G16B16_Uint",
"R16G16B16_Sint",
"R16G16B16_Sfloat",
"R16G16B16A16_Unorm",
"R16G16B16A16_Snorm",
"R16G16B16A16_Uscaled",
"R16G16B16A16_Sscaled",
"R16G16B16A16_Uint",
"R16G16B16A16_Sint",
"R16G16B16A16_Sfloat",
"R32_Uint",
"R32_Sint",
"R32_Sfloat",
"R32G32_Uint",
"R32G32_Sint",
"R32G32_Sfloat",
"R32G32B32_Uint",
"R32G32B32_Sint",
"R32G32B32_Sfloat",
"R32G32B32A32_Uint",
"R32G32B32A32_Sint",
"R32G32B32A32_Sfloat",
"R64_Uint",
"R64_Sint",
"R64_Sfloat",
"R64G64_Uint",
"R64G64_Sint",
"R64G64_Sfloat",
"R64G64B64_Uint",
"R64G64B64_Sint",
"R64G64B64_Sfloat",
"R64G64B64A64_Uint",
"R64G64B64A64_Sint",
"R64G64B64A64_Sfloat",
"B10G11R11_Ufloat_Pack32",
"E5B9G9R9_Ufloat_Pack32",
"D16_Unorm",
"X8_D24_Unorm_Pack32",
"D32_Sfloat",
"S8_Uint",
"D16_Unorm_S8_Uint",
"D24_Unorm_S8_Uint",
"D32_Sfloat_S8_Uint",
"Bc1_Rgb_Unorm_Block",
"Bc1_Rgb_Srgb_Block",
"Bc1_Rgba_Unorm_Block",
"Bc1_Rgba_Srgb_Block",
"Bc2_Unorm_Block",
"Bc2_Srgb_Block",
"Bc3_Unorm_Block",
"Bc3_Srgb_Block",
"Bc4_Unorm_Block",
"Bc4_Snorm_Block",
"Bc5_Unorm_Block",
"Bc5_Snorm_Block",
"Bc6H_Ufloat_Block",
"Bc6H_Sfloat_Block",
"Bc7_Unorm_Block",
"Bc7_Srgb_Block",
"Etc2_R8G8B8_Unorm_Block",
"Etc2_R8G8B8_Srgb_Block",
"Etc2_R8G8B8A1_Unorm_Block",
"Etc2_R8G8B8A1_Srgb_Block",
"Etc2_R8G8B8A8_Unorm_Block",
"Etc2_R8G8B8A8_Srgb_Block",
"Eac_R11_Unorm_Block",
"Eac_R11_Snorm_Block",
"Eac_R11G11_Unorm_Block",
"Eac_R11G11_Snorm_Block",
"Astc_4X4_Unorm_Block",
"Astc_4X4_Srgb_Block",
"Astc_5X4_Unorm_Block",
"Astc_5X4_Srgb_Block",
"Astc_5X5_Unorm_Block",
"Astc_5X5_Srgb_Block",
"Astc_6X5_Unorm_Block",
"Astc_6X5_Srgb_Block",
"Astc_6X6_Unorm_Block",
"Astc_6X6_Srgb_Block",
"Astc_8X5_Unorm_Block",
"Astc_8X5_Srgb_Block",
"Astc_8X6_Unorm_Block",
"Astc_8X6_Srgb_Block",
"Astc_8X8_Unorm_Block",
"Astc_8X8_Srgb_Block",
"Astc_10X5_Unorm_Block",
"Astc_10X5_Srgb_Block",
"Astc_10X6_Unorm_Block",
"Astc_10X6_Srgb_Block",
"Astc_10X8_Unorm_Block",
"Astc_10X8_Srgb_Block",
"Astc_10X10_Unorm_Block",
"Astc_10X10_Srgb_Block",
"Astc_12X10_Unorm_Block",
"Astc_12X10_Srgb_Block",
"Astc_12X12_Unorm_Block",
"Astc_12X12_Srgb_Block",
"G8B8G8R8_422_Unorm",
"B8G8R8G8_422_Unorm",
"G8_B8_R8_3Plane_420_Unorm",
"G8_B8R8_2Plane_420_Unorm",
"G8_B8_R8_3Plane_422_Unorm",
"G8_B8R8_2Plane_422_Unorm",
"G8_B8_R8_3Plane_444_Unorm",
"R10X6_Unorm_Pack16",
"R10X6G10X6_Unorm_2Pack16",
"R10X6G10X6B10X6A10X6_Unorm_4Pack16",
"G10X6B10X6G10X6R10X6_422_Unorm_4Pack16",
"B10X6G10X6R10X6G10X6_422_Unorm_4Pack16",
"G10X6_B10X6_R10X6_3Plane_420_Unorm_3Pack16",
"G10X6_B10X6R10X6_2Plane_420_Unorm_3Pack16",
"G10X6_B10X6_R10X6_3Plane_422_Unorm_3Pack16",
"G10X6_B10X6R10X6_2Plane_422_Unorm_3Pack16",
"G10X6_B10X6_R10X6_3Plane_444_Unorm_3Pack16",
"R12X4_Unorm_Pack16",
"R12X4G12X4_Unorm_2Pack16",
"R12X4G12X4B12X4A12X4_Unorm_4Pack16",
"G12X4B12X4G12X4R12X4_422_Unorm_4Pack16",
"B12X4G12X4R12X4G12X4_422_Unorm_4Pack16",
"G12X4_B12X4_R12X4_3Plane_420_Unorm_3Pack16",
"G12X4_B12X4R12X4_2Plane_420_Unorm_3Pack16",
"G12X4_B12X4_R12X4_3Plane_422_Unorm_3Pack16",
"G12X4_B12X4R12X4_2Plane_422_Unorm_3Pack16",
"G12X4_B12X4_R12X4_3Plane_444_Unorm_3Pack16",
"G16B16G16R16_422_Unorm",
"B16G16R16G16_422_Unorm",
"G16_B16_R16_3Plane_420_Unorm",
"G16_B16R16_2Plane_420_Unorm",
"G16_B16_R16_3Plane_422_Unorm",
"G16_B16R16_2Plane_422_Unorm",
"G16_B16_R16_3Plane_444_Unorm",
};
int RenderingDeviceVulkan::get_format_vertex_size(DataFormat p_format) {
switch (p_format) {
case DATA_FORMAT_R8_UNORM:
case DATA_FORMAT_R8_SNORM:
case DATA_FORMAT_R8_UINT:
case DATA_FORMAT_R8_SINT:
case DATA_FORMAT_R8G8_UNORM:
case DATA_FORMAT_R8G8_SNORM:
case DATA_FORMAT_R8G8_UINT:
case DATA_FORMAT_R8G8_SINT:
case DATA_FORMAT_R8G8B8_UNORM:
case DATA_FORMAT_R8G8B8_SNORM:
case DATA_FORMAT_R8G8B8_UINT:
case DATA_FORMAT_R8G8B8_SINT:
case DATA_FORMAT_B8G8R8_UNORM:
case DATA_FORMAT_B8G8R8_SNORM:
case DATA_FORMAT_B8G8R8_UINT:
case DATA_FORMAT_B8G8R8_SINT:
case DATA_FORMAT_R8G8B8A8_UNORM:
case DATA_FORMAT_R8G8B8A8_SNORM:
case DATA_FORMAT_R8G8B8A8_UINT:
case DATA_FORMAT_R8G8B8A8_SINT:
case DATA_FORMAT_B8G8R8A8_UNORM:
case DATA_FORMAT_B8G8R8A8_SNORM:
case DATA_FORMAT_B8G8R8A8_UINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_B8G8R8A8_SINT:
Refactored Mesh internals and formats. -Changed how mesh data is organized, hoping to make it more efficient on Vulkan and GLES. -Removed compression, it now always uses the most efficient format. -Added support for custom arrays (up to 8 custom formats) -Added support for 8 weights in skeleton data. -Added a simple optional versioning system for imported assets, to reimport if binary is newer -Fixes #43979 (I needed to test) WARNING: -NOT backwards compatible with previous 4.x-devel, will most likely never be, but it will force reimport scenes due to version change. -NOT backwards compatible with 3.x scenes, this will be eventually re-added. -Skeletons not working any longer, will fix in next PR.
2020-12-02 01:40:47 +00:00
case DATA_FORMAT_A2B10G10R10_UNORM_PACK32:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R16_UNORM:
case DATA_FORMAT_R16_SNORM:
case DATA_FORMAT_R16_UINT:
case DATA_FORMAT_R16_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R16_SFLOAT:
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R16G16_UNORM:
case DATA_FORMAT_R16G16_SNORM:
case DATA_FORMAT_R16G16_UINT:
case DATA_FORMAT_R16G16_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R16G16_SFLOAT:
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R16G16B16_UNORM:
case DATA_FORMAT_R16G16B16_SNORM:
case DATA_FORMAT_R16G16B16_UINT:
case DATA_FORMAT_R16G16B16_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R16G16B16_SFLOAT:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R16G16B16A16_UNORM:
case DATA_FORMAT_R16G16B16A16_SNORM:
case DATA_FORMAT_R16G16B16A16_UINT:
case DATA_FORMAT_R16G16B16A16_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R16G16B16A16_SFLOAT:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R32_UINT:
case DATA_FORMAT_R32_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R32_SFLOAT:
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R32G32_UINT:
case DATA_FORMAT_R32G32_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R32G32_SFLOAT:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R32G32B32_UINT:
case DATA_FORMAT_R32G32B32_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R32G32B32_SFLOAT:
return 12;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R32G32B32A32_UINT:
case DATA_FORMAT_R32G32B32A32_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R32G32B32A32_SFLOAT:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R64_UINT:
case DATA_FORMAT_R64_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R64_SFLOAT:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R64G64_UINT:
case DATA_FORMAT_R64G64_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R64G64_SFLOAT:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R64G64B64_UINT:
case DATA_FORMAT_R64G64B64_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R64G64B64_SFLOAT:
return 24;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R64G64B64A64_UINT:
case DATA_FORMAT_R64G64B64A64_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R64G64B64A64_SFLOAT:
return 32;
default:
return 0;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
uint32_t RenderingDeviceVulkan::get_image_format_pixel_size(DataFormat p_format) {
switch (p_format) {
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R4G4_UNORM_PACK8:
return 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R4G4B4A4_UNORM_PACK16:
case DATA_FORMAT_B4G4R4A4_UNORM_PACK16:
case DATA_FORMAT_R5G6B5_UNORM_PACK16:
case DATA_FORMAT_B5G6R5_UNORM_PACK16:
case DATA_FORMAT_R5G5B5A1_UNORM_PACK16:
case DATA_FORMAT_B5G5R5A1_UNORM_PACK16:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_A1R5G5B5_UNORM_PACK16:
return 2;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R8_UNORM:
case DATA_FORMAT_R8_SNORM:
case DATA_FORMAT_R8_USCALED:
case DATA_FORMAT_R8_SSCALED:
case DATA_FORMAT_R8_UINT:
case DATA_FORMAT_R8_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R8_SRGB:
return 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R8G8_UNORM:
case DATA_FORMAT_R8G8_SNORM:
case DATA_FORMAT_R8G8_USCALED:
case DATA_FORMAT_R8G8_SSCALED:
case DATA_FORMAT_R8G8_UINT:
case DATA_FORMAT_R8G8_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R8G8_SRGB:
return 2;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R8G8B8_UNORM:
case DATA_FORMAT_R8G8B8_SNORM:
case DATA_FORMAT_R8G8B8_USCALED:
case DATA_FORMAT_R8G8B8_SSCALED:
case DATA_FORMAT_R8G8B8_UINT:
case DATA_FORMAT_R8G8B8_SINT:
case DATA_FORMAT_R8G8B8_SRGB:
case DATA_FORMAT_B8G8R8_UNORM:
case DATA_FORMAT_B8G8R8_SNORM:
case DATA_FORMAT_B8G8R8_USCALED:
case DATA_FORMAT_B8G8R8_SSCALED:
case DATA_FORMAT_B8G8R8_UINT:
case DATA_FORMAT_B8G8R8_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_B8G8R8_SRGB:
return 3;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R8G8B8A8_UNORM:
case DATA_FORMAT_R8G8B8A8_SNORM:
case DATA_FORMAT_R8G8B8A8_USCALED:
case DATA_FORMAT_R8G8B8A8_SSCALED:
case DATA_FORMAT_R8G8B8A8_UINT:
case DATA_FORMAT_R8G8B8A8_SINT:
case DATA_FORMAT_R8G8B8A8_SRGB:
case DATA_FORMAT_B8G8R8A8_UNORM:
case DATA_FORMAT_B8G8R8A8_SNORM:
case DATA_FORMAT_B8G8R8A8_USCALED:
case DATA_FORMAT_B8G8R8A8_SSCALED:
case DATA_FORMAT_B8G8R8A8_UINT:
case DATA_FORMAT_B8G8R8A8_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_B8G8R8A8_SRGB:
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_A8B8G8R8_UNORM_PACK32:
case DATA_FORMAT_A8B8G8R8_SNORM_PACK32:
case DATA_FORMAT_A8B8G8R8_USCALED_PACK32:
case DATA_FORMAT_A8B8G8R8_SSCALED_PACK32:
case DATA_FORMAT_A8B8G8R8_UINT_PACK32:
case DATA_FORMAT_A8B8G8R8_SINT_PACK32:
case DATA_FORMAT_A8B8G8R8_SRGB_PACK32:
case DATA_FORMAT_A2R10G10B10_UNORM_PACK32:
case DATA_FORMAT_A2R10G10B10_SNORM_PACK32:
case DATA_FORMAT_A2R10G10B10_USCALED_PACK32:
case DATA_FORMAT_A2R10G10B10_SSCALED_PACK32:
case DATA_FORMAT_A2R10G10B10_UINT_PACK32:
case DATA_FORMAT_A2R10G10B10_SINT_PACK32:
case DATA_FORMAT_A2B10G10R10_UNORM_PACK32:
case DATA_FORMAT_A2B10G10R10_SNORM_PACK32:
case DATA_FORMAT_A2B10G10R10_USCALED_PACK32:
case DATA_FORMAT_A2B10G10R10_SSCALED_PACK32:
case DATA_FORMAT_A2B10G10R10_UINT_PACK32:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_A2B10G10R10_SINT_PACK32:
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R16_UNORM:
case DATA_FORMAT_R16_SNORM:
case DATA_FORMAT_R16_USCALED:
case DATA_FORMAT_R16_SSCALED:
case DATA_FORMAT_R16_UINT:
case DATA_FORMAT_R16_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R16_SFLOAT:
return 2;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R16G16_UNORM:
case DATA_FORMAT_R16G16_SNORM:
case DATA_FORMAT_R16G16_USCALED:
case DATA_FORMAT_R16G16_SSCALED:
case DATA_FORMAT_R16G16_UINT:
case DATA_FORMAT_R16G16_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R16G16_SFLOAT:
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R16G16B16_UNORM:
case DATA_FORMAT_R16G16B16_SNORM:
case DATA_FORMAT_R16G16B16_USCALED:
case DATA_FORMAT_R16G16B16_SSCALED:
case DATA_FORMAT_R16G16B16_UINT:
case DATA_FORMAT_R16G16B16_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R16G16B16_SFLOAT:
return 6;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R16G16B16A16_UNORM:
case DATA_FORMAT_R16G16B16A16_SNORM:
case DATA_FORMAT_R16G16B16A16_USCALED:
case DATA_FORMAT_R16G16B16A16_SSCALED:
case DATA_FORMAT_R16G16B16A16_UINT:
case DATA_FORMAT_R16G16B16A16_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R16G16B16A16_SFLOAT:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R32_UINT:
case DATA_FORMAT_R32_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R32_SFLOAT:
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R32G32_UINT:
case DATA_FORMAT_R32G32_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R32G32_SFLOAT:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R32G32B32_UINT:
case DATA_FORMAT_R32G32B32_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R32G32B32_SFLOAT:
return 12;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R32G32B32A32_UINT:
case DATA_FORMAT_R32G32B32A32_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R32G32B32A32_SFLOAT:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R64_UINT:
case DATA_FORMAT_R64_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R64_SFLOAT:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R64G64_UINT:
case DATA_FORMAT_R64G64_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R64G64_SFLOAT:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R64G64B64_UINT:
case DATA_FORMAT_R64G64B64_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R64G64B64_SFLOAT:
return 24;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R64G64B64A64_UINT:
case DATA_FORMAT_R64G64B64A64_SINT:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_R64G64B64A64_SFLOAT:
return 32;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_B10G11R11_UFLOAT_PACK32:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32:
return 4;
case DATA_FORMAT_D16_UNORM:
return 2;
case DATA_FORMAT_X8_D24_UNORM_PACK32:
return 4;
case DATA_FORMAT_D32_SFLOAT:
return 4;
case DATA_FORMAT_S8_UINT:
return 1;
case DATA_FORMAT_D16_UNORM_S8_UINT:
return 4;
case DATA_FORMAT_D24_UNORM_S8_UINT:
return 4;
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
case DATA_FORMAT_D32_SFLOAT_S8_UINT:
return 5; //?
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_BC1_RGB_UNORM_BLOCK:
case DATA_FORMAT_BC1_RGB_SRGB_BLOCK:
case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK:
case DATA_FORMAT_BC1_RGBA_SRGB_BLOCK:
case DATA_FORMAT_BC2_UNORM_BLOCK:
case DATA_FORMAT_BC2_SRGB_BLOCK:
case DATA_FORMAT_BC3_UNORM_BLOCK:
case DATA_FORMAT_BC3_SRGB_BLOCK:
case DATA_FORMAT_BC4_UNORM_BLOCK:
case DATA_FORMAT_BC4_SNORM_BLOCK:
case DATA_FORMAT_BC5_UNORM_BLOCK:
case DATA_FORMAT_BC5_SNORM_BLOCK:
case DATA_FORMAT_BC6H_UFLOAT_BLOCK:
case DATA_FORMAT_BC6H_SFLOAT_BLOCK:
case DATA_FORMAT_BC7_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_BC7_SRGB_BLOCK:
return 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8_SRGB_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK:
return 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_EAC_R11_UNORM_BLOCK:
case DATA_FORMAT_EAC_R11_SNORM_BLOCK:
case DATA_FORMAT_EAC_R11G11_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_EAC_R11G11_SNORM_BLOCK:
return 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK:
case DATA_FORMAT_ASTC_4x4_SRGB_BLOCK:
case DATA_FORMAT_ASTC_5x4_UNORM_BLOCK:
case DATA_FORMAT_ASTC_5x4_SRGB_BLOCK:
case DATA_FORMAT_ASTC_5x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_5x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_6x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_6x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_6x6_UNORM_BLOCK:
case DATA_FORMAT_ASTC_6x6_SRGB_BLOCK:
case DATA_FORMAT_ASTC_8x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_8x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_8x6_UNORM_BLOCK:
case DATA_FORMAT_ASTC_8x6_SRGB_BLOCK:
case DATA_FORMAT_ASTC_8x8_UNORM_BLOCK:
case DATA_FORMAT_ASTC_8x8_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x6_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x6_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x8_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x8_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x10_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x10_SRGB_BLOCK:
case DATA_FORMAT_ASTC_12x10_UNORM_BLOCK:
case DATA_FORMAT_ASTC_12x10_SRGB_BLOCK:
case DATA_FORMAT_ASTC_12x12_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_ASTC_12x12_SRGB_BLOCK:
return 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_G8B8G8R8_422_UNORM:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_B8G8R8G8_422_UNORM:
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_G8_B8_R8_3PLANE_420_UNORM:
case DATA_FORMAT_G8_B8R8_2PLANE_420_UNORM:
case DATA_FORMAT_G8_B8_R8_3PLANE_422_UNORM:
case DATA_FORMAT_G8_B8R8_2PLANE_422_UNORM:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_G8_B8_R8_3PLANE_444_UNORM:
return 4;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_R10X6_UNORM_PACK16:
case DATA_FORMAT_R10X6G10X6_UNORM_2PACK16:
case DATA_FORMAT_R10X6G10X6B10X6A10X6_UNORM_4PACK16:
case DATA_FORMAT_G10X6B10X6G10X6R10X6_422_UNORM_4PACK16:
case DATA_FORMAT_B10X6G10X6R10X6G10X6_422_UNORM_4PACK16:
case DATA_FORMAT_G10X6_B10X6_R10X6_3PLANE_420_UNORM_3PACK16:
case DATA_FORMAT_G10X6_B10X6R10X6_2PLANE_420_UNORM_3PACK16:
case DATA_FORMAT_G10X6_B10X6_R10X6_3PLANE_422_UNORM_3PACK16:
case DATA_FORMAT_G10X6_B10X6R10X6_2PLANE_422_UNORM_3PACK16:
case DATA_FORMAT_G10X6_B10X6_R10X6_3PLANE_444_UNORM_3PACK16:
case DATA_FORMAT_R12X4_UNORM_PACK16:
case DATA_FORMAT_R12X4G12X4_UNORM_2PACK16:
case DATA_FORMAT_R12X4G12X4B12X4A12X4_UNORM_4PACK16:
case DATA_FORMAT_G12X4B12X4G12X4R12X4_422_UNORM_4PACK16:
case DATA_FORMAT_B12X4G12X4R12X4G12X4_422_UNORM_4PACK16:
case DATA_FORMAT_G12X4_B12X4_R12X4_3PLANE_420_UNORM_3PACK16:
case DATA_FORMAT_G12X4_B12X4R12X4_2PLANE_420_UNORM_3PACK16:
case DATA_FORMAT_G12X4_B12X4_R12X4_3PLANE_422_UNORM_3PACK16:
case DATA_FORMAT_G12X4_B12X4R12X4_2PLANE_422_UNORM_3PACK16:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_G12X4_B12X4_R12X4_3PLANE_444_UNORM_3PACK16:
return 2;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_G16B16G16R16_422_UNORM:
case DATA_FORMAT_B16G16R16G16_422_UNORM:
case DATA_FORMAT_G16_B16_R16_3PLANE_420_UNORM:
case DATA_FORMAT_G16_B16R16_2PLANE_420_UNORM:
case DATA_FORMAT_G16_B16_R16_3PLANE_422_UNORM:
case DATA_FORMAT_G16_B16R16_2PLANE_422_UNORM:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_G16_B16_R16_3PLANE_444_UNORM:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
default: {
ERR_PRINT("Format not handled, bug");
}
}
return 1;
}
// https://www.khronos.org/registry/DataFormat/specs/1.1/dataformat.1.1.pdf
void RenderingDeviceVulkan::get_compressed_image_format_block_dimensions(DataFormat p_format, uint32_t &r_w, uint32_t &r_h) {
switch (p_format) {
case DATA_FORMAT_BC1_RGB_UNORM_BLOCK:
case DATA_FORMAT_BC1_RGB_SRGB_BLOCK:
case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK:
case DATA_FORMAT_BC1_RGBA_SRGB_BLOCK:
case DATA_FORMAT_BC2_UNORM_BLOCK:
case DATA_FORMAT_BC2_SRGB_BLOCK:
case DATA_FORMAT_BC3_UNORM_BLOCK:
case DATA_FORMAT_BC3_SRGB_BLOCK:
case DATA_FORMAT_BC4_UNORM_BLOCK:
case DATA_FORMAT_BC4_SNORM_BLOCK:
case DATA_FORMAT_BC5_UNORM_BLOCK:
case DATA_FORMAT_BC5_SNORM_BLOCK:
case DATA_FORMAT_BC6H_UFLOAT_BLOCK:
case DATA_FORMAT_BC6H_SFLOAT_BLOCK:
case DATA_FORMAT_BC7_UNORM_BLOCK:
case DATA_FORMAT_BC7_SRGB_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8_SRGB_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK:
case DATA_FORMAT_EAC_R11_UNORM_BLOCK:
case DATA_FORMAT_EAC_R11_SNORM_BLOCK:
case DATA_FORMAT_EAC_R11G11_UNORM_BLOCK:
case DATA_FORMAT_EAC_R11G11_SNORM_BLOCK:
case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK: //again, not sure about astc
case DATA_FORMAT_ASTC_4x4_SRGB_BLOCK:
case DATA_FORMAT_ASTC_5x4_UNORM_BLOCK:
case DATA_FORMAT_ASTC_5x4_SRGB_BLOCK:
case DATA_FORMAT_ASTC_5x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_5x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_6x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_6x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_6x6_UNORM_BLOCK:
case DATA_FORMAT_ASTC_6x6_SRGB_BLOCK:
case DATA_FORMAT_ASTC_8x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_8x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_8x6_UNORM_BLOCK:
case DATA_FORMAT_ASTC_8x6_SRGB_BLOCK:
case DATA_FORMAT_ASTC_8x8_UNORM_BLOCK:
case DATA_FORMAT_ASTC_8x8_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x6_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x6_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x8_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x8_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x10_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x10_SRGB_BLOCK:
case DATA_FORMAT_ASTC_12x10_UNORM_BLOCK:
case DATA_FORMAT_ASTC_12x10_SRGB_BLOCK:
case DATA_FORMAT_ASTC_12x12_UNORM_BLOCK:
case DATA_FORMAT_ASTC_12x12_SRGB_BLOCK:
r_w = 4;
r_h = 4;
return;
default: {
r_w = 1;
r_h = 1;
}
}
}
uint32_t RenderingDeviceVulkan::get_compressed_image_format_block_byte_size(DataFormat p_format) {
switch (p_format) {
case DATA_FORMAT_BC1_RGB_UNORM_BLOCK:
case DATA_FORMAT_BC1_RGB_SRGB_BLOCK:
case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_BC1_RGBA_SRGB_BLOCK:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_BC2_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_BC2_SRGB_BLOCK:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_BC3_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_BC3_SRGB_BLOCK:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_BC4_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_BC4_SNORM_BLOCK:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_BC5_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_BC5_SNORM_BLOCK:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_BC6H_UFLOAT_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_BC6H_SFLOAT_BLOCK:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_BC7_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_BC7_SRGB_BLOCK:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_ETC2_R8G8B8_SRGB_BLOCK:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_EAC_R11_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_EAC_R11_SNORM_BLOCK:
return 8;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_EAC_R11G11_UNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case DATA_FORMAT_EAC_R11G11_SNORM_BLOCK:
return 16;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK: //again, not sure about astc
case DATA_FORMAT_ASTC_4x4_SRGB_BLOCK:
case DATA_FORMAT_ASTC_5x4_UNORM_BLOCK:
case DATA_FORMAT_ASTC_5x4_SRGB_BLOCK:
case DATA_FORMAT_ASTC_5x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_5x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_6x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_6x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_6x6_UNORM_BLOCK:
case DATA_FORMAT_ASTC_6x6_SRGB_BLOCK:
case DATA_FORMAT_ASTC_8x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_8x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_8x6_UNORM_BLOCK:
case DATA_FORMAT_ASTC_8x6_SRGB_BLOCK:
case DATA_FORMAT_ASTC_8x8_UNORM_BLOCK:
case DATA_FORMAT_ASTC_8x8_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x5_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x5_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x6_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x6_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x8_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x8_SRGB_BLOCK:
case DATA_FORMAT_ASTC_10x10_UNORM_BLOCK:
case DATA_FORMAT_ASTC_10x10_SRGB_BLOCK:
case DATA_FORMAT_ASTC_12x10_UNORM_BLOCK:
case DATA_FORMAT_ASTC_12x10_SRGB_BLOCK:
case DATA_FORMAT_ASTC_12x12_UNORM_BLOCK:
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
case DATA_FORMAT_ASTC_12x12_SRGB_BLOCK:
return 8; //wrong
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
default: {
}
}
return 1;
}
uint32_t RenderingDeviceVulkan::get_compressed_image_format_pixel_rshift(DataFormat p_format) {
switch (p_format) {
case DATA_FORMAT_BC1_RGB_UNORM_BLOCK: //these formats are half byte size, so rshift is 1
case DATA_FORMAT_BC1_RGB_SRGB_BLOCK:
case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK:
case DATA_FORMAT_BC1_RGBA_SRGB_BLOCK:
case DATA_FORMAT_BC4_UNORM_BLOCK:
case DATA_FORMAT_BC4_SNORM_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8_SRGB_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK:
case DATA_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK:
case DATA_FORMAT_EAC_R11_UNORM_BLOCK:
case DATA_FORMAT_EAC_R11_SNORM_BLOCK:
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
return 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
default: {
}
}
return 0;
}
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
bool RenderingDeviceVulkan::format_has_stencil(DataFormat p_format) {
switch (p_format) {
case DATA_FORMAT_S8_UINT:
case DATA_FORMAT_D16_UNORM_S8_UINT:
case DATA_FORMAT_D24_UNORM_S8_UINT:
case DATA_FORMAT_D32_SFLOAT_S8_UINT: {
return true;
}
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
default: {
}
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
}
return false;
}
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
uint32_t RenderingDeviceVulkan::get_image_format_required_size(DataFormat p_format, uint32_t p_width, uint32_t p_height, uint32_t p_depth, uint32_t p_mipmaps, uint32_t *r_blockw, uint32_t *r_blockh, uint32_t *r_depth) {
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
ERR_FAIL_COND_V(p_mipmaps == 0, 0);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
uint32_t w = p_width;
uint32_t h = p_height;
uint32_t d = p_depth;
uint32_t size = 0;
uint32_t pixel_size = get_image_format_pixel_size(p_format);
uint32_t pixel_rshift = get_compressed_image_format_pixel_rshift(p_format);
uint32_t blockw, blockh;
get_compressed_image_format_block_dimensions(p_format, blockw, blockh);
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
for (uint32_t i = 0; i < p_mipmaps; i++) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
uint32_t bw = w % blockw != 0 ? w + (blockw - w % blockw) : w;
uint32_t bh = h % blockh != 0 ? h + (blockh - h % blockh) : h;
uint32_t s = bw * bh;
s *= pixel_size;
s >>= pixel_rshift;
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
size += s * d;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (r_blockw) {
*r_blockw = bw;
}
if (r_blockh) {
*r_blockh = bh;
}
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
if (r_depth) {
*r_depth = d;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
w = MAX(blockw, w >> 1);
h = MAX(blockh, h >> 1);
Change some math macros to constexpr Changes `MAX`, `MIN`, `ABS`, `CLAMP` and `SIGN`.
2022-03-08 14:10:48 +00:00
d = MAX(1u, d >> 1);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return size;
}
uint32_t RenderingDeviceVulkan::get_image_required_mipmaps(uint32_t p_width, uint32_t p_height, uint32_t p_depth) {
//formats and block size don't really matter here since they can all go down to 1px (even if block is larger)
Change some math macros to constexpr Changes `MAX`, `MIN`, `ABS`, `CLAMP` and `SIGN`.
2022-03-08 14:10:48 +00:00
uint32_t w = p_width;
uint32_t h = p_height;
uint32_t d = p_depth;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Change some math macros to constexpr Changes `MAX`, `MIN`, `ABS`, `CLAMP` and `SIGN`.
2022-03-08 14:10:48 +00:00
uint32_t mipmaps = 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
while (true) {
if (w == 1 && h == 1 && d == 1) {
break;
}
Change some math macros to constexpr Changes `MAX`, `MIN`, `ABS`, `CLAMP` and `SIGN`.
2022-03-08 14:10:48 +00:00
w = MAX(1u, w >> 1);
h = MAX(1u, h >> 1);
d = MAX(1u, d >> 1);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
mipmaps++;
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
return mipmaps;
}
///////////////////////
const VkCompareOp RenderingDeviceVulkan::compare_operators[RenderingDevice::COMPARE_OP_MAX] = {
VK_COMPARE_OP_NEVER,
VK_COMPARE_OP_LESS,
VK_COMPARE_OP_EQUAL,
VK_COMPARE_OP_LESS_OR_EQUAL,
VK_COMPARE_OP_GREATER,
VK_COMPARE_OP_NOT_EQUAL,
VK_COMPARE_OP_GREATER_OR_EQUAL,
VK_COMPARE_OP_ALWAYS
};
const VkStencilOp RenderingDeviceVulkan::stencil_operations[RenderingDevice::STENCIL_OP_MAX] = {
VK_STENCIL_OP_KEEP,
VK_STENCIL_OP_ZERO,
VK_STENCIL_OP_REPLACE,
VK_STENCIL_OP_INCREMENT_AND_CLAMP,
VK_STENCIL_OP_DECREMENT_AND_CLAMP,
VK_STENCIL_OP_INVERT,
VK_STENCIL_OP_INCREMENT_AND_WRAP,
VK_STENCIL_OP_DECREMENT_AND_WRAP
};
const VkSampleCountFlagBits RenderingDeviceVulkan::rasterization_sample_count[RenderingDevice::TEXTURE_SAMPLES_MAX] = {
VK_SAMPLE_COUNT_1_BIT,
VK_SAMPLE_COUNT_2_BIT,
VK_SAMPLE_COUNT_4_BIT,
VK_SAMPLE_COUNT_8_BIT,
VK_SAMPLE_COUNT_16_BIT,
VK_SAMPLE_COUNT_32_BIT,
VK_SAMPLE_COUNT_64_BIT,
};
const VkLogicOp RenderingDeviceVulkan::logic_operations[RenderingDevice::LOGIC_OP_MAX] = {
VK_LOGIC_OP_CLEAR,
VK_LOGIC_OP_AND,
VK_LOGIC_OP_AND_REVERSE,
VK_LOGIC_OP_COPY,
VK_LOGIC_OP_AND_INVERTED,
VK_LOGIC_OP_NO_OP,
VK_LOGIC_OP_XOR,
VK_LOGIC_OP_OR,
VK_LOGIC_OP_NOR,
VK_LOGIC_OP_EQUIVALENT,
VK_LOGIC_OP_INVERT,
VK_LOGIC_OP_OR_REVERSE,
VK_LOGIC_OP_COPY_INVERTED,
VK_LOGIC_OP_OR_INVERTED,
VK_LOGIC_OP_NAND,
VK_LOGIC_OP_SET
};
const VkBlendFactor RenderingDeviceVulkan::blend_factors[RenderingDevice::BLEND_FACTOR_MAX] = {
VK_BLEND_FACTOR_ZERO,
VK_BLEND_FACTOR_ONE,
VK_BLEND_FACTOR_SRC_COLOR,
VK_BLEND_FACTOR_ONE_MINUS_SRC_COLOR,
VK_BLEND_FACTOR_DST_COLOR,
VK_BLEND_FACTOR_ONE_MINUS_DST_COLOR,
VK_BLEND_FACTOR_SRC_ALPHA,
VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA,
VK_BLEND_FACTOR_DST_ALPHA,
VK_BLEND_FACTOR_ONE_MINUS_DST_ALPHA,
VK_BLEND_FACTOR_CONSTANT_COLOR,
VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_COLOR,
VK_BLEND_FACTOR_CONSTANT_ALPHA,
VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_ALPHA,
VK_BLEND_FACTOR_SRC_ALPHA_SATURATE,
VK_BLEND_FACTOR_SRC1_COLOR,
VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR,
VK_BLEND_FACTOR_SRC1_ALPHA,
VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA
};
const VkBlendOp RenderingDeviceVulkan::blend_operations[RenderingDevice::BLEND_OP_MAX] = {
VK_BLEND_OP_ADD,
VK_BLEND_OP_SUBTRACT,
VK_BLEND_OP_REVERSE_SUBTRACT,
VK_BLEND_OP_MIN,
VK_BLEND_OP_MAX
};
const VkSamplerAddressMode RenderingDeviceVulkan::address_modes[RenderingDevice::SAMPLER_REPEAT_MODE_MAX] = {
VK_SAMPLER_ADDRESS_MODE_REPEAT,
VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT,
VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE,
VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER,
VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE
};
const VkBorderColor RenderingDeviceVulkan::sampler_border_colors[RenderingDevice::SAMPLER_BORDER_COLOR_MAX] = {
VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK,
VK_BORDER_COLOR_INT_TRANSPARENT_BLACK,
VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK,
VK_BORDER_COLOR_INT_OPAQUE_BLACK,
VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE,
VK_BORDER_COLOR_INT_OPAQUE_WHITE
};
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
const VkImageType RenderingDeviceVulkan::vulkan_image_type[RenderingDevice::TEXTURE_TYPE_MAX] = {
VK_IMAGE_TYPE_1D,
VK_IMAGE_TYPE_2D,
VK_IMAGE_TYPE_3D,
VK_IMAGE_TYPE_2D,
VK_IMAGE_TYPE_1D,
VK_IMAGE_TYPE_2D,
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
VK_IMAGE_TYPE_2D
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
};
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
/***************************/
/**** BUFFER MANAGEMENT ****/
/***************************/
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
Error RenderingDeviceVulkan::_buffer_allocate(Buffer *p_buffer, uint32_t p_size, uint32_t p_usage, VmaMemoryUsage p_mem_usage, VmaAllocationCreateFlags p_mem_flags) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkBufferCreateInfo bufferInfo;
bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
bufferInfo.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
bufferInfo.flags = 0;
bufferInfo.size = p_size;
bufferInfo.usage = p_usage;
bufferInfo.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
bufferInfo.queueFamilyIndexCount = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
bufferInfo.pQueueFamilyIndices = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VmaAllocationCreateInfo allocInfo;
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
allocInfo.flags = p_mem_flags;
allocInfo.usage = p_mem_usage;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
allocInfo.requiredFlags = 0;
allocInfo.preferredFlags = 0;
allocInfo.memoryTypeBits = 0;
vk_mem_alloc: Update to upstream + Adapt approach to small objects pooling This updates VMA and instead of using the custom small pool approach from 4e6c9d3ae979f2eb0151cf581fe61d2f3194ea72, lazily creates pools for the relevant memory type indices, which doesn't require patching VMA. Also, patches already merged upstream or not needed any longer are removed.
2022-02-24 11:47:34 +00:00
allocInfo.pool = nullptr;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
allocInfo.pUserData = nullptr;
vk_mem_alloc: Update to upstream + Adapt approach to small objects pooling This updates VMA and instead of using the custom small pool approach from 4e6c9d3ae979f2eb0151cf581fe61d2f3194ea72, lazily creates pools for the relevant memory type indices, which doesn't require patching VMA. Also, patches already merged upstream or not needed any longer are removed.
2022-02-24 11:47:34 +00:00
if (p_size <= SMALL_ALLOCATION_MAX_SIZE) {
uint32_t mem_type_index = 0;
vmaFindMemoryTypeIndexForBufferInfo(allocator, &bufferInfo, &allocInfo, &mem_type_index);
allocInfo.pool = _find_or_create_small_allocs_pool(mem_type_index);
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult err = vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &p_buffer->buffer, &p_buffer->allocation, nullptr);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(err, ERR_CANT_CREATE, "Can't create buffer of size: " + itos(p_size) + ", error " + itos(err) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
p_buffer->size = p_size;
p_buffer->buffer_info.buffer = p_buffer->buffer;
p_buffer->buffer_info.offset = 0;
p_buffer->buffer_info.range = p_size;
Reimplement skeletons and blend shapes Uses compute shaders, which only once, on demand, and all in parallel.
2020-12-16 14:07:08 +00:00
p_buffer->usage = p_usage;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
buffer_memory += p_size;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
return OK;
}
Error RenderingDeviceVulkan::_buffer_free(Buffer *p_buffer) {
ERR_FAIL_COND_V(p_buffer->size == 0, ERR_INVALID_PARAMETER);
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
buffer_memory -= p_buffer->size;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vmaDestroyBuffer(allocator, p_buffer->buffer, p_buffer->allocation);
Vulkan: Use VK_NULL_HANDLE for handles, fixes 32-bit builds Some Vulkan types are defined as "non dispatchable handles" and use a different typedef on 32-bit and 64-bit systems (struct pointer on 64-bit, `uint64_t` otherwise). https://github.com/KhronosGroup/Vulkan-Headers/blob/0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5c/include/vulkan/vulkan_core.h#L59-L65 Contrarily to `NULL`, `nullptr` can't be converted to `uint64_t` so build was now failing on 32-bit after converting the codebase from using `NULL` to `nullptr`. Fixes #37620.
2020-04-06 08:17:42 +00:00
p_buffer->buffer = VK_NULL_HANDLE;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
p_buffer->allocation = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
p_buffer->size = 0;
return OK;
}
Error RenderingDeviceVulkan::_insert_staging_block() {
VkBufferCreateInfo bufferInfo;
bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
bufferInfo.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
bufferInfo.flags = 0;
bufferInfo.size = staging_buffer_block_size;
bufferInfo.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT;
bufferInfo.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
bufferInfo.queueFamilyIndexCount = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
bufferInfo.pQueueFamilyIndices = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VmaAllocationCreateInfo allocInfo;
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
allocInfo.flags = VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
allocInfo.usage = VMA_MEMORY_USAGE_AUTO_PREFER_HOST;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
allocInfo.requiredFlags = 0;
allocInfo.preferredFlags = 0;
allocInfo.memoryTypeBits = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
allocInfo.pool = nullptr;
allocInfo.pUserData = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
StagingBufferBlock block;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult err = vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &block.buffer, &block.allocation, nullptr);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(err, ERR_CANT_CREATE, "vmaCreateBuffer failed with error " + itos(err) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
block.frame_used = 0;
block.fill_amount = 0;
staging_buffer_blocks.insert(staging_buffer_current, block);
return OK;
}
Error RenderingDeviceVulkan::_staging_buffer_allocate(uint32_t p_amount, uint32_t p_required_align, uint32_t &r_alloc_offset, uint32_t &r_alloc_size, bool p_can_segment, bool p_on_draw_command_buffer) {
//determine a block to use
r_alloc_size = p_amount;
while (true) {
r_alloc_offset = 0;
//see if we can use current block
if (staging_buffer_blocks[staging_buffer_current].frame_used == frames_drawn) {
//we used this block this frame, let's see if there is still room
uint32_t write_from = staging_buffer_blocks[staging_buffer_current].fill_amount;
{
uint32_t align_remainder = write_from % p_required_align;
if (align_remainder != 0) {
write_from += p_required_align - align_remainder;
}
}
int32_t available_bytes = int32_t(staging_buffer_block_size) - int32_t(write_from);
if ((int32_t)p_amount < available_bytes) {
//all is good, we should be ok, all will fit
r_alloc_offset = write_from;
} else if (p_can_segment && available_bytes >= (int32_t)p_required_align) {
//ok all won't fit but at least we can fit a chunkie
//all is good, update what needs to be written to
r_alloc_offset = write_from;
r_alloc_size = available_bytes - (available_bytes % p_required_align);
} else {
//can't fit it into this buffer.
//will need to try next buffer
staging_buffer_current = (staging_buffer_current + 1) % staging_buffer_blocks.size();
// before doing anything, though, let's check that we didn't manage to fill all blocks
// possible in a single frame
if (staging_buffer_blocks[staging_buffer_current].frame_used == frames_drawn) {
//guess we did.. ok, let's see if we can insert a new block..
fix: 3 LGTM alerts/warnings
2021-03-23 02:52:05 +00:00
if ((uint64_t)staging_buffer_blocks.size() * staging_buffer_block_size < staging_buffer_max_size) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//we can, so we are safe
Error err = _insert_staging_block();
if (err) {
return err;
}
//claim for this frame
staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
} else {
// Ok, worst case scenario, all the staging buffers belong to this frame
// and this frame is not even done.
// If this is the main thread, it means the user is likely loading a lot of resources at once,
// otherwise, the thread should just be blocked until the next frame (currently unimplemented)
if (false) { //separate thread from render
//block_until_next_frame()
continue;
} else {
//flush EVERYTHING including setup commands. IF not immediate, also need to flush the draw commands
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
_flush(true);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//clear the whole staging buffer
for (int i = 0; i < staging_buffer_blocks.size(); i++) {
staging_buffer_blocks.write[i].frame_used = 0;
staging_buffer_blocks.write[i].fill_amount = 0;
}
//claim current
staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
}
}
} else {
//not from current frame, so continue and try again
continue;
}
}
} else if (staging_buffer_blocks[staging_buffer_current].frame_used <= frames_drawn - frame_count) {
//this is an old block, which was already processed, let's reuse
staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
staging_buffer_blocks.write[staging_buffer_current].fill_amount = 0;
Fix some issues found by cppcheck.
2022-04-05 10:40:26 +00:00
} else {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//this block may still be in use, let's not touch it unless we have to, so.. can we create a new one?
fix: 3 LGTM alerts/warnings
2021-03-23 02:52:05 +00:00
if ((uint64_t)staging_buffer_blocks.size() * staging_buffer_block_size < staging_buffer_max_size) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//we are still allowed to create a new block, so let's do that and insert it for current pos
Error err = _insert_staging_block();
if (err) {
return err;
}
//claim for this frame
staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
} else {
// oops, we are out of room and we can't create more.
// let's flush older frames.
// The logic here is that if a game is loading a lot of data from the main thread, it will need to be stalled anyway.
// If loading from a separate thread, we can block that thread until next frame when more room is made (not currently implemented, though).
if (false) {
//separate thread from render
//block_until_next_frame()
continue; //and try again
} else {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
_flush(false);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (int i = 0; i < staging_buffer_blocks.size(); i++) {
//clear all blocks but the ones from this frame
int block_idx = (i + staging_buffer_current) % staging_buffer_blocks.size();
if (staging_buffer_blocks[block_idx].frame_used == frames_drawn) {
break; //ok, we reached something from this frame, abort
}
staging_buffer_blocks.write[block_idx].frame_used = 0;
staging_buffer_blocks.write[block_idx].fill_amount = 0;
}
//claim for current frame
staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
}
}
}
//all was good, break
break;
}
staging_buffer_used = true;
return OK;
}
Error RenderingDeviceVulkan::_buffer_update(Buffer *p_buffer, size_t p_offset, const uint8_t *p_data, size_t p_data_size, bool p_use_draw_command_buffer, uint32_t p_required_align) {
//submitting may get chunked for various reasons, so convert this to a task
size_t to_submit = p_data_size;
size_t submit_from = 0;
while (to_submit > 0) {
uint32_t block_write_offset;
uint32_t block_write_amount;
Error err = _staging_buffer_allocate(MIN(to_submit, staging_buffer_block_size), p_required_align, block_write_offset, block_write_amount, p_use_draw_command_buffer);
if (err) {
return err;
}
//map staging buffer (It's CPU and coherent)
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
void *data_ptr = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
{
VkResult vkerr = vmaMapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation, &data_ptr);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(vkerr, ERR_CANT_CREATE, "vmaMapMemory failed with error " + itos(vkerr) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//copy to staging buffer
Core: Drop custom `copymem`/`zeromem` defines We've been using standard C library functions `memcpy`/`memset` for these since 2016 with 67f65f66391327b2967a20a89c3627e1dd6e84eb. There was still the possibility for third-party platform ports to override the definitions with a custom header, but this doesn't seem useful anymore.
2021-04-27 14:19:21 +00:00
memcpy(((uint8_t *)data_ptr) + block_write_offset, p_data + submit_from, block_write_amount);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//unmap
vmaUnmapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation);
//insert a command to copy this
VkBufferCopy region;
region.srcOffset = block_write_offset;
region.dstOffset = submit_from + p_offset;
region.size = block_write_amount;
vkCmdCopyBuffer(p_use_draw_command_buffer ? frames[frame].draw_command_buffer : frames[frame].setup_command_buffer, staging_buffer_blocks[staging_buffer_current].buffer, p_buffer->buffer, 1, &region);
staging_buffer_blocks.write[staging_buffer_current].fill_amount = block_write_offset + block_write_amount;
to_submit -= block_write_amount;
submit_from += block_write_amount;
}
return OK;
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
void RenderingDeviceVulkan::_memory_barrier(VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_sccess, bool p_sync_with_draw) {
VkMemoryBarrier mem_barrier;
mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
mem_barrier.pNext = nullptr;
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
mem_barrier.srcAccessMask = p_src_access;
mem_barrier.dstAccessMask = p_dst_sccess;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (p_src_stage_mask == 0 || p_dst_stage_mask == 0) {
return; //no barrier, since this is invalid
}
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkCmdPipelineBarrier(p_sync_with_draw ? frames[frame].draw_command_buffer : frames[frame].setup_command_buffer, p_src_stage_mask, p_dst_stage_mask, 0, 1, &mem_barrier, 0, nullptr, 0, nullptr);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
void RenderingDeviceVulkan::_full_barrier(bool p_sync_with_draw) {
//used for debug
_memory_barrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
VK_ACCESS_INDEX_READ_BIT |
VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
VK_ACCESS_UNIFORM_READ_BIT |
VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
VK_ACCESS_SHADER_READ_BIT |
VK_ACCESS_SHADER_WRITE_BIT |
VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
VK_ACCESS_TRANSFER_READ_BIT |
VK_ACCESS_TRANSFER_WRITE_BIT |
VK_ACCESS_HOST_READ_BIT |
VK_ACCESS_HOST_WRITE_BIT,
VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
VK_ACCESS_INDEX_READ_BIT |
VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
VK_ACCESS_UNIFORM_READ_BIT |
VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
VK_ACCESS_SHADER_READ_BIT |
VK_ACCESS_SHADER_WRITE_BIT |
VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
VK_ACCESS_TRANSFER_READ_BIT |
VK_ACCESS_TRANSFER_WRITE_BIT |
VK_ACCESS_HOST_READ_BIT |
VK_ACCESS_HOST_WRITE_BIT,
p_sync_with_draw);
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
void RenderingDeviceVulkan::_buffer_memory_barrier(VkBuffer buffer, uint64_t p_from, uint64_t p_size, VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_sccess, bool p_sync_with_draw) {
VkBufferMemoryBarrier buffer_mem_barrier;
buffer_mem_barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
buffer_mem_barrier.pNext = nullptr;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
buffer_mem_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
buffer_mem_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
buffer_mem_barrier.srcAccessMask = p_src_access;
buffer_mem_barrier.dstAccessMask = p_dst_sccess;
buffer_mem_barrier.buffer = buffer;
buffer_mem_barrier.offset = p_from;
buffer_mem_barrier.size = p_size;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkCmdPipelineBarrier(p_sync_with_draw ? frames[frame].draw_command_buffer : frames[frame].setup_command_buffer, p_src_stage_mask, p_dst_stage_mask, 0, 0, nullptr, 1, &buffer_mem_barrier, 0, nullptr);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
/*****************/
/**** TEXTURE ****/
/*****************/
Style: Set clang-format Standard to Cpp11 For us, it practically only changes the fact that `A<A<int>>` is now used instead of the C++03 compatible `A<A<int> >`. Note: clang-format 10+ changed the `Standard` arguments to fully specified `c++11`, `c++14`, etc. versions, but we can't use `c++17` now if we want to preserve compatibility with clang-format 8 and 9. `Cpp11` is still supported as deprecated alias for `Latest`.
2020-03-17 06:33:00 +00:00
RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const TextureView &p_view, const Vector<Vector<uint8_t>> &p_data) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
VkImageCreateInfo image_create_info;
image_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.flags = 0;
Display Server support
2020-03-27 16:30:18 +00:00
#ifndef _MSC_VER
#warning TODO check for support via RenderingDevice to enable on mobile when possible
#endif
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
Vulkan rendering for Android
2020-03-05 18:00:28 +00:00
#ifndef ANDROID_ENABLED
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
// vkCreateImage fails with format list on Android (VK_ERROR_OUT_OF_HOST_MEMORY)
VkImageFormatListCreateInfoKHR format_list_create_info; //keep out of the if, needed for creation
Vector<VkFormat> allowed_formats; //keep out of the if, needed for creation
#endif
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
if (p_format.shareable_formats.size()) {
image_create_info.flags |= VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT;
Vulkan rendering for Android
2020-03-05 18:00:28 +00:00
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
#ifndef ANDROID_ENABLED
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
for (int i = 0; i < p_format.shareable_formats.size(); i++) {
allowed_formats.push_back(vulkan_formats[p_format.shareable_formats[i]]);
}
format_list_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_FORMAT_LIST_CREATE_INFO_KHR;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
format_list_create_info.pNext = nullptr;
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
format_list_create_info.viewFormatCount = allowed_formats.size();
format_list_create_info.pViewFormats = allowed_formats.ptr();
image_create_info.pNext = &format_list_create_info;
ERR_FAIL_COND_V_MSG(p_format.shareable_formats.find(p_format.format) == -1, RID(),
"If supplied a list of shareable formats, the current format must be present in the list");
ERR_FAIL_COND_V_MSG(p_view.format_override != DATA_FORMAT_MAX && p_format.shareable_formats.find(p_view.format_override) == -1, RID(),
"If supplied a list of shareable formats, the current view format override must be present in the list");
Vulkan rendering for Android
2020-03-05 18:00:28 +00:00
#endif
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
}
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
if (p_format.texture_type == TEXTURE_TYPE_CUBE || p_format.texture_type == TEXTURE_TYPE_CUBE_ARRAY) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.flags |= VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT;
}
/*if (p_format.type == TEXTURE_TYPE_2D || p_format.type == TEXTURE_TYPE_2D_ARRAY) {
image_create_info.flags |= VK_IMAGE_CREATE_2D_ARRAY_COMPATIBLE_BIT;
}*/
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
ERR_FAIL_INDEX_V(p_format.texture_type, TEXTURE_TYPE_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
image_create_info.imageType = vulkan_image_type[p_format.texture_type];
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(p_format.width < 1, RID(), "Width must be equal or greater than 1 for all textures");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.format = vulkan_formats[p_format.format];
image_create_info.extent.width = p_format.width;
if (image_create_info.imageType == VK_IMAGE_TYPE_3D || image_create_info.imageType == VK_IMAGE_TYPE_2D) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(p_format.height < 1, RID(), "Height must be equal or greater than 1 for 2D and 3D textures");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.extent.height = p_format.height;
} else {
image_create_info.extent.height = 1;
}
if (image_create_info.imageType == VK_IMAGE_TYPE_3D) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(p_format.depth < 1, RID(), "Depth must be equal or greater than 1 for 3D textures");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.extent.depth = p_format.depth;
} else {
image_create_info.extent.depth = 1;
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(p_format.mipmaps < 1, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.mipLevels = p_format.mipmaps;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
if (p_format.texture_type == TEXTURE_TYPE_1D_ARRAY || p_format.texture_type == TEXTURE_TYPE_2D_ARRAY || p_format.texture_type == TEXTURE_TYPE_CUBE_ARRAY || p_format.texture_type == TEXTURE_TYPE_CUBE) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(p_format.array_layers < 1, RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Amount of layers must be equal or greater than 1 for arrays and cubemaps.");
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
ERR_FAIL_COND_V_MSG((p_format.texture_type == TEXTURE_TYPE_CUBE_ARRAY || p_format.texture_type == TEXTURE_TYPE_CUBE) && (p_format.array_layers % 6) != 0, RID(),
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
"Cubemap and cubemap array textures must provide a layer number that is multiple of 6");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.arrayLayers = p_format.array_layers;
} else {
image_create_info.arrayLayers = 1;
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_format.samples, TEXTURE_SAMPLES_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.samples = rasterization_sample_count[p_format.samples];
image_create_info.tiling = (p_format.usage_bits & TEXTURE_USAGE_CPU_READ_BIT) ? VK_IMAGE_TILING_LINEAR : VK_IMAGE_TILING_OPTIMAL;
//usage
image_create_info.usage = 0;
if (p_format.usage_bits & TEXTURE_USAGE_SAMPLING_BIT) {
image_create_info.usage |= VK_IMAGE_USAGE_SAMPLED_BIT;
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
if (p_format.usage_bits & TEXTURE_USAGE_STORAGE_BIT) {
image_create_info.usage |= VK_IMAGE_USAGE_STORAGE_BIT;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_format.usage_bits & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
image_create_info.usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
}
if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
image_create_info.usage |= VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
}
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
if (p_format.usage_bits & TEXTURE_USAGE_INPUT_ATTACHMENT_BIT) {
image_create_info.usage |= VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_format.usage_bits & TEXTURE_USAGE_CAN_UPDATE_BIT) {
image_create_info.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
}
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
if (p_format.usage_bits & TEXTURE_USAGE_CAN_COPY_FROM_BIT) {
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_create_info.usage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
if (p_format.usage_bits & TEXTURE_USAGE_CAN_COPY_TO_BIT) {
image_create_info.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
image_create_info.queueFamilyIndexCount = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_create_info.pQueueFamilyIndices = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_create_info.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
uint32_t required_mipmaps = get_image_required_mipmaps(image_create_info.extent.width, image_create_info.extent.height, image_create_info.extent.depth);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(required_mipmaps < image_create_info.mipLevels, RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Too many mipmaps requested for texture format and dimensions (" + itos(image_create_info.mipLevels) + "), maximum allowed: (" + itos(required_mipmaps) + ").");
if (p_data.size()) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!(p_format.usage_bits & TEXTURE_USAGE_CAN_UPDATE_BIT), RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Texture needs the TEXTURE_USAGE_CAN_UPDATE_BIT usage flag in order to be updated at initialization or later");
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
int expected_images = image_create_info.arrayLayers;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(p_data.size() != expected_images, RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Default supplied data for image format is of invalid length (" + itos(p_data.size()) + "), should be (" + itos(expected_images) + ").");
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
for (uint32_t i = 0; i < image_create_info.arrayLayers; i++) {
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
uint32_t required_size = get_image_format_required_size(p_format.format, image_create_info.extent.width, image_create_info.extent.height, image_create_info.extent.depth, image_create_info.mipLevels);
ERR_FAIL_COND_V_MSG((uint32_t)p_data[i].size() != required_size, RID(),
"Data for slice index " + itos(i) + " (mapped to layer " + itos(i) + ") differs in size (supplied: " + itos(p_data[i].size()) + ") than what is required by the format (" + itos(required_size) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
{
//validate that this image is supported for the intended use
VkFormatProperties properties;
vkGetPhysicalDeviceFormatProperties(context->get_physical_device(), image_create_info.format, &properties);
VkFormatFeatureFlags flags;
String format_text = "'" + String(named_formats[p_format.format]) + "'";
if (p_format.usage_bits & TEXTURE_USAGE_CPU_READ_BIT) {
flags = properties.linearTilingFeatures;
format_text += " (with CPU read bit)";
} else {
flags = properties.optimalTilingFeatures;
}
if (p_format.usage_bits & TEXTURE_USAGE_SAMPLING_BIT && !(flags & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT)) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as sampling texture.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
if (p_format.usage_bits & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT)) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as color attachment.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
Normalmapping and Specularmapping working in 2D engine Added support for Sprite, AnimatedSprite and Polygon2D (should add for tileset eventually).
2019-07-05 01:54:32 +00:00
printf("vkformat: %x\n", image_create_info.format);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as depth-stencil attachment.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
if (p_format.usage_bits & TEXTURE_USAGE_STORAGE_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT)) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as storage image.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
if (p_format.usage_bits & TEXTURE_USAGE_STORAGE_ATOMIC_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_ATOMIC_BIT)) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as atomic storage image.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
//some view validation
if (p_view.format_override != DATA_FORMAT_MAX) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_view.format_override, DATA_FORMAT_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_view.swizzle_r, TEXTURE_SWIZZLE_MAX, RID());
ERR_FAIL_INDEX_V(p_view.swizzle_g, TEXTURE_SWIZZLE_MAX, RID());
ERR_FAIL_INDEX_V(p_view.swizzle_b, TEXTURE_SWIZZLE_MAX, RID());
ERR_FAIL_INDEX_V(p_view.swizzle_a, TEXTURE_SWIZZLE_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//allocate memory
Add a separate pool for small allocations in Vulkan RD
2022-02-12 11:47:08 +00:00
uint32_t width, height;
uint32_t image_size = get_image_format_required_size(p_format.format, p_format.width, p_format.height, p_format.depth, p_format.mipmaps, &width, &height);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VmaAllocationCreateInfo allocInfo;
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
allocInfo.flags = (p_format.usage_bits & TEXTURE_USAGE_CPU_READ_BIT) ? VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT : 0;
vk_mem_alloc: Update to upstream + Adapt approach to small objects pooling This updates VMA and instead of using the custom small pool approach from 4e6c9d3ae979f2eb0151cf581fe61d2f3194ea72, lazily creates pools for the relevant memory type indices, which doesn't require patching VMA. Also, patches already merged upstream or not needed any longer are removed.
2022-02-24 11:47:34 +00:00
allocInfo.pool = nullptr;
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
allocInfo.usage = VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
allocInfo.requiredFlags = 0;
allocInfo.preferredFlags = 0;
allocInfo.memoryTypeBits = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
allocInfo.pUserData = nullptr;
vk_mem_alloc: Update to upstream + Adapt approach to small objects pooling This updates VMA and instead of using the custom small pool approach from 4e6c9d3ae979f2eb0151cf581fe61d2f3194ea72, lazily creates pools for the relevant memory type indices, which doesn't require patching VMA. Also, patches already merged upstream or not needed any longer are removed.
2022-02-24 11:47:34 +00:00
if (image_size <= SMALL_ALLOCATION_MAX_SIZE) {
uint32_t mem_type_index = 0;
vmaFindMemoryTypeIndexForImageInfo(allocator, &image_create_info, &allocInfo, &mem_type_index);
allocInfo.pool = _find_or_create_small_allocs_pool(mem_type_index);
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Texture texture;
VkResult err = vmaCreateImage(allocator, &image_create_info, &allocInfo, &texture.image, &texture.allocation, &texture.allocation_info);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(err, RID(), "vmaCreateImage failed with error " + itos(err) + ".");
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
image_memory += texture.allocation_info.size;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
texture.type = p_format.texture_type;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
texture.format = p_format.format;
texture.width = image_create_info.extent.width;
texture.height = image_create_info.extent.height;
texture.depth = image_create_info.extent.depth;
texture.layers = image_create_info.arrayLayers;
texture.mipmaps = image_create_info.mipLevels;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
texture.base_mipmap = 0;
texture.base_layer = 0;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
texture.usage_flags = p_format.usage_bits;
texture.samples = p_format.samples;
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
texture.allowed_shared_formats = p_format.shareable_formats;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
//set base layout based on usage priority
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
if (p_format.usage_bits & TEXTURE_USAGE_SAMPLING_BIT) {
//first priority, readable
texture.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
Improved validation of attachments
2019-06-07 18:20:01 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
} else if (p_format.usage_bits & TEXTURE_USAGE_STORAGE_BIT) {
//second priority, storage
texture.layout = VK_IMAGE_LAYOUT_GENERAL;
Improved validation of attachments
2019-06-07 18:20:01 +00:00
} else if (p_format.usage_bits & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
//third priority, color or depth
Improved validation of attachments
2019-06-07 18:20:01 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
texture.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
} else if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
texture.layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
} else {
texture.layout = VK_IMAGE_LAYOUT_GENERAL;
}
if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
texture.read_aspect_mask = VK_IMAGE_ASPECT_DEPTH_BIT;
texture.barrier_aspect_mask = VK_IMAGE_ASPECT_DEPTH_BIT;
if (format_has_stencil(p_format.format)) {
texture.barrier_aspect_mask |= VK_IMAGE_ASPECT_STENCIL_BIT;
Improved validation of attachments
2019-06-07 18:20:01 +00:00
}
} else {
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
texture.read_aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
texture.barrier_aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
texture.bound = false;
//create view
VkImageViewCreateInfo image_view_create_info;
image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_view_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_view_create_info.flags = 0;
image_view_create_info.image = texture.image;
static const VkImageViewType view_types[TEXTURE_TYPE_MAX] = {
VK_IMAGE_VIEW_TYPE_1D,
VK_IMAGE_VIEW_TYPE_2D,
VK_IMAGE_VIEW_TYPE_3D,
VK_IMAGE_VIEW_TYPE_CUBE,
VK_IMAGE_VIEW_TYPE_1D_ARRAY,
VK_IMAGE_VIEW_TYPE_2D_ARRAY,
VK_IMAGE_VIEW_TYPE_CUBE_ARRAY,
};
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
image_view_create_info.viewType = view_types[p_format.texture_type];
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_view.format_override == DATA_FORMAT_MAX) {
image_view_create_info.format = image_create_info.format;
} else {
image_view_create_info.format = vulkan_formats[p_view.format_override];
}
static const VkComponentSwizzle component_swizzles[TEXTURE_SWIZZLE_MAX] = {
VK_COMPONENT_SWIZZLE_IDENTITY,
VK_COMPONENT_SWIZZLE_ZERO,
VK_COMPONENT_SWIZZLE_ONE,
VK_COMPONENT_SWIZZLE_R,
VK_COMPONENT_SWIZZLE_G,
VK_COMPONENT_SWIZZLE_B,
VK_COMPONENT_SWIZZLE_A
};
image_view_create_info.components.r = component_swizzles[p_view.swizzle_r];
image_view_create_info.components.g = component_swizzles[p_view.swizzle_g];
image_view_create_info.components.b = component_swizzles[p_view.swizzle_b];
image_view_create_info.components.a = component_swizzles[p_view.swizzle_a];
image_view_create_info.subresourceRange.baseMipLevel = 0;
image_view_create_info.subresourceRange.levelCount = image_create_info.mipLevels;
image_view_create_info.subresourceRange.baseArrayLayer = 0;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
image_view_create_info.subresourceRange.layerCount = image_create_info.arrayLayers;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
} else {
image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
}
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
err = vkCreateImageView(device, &image_view_create_info, nullptr, &texture.view);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (err) {
vmaDestroyImage(allocator, texture.image, texture.allocation);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_V_MSG(RID(), "vkCreateImageView failed with error " + itos(err) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//barrier to set layout
{
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.srcAccessMask = 0;
image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_UNDEFINED;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.newLayout = texture.layout;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = texture.image;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
image_memory_barrier.subresourceRange.aspectMask = texture.barrier_aspect_mask;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = 0;
image_memory_barrier.subresourceRange.levelCount = image_create_info.mipLevels;
image_memory_barrier.subresourceRange.baseArrayLayer = 0;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
image_memory_barrier.subresourceRange.layerCount = image_create_info.arrayLayers;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
vkCmdPipelineBarrier(frames[frame].setup_command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID id = texture_owner.make_rid(texture);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_data.size()) {
for (uint32_t i = 0; i < image_create_info.arrayLayers; i++) {
Fix threaded update for textures * Previews and other stuff now works again. * Not the best solution, will have to be improved in the future usinc async queues where supported.
2021-07-07 21:06:06 +00:00
_texture_update(id, i, p_data[i], RD::BARRIER_MASK_ALL, true);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
return id;
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID RenderingDeviceVulkan::texture_create_shared(const TextureView &p_view, RID p_with_texture) {
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *src_texture = texture_owner.get_or_null(p_with_texture);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!src_texture, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
if (src_texture->owner.is_valid()) { //ahh this is a share
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
p_with_texture = src_texture->owner;
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
src_texture = texture_owner.get_or_null(src_texture->owner);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!src_texture, RID()); //this is a bug
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//create view
Texture texture = *src_texture;
VkImageViewCreateInfo image_view_create_info;
image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_view_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_view_create_info.flags = 0;
image_view_create_info.image = texture.image;
static const VkImageViewType view_types[TEXTURE_TYPE_MAX] = {
VK_IMAGE_VIEW_TYPE_1D,
VK_IMAGE_VIEW_TYPE_2D,
VK_IMAGE_VIEW_TYPE_3D,
VK_IMAGE_VIEW_TYPE_CUBE,
VK_IMAGE_VIEW_TYPE_1D_ARRAY,
VK_IMAGE_VIEW_TYPE_2D_ARRAY,
VK_IMAGE_VIEW_TYPE_CUBE_ARRAY,
};
image_view_create_info.viewType = view_types[texture.type];
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
if (p_view.format_override == DATA_FORMAT_MAX || p_view.format_override == texture.format) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_view_create_info.format = vulkan_formats[texture.format];
} else {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_view.format_override, DATA_FORMAT_MAX, RID());
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
ERR_FAIL_COND_V_MSG(texture.allowed_shared_formats.find(p_view.format_override) == -1, RID(),
"Format override is not in the list of allowed shareable formats for original texture.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_view_create_info.format = vulkan_formats[p_view.format_override];
}
static const VkComponentSwizzle component_swizzles[TEXTURE_SWIZZLE_MAX] = {
VK_COMPONENT_SWIZZLE_IDENTITY,
VK_COMPONENT_SWIZZLE_ZERO,
VK_COMPONENT_SWIZZLE_ONE,
VK_COMPONENT_SWIZZLE_R,
VK_COMPONENT_SWIZZLE_G,
VK_COMPONENT_SWIZZLE_B,
VK_COMPONENT_SWIZZLE_A
};
image_view_create_info.components.r = component_swizzles[p_view.swizzle_r];
image_view_create_info.components.g = component_swizzles[p_view.swizzle_g];
image_view_create_info.components.b = component_swizzles[p_view.swizzle_b];
image_view_create_info.components.a = component_swizzles[p_view.swizzle_a];
image_view_create_info.subresourceRange.baseMipLevel = 0;
image_view_create_info.subresourceRange.levelCount = texture.mipmaps;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
image_view_create_info.subresourceRange.layerCount = texture.layers;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_view_create_info.subresourceRange.baseArrayLayer = 0;
if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
} else {
image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
VkImageViewUsageCreateInfo usage_info;
usage_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_USAGE_CREATE_INFO;
usage_info.pNext = nullptr;
if (p_view.format_override != DATA_FORMAT_MAX) {
//need to validate usage with vulkan
usage_info.usage = 0;
if (texture.usage_flags & TEXTURE_USAGE_SAMPLING_BIT) {
usage_info.usage |= VK_IMAGE_USAGE_SAMPLED_BIT;
}
if (texture.usage_flags & TEXTURE_USAGE_STORAGE_BIT) {
if (texture_is_format_supported_for_usage(p_view.format_override, TEXTURE_USAGE_STORAGE_BIT)) {
usage_info.usage |= VK_IMAGE_USAGE_STORAGE_BIT;
}
}
if (texture.usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
if (texture_is_format_supported_for_usage(p_view.format_override, TEXTURE_USAGE_COLOR_ATTACHMENT_BIT)) {
usage_info.usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
}
}
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
if (texture.usage_flags & TEXTURE_USAGE_INPUT_ATTACHMENT_BIT) {
usage_info.usage |= VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT;
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
usage_info.usage |= VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
}
if (texture.usage_flags & TEXTURE_USAGE_CAN_UPDATE_BIT) {
usage_info.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
}
if (texture.usage_flags & TEXTURE_USAGE_CAN_COPY_FROM_BIT) {
usage_info.usage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
}
if (texture.usage_flags & TEXTURE_USAGE_CAN_COPY_TO_BIT) {
usage_info.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
}
image_view_create_info.pNext = &usage_info;
}
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult err = vkCreateImageView(device, &image_view_create_info, nullptr, &texture.view);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateImageView failed with error " + itos(err) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
texture.owner = p_with_texture;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID id = texture_owner.make_rid(texture);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_add_dependency(id, p_with_texture);
return id;
}
Implementing OpenXR driver
2021-12-14 01:44:12 +00:00
RID RenderingDeviceVulkan::texture_create_from_extension(TextureType p_type, DataFormat p_format, TextureSamples p_samples, uint64_t p_flags, uint64_t p_image, uint64_t p_width, uint64_t p_height, uint64_t p_depth, uint64_t p_layers) {
_THREAD_SAFE_METHOD_
// This method creates a texture object using a VkImage created by an extension, module or other external source (OpenXR uses this).
VkImage image = (VkImage)p_image;
Texture texture;
texture.image = image;
// if we leave texture.allocation as a nullptr, would that be enough to detect we don't "own" the image?
// also leave texture.allocation_info alone
// we'll set texture.view later on
texture.type = p_type;
texture.format = p_format;
texture.samples = p_samples;
texture.width = p_width;
texture.height = p_height;
texture.depth = p_depth;
texture.layers = p_layers;
texture.mipmaps = 0; // maybe make this settable too?
texture.usage_flags = p_flags;
texture.base_mipmap = 0;
texture.base_layer = 0;
texture.allowed_shared_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_UNORM);
texture.allowed_shared_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_SRGB);
// Do we need to do something with texture.layout ?
if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
texture.read_aspect_mask = VK_IMAGE_ASPECT_DEPTH_BIT;
texture.barrier_aspect_mask = VK_IMAGE_ASPECT_DEPTH_BIT;
// if (format_has_stencil(p_format.format)) {
// texture.barrier_aspect_mask |= VK_IMAGE_ASPECT_STENCIL_BIT;
// }
} else {
texture.read_aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
texture.barrier_aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
}
// Create a view for us to use
VkImageViewCreateInfo image_view_create_info;
image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
image_view_create_info.pNext = nullptr;
image_view_create_info.flags = 0;
image_view_create_info.image = texture.image;
static const VkImageViewType view_types[TEXTURE_TYPE_MAX] = {
VK_IMAGE_VIEW_TYPE_1D,
VK_IMAGE_VIEW_TYPE_2D,
VK_IMAGE_VIEW_TYPE_3D,
VK_IMAGE_VIEW_TYPE_CUBE,
VK_IMAGE_VIEW_TYPE_1D_ARRAY,
VK_IMAGE_VIEW_TYPE_2D_ARRAY,
VK_IMAGE_VIEW_TYPE_CUBE_ARRAY,
};
image_view_create_info.viewType = view_types[texture.type];
image_view_create_info.format = vulkan_formats[texture.format];
static const VkComponentSwizzle component_swizzles[TEXTURE_SWIZZLE_MAX] = {
VK_COMPONENT_SWIZZLE_IDENTITY,
VK_COMPONENT_SWIZZLE_ZERO,
VK_COMPONENT_SWIZZLE_ONE,
VK_COMPONENT_SWIZZLE_R,
VK_COMPONENT_SWIZZLE_G,
VK_COMPONENT_SWIZZLE_B,
VK_COMPONENT_SWIZZLE_A
};
// hardcode for now, maybe make this settable from outside..
image_view_create_info.components.r = component_swizzles[TEXTURE_SWIZZLE_R];
image_view_create_info.components.g = component_swizzles[TEXTURE_SWIZZLE_G];
image_view_create_info.components.b = component_swizzles[TEXTURE_SWIZZLE_B];
image_view_create_info.components.a = component_swizzles[TEXTURE_SWIZZLE_A];
image_view_create_info.subresourceRange.baseMipLevel = 0;
image_view_create_info.subresourceRange.levelCount = texture.mipmaps;
image_view_create_info.subresourceRange.baseArrayLayer = 0;
image_view_create_info.subresourceRange.layerCount = texture.layers;
if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
} else {
image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
}
VkResult err = vkCreateImageView(device, &image_view_create_info, nullptr, &texture.view);
if (err) {
// vmaDestroyImage(allocator, texture.image, texture.allocation);
ERR_FAIL_V_MSG(RID(), "vkCreateImageView failed with error " + itos(err) + ".");
}
//barrier to set layout
{
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
image_memory_barrier.pNext = nullptr;
image_memory_barrier.srcAccessMask = 0;
image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_UNDEFINED;
image_memory_barrier.newLayout = texture.layout;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = texture.image;
image_memory_barrier.subresourceRange.aspectMask = texture.barrier_aspect_mask;
image_memory_barrier.subresourceRange.baseMipLevel = 0;
image_memory_barrier.subresourceRange.levelCount = texture.mipmaps;
image_memory_barrier.subresourceRange.baseArrayLayer = 0;
image_memory_barrier.subresourceRange.layerCount = texture.layers;
vkCmdPipelineBarrier(frames[frame].setup_command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
}
RID id = texture_owner.make_rid(texture);
return id;
}
Added SSIL post processing effect
2021-08-03 07:07:32 +00:00
RID RenderingDeviceVulkan::texture_create_shared_from_slice(const TextureView &p_view, RID p_with_texture, uint32_t p_layer, uint32_t p_mipmap, uint32_t p_mipmaps, TextureSliceType p_slice_type) {
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *src_texture = texture_owner.get_or_null(p_with_texture);
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
ERR_FAIL_COND_V(!src_texture, RID());
if (src_texture->owner.is_valid()) { //ahh this is a share
p_with_texture = src_texture->owner;
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
src_texture = texture_owner.get_or_null(src_texture->owner);
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
ERR_FAIL_COND_V(!src_texture, RID()); //this is a bug
}
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V_MSG(p_slice_type == TEXTURE_SLICE_CUBEMAP && (src_texture->type != TEXTURE_TYPE_CUBE && src_texture->type != TEXTURE_TYPE_CUBE_ARRAY), RID(),
"Can only create a cubemap slice from a cubemap or cubemap array mipmap");
GIProbes working.
2019-10-03 20:39:08 +00:00
ERR_FAIL_COND_V_MSG(p_slice_type == TEXTURE_SLICE_3D && src_texture->type != TEXTURE_TYPE_3D, RID(),
"Can only create a 3D slice from a 3D texture");
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
ERR_FAIL_COND_V_MSG(p_slice_type == TEXTURE_SLICE_2D_ARRAY && (src_texture->type != TEXTURE_TYPE_2D_ARRAY), RID(),
"Can only create an array slice from a 2D array mipmap");
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
//create view
Fix various GCC compilation warnings after Vulkan merge Part of #36132.
2020-02-13 14:53:32 +00:00
ERR_FAIL_UNSIGNED_INDEX_V(p_mipmap, src_texture->mipmaps, RID());
Added SSIL post processing effect
2021-08-03 07:07:32 +00:00
ERR_FAIL_COND_V(p_mipmap + p_mipmaps > src_texture->mipmaps, RID());
Fix various GCC compilation warnings after Vulkan merge Part of #36132.
2020-02-13 14:53:32 +00:00
ERR_FAIL_UNSIGNED_INDEX_V(p_layer, src_texture->layers, RID());
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
int slice_layers = 1;
if (p_slice_type == TEXTURE_SLICE_2D_ARRAY) {
ERR_FAIL_COND_V_MSG(p_layer != 0, RID(), "layer must be 0 when obtaining a 2D array mipmap slice");
slice_layers = src_texture->layers;
} else if (p_slice_type == TEXTURE_SLICE_CUBEMAP) {
slice_layers = 6;
}
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
Texture texture = *src_texture;
get_image_format_required_size(texture.format, texture.width, texture.height, texture.depth, p_mipmap + 1, &texture.width, &texture.height);
Added SSIL post processing effect
2021-08-03 07:07:32 +00:00
texture.mipmaps = p_mipmaps;
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
texture.layers = slice_layers;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
texture.base_mipmap = p_mipmap;
texture.base_layer = p_layer;
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
VkImageViewCreateInfo image_view_create_info;
image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_view_create_info.pNext = nullptr;
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
image_view_create_info.flags = 0;
image_view_create_info.image = texture.image;
static const VkImageViewType view_types[TEXTURE_TYPE_MAX] = {
VK_IMAGE_VIEW_TYPE_1D,
VK_IMAGE_VIEW_TYPE_2D,
VK_IMAGE_VIEW_TYPE_2D,
VK_IMAGE_VIEW_TYPE_2D,
VK_IMAGE_VIEW_TYPE_1D,
VK_IMAGE_VIEW_TYPE_2D,
VK_IMAGE_VIEW_TYPE_2D,
};
SSAO cleanup and fixes
2020-12-31 18:33:38 +00:00
image_view_create_info.viewType = view_types[texture.type];
if (p_slice_type == TEXTURE_SLICE_CUBEMAP) {
image_view_create_info.viewType = VK_IMAGE_VIEW_TYPE_CUBE;
} else if (p_slice_type == TEXTURE_SLICE_3D) {
image_view_create_info.viewType = VK_IMAGE_VIEW_TYPE_3D;
} else if (p_slice_type == TEXTURE_SLICE_2D_ARRAY) {
image_view_create_info.viewType = VK_IMAGE_VIEW_TYPE_2D_ARRAY;
}
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
if (p_view.format_override == DATA_FORMAT_MAX || p_view.format_override == texture.format) {
image_view_create_info.format = vulkan_formats[texture.format];
} else {
ERR_FAIL_INDEX_V(p_view.format_override, DATA_FORMAT_MAX, RID());
ERR_FAIL_COND_V_MSG(texture.allowed_shared_formats.find(p_view.format_override) == -1, RID(),
"Format override is not in the list of allowed shareable formats for original texture.");
image_view_create_info.format = vulkan_formats[p_view.format_override];
}
static const VkComponentSwizzle component_swizzles[TEXTURE_SWIZZLE_MAX] = {
VK_COMPONENT_SWIZZLE_IDENTITY,
VK_COMPONENT_SWIZZLE_ZERO,
VK_COMPONENT_SWIZZLE_ONE,
VK_COMPONENT_SWIZZLE_R,
VK_COMPONENT_SWIZZLE_G,
VK_COMPONENT_SWIZZLE_B,
VK_COMPONENT_SWIZZLE_A
};
image_view_create_info.components.r = component_swizzles[p_view.swizzle_r];
image_view_create_info.components.g = component_swizzles[p_view.swizzle_g];
image_view_create_info.components.b = component_swizzles[p_view.swizzle_b];
image_view_create_info.components.a = component_swizzles[p_view.swizzle_a];
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
if (p_slice_type == TEXTURE_SLICE_CUBEMAP) {
ERR_FAIL_COND_V_MSG(p_layer >= src_texture->layers, RID(),
"Specified layer is invalid for cubemap");
ERR_FAIL_COND_V_MSG((p_layer % 6) != 0, RID(),
"Specified layer must be a multiple of 6.");
}
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
image_view_create_info.subresourceRange.baseMipLevel = p_mipmap;
Added SSIL post processing effect
2021-08-03 07:07:32 +00:00
image_view_create_info.subresourceRange.levelCount = p_mipmaps;
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
image_view_create_info.subresourceRange.layerCount = slice_layers;
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
image_view_create_info.subresourceRange.baseArrayLayer = p_layer;
if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
} else {
image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
}
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult err = vkCreateImageView(device, &image_view_create_info, nullptr, &texture.view);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateImageView failed with error " + itos(err) + ".");
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
texture.owner = p_with_texture;
RID id = texture_owner.make_rid(texture);
_add_dependency(id, p_with_texture);
return id;
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
Error RenderingDeviceVulkan::texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, uint32_t p_post_barrier) {
Fix threaded update for textures * Previews and other stuff now works again. * Not the best solution, will have to be improved in the future usinc async queues where supported.
2021-07-07 21:06:06 +00:00
return _texture_update(p_texture, p_layer, p_data, p_post_barrier, false);
}
Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, uint32_t p_post_barrier, bool p_use_setup_queue) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Fix threaded update for textures * Previews and other stuff now works again. * Not the best solution, will have to be improved in the future usinc async queues where supported.
2021-07-07 21:06:06 +00:00
ERR_FAIL_COND_V_MSG((draw_list || compute_list) && !p_use_setup_queue, ERR_INVALID_PARAMETER,
"Updating textures is forbidden during creation of a draw or compute list");
Bugfixes and ability to better specify filter and repeat modes everywhere. Removes antialiased flag for draw_* methods.
2019-06-25 01:24:07 +00:00
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(p_texture);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_V(!texture, ERR_INVALID_PARAMETER);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
if (texture->owner != RID()) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
p_texture = texture->owner;
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
texture = texture_owner.get_or_null(texture->owner);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_V(!texture, ERR_BUG); //this is a bug
}
ERR_FAIL_COND_V_MSG(texture->bound, ERR_CANT_ACQUIRE_RESOURCE,
"Texture can't be updated while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_CAN_UPDATE_BIT), ERR_INVALID_PARAMETER,
"Texture requires the TEXTURE_USAGE_CAN_UPDATE_BIT in order to be updatable.");
uint32_t layer_count = texture->layers;
if (texture->type == TEXTURE_TYPE_CUBE || texture->type == TEXTURE_TYPE_CUBE_ARRAY) {
layer_count *= 6;
}
ERR_FAIL_COND_V(p_layer >= layer_count, ERR_INVALID_PARAMETER);
uint32_t width, height;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
uint32_t image_size = get_image_format_required_size(texture->format, texture->width, texture->height, texture->depth, texture->mipmaps, &width, &height);
uint32_t required_size = image_size;
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
uint32_t required_align = get_compressed_image_format_block_byte_size(texture->format);
if (required_align == 1) {
required_align = get_image_format_pixel_size(texture->format);
}
if ((required_align % 4) != 0) { //alignment rules are really strange
required_align *= 4;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_V_MSG(required_size != (uint32_t)p_data.size(), ERR_INVALID_PARAMETER,
"Required size for texture update (" + itos(required_size) + ") does not match data supplied size (" + itos(p_data.size()) + ").");
uint32_t region_size = texture_upload_region_size_px;
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint8_t *r = p_data.ptr();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Fix threaded update for textures * Previews and other stuff now works again. * Not the best solution, will have to be improved in the future usinc async queues where supported.
2021-07-07 21:06:06 +00:00
VkCommandBuffer command_buffer = p_use_setup_queue ? frames[frame].setup_command_buffer : frames[frame].draw_command_buffer;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//barrier to transfer
{
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.srcAccessMask = 0;
image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.oldLayout = texture->layout;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = texture->image;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
image_memory_barrier.subresourceRange.aspectMask = texture->barrier_aspect_mask;
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = 0;
image_memory_barrier.subresourceRange.levelCount = texture->mipmaps;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.subresourceRange.baseArrayLayer = p_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
uint32_t mipmap_offset = 0;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
uint32_t logic_width = texture->width;
uint32_t logic_height = texture->height;
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
for (uint32_t mm_i = 0; mm_i < texture->mipmaps; mm_i++) {
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
uint32_t depth;
uint32_t image_total = get_image_format_required_size(texture->format, texture->width, texture->height, texture->depth, mm_i + 1, &width, &height, &depth);
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint8_t *read_ptr_mipmap = r + mipmap_offset;
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
image_size = image_total - mipmap_offset;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
for (uint32_t z = 0; z < depth; z++) { //for 3D textures, depth may be > 0
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
const uint8_t *read_ptr = read_ptr_mipmap + image_size * z / depth;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
for (uint32_t x = 0; x < width; x += region_size) {
for (uint32_t y = 0; y < height; y += region_size) {
uint32_t region_w = MIN(region_size, width - x);
uint32_t region_h = MIN(region_size, height - y);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
uint32_t region_logic_w = MIN(region_size, logic_width - x);
uint32_t region_logic_h = MIN(region_size, logic_height - y);
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
uint32_t pixel_size = get_image_format_pixel_size(texture->format);
uint32_t to_allocate = region_w * region_h * pixel_size;
to_allocate >>= get_compressed_image_format_pixel_rshift(texture->format);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
uint32_t alloc_offset, alloc_size;
Fix threaded update for textures * Previews and other stuff now works again. * Not the best solution, will have to be improved in the future usinc async queues where supported.
2021-07-07 21:06:06 +00:00
Error err = _staging_buffer_allocate(to_allocate, required_align, alloc_offset, alloc_size, false, !p_use_setup_queue);
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
ERR_FAIL_COND_V(err, ERR_CANT_CREATE);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
uint8_t *write_ptr;
{ //map
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
void *data_ptr = nullptr;
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
VkResult vkerr = vmaMapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation, &data_ptr);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(vkerr, ERR_CANT_CREATE, "vmaMapMemory failed with error " + itos(vkerr) + ".");
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
write_ptr = (uint8_t *)data_ptr;
write_ptr += alloc_offset;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
uint32_t block_w, block_h;
get_compressed_image_format_block_dimensions(texture->format, block_w, block_h);
ERR_FAIL_COND_V(region_w % block_w, ERR_BUG);
ERR_FAIL_COND_V(region_h % block_h, ERR_BUG);
if (block_w != 1 || block_h != 1) {
//compressed image (blocks)
//must copy a block region
uint32_t block_size = get_compressed_image_format_block_byte_size(texture->format);
//re-create current variables in blocky format
uint32_t xb = x / block_w;
uint32_t yb = y / block_h;
uint32_t wb = width / block_w;
//uint32_t hb = height / block_h;
uint32_t region_wb = region_w / block_w;
uint32_t region_hb = region_h / block_h;
for (uint32_t xr = 0; xr < region_wb; xr++) {
for (uint32_t yr = 0; yr < region_hb; yr++) {
uint32_t src_offset = ((yr + yb) * wb + xr + xb) * block_size;
uint32_t dst_offset = (yr * region_wb + xr) * block_size;
//copy block
for (uint32_t i = 0; i < block_size; i++) {
write_ptr[dst_offset + i] = read_ptr[src_offset + i];
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
} else {
//regular image (pixels)
//must copy a pixel region
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
for (uint32_t xr = 0; xr < region_w; xr++) {
for (uint32_t yr = 0; yr < region_h; yr++) {
uint32_t src_offset = ((yr + y) * width + xr + x) * pixel_size;
uint32_t dst_offset = (yr * region_w + xr) * pixel_size;
//copy block
for (uint32_t i = 0; i < pixel_size; i++) {
write_ptr[dst_offset + i] = read_ptr[src_offset + i];
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
}
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
{ //unmap
vmaUnmapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation);
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
VkBufferImageCopy buffer_image_copy;
buffer_image_copy.bufferOffset = alloc_offset;
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
buffer_image_copy.bufferRowLength = 0; //tightly packed
buffer_image_copy.bufferImageHeight = 0; //tightly packed
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
buffer_image_copy.imageSubresource.aspectMask = texture->read_aspect_mask;
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
buffer_image_copy.imageSubresource.mipLevel = mm_i;
buffer_image_copy.imageSubresource.baseArrayLayer = p_layer;
buffer_image_copy.imageSubresource.layerCount = 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
buffer_image_copy.imageOffset.x = x;
buffer_image_copy.imageOffset.y = y;
buffer_image_copy.imageOffset.z = z;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
buffer_image_copy.imageExtent.width = region_logic_w;
buffer_image_copy.imageExtent.height = region_logic_h;
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
buffer_image_copy.imageExtent.depth = 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
vkCmdCopyBufferToImage(command_buffer, staging_buffer_blocks[staging_buffer_current].buffer, texture->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &buffer_image_copy);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
staging_buffer_blocks.write[staging_buffer_current].fill_amount += alloc_size;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
mipmap_offset = image_total;
Change some math macros to constexpr Changes `MAX`, `MIN`, `ABS`, `CLAMP` and `SIGN`.
2022-03-08 14:10:48 +00:00
logic_width = MAX(1u, logic_width >> 1);
logic_height = MAX(1u, logic_height >> 1);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//barrier to restore layout
{
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
uint32_t barrier_flags = 0;
uint32_t access_flags = 0;
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT;
}
if (p_post_barrier & BARRIER_MASK_RASTER) {
barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT;
}
if (p_post_barrier & BARRIER_MASK_TRANSFER) {
barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (barrier_flags == 0) {
barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
image_memory_barrier.dstAccessMask = access_flags;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.newLayout = texture->layout;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = texture->image;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
image_memory_barrier.subresourceRange.aspectMask = texture->barrier_aspect_mask;
Changed my mind on Vulkan image API, images should now include mipmaps This should make it easier to obtain the data directly from an Image
2019-06-11 20:21:39 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = 0;
image_memory_barrier.subresourceRange.levelCount = texture->mipmaps;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_memory_barrier.subresourceRange.baseArrayLayer = p_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Fix misuse of Vulkan enum value
2021-12-07 18:33:56 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (texture->used_in_frame != frames_drawn) {
texture->used_in_raster = false;
texture->used_in_compute = false;
texture->used_in_frame = frames_drawn;
}
texture->used_in_transfer = true;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
return OK;
}
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
Vector<uint8_t> RenderingDeviceVulkan::_texture_get_data_from_image(Texture *tex, VkImage p_image, VmaAllocation p_allocation, uint32_t p_layer, bool p_2d) {
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
uint32_t width, height, depth;
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
uint32_t image_size = get_image_format_required_size(tex->format, tex->width, tex->height, p_2d ? 1 : tex->depth, tex->mipmaps, &width, &height, &depth);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
Vector<uint8_t> image_data;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_data.resize(image_size);
void *img_mem;
vmaMapMemory(allocator, p_allocation, &img_mem);
uint32_t blockw, blockh;
get_compressed_image_format_block_dimensions(tex->format, blockw, blockh);
uint32_t block_size = get_compressed_image_format_block_byte_size(tex->format);
uint32_t pixel_size = get_image_format_pixel_size(tex->format);
{
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
uint8_t *w = image_data.ptrw();
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
uint32_t mipmap_offset = 0;
for (uint32_t mm_i = 0; mm_i < tex->mipmaps; mm_i++) {
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
uint32_t image_total = get_image_format_required_size(tex->format, tex->width, tex->height, p_2d ? 1 : tex->depth, mm_i + 1, &width, &height, &depth);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
uint8_t *write_ptr_mipmap = w + mipmap_offset;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_size = image_total - mipmap_offset;
VkImageSubresource image_sub_resorce;
image_sub_resorce.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
image_sub_resorce.arrayLayer = p_layer;
image_sub_resorce.mipLevel = mm_i;
VkSubresourceLayout layout;
vkGetImageSubresourceLayout(device, p_image, &image_sub_resorce, &layout);
for (uint32_t z = 0; z < depth; z++) {
uint8_t *write_ptr = write_ptr_mipmap + z * image_size / depth;
const uint8_t *slice_read_ptr = ((uint8_t *)img_mem) + layout.offset + z * layout.depthPitch;
if (block_size > 1) {
//compressed
uint32_t line_width = (block_size * (width / blockw));
for (uint32_t y = 0; y < height / blockh; y++) {
const uint8_t *rptr = slice_read_ptr + y * layout.rowPitch;
uint8_t *wptr = write_ptr + y * line_width;
Core: Drop custom `copymem`/`zeromem` defines We've been using standard C library functions `memcpy`/`memset` for these since 2016 with 67f65f66391327b2967a20a89c3627e1dd6e84eb. There was still the possibility for third-party platform ports to override the definitions with a custom header, but this doesn't seem useful anymore.
2021-04-27 14:19:21 +00:00
memcpy(wptr, rptr, line_width);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
}
} else {
//uncompressed
for (uint32_t y = 0; y < height; y++) {
const uint8_t *rptr = slice_read_ptr + y * layout.rowPitch;
uint8_t *wptr = write_ptr + y * pixel_size * width;
Core: Drop custom `copymem`/`zeromem` defines We've been using standard C library functions `memcpy`/`memset` for these since 2016 with 67f65f66391327b2967a20a89c3627e1dd6e84eb. There was still the possibility for third-party platform ports to override the definitions with a custom header, but this doesn't seem useful anymore.
2021-04-27 14:19:21 +00:00
memcpy(wptr, rptr, (uint64_t)pixel_size * width);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
}
}
}
mipmap_offset = image_total;
}
}
vmaUnmapMemory(allocator, p_allocation);
return image_data;
}
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
Vector<uint8_t> RenderingDeviceVulkan::texture_get_data(RID p_texture, uint32_t p_layer) {
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *tex = texture_owner.get_or_null(p_texture);
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
ERR_FAIL_COND_V(!tex, Vector<uint8_t>());
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
ERR_FAIL_COND_V_MSG(tex->bound, Vector<uint8_t>(),
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
"Texture can't be retrieved while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
ERR_FAIL_COND_V_MSG(!(tex->usage_flags & TEXTURE_USAGE_CAN_COPY_FROM_BIT), Vector<uint8_t>(),
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
"Texture requires the TEXTURE_USAGE_CAN_COPY_FROM_BIT in order to be retrieved.");
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
uint32_t layer_count = tex->layers;
if (tex->type == TEXTURE_TYPE_CUBE || tex->type == TEXTURE_TYPE_CUBE_ARRAY) {
layer_count *= 6;
}
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
ERR_FAIL_COND_V(p_layer >= layer_count, Vector<uint8_t>());
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
if (tex->usage_flags & TEXTURE_USAGE_CPU_READ_BIT) {
//does not need anything fancy, map and read.
return _texture_get_data_from_image(tex, tex->image, tex->allocation, p_layer);
} else {
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
//compute total image size
uint32_t width, height, depth;
uint32_t buffer_size = get_image_format_required_size(tex->format, tex->width, tex->height, tex->depth, tex->mipmaps, &width, &height, &depth);
//allocate buffer
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
VkCommandBuffer command_buffer = frames[frame].draw_command_buffer; //makes more sense to retrieve
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
Buffer tmp_buffer;
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
_buffer_allocate(&tmp_buffer, buffer_size, VK_BUFFER_USAGE_TRANSFER_DST_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_HOST, VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
{ //Source image barrier
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_memory_barrier.srcAccessMask = 0;
image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.oldLayout = tex->layout;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = tex->image;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
image_memory_barrier.subresourceRange.aspectMask = tex->barrier_aspect_mask;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = 0;
image_memory_barrier.subresourceRange.levelCount = tex->mipmaps;
image_memory_barrier.subresourceRange.baseArrayLayer = p_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
}
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
uint32_t computed_w = tex->width;
uint32_t computed_h = tex->height;
uint32_t computed_d = tex->depth;
uint32_t prev_size = 0;
uint32_t offset = 0;
for (uint32_t i = 0; i < tex->mipmaps; i++) {
VkBufferImageCopy buffer_image_copy;
uint32_t image_size = get_image_format_required_size(tex->format, tex->width, tex->height, tex->depth, i + 1);
uint32_t size = image_size - prev_size;
prev_size = image_size;
buffer_image_copy.bufferOffset = offset;
buffer_image_copy.bufferImageHeight = 0;
buffer_image_copy.bufferRowLength = 0;
buffer_image_copy.imageSubresource.aspectMask = tex->read_aspect_mask;
buffer_image_copy.imageSubresource.baseArrayLayer = p_layer;
buffer_image_copy.imageSubresource.layerCount = 1;
buffer_image_copy.imageSubresource.mipLevel = i;
buffer_image_copy.imageOffset.x = 0;
buffer_image_copy.imageOffset.y = 0;
buffer_image_copy.imageOffset.z = 0;
buffer_image_copy.imageExtent.width = computed_w;
buffer_image_copy.imageExtent.height = computed_h;
buffer_image_copy.imageExtent.depth = computed_d;
vkCmdCopyImageToBuffer(command_buffer, tex->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, tmp_buffer.buffer, 1, &buffer_image_copy);
Change some math macros to constexpr Changes `MAX`, `MIN`, `ABS`, `CLAMP` and `SIGN`.
2022-03-08 14:10:48 +00:00
computed_w = MAX(1u, computed_w >> 1);
computed_h = MAX(1u, computed_h >> 1);
computed_d = MAX(1u, computed_d >> 1);
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
offset += size;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
}
{ //restore src
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
if (tex->usage_flags & TEXTURE_USAGE_STORAGE_BIT) {
image_memory_barrier.dstAccessMask |= VK_ACCESS_SHADER_WRITE_BIT;
}
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.newLayout = tex->layout;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = tex->image;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
image_memory_barrier.subresourceRange.aspectMask = tex->barrier_aspect_mask;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = 0;
image_memory_barrier.subresourceRange.levelCount = tex->mipmaps;
image_memory_barrier.subresourceRange.baseArrayLayer = p_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
}
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
_flush(true);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
void *buffer_mem;
VkResult vkerr = vmaMapMemory(allocator, tmp_buffer.allocation, &buffer_mem);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(vkerr, Vector<uint8_t>(), "vmaMapMemory failed with error " + itos(vkerr) + ".");
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
Vector<uint8_t> buffer_data;
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
{
buffer_data.resize(buffer_size);
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
uint8_t *w = buffer_data.ptrw();
Core: Drop custom `copymem`/`zeromem` defines We've been using standard C library functions `memcpy`/`memset` for these since 2016 with 67f65f66391327b2967a20a89c3627e1dd6e84eb. There was still the possibility for third-party platform ports to override the definitions with a custom header, but this doesn't seem useful anymore.
2021-04-27 14:19:21 +00:00
memcpy(w, buffer_mem, buffer_size);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
}
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
vmaUnmapMemory(allocator, tmp_buffer.allocation);
_buffer_free(&tmp_buffer);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
return buffer_data;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
}
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
bool RenderingDeviceVulkan::texture_is_shared(RID p_texture) {
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *tex = texture_owner.get_or_null(p_texture);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
ERR_FAIL_COND_V(!tex, false);
return tex->owner.is_valid();
}
bool RenderingDeviceVulkan::texture_is_valid(RID p_texture) {
return texture_owner.owns(p_texture);
}
Optionally scale 3D render content
2021-08-19 01:52:06 +00:00
Size2i RenderingDeviceVulkan::texture_size(RID p_texture) {
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *tex = texture_owner.get_or_null(p_texture);
Optionally scale 3D render content
2021-08-19 01:52:06 +00:00
ERR_FAIL_COND_V(!tex, Size2i());
return Size2i(tex->width, tex->height);
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, const Vector3 &p_from, const Vector3 &p_to, const Vector3 &p_size, uint32_t p_src_mipmap, uint32_t p_dst_mipmap, uint32_t p_src_layer, uint32_t p_dst_layer, uint32_t p_post_barrier) {
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *src_tex = texture_owner.get_or_null(p_from_texture);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
ERR_FAIL_COND_V(!src_tex, ERR_INVALID_PARAMETER);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
ERR_FAIL_COND_V_MSG(src_tex->bound, ERR_INVALID_PARAMETER,
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
"Source texture can't be copied while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
ERR_FAIL_COND_V_MSG(!(src_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_FROM_BIT), ERR_INVALID_PARAMETER,
"Source texture requires the TEXTURE_USAGE_CAN_COPY_FROM_BIT in order to be retrieved.");
uint32_t src_layer_count = src_tex->layers;
uint32_t src_width, src_height, src_depth;
get_image_format_required_size(src_tex->format, src_tex->width, src_tex->height, src_tex->depth, p_src_mipmap + 1, &src_width, &src_height, &src_depth);
if (src_tex->type == TEXTURE_TYPE_CUBE || src_tex->type == TEXTURE_TYPE_CUBE_ARRAY) {
src_layer_count *= 6;
}
ERR_FAIL_COND_V(p_from.x < 0 || p_from.x + p_size.x > src_width, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_from.y < 0 || p_from.y + p_size.y > src_height, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_from.z < 0 || p_from.z + p_size.z > src_depth, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_src_mipmap >= src_tex->mipmaps, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_src_layer >= src_layer_count, ERR_INVALID_PARAMETER);
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *dst_tex = texture_owner.get_or_null(p_to_texture);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
ERR_FAIL_COND_V(!dst_tex, ERR_INVALID_PARAMETER);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
ERR_FAIL_COND_V_MSG(dst_tex->bound, ERR_INVALID_PARAMETER,
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
"Destination texture can't be copied while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
ERR_FAIL_COND_V_MSG(!(dst_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_TO_BIT), ERR_INVALID_PARAMETER,
"Destination texture requires the TEXTURE_USAGE_CAN_COPY_TO_BIT in order to be retrieved.");
uint32_t dst_layer_count = dst_tex->layers;
uint32_t dst_width, dst_height, dst_depth;
get_image_format_required_size(dst_tex->format, dst_tex->width, dst_tex->height, dst_tex->depth, p_dst_mipmap + 1, &dst_width, &dst_height, &dst_depth);
if (dst_tex->type == TEXTURE_TYPE_CUBE || dst_tex->type == TEXTURE_TYPE_CUBE_ARRAY) {
dst_layer_count *= 6;
}
ERR_FAIL_COND_V(p_to.x < 0 || p_to.x + p_size.x > dst_width, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_to.y < 0 || p_to.y + p_size.y > dst_height, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_to.z < 0 || p_to.z + p_size.z > dst_depth, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_dst_mipmap >= dst_tex->mipmaps, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_dst_layer >= dst_layer_count, ERR_INVALID_PARAMETER);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ERR_FAIL_COND_V_MSG(src_tex->read_aspect_mask != dst_tex->read_aspect_mask, ERR_INVALID_PARAMETER,
"Source and destination texture must be of the same type (color or depth).");
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
{
//PRE Copy the image
{ //Source
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.srcAccessMask = 0;
image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.oldLayout = src_tex->layout;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = src_tex->image;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
image_memory_barrier.subresourceRange.aspectMask = src_tex->barrier_aspect_mask;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = p_src_mipmap;
image_memory_barrier.subresourceRange.levelCount = 1;
image_memory_barrier.subresourceRange.baseArrayLayer = p_src_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
}
{ //Dest
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.srcAccessMask = 0;
image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.oldLayout = dst_tex->layout;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = dst_tex->image;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
image_memory_barrier.subresourceRange.aspectMask = dst_tex->read_aspect_mask;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = p_dst_mipmap;
image_memory_barrier.subresourceRange.levelCount = 1;
image_memory_barrier.subresourceRange.baseArrayLayer = p_dst_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
}
//COPY
{
VkImageCopy image_copy_region;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
image_copy_region.srcSubresource.aspectMask = src_tex->read_aspect_mask;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_copy_region.srcSubresource.baseArrayLayer = p_src_layer;
image_copy_region.srcSubresource.layerCount = 1;
image_copy_region.srcSubresource.mipLevel = p_src_mipmap;
image_copy_region.srcOffset.x = p_from.x;
image_copy_region.srcOffset.y = p_from.y;
image_copy_region.srcOffset.z = p_from.z;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
image_copy_region.dstSubresource.aspectMask = dst_tex->read_aspect_mask;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_copy_region.dstSubresource.baseArrayLayer = p_dst_layer;
image_copy_region.dstSubresource.layerCount = 1;
image_copy_region.dstSubresource.mipLevel = p_dst_mipmap;
image_copy_region.dstOffset.x = p_to.x;
image_copy_region.dstOffset.y = p_to.y;
image_copy_region.dstOffset.z = p_to.z;
image_copy_region.extent.width = p_size.x;
image_copy_region.extent.height = p_size.y;
image_copy_region.extent.depth = p_size.z;
vkCmdCopyImage(command_buffer, src_tex->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_tex->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &image_copy_region);
}
// RESTORE LAYOUT for SRC and DST
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
uint32_t barrier_flags = 0;
uint32_t access_flags = 0;
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (p_post_barrier & BARRIER_MASK_RASTER) {
barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (p_post_barrier & BARRIER_MASK_TRANSFER) {
barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (barrier_flags == 0) {
barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
}
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
{ //restore src
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
image_memory_barrier.dstAccessMask = access_flags;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.newLayout = src_tex->layout;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = src_tex->image;
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
image_memory_barrier.subresourceRange.aspectMask = src_tex->barrier_aspect_mask;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = p_src_mipmap;
image_memory_barrier.subresourceRange.levelCount = src_tex->mipmaps;
image_memory_barrier.subresourceRange.baseArrayLayer = p_src_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
vkCmdPipelineBarrier(command_buffer, VK_ACCESS_TRANSFER_WRITE_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
}
{ //make dst readable
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
image_memory_barrier.dstAccessMask = access_flags;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.newLayout = dst_tex->layout;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = dst_tex->image;
image_memory_barrier.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
image_memory_barrier.subresourceRange.baseMipLevel = p_src_mipmap;
image_memory_barrier.subresourceRange.levelCount = 1;
image_memory_barrier.subresourceRange.baseArrayLayer = p_src_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Implement MSAA
2020-04-12 18:33:57 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Implement MSAA
2020-04-12 18:33:57 +00:00
}
}
return OK;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID p_to_texture, uint32_t p_post_barrier) {
Implement MSAA
2020-04-12 18:33:57 +00:00
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *src_tex = texture_owner.get_or_null(p_from_texture);
Implement MSAA
2020-04-12 18:33:57 +00:00
ERR_FAIL_COND_V(!src_tex, ERR_INVALID_PARAMETER);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
ERR_FAIL_COND_V_MSG(src_tex->bound, ERR_INVALID_PARAMETER,
Implement MSAA
2020-04-12 18:33:57 +00:00
"Source texture can't be copied while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
ERR_FAIL_COND_V_MSG(!(src_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_FROM_BIT), ERR_INVALID_PARAMETER,
"Source texture requires the TEXTURE_USAGE_CAN_COPY_FROM_BIT in order to be retrieved.");
ERR_FAIL_COND_V_MSG(src_tex->type != TEXTURE_TYPE_2D, ERR_INVALID_PARAMETER, "Source texture must be 2D (or a slice of a 3D/Cube texture)");
ERR_FAIL_COND_V_MSG(src_tex->samples == TEXTURE_SAMPLES_1, ERR_INVALID_PARAMETER, "Source texture must be multisampled.");
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *dst_tex = texture_owner.get_or_null(p_to_texture);
Implement MSAA
2020-04-12 18:33:57 +00:00
ERR_FAIL_COND_V(!dst_tex, ERR_INVALID_PARAMETER);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
ERR_FAIL_COND_V_MSG(dst_tex->bound, ERR_INVALID_PARAMETER,
Implement MSAA
2020-04-12 18:33:57 +00:00
"Destination texture can't be copied while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
ERR_FAIL_COND_V_MSG(!(dst_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_TO_BIT), ERR_INVALID_PARAMETER,
"Destination texture requires the TEXTURE_USAGE_CAN_COPY_TO_BIT in order to be retrieved.");
ERR_FAIL_COND_V_MSG(dst_tex->type != TEXTURE_TYPE_2D, ERR_INVALID_PARAMETER, "Destination texture must be 2D (or a slice of a 3D/Cube texture).");
ERR_FAIL_COND_V_MSG(dst_tex->samples != TEXTURE_SAMPLES_1, ERR_INVALID_PARAMETER, "Destination texture must not be multisampled.");
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
ERR_FAIL_COND_V_MSG(src_tex->format != dst_tex->format, ERR_INVALID_PARAMETER, "Source and Destination textures must be the same format.");
ERR_FAIL_COND_V_MSG(src_tex->width != dst_tex->width && src_tex->height != dst_tex->height && src_tex->depth != dst_tex->depth, ERR_INVALID_PARAMETER, "Source and Destination textures must have the same dimensions.");
Implement MSAA
2020-04-12 18:33:57 +00:00
ERR_FAIL_COND_V_MSG(src_tex->read_aspect_mask != dst_tex->read_aspect_mask, ERR_INVALID_PARAMETER,
"Source and destination texture must be of the same type (color or depth).");
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;
Implement MSAA
2020-04-12 18:33:57 +00:00
{
//PRE Copy the image
{ //Source
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
image_memory_barrier.pNext = nullptr;
image_memory_barrier.srcAccessMask = 0;
image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
image_memory_barrier.oldLayout = src_tex->layout;
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = src_tex->image;
image_memory_barrier.subresourceRange.aspectMask = src_tex->barrier_aspect_mask;
image_memory_barrier.subresourceRange.baseMipLevel = src_tex->base_mipmap;
image_memory_barrier.subresourceRange.levelCount = 1;
image_memory_barrier.subresourceRange.baseArrayLayer = src_tex->base_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Implement MSAA
2020-04-12 18:33:57 +00:00
}
{ //Dest
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
image_memory_barrier.pNext = nullptr;
image_memory_barrier.srcAccessMask = 0;
image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
image_memory_barrier.oldLayout = dst_tex->layout;
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = dst_tex->image;
image_memory_barrier.subresourceRange.aspectMask = dst_tex->read_aspect_mask;
image_memory_barrier.subresourceRange.baseMipLevel = dst_tex->base_mipmap;
image_memory_barrier.subresourceRange.levelCount = 1;
image_memory_barrier.subresourceRange.baseArrayLayer = dst_tex->base_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
}
//COPY
{
VkImageResolve image_copy_region;
image_copy_region.srcSubresource.aspectMask = src_tex->read_aspect_mask;
image_copy_region.srcSubresource.baseArrayLayer = src_tex->base_layer;
image_copy_region.srcSubresource.layerCount = 1;
image_copy_region.srcSubresource.mipLevel = src_tex->base_mipmap;
image_copy_region.srcOffset.x = 0;
image_copy_region.srcOffset.y = 0;
image_copy_region.srcOffset.z = 0;
image_copy_region.dstSubresource.aspectMask = dst_tex->read_aspect_mask;
image_copy_region.dstSubresource.baseArrayLayer = dst_tex->base_layer;
image_copy_region.dstSubresource.layerCount = 1;
image_copy_region.dstSubresource.mipLevel = dst_tex->base_mipmap;
image_copy_region.dstOffset.x = 0;
image_copy_region.dstOffset.y = 0;
image_copy_region.dstOffset.z = 0;
image_copy_region.extent.width = src_tex->width;
image_copy_region.extent.height = src_tex->height;
image_copy_region.extent.depth = src_tex->depth;
vkCmdResolveImage(command_buffer, src_tex->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_tex->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &image_copy_region);
}
// RESTORE LAYOUT for SRC and DST
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
uint32_t barrier_flags = 0;
uint32_t access_flags = 0;
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (p_post_barrier & BARRIER_MASK_RASTER) {
barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (p_post_barrier & BARRIER_MASK_TRANSFER) {
barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (barrier_flags == 0) {
barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
}
Implement MSAA
2020-04-12 18:33:57 +00:00
{ //restore src
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
image_memory_barrier.pNext = nullptr;
image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
image_memory_barrier.dstAccessMask = access_flags;
Implement MSAA
2020-04-12 18:33:57 +00:00
image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
image_memory_barrier.newLayout = src_tex->layout;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = src_tex->image;
image_memory_barrier.subresourceRange.aspectMask = src_tex->barrier_aspect_mask;
image_memory_barrier.subresourceRange.baseMipLevel = src_tex->base_mipmap;
image_memory_barrier.subresourceRange.levelCount = 1;
image_memory_barrier.subresourceRange.baseArrayLayer = src_tex->base_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
vkCmdPipelineBarrier(command_buffer, VK_ACCESS_TRANSFER_WRITE_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Implement MSAA
2020-04-12 18:33:57 +00:00
}
{ //make dst readable
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
image_memory_barrier.pNext = nullptr;
image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
image_memory_barrier.dstAccessMask = access_flags;
Implement MSAA
2020-04-12 18:33:57 +00:00
image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
image_memory_barrier.newLayout = dst_tex->layout;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = dst_tex->image;
image_memory_barrier.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
image_memory_barrier.subresourceRange.baseMipLevel = dst_tex->base_mipmap;
image_memory_barrier.subresourceRange.levelCount = 1;
image_memory_barrier.subresourceRange.baseArrayLayer = dst_tex->base_layer;
image_memory_barrier.subresourceRange.layerCount = 1;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
}
}
return OK;
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
Error RenderingDeviceVulkan::texture_clear(RID p_texture, const Color &p_color, uint32_t p_base_mipmap, uint32_t p_mipmaps, uint32_t p_base_layer, uint32_t p_layers, uint32_t p_post_barrier) {
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *src_tex = texture_owner.get_or_null(p_texture);
GIProbes working.
2019-10-03 20:39:08 +00:00
ERR_FAIL_COND_V(!src_tex, ERR_INVALID_PARAMETER);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
ERR_FAIL_COND_V_MSG(src_tex->bound, ERR_INVALID_PARAMETER,
GIProbes working.
2019-10-03 20:39:08 +00:00
"Source texture can't be cleared while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
ERR_FAIL_COND_V(p_layers == 0, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_mipmaps == 0, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V_MSG(!(src_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_TO_BIT), ERR_INVALID_PARAMETER,
"Source texture requires the TEXTURE_USAGE_CAN_COPY_TO_BIT in order to be cleared.");
uint32_t src_layer_count = src_tex->layers;
if (src_tex->type == TEXTURE_TYPE_CUBE || src_tex->type == TEXTURE_TYPE_CUBE_ARRAY) {
src_layer_count *= 6;
}
Implement decals Also implemented decal atlas, so projectors and other stuff can be added. Sidenote: Had to make RID hashable, so some unrelated includes changed in order to include it in hashfuncs.h
2020-04-14 03:05:21 +00:00
ERR_FAIL_COND_V(p_base_mipmap + p_mipmaps > src_tex->mipmaps, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(p_base_layer + p_layers > src_layer_count, ERR_INVALID_PARAMETER);
GIProbes working.
2019-10-03 20:39:08 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;
GIProbes working.
2019-10-03 20:39:08 +00:00
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
VkImageLayout clear_layout = (src_tex->layout == VK_IMAGE_LAYOUT_GENERAL) ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
GIProbes working.
2019-10-03 20:39:08 +00:00
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
// NOTE: Perhaps the valid stages/accesses for a given owner should be a property of the owner. (Here and places like _get_buffer_from_owner)
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
const VkPipelineStageFlags valid_texture_stages = VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
constexpr VkAccessFlags read_access = VK_ACCESS_SHADER_READ_BIT;
constexpr VkAccessFlags read_write_access = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
const VkAccessFlags valid_texture_access = (src_tex->usage_flags & TEXTURE_USAGE_STORAGE_BIT) ? read_write_access : read_access;
{ // Barrier from previous access with optional layout change (see clear_layout logic above)
GIProbes working.
2019-10-03 20:39:08 +00:00
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
image_memory_barrier.srcAccessMask = valid_texture_access;
GIProbes working.
2019-10-03 20:39:08 +00:00
image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
image_memory_barrier.oldLayout = src_tex->layout;
Implement CanvasGroup and CanvasItem clipping -Allows merging several 2D objects into a single draw operation -Use current node to clip children nodes -Further fixes to Vulkan barriers -Changed font texture generation to white, fixes dark eges when blurred -Other small misc fixes to backbuffer code.
2020-10-28 18:34:27 +00:00
image_memory_barrier.newLayout = clear_layout;
GIProbes working.
2019-10-03 20:39:08 +00:00
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = src_tex->image;
image_memory_barrier.subresourceRange.aspectMask = src_tex->read_aspect_mask;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = src_tex->base_mipmap + p_base_mipmap;
GIProbes working.
2019-10-03 20:39:08 +00:00
image_memory_barrier.subresourceRange.levelCount = p_mipmaps;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseArrayLayer = src_tex->base_layer + p_base_layer;
GIProbes working.
2019-10-03 20:39:08 +00:00
image_memory_barrier.subresourceRange.layerCount = p_layers;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
vkCmdPipelineBarrier(command_buffer, valid_texture_stages, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
GIProbes working.
2019-10-03 20:39:08 +00:00
}
VkClearColorValue clear_color;
clear_color.float32[0] = p_color.r;
clear_color.float32[1] = p_color.g;
clear_color.float32[2] = p_color.b;
clear_color.float32[3] = p_color.a;
VkImageSubresourceRange range;
range.aspectMask = src_tex->read_aspect_mask;
Implement decals Also implemented decal atlas, so projectors and other stuff can be added. Sidenote: Had to make RID hashable, so some unrelated includes changed in order to include it in hashfuncs.h
2020-04-14 03:05:21 +00:00
range.baseArrayLayer = src_tex->base_layer + p_base_layer;
GIProbes working.
2019-10-03 20:39:08 +00:00
range.layerCount = p_layers;
Implement decals Also implemented decal atlas, so projectors and other stuff can be added. Sidenote: Had to make RID hashable, so some unrelated includes changed in order to include it in hashfuncs.h
2020-04-14 03:05:21 +00:00
range.baseMipLevel = src_tex->base_mipmap + p_base_mipmap;
GIProbes working.
2019-10-03 20:39:08 +00:00
range.levelCount = p_mipmaps;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
vkCmdClearColorImage(command_buffer, src_tex->image, clear_layout, &clear_color, 1, &range);
GIProbes working.
2019-10-03 20:39:08 +00:00
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
{ // Barrier to post clear accesses (changing back the layout if needed)
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
uint32_t barrier_flags = 0;
uint32_t access_flags = 0;
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (p_post_barrier & BARRIER_MASK_RASTER) {
barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (p_post_barrier & BARRIER_MASK_TRANSFER) {
barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (barrier_flags == 0) {
barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
GIProbes working.
2019-10-03 20:39:08 +00:00
image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
image_memory_barrier.dstAccessMask = access_flags;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
image_memory_barrier.oldLayout = clear_layout;
GIProbes working.
2019-10-03 20:39:08 +00:00
image_memory_barrier.newLayout = src_tex->layout;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = src_tex->image;
image_memory_barrier.subresourceRange.aspectMask = src_tex->read_aspect_mask;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = src_tex->base_mipmap + p_base_mipmap;
GIProbes working.
2019-10-03 20:39:08 +00:00
image_memory_barrier.subresourceRange.levelCount = p_mipmaps;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseArrayLayer = src_tex->base_layer + p_base_layer;
GIProbes working.
2019-10-03 20:39:08 +00:00
image_memory_barrier.subresourceRange.layerCount = p_layers;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
GIProbes working.
2019-10-03 20:39:08 +00:00
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (src_tex->used_in_frame != frames_drawn) {
src_tex->used_in_raster = false;
src_tex->used_in_compute = false;
src_tex->used_in_frame = frames_drawn;
}
src_tex->used_in_transfer = true;
GIProbes working.
2019-10-03 20:39:08 +00:00
return OK;
}
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
bool RenderingDeviceVulkan::texture_is_format_supported_for_usage(DataFormat p_format, uint32_t p_usage) const {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_INDEX_V(p_format, DATA_FORMAT_MAX, false);
_THREAD_SAFE_METHOD_
//validate that this image is supported for the intended use
VkFormatProperties properties;
vkGetPhysicalDeviceFormatProperties(context->get_physical_device(), vulkan_formats[p_format], &properties);
VkFormatFeatureFlags flags;
if (p_usage & TEXTURE_USAGE_CPU_READ_BIT) {
flags = properties.linearTilingFeatures;
} else {
flags = properties.optimalTilingFeatures;
}
if (p_usage & TEXTURE_USAGE_SAMPLING_BIT && !(flags & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT)) {
return false;
}
if (p_usage & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT)) {
return false;
}
if (p_usage & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
return false;
}
if (p_usage & TEXTURE_USAGE_STORAGE_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT)) {
return false;
}
if (p_usage & TEXTURE_USAGE_STORAGE_ATOMIC_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_ATOMIC_BIT)) {
return false;
}
return true;
}
/********************/
/**** ATTACHMENT ****/
/********************/
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentFormat> &p_attachments, const Vector<FramebufferPass> &p_passes, InitialAction p_initial_action, FinalAction p_final_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, uint32_t p_view_count, Vector<TextureSamples> *r_samples) {
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
// Set up dependencies from/to external equivalent to the default (implicit) one, and then amend them
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
const VkPipelineStageFlags default_access_mask = VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
clang-format: Disable alignment of operands, too unreliable Sets `AlignOperands` to `DontAlign`. `clang-format` developers seem to mostly care about space-based indentation and every other version of clang-format breaks the bad mismatch of tabs and spaces that it seems to use for operand alignment. So it's better without, so that it respects our two-tabs `ContinuationIndentWidth`.
2021-10-28 13:19:35 +00:00
VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT; // From Section 7.1 of Vulkan API Spec v1.1.148
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
VkPipelineStageFlags reading_stages = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_TRANSFER_BIT;
VkSubpassDependency dependencies[2] = { { VK_SUBPASS_EXTERNAL, 0, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, 0, default_access_mask, 0 },
{ 0, VK_SUBPASS_EXTERNAL, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, default_access_mask, 0, 0 } };
VkSubpassDependency &dependency_from_external = dependencies[0];
VkSubpassDependency &dependency_to_external = dependencies[1];
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
LocalVector<int32_t> attachment_last_pass;
attachment_last_pass.resize(p_attachments.size());
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
Vector<VkAttachmentDescription> attachments;
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
Vector<int> attachment_remap;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
for (int i = 0; i < p_attachments.size(); i++) {
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (p_attachments[i].usage_flags == AttachmentFormat::UNUSED_ATTACHMENT) {
attachment_remap.push_back(VK_ATTACHMENT_UNUSED);
continue;
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_INDEX_V(p_attachments[i].format, DATA_FORMAT_MAX, VK_NULL_HANDLE);
ERR_FAIL_INDEX_V(p_attachments[i].samples, TEXTURE_SAMPLES_MAX, VK_NULL_HANDLE);
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
ERR_FAIL_COND_V_MSG(!(p_attachments[i].usage_flags & (TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | TEXTURE_USAGE_INPUT_ATTACHMENT_BIT)),
VK_NULL_HANDLE, "Texture format for index (" + itos(i) + ") requires an attachment (color, depth, input or stencil) bit set.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Vulkan: Fix struct init for VkClearAttachment The changes from #38835 were not sufficient to fix #38829, as VkClearAttachment still had uninitialized member structs in its VkClearColor member struct. The struct rabbit hole goes deep and trying to do validation as done in #38829 doesn't appear realistic.
2020-07-22 14:31:17 +00:00
VkAttachmentDescription description = {};
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
description.flags = 0;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
description.format = vulkan_formats[p_attachments[i].format];
description.samples = rasterization_sample_count[p_attachments[i].samples];
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
bool is_sampled = p_attachments[i].usage_flags & TEXTURE_USAGE_SAMPLING_BIT;
bool is_storage = p_attachments[i].usage_flags & TEXTURE_USAGE_STORAGE_BIT;
bool is_depth = p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
// For each UNDEFINED, assume the prior use was a *read*, as we'd be discarding the output of a write
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
// Also, each UNDEFINED will do an immediate layout transition (write), s.t. we must ensure execution synchronization vs.
Style: Cleanup uses of double spaces between words Or after punctuation. Tried to leave third-party stuff alone, unless it has been heavily modified for Godot.
2021-06-07 08:17:32 +00:00
// the read. If this is a performance issue, one could track the actual last accessor of each resource, adding only that
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
// stage
Fixes and optimizations to mobile renderer * Only apply final actions to attachments used in the last pass. * Fixes to draw list final action (was using continue instead of read/drop). * Profiling regions inside draw lists now properly throw errors. * Ability to enable gpu profile printing from project settings. (used to debug).
2021-08-11 00:50:28 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
switch (is_depth ? p_initial_depth_action : p_initial_action) {
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
case INITIAL_ACTION_CLEAR_REGION:
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case INITIAL_ACTION_CLEAR: {
Fixes and optimizations to mobile renderer * Only apply final actions to attachments used in the last pass. * Fixes to draw list final action (was using continue instead of read/drop). * Profiling regions inside draw lists now properly throw errors. * Ability to enable gpu profile printing from project settings. (used to debug).
2021-08-11 00:50:28 +00:00
if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
description.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
description.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL);
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
dependency_from_external.srcStageMask |= reading_stages;
} else {
description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there
dependency_from_external.srcStageMask |= reading_stages;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
GIProbes working.
2019-10-03 20:39:08 +00:00
case INITIAL_ACTION_KEEP: {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL);
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
dependency_from_external.srcStageMask |= reading_stages;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
dependency_from_external.srcStageMask |= reading_stages;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
} break;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
case INITIAL_ACTION_DROP: {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
dependency_from_external.srcStageMask |= reading_stages;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
} else {
description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
dependency_from_external.srcStageMask |= reading_stages;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
}
} break;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
case INITIAL_ACTION_CLEAR_REGION_CONTINUE:
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case INITIAL_ACTION_CONTINUE: {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
description.initialLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
description.initialLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
} else {
description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
dependency_from_external.srcStageMask |= reading_stages;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
} break;
default: {
ERR_FAIL_V(VK_NULL_HANDLE); //should never reach here
}
}
Fixes and optimizations to mobile renderer * Only apply final actions to attachments used in the last pass. * Fixes to draw list final action (was using continue instead of read/drop). * Profiling regions inside draw lists now properly throw errors. * Ability to enable gpu profile printing from project settings. (used to debug).
2021-08-11 00:50:28 +00:00
bool used_last = false;
{
int last_pass = p_passes.size() - 1;
if (is_depth) {
//likely missing depth resolve?
if (p_passes[last_pass].depth_attachment == i) {
used_last = true;
}
} else {
if (p_passes[last_pass].resolve_attachments.size()) {
//if using resolve attachments, check resolve attachments
for (int j = 0; j < p_passes[last_pass].resolve_attachments.size(); j++) {
if (p_passes[last_pass].resolve_attachments[j] == i) {
used_last = true;
break;
}
}
} else {
for (int j = 0; j < p_passes[last_pass].color_attachments.size(); j++) {
if (p_passes[last_pass].color_attachments[j] == i) {
used_last = true;
break;
}
}
}
}
if (!used_last) {
for (int j = 0; j < p_passes[last_pass].preserve_attachments.size(); j++) {
if (p_passes[last_pass].preserve_attachments[j] == i) {
used_last = true;
break;
}
}
}
}
FinalAction final_action = p_final_action;
FinalAction final_depth_action = p_final_depth_action;
if (!used_last) {
if (is_depth) {
final_depth_action = FINAL_ACTION_DISCARD;
} else {
final_action = FINAL_ACTION_DISCARD;
}
}
switch (is_depth ? final_depth_action : final_action) {
GIProbes working.
2019-10-03 20:39:08 +00:00
case FINAL_ACTION_READ: {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
description.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
description.finalLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
update_external_dependency_for_store(dependency_to_external, is_sampled, is_storage, false);
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
description.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
description.finalLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL);
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
update_external_dependency_for_store(dependency_to_external, is_sampled, is_storage, true);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
// TODO: What does this mean about the next usage (and thus appropriate dependency masks
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
GIProbes working.
2019-10-03 20:39:08 +00:00
} break;
case FINAL_ACTION_DISCARD: {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
GIProbes working.
2019-10-03 20:39:08 +00:00
description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
description.finalLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
GIProbes working.
2019-10-03 20:39:08 +00:00
description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
description.finalLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL);
} else {
description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
GIProbes working.
2019-10-03 20:39:08 +00:00
} break;
case FINAL_ACTION_CONTINUE: {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
GIProbes working.
2019-10-03 20:39:08 +00:00
description.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
description.finalLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
GIProbes working.
2019-10-03 20:39:08 +00:00
description.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE;
description.finalLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
} else {
description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; //don't care what is there
}
} break;
default: {
ERR_FAIL_V(VK_NULL_HANDLE); //should never reach here
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
attachment_last_pass[i] = -1;
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
attachment_remap.push_back(attachments.size());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
attachments.push_back(description);
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
}
LocalVector<VkSubpassDescription> subpasses;
LocalVector<LocalVector<VkAttachmentReference>> color_reference_array;
LocalVector<LocalVector<VkAttachmentReference>> input_reference_array;
LocalVector<LocalVector<VkAttachmentReference>> resolve_reference_array;
LocalVector<LocalVector<uint32_t>> preserve_reference_array;
LocalVector<VkAttachmentReference> depth_reference_array;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
subpasses.resize(p_passes.size());
color_reference_array.resize(p_passes.size());
input_reference_array.resize(p_passes.size());
resolve_reference_array.resize(p_passes.size());
preserve_reference_array.resize(p_passes.size());
depth_reference_array.resize(p_passes.size());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
LocalVector<VkSubpassDependency> subpass_dependencies;
for (int i = 0; i < p_passes.size(); i++) {
const FramebufferPass *pass = &p_passes[i];
LocalVector<VkAttachmentReference> &color_references = color_reference_array[i];
TextureSamples texture_samples = TEXTURE_SAMPLES_1;
bool is_multisample_first = true;
for (int j = 0; j < pass->color_attachments.size(); j++) {
int32_t attachment = pass->color_attachments[j];
VkAttachmentReference reference;
if (attachment == FramebufferPass::ATTACHMENT_UNUSED) {
reference.attachment = VK_ATTACHMENT_UNUSED;
reference.layout = VK_IMAGE_LAYOUT_UNDEFINED;
} else {
ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), color attachment (" + itos(j) + ").");
ERR_FAIL_COND_V_MSG(!(p_attachments[attachment].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it's marked as depth, but it's not usable as color attachment.");
ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass.");
if (is_multisample_first) {
texture_samples = p_attachments[attachment].samples;
is_multisample_first = false;
} else {
ERR_FAIL_COND_V_MSG(texture_samples != p_attachments[attachment].samples, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), if an attachment is marked as multisample, all of them should be multisample and use the same number of samples.");
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
reference.attachment = attachment_remap[attachment];
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
reference.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
attachment_last_pass[attachment] = i;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
color_references.push_back(reference);
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
}
LocalVector<VkAttachmentReference> &input_references = input_reference_array[i];
for (int j = 0; j < pass->input_attachments.size(); j++) {
int32_t attachment = pass->input_attachments[j];
VkAttachmentReference reference;
if (attachment == FramebufferPass::ATTACHMENT_UNUSED) {
reference.attachment = VK_ATTACHMENT_UNUSED;
reference.layout = VK_IMAGE_LAYOUT_UNDEFINED;
} else {
ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), input attachment (" + itos(j) + ").");
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
ERR_FAIL_COND_V_MSG(!(p_attachments[attachment].usage_flags & TEXTURE_USAGE_INPUT_ATTACHMENT_BIT), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it isn't marked as an input texture.");
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass.");
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
reference.attachment = attachment_remap[attachment];
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
reference.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
attachment_last_pass[attachment] = i;
}
input_references.push_back(reference);
}
LocalVector<VkAttachmentReference> &resolve_references = resolve_reference_array[i];
if (pass->resolve_attachments.size() > 0) {
ERR_FAIL_COND_V_MSG(pass->resolve_attachments.size() != pass->color_attachments.size(), VK_NULL_HANDLE, "The amount of resolve attachments (" + itos(pass->resolve_attachments.size()) + ") must match the number of color attachments (" + itos(pass->color_attachments.size()) + ").");
ERR_FAIL_COND_V_MSG(texture_samples == TEXTURE_SAMPLES_1, VK_NULL_HANDLE, "Resolve attachments specified, but color attachments are not multisample.");
}
for (int j = 0; j < pass->resolve_attachments.size(); j++) {
int32_t attachment = pass->resolve_attachments[j];
VkAttachmentReference reference;
if (attachment == FramebufferPass::ATTACHMENT_UNUSED) {
reference.attachment = VK_ATTACHMENT_UNUSED;
reference.layout = VK_IMAGE_LAYOUT_UNDEFINED;
} else {
ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), resolve attachment (" + itos(j) + ").");
ERR_FAIL_COND_V_MSG(pass->color_attachments[j] == FramebufferPass::ATTACHMENT_UNUSED, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), resolve attachment (" + itos(j) + "), the respective color attachment is marked as unused.");
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
ERR_FAIL_COND_V_MSG(!(p_attachments[attachment].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), resolve attachment, it isn't marked as a color texture.");
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass.");
bool multisample = p_attachments[attachment].samples > TEXTURE_SAMPLES_1;
ERR_FAIL_COND_V_MSG(multisample, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), resolve attachments can't be multisample.");
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
reference.attachment = attachment_remap[attachment];
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
reference.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; // VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
attachment_last_pass[attachment] = i;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
resolve_references.push_back(reference);
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
}
VkAttachmentReference &depth_stencil_reference = depth_reference_array[i];
if (pass->depth_attachment != FramebufferPass::ATTACHMENT_UNUSED) {
int32_t attachment = pass->depth_attachment;
ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer depth format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), depth attachment.");
ERR_FAIL_COND_V_MSG(!(p_attachments[attachment].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT), VK_NULL_HANDLE, "Invalid framebuffer depth format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it's marked as depth, but it's not a depth attachment.");
ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer depth format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass.");
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
depth_stencil_reference.attachment = attachment_remap[attachment];
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
depth_stencil_reference.layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
attachment_last_pass[attachment] = i;
Fix Context Validation Layer Errors * Multisampling was wrongly selected, possibly fixes #49937 * Image semaphore acquisition is now per window, possibly fixes #41614 Please make sure to test the above two issues again, since I can't reproduce either anyway.
2021-06-28 20:43:10 +00:00
if (is_multisample_first) {
texture_samples = p_attachments[attachment].samples;
is_multisample_first = false;
} else {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND_V_MSG(texture_samples != p_attachments[attachment].samples, VK_NULL_HANDLE, "Invalid framebuffer depth format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), if an attachment is marked as multisample, all of them should be multisample and use the same number of samples including the depth.");
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
depth_stencil_reference.attachment = VK_ATTACHMENT_UNUSED;
depth_stencil_reference.layout = VK_IMAGE_LAYOUT_UNDEFINED;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
Fixes and optimizations to mobile renderer * Only apply final actions to attachments used in the last pass. * Fixes to draw list final action (was using continue instead of read/drop). * Profiling regions inside draw lists now properly throw errors. * Ability to enable gpu profile printing from project settings. (used to debug).
2021-08-11 00:50:28 +00:00
LocalVector<uint32_t> &preserve_references = preserve_reference_array[i];
for (int j = 0; j < pass->preserve_attachments.size(); j++) {
int32_t attachment = pass->preserve_attachments[j];
ERR_FAIL_COND_V_MSG(attachment == FramebufferPass::ATTACHMENT_UNUSED, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), preserve attachment (" + itos(j) + "). Preserve attachments can't be unused.");
ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), preserve attachment (" + itos(j) + ").");
if (attachment_last_pass[attachment] != i) {
//preserve can still be used to keep depth or color from being discarded after use
attachment_last_pass[attachment] = i;
preserve_references.push_back(attachment);
}
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
VkSubpassDescription &subpass = subpasses[i];
subpass.flags = 0;
subpass.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
subpass.inputAttachmentCount = input_references.size();
if (input_references.size()) {
subpass.pInputAttachments = input_references.ptr();
} else {
subpass.pInputAttachments = nullptr;
}
subpass.colorAttachmentCount = color_references.size();
if (color_references.size()) {
subpass.pColorAttachments = color_references.ptr();
} else {
subpass.pColorAttachments = nullptr;
}
if (depth_stencil_reference.attachment != VK_ATTACHMENT_UNUSED) {
subpass.pDepthStencilAttachment = &depth_stencil_reference;
} else {
subpass.pDepthStencilAttachment = nullptr;
}
if (resolve_references.size()) {
subpass.pResolveAttachments = resolve_references.ptr();
} else {
subpass.pResolveAttachments = nullptr;
}
subpass.preserveAttachmentCount = preserve_references.size();
if (preserve_references.size()) {
subpass.pPreserveAttachments = preserve_references.ptr();
} else {
subpass.pPreserveAttachments = nullptr;
}
if (r_samples) {
r_samples->push_back(texture_samples);
}
if (i > 0) {
VkSubpassDependency dependency;
dependency.srcSubpass = i - 1;
dependency.dstSubpass = i;
dependency.srcStageMask = 0;
dependency.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
dependency.dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
dependency.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
dependency.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_INPUT_ATTACHMENT_READ_BIT;
dependency.dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT;
subpass_dependencies.push_back(dependency);
}
/*
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
// NOTE: Big Mallet Approach -- any layout transition causes a full barrier
if (reference.layout != description.initialLayout) {
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
// NOTE: this should be smarter based on the texture's knowledge of its previous role
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
dependency_from_external.srcStageMask |= VK_PIPELINE_STAGE_ALL_COMMANDS_BIT;
dependency_from_external.srcAccessMask |= VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT;
}
if (reference.layout != description.finalLayout) {
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
// NOTE: this should be smarter based on the texture's knowledge of its subsequent role
Synchronization - Clean subpass dependency init Fix incorrect subpass dependency initialization
2020-10-29 21:42:30 +00:00
dependency_to_external.dstStageMask |= VK_PIPELINE_STAGE_ALL_COMMANDS_BIT;
dependency_to_external.dstAccessMask |= VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
*/
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
VkRenderPassCreateInfo render_pass_create_info;
render_pass_create_info.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
render_pass_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
render_pass_create_info.flags = 0;
render_pass_create_info.attachmentCount = attachments.size();
render_pass_create_info.pAttachments = attachments.ptr();
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
render_pass_create_info.subpassCount = subpasses.size();
render_pass_create_info.pSubpasses = subpasses.ptr();
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
// Commenting this because it seems it just avoids raster and compute to work at the same time.
// Other barriers seem to be protecting the render pass fine.
// render_pass_create_info.dependencyCount = 2;
// render_pass_create_info.pDependencies = dependencies;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
render_pass_create_info.dependencyCount = subpass_dependencies.size();
if (subpass_dependencies.size()) {
render_pass_create_info.pDependencies = subpass_dependencies.ptr();
} else {
render_pass_create_info.pDependencies = nullptr;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
// These are only used if we use multiview but we need to define them in scope.
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
const uint32_t view_mask = (1 << p_view_count) - 1;
const uint32_t correlation_mask = (1 << p_view_count) - 1;
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
Vector<uint32_t> view_masks;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
VkRenderPassMultiviewCreateInfo render_pass_multiview_create_info;
if (p_view_count > 1) {
const VulkanContext::MultiviewCapabilities capabilities = context->get_multiview_capabilities();
// For now this only works with multiview!
ERR_FAIL_COND_V_MSG(!capabilities.is_supported, VK_NULL_HANDLE, "Multiview not supported");
// Make sure we limit this to the number of views we support.
ERR_FAIL_COND_V_MSG(p_view_count > capabilities.max_view_count, VK_NULL_HANDLE, "Hardware does not support requested number of views for Multiview render pass");
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
// Set view masks for each subpass
for (uint32_t i = 0; i < subpasses.size(); i++) {
view_masks.push_back(view_mask);
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
}
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
render_pass_multiview_create_info.sType = VK_STRUCTURE_TYPE_RENDER_PASS_MULTIVIEW_CREATE_INFO;
render_pass_multiview_create_info.pNext = nullptr;
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
render_pass_multiview_create_info.subpassCount = subpasses.size();
render_pass_multiview_create_info.pViewMasks = view_masks.ptr();
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
render_pass_multiview_create_info.dependencyCount = 0;
render_pass_multiview_create_info.pViewOffsets = nullptr;
render_pass_multiview_create_info.correlationMaskCount = 1;
render_pass_multiview_create_info.pCorrelationMasks = &correlation_mask;
render_pass_create_info.pNext = &render_pass_multiview_create_info;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkRenderPass render_pass;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult res = vkCreateRenderPass(device, &render_pass_create_info, nullptr, &render_pass);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(res, VK_NULL_HANDLE, "vkCreateRenderPass failed with error " + itos(res) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
return render_pass;
}
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_create(const Vector<AttachmentFormat> &p_format, uint32_t p_view_count) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
FramebufferPass pass;
for (int i = 0; i < p_format.size(); i++) {
if (p_format[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
pass.depth_attachment = i;
} else {
pass.color_attachments.push_back(i);
}
}
Vector<FramebufferPass> passes;
passes.push_back(pass);
return framebuffer_format_create_multipass(p_format, passes, p_view_count);
}
RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_create_multipass(const Vector<AttachmentFormat> &p_attachments, Vector<FramebufferPass> &p_passes, uint32_t p_view_count) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
FramebufferFormatKey key;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
key.attachments = p_attachments;
key.passes = p_passes;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
key.view_count = p_view_count;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
const Map<FramebufferFormatKey, FramebufferFormatID>::Element *E = framebuffer_format_cache.find(key);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (E) {
//exists, return
return E->get();
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
Vector<TextureSamples> samples;
VkRenderPass render_pass = _render_pass_create(p_attachments, p_passes, INITIAL_ACTION_CLEAR, FINAL_ACTION_READ, INITIAL_ACTION_CLEAR, FINAL_ACTION_READ, p_view_count, &samples); //actions don't matter for this use case
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Improved validation of attachments
2019-06-07 18:20:01 +00:00
if (render_pass == VK_NULL_HANDLE) { //was likely invalid
return INVALID_ID;
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
FramebufferFormatID id = FramebufferFormatID(framebuffer_format_cache.size()) | (FramebufferFormatID(ID_TYPE_FRAMEBUFFER_FORMAT) << FramebufferFormatID(ID_BASE_SHIFT));
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
E = framebuffer_format_cache.insert(key, id);
FramebufferFormat fb_format;
fb_format.E = E;
fb_format.render_pass = render_pass;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
fb_format.pass_samples = samples;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
fb_format.view_count = p_view_count;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
framebuffer_formats[id] = fb_format;
return id;
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_create_empty(TextureSamples p_samples) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
FramebufferFormatKey key;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
key.passes.push_back(FramebufferPass());
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const Map<FramebufferFormatKey, FramebufferFormatID>::Element *E = framebuffer_format_cache.find(key);
if (E) {
//exists, return
return E->get();
}
VkSubpassDescription subpass;
subpass.flags = 0;
subpass.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
subpass.inputAttachmentCount = 0; //unsupported for now
subpass.pInputAttachments = nullptr;
subpass.colorAttachmentCount = 0;
subpass.pColorAttachments = nullptr;
subpass.pDepthStencilAttachment = nullptr;
subpass.pResolveAttachments = nullptr;
subpass.preserveAttachmentCount = 0;
subpass.pPreserveAttachments = nullptr;
VkRenderPassCreateInfo render_pass_create_info;
render_pass_create_info.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO;
render_pass_create_info.pNext = nullptr;
render_pass_create_info.flags = 0;
render_pass_create_info.attachmentCount = 0;
render_pass_create_info.pAttachments = nullptr;
render_pass_create_info.subpassCount = 1;
render_pass_create_info.pSubpasses = &subpass;
render_pass_create_info.dependencyCount = 0;
render_pass_create_info.pDependencies = nullptr;
VkRenderPass render_pass;
VkResult res = vkCreateRenderPass(device, &render_pass_create_info, nullptr, &render_pass);
Use "volk" instead of statically linked Vulkan loader.
2021-08-12 11:24:54 +00:00
ERR_FAIL_COND_V_MSG(res, 0, "vkCreateRenderPass for empty fb failed with error " + itos(res) + ".");
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (render_pass == VK_NULL_HANDLE) { //was likely invalid
return INVALID_ID;
}
FramebufferFormatID id = FramebufferFormatID(framebuffer_format_cache.size()) | (FramebufferFormatID(ID_TYPE_FRAMEBUFFER_FORMAT) << FramebufferFormatID(ID_BASE_SHIFT));
E = framebuffer_format_cache.insert(key, id);
FramebufferFormat fb_format;
fb_format.E = E;
fb_format.render_pass = render_pass;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
fb_format.pass_samples.push_back(p_samples);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
framebuffer_formats[id] = fb_format;
return id;
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
RenderingDevice::TextureSamples RenderingDeviceVulkan::framebuffer_format_get_texture_samples(FramebufferFormatID p_format, uint32_t p_pass) {
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
Map<FramebufferFormatID, FramebufferFormat>::Element *E = framebuffer_formats.find(p_format);
ERR_FAIL_COND_V(!E, TEXTURE_SAMPLES_1);
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND_V(p_pass >= uint32_t(E->get().pass_samples.size()), TEXTURE_SAMPLES_1);
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
return E->get().pass_samples[p_pass];
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
/***********************/
/**** RENDER TARGET ****/
/***********************/
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RID RenderingDeviceVulkan::framebuffer_create_empty(const Size2i &p_size, TextureSamples p_samples, FramebufferFormatID p_format_check) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
_THREAD_SAFE_METHOD_
Framebuffer framebuffer;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
framebuffer.format_id = framebuffer_format_create_empty(p_samples);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
ERR_FAIL_COND_V(p_format_check != INVALID_FORMAT_ID && framebuffer.format_id != p_format_check, RID());
framebuffer.size = p_size;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
framebuffer.view_count = 1;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
return framebuffer_owner.make_rid(framebuffer);
}
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
RID RenderingDeviceVulkan::framebuffer_create(const Vector<RID> &p_texture_attachments, FramebufferFormatID p_format_check, uint32_t p_view_count) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
FramebufferPass pass;
for (int i = 0; i < p_texture_attachments.size(); i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(p_texture_attachments[i]);
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
ERR_FAIL_COND_V_MSG(texture && texture->layers != p_view_count, RID(), "Layers of our texture doesn't match view count for this framebuffer");
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (texture && texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
pass.depth_attachment = i;
} else {
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
pass.color_attachments.push_back(texture ? i : FramebufferPass::ATTACHMENT_UNUSED);
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
}
}
Vector<FramebufferPass> passes;
passes.push_back(pass);
return framebuffer_create_multipass(p_texture_attachments, passes, p_format_check, p_view_count);
}
RID RenderingDeviceVulkan::framebuffer_create_multipass(const Vector<RID> &p_texture_attachments, Vector<FramebufferPass> &p_passes, FramebufferFormatID p_format_check, uint32_t p_view_count) {
_THREAD_SAFE_METHOD_
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Vector<AttachmentFormat> attachments;
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
attachments.resize(p_texture_attachments.size());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Size2i size;
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
bool size_set = false;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (int i = 0; i < p_texture_attachments.size(); i++) {
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
AttachmentFormat af;
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(p_texture_attachments[i]);
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (!texture) {
af.usage_flags = AttachmentFormat::UNUSED_ATTACHMENT;
} else {
ERR_FAIL_COND_V_MSG(texture->layers != p_view_count, RID(), "Layers of our texture doesn't match view count for this framebuffer");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (!size_set) {
size.width = texture->width;
size.height = texture->height;
size_set = true;
} else {
ERR_FAIL_COND_V_MSG((uint32_t)size.width != texture->width || (uint32_t)size.height != texture->height, RID(),
"All textures in a framebuffer should be the same size.");
}
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
af.format = texture->format;
af.samples = texture->samples;
af.usage_flags = texture->usage_flags;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
attachments.write[i] = af;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
ERR_FAIL_COND_V_MSG(!size_set, RID(), "All attachments unused.");
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
FramebufferFormatID format_id = framebuffer_format_create_multipass(attachments, p_passes, p_view_count);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (format_id == INVALID_ID) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
return RID();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(p_format_check != INVALID_ID && format_id != p_format_check, RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"The format used to check this framebuffer differs from the intended framebuffer format.");
Framebuffer framebuffer;
framebuffer.format_id = format_id;
framebuffer.texture_ids = p_texture_attachments;
framebuffer.size = size;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
framebuffer.view_count = p_view_count;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID id = framebuffer_owner.make_rid(framebuffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (int i = 0; i < p_texture_attachments.size(); i++) {
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (p_texture_attachments[i].is_valid()) {
_add_dependency(id, p_texture_attachments[i]);
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return id;
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_get_format(RID p_framebuffer) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_framebuffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_V(!framebuffer, INVALID_ID);
return framebuffer->format_id;
}
/*****************/
/**** SAMPLER ****/
/*****************/
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID RenderingDeviceVulkan::sampler_create(const SamplerState &p_state) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
VkSamplerCreateInfo sampler_create_info;
sampler_create_info.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
sampler_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
sampler_create_info.flags = 0;
sampler_create_info.magFilter = p_state.mag_filter == SAMPLER_FILTER_LINEAR ? VK_FILTER_LINEAR : VK_FILTER_NEAREST;
sampler_create_info.minFilter = p_state.min_filter == SAMPLER_FILTER_LINEAR ? VK_FILTER_LINEAR : VK_FILTER_NEAREST;
sampler_create_info.mipmapMode = p_state.mip_filter == SAMPLER_FILTER_LINEAR ? VK_SAMPLER_MIPMAP_MODE_LINEAR : VK_SAMPLER_MIPMAP_MODE_NEAREST;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_state.repeat_u, SAMPLER_REPEAT_MODE_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
sampler_create_info.addressModeU = address_modes[p_state.repeat_u];
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_state.repeat_v, SAMPLER_REPEAT_MODE_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
sampler_create_info.addressModeV = address_modes[p_state.repeat_v];
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_state.repeat_w, SAMPLER_REPEAT_MODE_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
sampler_create_info.addressModeW = address_modes[p_state.repeat_w];
sampler_create_info.mipLodBias = p_state.lod_bias;
sampler_create_info.anisotropyEnable = p_state.use_anisotropy;
sampler_create_info.maxAnisotropy = p_state.anisotropy_max;
sampler_create_info.compareEnable = p_state.enable_compare;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_state.compare_op, COMPARE_OP_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
sampler_create_info.compareOp = compare_operators[p_state.compare_op];
sampler_create_info.minLod = p_state.min_lod;
sampler_create_info.maxLod = p_state.max_lod;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_state.border_color, SAMPLER_BORDER_COLOR_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
sampler_create_info.borderColor = sampler_border_colors[p_state.border_color];
sampler_create_info.unnormalizedCoordinates = p_state.unnormalized_uvw;
VkSampler sampler;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult res = vkCreateSampler(device, &sampler_create_info, nullptr, &sampler);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(res, RID(), "vkCreateSampler failed with error " + itos(res) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
return sampler_owner.make_rid(sampler);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
/**********************/
/**** VERTEX ARRAY ****/
/**********************/
Reimplement skeletons and blend shapes Uses compute shaders, which only once, on demand, and all in parallel.
2020-12-16 14:07:08 +00:00
RID RenderingDeviceVulkan::vertex_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data, bool p_use_as_storage) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(p_data.size() && (uint32_t)p_data.size() != p_size_bytes, RID());
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Reimplement skeletons and blend shapes Uses compute shaders, which only once, on demand, and all in parallel.
2020-12-16 14:07:08 +00:00
uint32_t usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT;
if (p_use_as_storage) {
usage |= VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Buffer buffer;
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
_buffer_allocate(&buffer, p_size_bytes, usage, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_data.size()) {
uint64_t data_size = p_data.size();
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint8_t *r = p_data.ptr();
_buffer_update(&buffer, 0, r, data_size);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
_buffer_memory_barrier(buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT, false);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
return vertex_buffer_owner.make_rid(buffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
// Internally reference counted, this ID is warranted to be unique for the same description, but needs to be freed as many times as it was allocated
Add proper type to most public API uses of Array
2020-04-21 15:16:45 +00:00
RenderingDevice::VertexFormatID RenderingDeviceVulkan::vertex_format_create(const Vector<VertexAttribute> &p_vertex_formats) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
VertexDescriptionKey key;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
key.vertex_formats = p_vertex_formats;
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
VertexFormatID *idptr = vertex_format_cache.getptr(key);
if (idptr) {
return *idptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//does not exist, create one and cache it
VertexDescriptionCache vdcache;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
vdcache.bindings = memnew_arr(VkVertexInputBindingDescription, p_vertex_formats.size());
vdcache.attributes = memnew_arr(VkVertexInputAttributeDescription, p_vertex_formats.size());
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Set<int> used_locations;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
for (int i = 0; i < p_vertex_formats.size(); i++) {
ERR_CONTINUE(p_vertex_formats[i].format >= DATA_FORMAT_MAX);
ERR_FAIL_COND_V(used_locations.has(p_vertex_formats[i].location), INVALID_ID);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(get_format_vertex_size(p_vertex_formats[i].format) == 0, INVALID_ID,
Refactored Mesh internals and formats. -Changed how mesh data is organized, hoping to make it more efficient on Vulkan and GLES. -Removed compression, it now always uses the most efficient format. -Added support for custom arrays (up to 8 custom formats) -Added support for 8 weights in skeleton data. -Added a simple optional versioning system for imported assets, to reimport if binary is newer -Fixes #43979 (I needed to test) WARNING: -NOT backwards compatible with previous 4.x-devel, will most likely never be, but it will force reimport scenes due to version change. -NOT backwards compatible with 3.x scenes, this will be eventually re-added. -Skeletons not working any longer, will fix in next PR.
2020-12-02 01:40:47 +00:00
"Data format for attachment (" + itos(i) + "), '" + named_formats[p_vertex_formats[i].format] + "', is not valid for a vertex array.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vdcache.bindings[i].binding = i;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
vdcache.bindings[i].stride = p_vertex_formats[i].stride;
vdcache.bindings[i].inputRate = p_vertex_formats[i].frequency == VERTEX_FREQUENCY_INSTANCE ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vdcache.attributes[i].binding = i;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
vdcache.attributes[i].location = p_vertex_formats[i].location;
vdcache.attributes[i].format = vulkan_formats[p_vertex_formats[i].format];
vdcache.attributes[i].offset = p_vertex_formats[i].offset;
used_locations.insert(p_vertex_formats[i].location);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
vdcache.create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vdcache.create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vdcache.create_info.flags = 0;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
vdcache.create_info.vertexAttributeDescriptionCount = p_vertex_formats.size();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vdcache.create_info.pVertexAttributeDescriptions = vdcache.attributes;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
vdcache.create_info.vertexBindingDescriptionCount = p_vertex_formats.size();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vdcache.create_info.pVertexBindingDescriptions = vdcache.bindings;
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
vdcache.vertex_formats = p_vertex_formats;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
VertexFormatID id = VertexFormatID(vertex_format_cache.size()) | (VertexFormatID(ID_TYPE_VERTEX_FORMAT) << ID_BASE_SHIFT);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
vertex_format_cache[key] = id;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
vertex_formats[id] = vdcache;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
return id;
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID RenderingDeviceVulkan::vertex_array_create(uint32_t p_vertex_count, VertexFormatID p_vertex_format, const Vector<RID> &p_src_buffers) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!vertex_formats.has(p_vertex_format), RID());
const VertexDescriptionCache &vd = vertex_formats[p_vertex_format];
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
ERR_FAIL_COND_V(vd.vertex_formats.size() != p_src_buffers.size(), RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (int i = 0; i < p_src_buffers.size(); i++) {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!vertex_buffer_owner.owns(p_src_buffers[i]), RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
VertexArray vertex_array;
vertex_array.vertex_count = p_vertex_count;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
vertex_array.description = p_vertex_format;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vertex_array.max_instances_allowed = 0xFFFFFFFF; //by default as many as you want
for (int i = 0; i < p_src_buffers.size(); i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Buffer *buffer = vertex_buffer_owner.get_or_null(p_src_buffers[i]);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//validate with buffer
{
Add proper type to most public API uses of Array
2020-04-21 15:16:45 +00:00
const VertexAttribute &atf = vd.vertex_formats[i];
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
uint32_t element_size = get_format_vertex_size(atf.format);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(element_size == 0, RID()); //should never happens since this was prevalidated
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (atf.frequency == VERTEX_FREQUENCY_VERTEX) {
//validate size for regular drawing
uint64_t total_size = uint64_t(atf.stride) * (p_vertex_count - 1) + atf.offset + element_size;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(total_size > buffer->size, RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Attachment (" + itos(i) + ") will read past the end of the buffer.");
} else {
//validate size for instances drawing
uint64_t available = buffer->size - atf.offset;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(available < element_size, RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Attachment (" + itos(i) + ") uses instancing, but it's just too small.");
uint32_t instances_allowed = available / atf.stride;
vertex_array.max_instances_allowed = MIN(instances_allowed, vertex_array.max_instances_allowed);
}
}
vertex_array.buffers.push_back(buffer->buffer);
vertex_array.offsets.push_back(0); //offset unused, but passing anyway
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID id = vertex_array_owner.make_rid(vertex_array);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (int i = 0; i < p_src_buffers.size(); i++) {
_add_dependency(id, p_src_buffers[i]);
}
return id;
}
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
RID RenderingDeviceVulkan::index_buffer_create(uint32_t p_index_count, IndexBufferFormat p_format, const Vector<uint8_t> &p_data, bool p_use_restart_indices) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(p_index_count == 0, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
IndexBuffer index_buffer;
index_buffer.index_type = (p_format == INDEX_BUFFER_FORMAT_UINT16) ? VK_INDEX_TYPE_UINT16 : VK_INDEX_TYPE_UINT32;
index_buffer.supports_restart_indices = p_use_restart_indices;
index_buffer.index_count = p_index_count;
uint32_t size_bytes = p_index_count * ((p_format == INDEX_BUFFER_FORMAT_UINT16) ? 2 : 4);
#ifdef DEBUG_ENABLED
if (p_data.size()) {
index_buffer.max_index = 0;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG((uint32_t)p_data.size() != size_bytes, RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Default index buffer initializer array size (" + itos(p_data.size()) + ") does not match format required size (" + itos(size_bytes) + ").");
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint8_t *r = p_data.ptr();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_format == INDEX_BUFFER_FORMAT_UINT16) {
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint16_t *index16 = (const uint16_t *)r;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (uint32_t i = 0; i < p_index_count; i++) {
if (p_use_restart_indices && index16[i] == 0xFFFF) {
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
continue; //restart index, ignore
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
index_buffer.max_index = MAX(index16[i], index_buffer.max_index);
}
} else {
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint32_t *index32 = (const uint32_t *)r;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (uint32_t i = 0; i < p_index_count; i++) {
if (p_use_restart_indices && index32[i] == 0xFFFFFFFF) {
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
continue; //restart index, ignore
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
index_buffer.max_index = MAX(index32[i], index_buffer.max_index);
}
}
} else {
index_buffer.max_index = 0xFFFFFFFF;
}
#else
index_buffer.max_index = 0xFFFFFFFF;
#endif
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
_buffer_allocate(&index_buffer, size_bytes, VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_INDEX_BUFFER_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_data.size()) {
uint64_t data_size = p_data.size();
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint8_t *r = p_data.ptr();
_buffer_update(&index_buffer, 0, r, data_size);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
_buffer_memory_barrier(index_buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_INDEX_READ_BIT, false);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
return index_buffer_owner.make_rid(index_buffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID RenderingDeviceVulkan::index_array_create(RID p_index_buffer, uint32_t p_index_offset, uint32_t p_index_count) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!index_buffer_owner.owns(p_index_buffer), RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
IndexBuffer *index_buffer = index_buffer_owner.get_or_null(p_index_buffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(p_index_count == 0, RID());
ERR_FAIL_COND_V(p_index_offset + p_index_count > index_buffer->index_count, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
IndexArray index_array;
index_array.max_index = index_buffer->max_index;
index_array.buffer = index_buffer->buffer;
index_array.offset = p_index_offset;
index_array.indices = p_index_count;
index_array.index_type = index_buffer->index_type;
index_array.supports_restart_indices = index_buffer->supports_restart_indices;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID id = index_array_owner.make_rid(index_array);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_add_dependency(id, p_index_buffer);
return id;
}
/****************/
/**** SHADER ****/
/****************/
static const char *shader_stage_names[RenderingDevice::SHADER_STAGE_MAX] = {
"Vertex",
"Fragment",
"TesselationControl",
"TesselationEvaluation",
"Compute"
};
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
static const char *shader_uniform_names[RenderingDevice::UNIFORM_TYPE_MAX] = {
"Sampler", "CombinedSampler", "Texture", "Image", "TextureBuffer", "SamplerTextureBuffer", "ImageBuffer", "UniformBuffer", "StorageBuffer", "InputAttachment"
};
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
static VkShaderStageFlagBits shader_stage_masks[RenderingDevice::SHADER_STAGE_MAX] = {
VK_SHADER_STAGE_VERTEX_BIT,
VK_SHADER_STAGE_FRAGMENT_BIT,
VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT,
VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT,
VK_SHADER_STAGE_COMPUTE_BIT,
};
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
String RenderingDeviceVulkan::_shader_uniform_debug(RID p_shader, int p_set) {
String ret;
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
const Shader *shader = shader_owner.get_or_null(p_shader);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
ERR_FAIL_COND_V(!shader, String());
for (int i = 0; i < shader->sets.size(); i++) {
if (p_set >= 0 && i != p_set) {
continue;
}
for (int j = 0; j < shader->sets[i].uniform_info.size(); j++) {
const UniformInfo &ui = shader->sets[i].uniform_info[j];
Replace String comparisons with "", String() to is_empty() Also: - Adds two stress tests to test_string.h - Changes to .empty() on std::strings
2021-12-09 09:42:46 +00:00
if (!ret.is_empty()) {
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
ret += "\n";
}
ret += "Set: " + itos(i) + " Binding: " + itos(ui.binding) + " Type: " + shader_uniform_names[ui.type] + " Length: " + itos(ui.length);
}
}
return ret;
}
Moved the shader source compilation code outside RenderingDevice and Vulkan
2019-07-28 22:58:32 +00:00
#if 0
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
bool RenderingDeviceVulkan::_uniform_add_binding(Vector<Vector<VkDescriptorSetLayoutBinding> > &bindings, Vector<Vector<UniformInfo> > &uniform_infos, const glslang::TObjectReflection &reflection, RenderingDevice::ShaderStage p_stage, Shader::PushConstant &push_constant, String *r_error) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkDescriptorSetLayoutBinding layout_binding;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
UniformInfo info;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
switch (reflection.getType()->getBasicType()) {
case glslang::EbtSampler: {
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: IsSampler");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (reflection.getType()->getSampler().dim == glslang::EsdBuffer) {
//texture buffers
if (reflection.getType()->getSampler().isCombined()) {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
info.type = UNIFORM_TYPE_SAMPLER_WITH_TEXTURE_BUFFER;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: SAMPLER: texel combined");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else if (reflection.getType()->getSampler().isTexture()) {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
info.type = UNIFORM_TYPE_TEXTURE_BUFFER;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: SAMPLER: texel alone");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else if (reflection.getType()->getSampler().isImage()) {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER;
info.type = UNIFORM_TYPE_IMAGE_BUFFER;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: SAMPLER: texel buffer");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
if (r_error) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' is of unsupported buffer type.";
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return false;
}
} else if (reflection.getType()->getSampler().isCombined()) {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
info.type = UNIFORM_TYPE_SAMPLER_WITH_TEXTURE;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: SAMPLER: combined");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else if (reflection.getType()->getSampler().isPureSampler()) {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLER;
info.type = UNIFORM_TYPE_SAMPLER;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: SAMPLER: sampler");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else if (reflection.getType()->getSampler().isTexture()) {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE;
info.type = UNIFORM_TYPE_TEXTURE;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: SAMPLER: image");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else if (reflection.getType()->getSampler().isImage()) {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
info.type = UNIFORM_TYPE_IMAGE;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: SAMPLER: storage image");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: sampler unknown");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (r_error) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' is of unsupported sampler type.";
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return false;
}
if (reflection.getType()->isArray()) {
layout_binding.descriptorCount = reflection.getType()->getArraySizes()->getCumulativeSize();
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: array of size: " + itos(layout_binding.descriptorCount));
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
layout_binding.descriptorCount = 1;
}
info.length = layout_binding.descriptorCount;
} break;
/*case glslang::EbtStruct: {
print_line("DEBUG: Struct");
} break;*/
case glslang::EbtBlock: {
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: Block");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (reflection.getType()->getQualifier().storage == glslang::EvqUniform) {
Added support for push constants
2019-06-08 20:10:52 +00:00
if (reflection.getType()->getQualifier().layoutPushConstant) {
uint32_t len = reflection.size;
if (push_constant.push_constant_size != 0 && push_constant.push_constant_size != len) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' push constants for different stages should all be the same size.";
Added support for push constants
2019-06-08 20:10:52 +00:00
return false;
}
push_constant.push_constant_size = len;
push_constant.push_constants_vk_stage |= shader_stage_masks[p_stage];
return true;
}
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: Uniform buffer");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
info.type = UNIFORM_TYPE_UNIFORM_BUFFER;
} else if (reflection.getType()->getQualifier().storage == glslang::EvqBuffer) {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
info.type = UNIFORM_TYPE_STORAGE_BUFFER;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: Storage buffer");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
if (r_error) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' is of unsupported block type: (" + itos(reflection.getType()->getQualifier().storage) + ").";
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return false;
}
if (reflection.getType()->isArray()) {
layout_binding.descriptorCount = reflection.getType()->getArraySizes()->getCumulativeSize();
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
//print_line("DEBUG: array of size: " + itos(layout_binding.descriptorCount));
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
layout_binding.descriptorCount = 1;
}
info.length = reflection.size;
} break;
/*case glslang::EbtReference: {
} break;*/
/*case glslang::EbtAtomicUint: {
} break;*/
default: {
Normalmapping and Specularmapping working in 2D engine Added support for Sprite, AnimatedSprite and Polygon2D (should add for tileset eventually).
2019-07-05 01:54:32 +00:00
if (reflection.getType()->getQualifier().hasOffset() || reflection.name.find(".") != std::string::npos) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//member of uniform block?
return true;
}
if (r_error) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' unsupported uniform type.";
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return false;
}
}
if (!reflection.getType()->getQualifier().hasBinding()) {
if (r_error) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' lacks a binding number.";
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return false;
}
uint32_t set = reflection.getType()->getQualifier().hasSet() ? reflection.getType()->getQualifier().layoutSet : 0;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
if (set >= MAX_UNIFORM_SETS) {
if (r_error) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' uses a set (" + itos(set) + ") index larger than what is supported (" + itos(MAX_UNIFORM_SETS) + ").";
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
}
return false;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (set >= limits.maxBoundDescriptorSets) {
if (r_error) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' uses a set (" + itos(set) + ") index larger than what is supported by the hardware (" + itos(limits.maxBoundDescriptorSets) + ").";
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return false;
}
uint32_t binding = reflection.getType()->getQualifier().layoutBinding;
if (set < (uint32_t)bindings.size()) {
//check if this already exists
for (int i = 0; i < bindings[set].size(); i++) {
if (bindings[set][i].binding == binding) {
//already exists, verify that it's the same type
if (bindings[set][i].descriptorType != layout_binding.descriptorType) {
if (r_error) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(binding) + " with different uniform type.";
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return false;
}
//also, verify that it's the same size
if (bindings[set][i].descriptorCount != layout_binding.descriptorCount || uniform_infos[set][i].length != info.length) {
if (r_error) {
[Complex Test Layouts] Change `String` to use UTF-32 encoding on all platforms.
2020-07-27 10:43:20 +00:00
*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(binding) + " with different uniform size.";
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return false;
}
//just append stage mask and return
bindings.write[set].write[i].stageFlags |= shader_stage_masks[p_stage];
uniform_infos.write[set].write[i].stages |= 1 << p_stage;
return true;
}
}
}
layout_binding.binding = binding;
layout_binding.stageFlags = shader_stage_masks[p_stage];
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
layout_binding.pImmutableSamplers = nullptr; //no support for this yet
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
info.stages = 1 << p_stage;
info.binding = binding;
if (set >= (uint32_t)bindings.size()) {
bindings.resize(set + 1);
uniform_infos.resize(set + 1);
}
Basic 2D engine is more or less working, needs more work for editor to be usable.
2019-06-19 20:03:19 +00:00
#if 0
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
print_line("stage: " + String(shader_stage_names[p_stage]) + " set: " + itos(set) + " binding: " + itos(info.binding) + " type:" + shader_uniform_names[info.type] + " length: " + itos(info.length));
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
bindings.write[set].push_back(layout_binding);
uniform_infos.write[set].push_back(info);
return true;
}
Moved the shader source compilation code outside RenderingDevice and Vulkan
2019-07-28 22:58:32 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Fixes to mobile renderer * Make sure shaders are named, to aid in debug in case of failure * SceneRenderRD was being wrongly initialized (virtual functions being called when derivative class not initialized). * Fixed some bugs resulting on the above being corrected.
2021-08-16 17:51:29 +00:00
//version 1: initial
//version 2: Added shader name
#define SHADER_BINARY_VERSION 2
Added support for push constants
2019-06-08 20:10:52 +00:00
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
String RenderingDeviceVulkan::shader_get_binary_cache_key() const {
return "Vulkan-SV" + itos(SHADER_BINARY_VERSION);
}
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
struct RenderingDeviceVulkanShaderBinaryDataBinding {
uint32_t type;
uint32_t binding;
uint32_t stages;
uint32_t length; //size of arrays (in total elements), or ubos (in bytes * total elements)
};
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
struct RenderingDeviceVulkanShaderBinarySpecializationConstant {
uint32_t type;
uint32_t constant_id;
union {
uint32_t int_value;
float float_value;
bool bool_value;
};
uint32_t stage_flags;
};
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
struct RenderingDeviceVulkanShaderBinaryData {
uint32_t vertex_input_mask;
uint32_t fragment_outputs;
uint32_t specialization_constant_count;
uint32_t is_compute;
uint32_t compute_local_size[3];
uint32_t set_count;
uint32_t push_constant_size;
uint32_t push_constants_vk_stage;
uint32_t stage_count;
Fixes to mobile renderer * Make sure shaders are named, to aid in debug in case of failure * SceneRenderRD was being wrongly initialized (virtual functions being called when derivative class not initialized). * Fixed some bugs resulting on the above being corrected.
2021-08-16 17:51:29 +00:00
uint32_t shader_name_len;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
};
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
Fixes to mobile renderer * Make sure shaders are named, to aid in debug in case of failure * SceneRenderRD was being wrongly initialized (virtual functions being called when derivative class not initialized). * Fixed some bugs resulting on the above being corrected.
2021-08-16 17:51:29 +00:00
Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Vector<ShaderStageSPIRVData> &p_spirv, const String &p_shader_name) {
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
RenderingDeviceVulkanShaderBinaryData binary_data;
binary_data.vertex_input_mask = 0;
binary_data.fragment_outputs = 0;
binary_data.specialization_constant_count = 0;
binary_data.is_compute = 0;
binary_data.compute_local_size[0] = 0;
binary_data.compute_local_size[1] = 0;
binary_data.compute_local_size[2] = 0;
binary_data.set_count = 0;
binary_data.push_constant_size = 0;
binary_data.push_constants_vk_stage = 0;
Vector<Vector<RenderingDeviceVulkanShaderBinaryDataBinding>> uniform_info; //set bindings
Vector<RenderingDeviceVulkanShaderBinarySpecializationConstant> specialization_constants;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
uint32_t stages_processed = 0;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
for (int i = 0; i < p_spirv.size(); i++) {
if (p_spirv[i].shader_stage == SHADER_STAGE_COMPUTE) {
binary_data.is_compute = true;
ERR_FAIL_COND_V_MSG(p_spirv.size() != 1, Vector<uint8_t>(),
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
"Compute shaders can only receive one stage, dedicated to compute.");
}
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(stages_processed & (1 << p_spirv[i].shader_stage), Vector<uint8_t>(),
"Stage " + String(shader_stage_names[p_spirv[i].shader_stage]) + " submitted more than once.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
{
SpvReflectShaderModule module;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
const uint8_t *spirv = p_spirv[i].spir_v.ptr();
SpvReflectResult result = spvReflectCreateShaderModule(p_spirv[i].spir_v.size(), spirv, &module);
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed parsing shader.");
if (binary_data.is_compute) {
binary_data.compute_local_size[0] = module.entry_points->local_size.x;
binary_data.compute_local_size[1] = module.entry_points->local_size.y;
binary_data.compute_local_size[2] = module.entry_points->local_size.z;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
uint32_t binding_count = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
result = spvReflectEnumerateDescriptorBindings(&module, &binding_count, nullptr);
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating descriptor bindings.");
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
uint32_t stage = p_spirv[i].shader_stage;
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
if (binding_count > 0) {
//Parse bindings
Vector<SpvReflectDescriptorBinding *> bindings;
bindings.resize(binding_count);
result = spvReflectEnumerateDescriptorBindings(&module, &binding_count, bindings.ptrw());
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed getting descriptor bindings.");
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
for (uint32_t j = 0; j < binding_count; j++) {
const SpvReflectDescriptorBinding &binding = *bindings[j];
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
RenderingDeviceVulkanShaderBinaryDataBinding info;
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
bool need_array_dimensions = false;
bool need_block_size = false;
switch (binding.descriptor_type) {
case SPV_REFLECT_DESCRIPTOR_TYPE_SAMPLER: {
info.type = UNIFORM_TYPE_SAMPLER;
need_array_dimensions = true;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER: {
info.type = UNIFORM_TYPE_SAMPLER_WITH_TEXTURE;
need_array_dimensions = true;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_SAMPLED_IMAGE: {
info.type = UNIFORM_TYPE_TEXTURE;
need_array_dimensions = true;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_IMAGE: {
info.type = UNIFORM_TYPE_IMAGE;
need_array_dimensions = true;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER: {
info.type = UNIFORM_TYPE_TEXTURE_BUFFER;
need_array_dimensions = true;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER: {
info.type = UNIFORM_TYPE_IMAGE_BUFFER;
need_array_dimensions = true;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_UNIFORM_BUFFER: {
info.type = UNIFORM_TYPE_UNIFORM_BUFFER;
need_block_size = true;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_BUFFER: {
info.type = UNIFORM_TYPE_STORAGE_BUFFER;
need_block_size = true;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC: {
ERR_PRINT("Dynamic uniform buffer not supported.");
continue;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC: {
ERR_PRINT("Dynamic storage buffer not supported.");
continue;
} break;
case SPV_REFLECT_DESCRIPTOR_TYPE_INPUT_ATTACHMENT: {
info.type = UNIFORM_TYPE_INPUT_ATTACHMENT;
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
need_array_dimensions = true;
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
} break;
spirv-reflect: Update to upstream commit c0ce03a (Jan 6, 2021)
2021-01-08 13:17:49 +00:00
case SPV_REFLECT_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR: {
ERR_PRINT("Acceleration structure not supported.");
continue;
} break;
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
}
if (need_array_dimensions) {
if (binding.array.dims_count == 0) {
info.length = 1;
} else {
for (uint32_t k = 0; k < binding.array.dims_count; k++) {
if (k == 0) {
info.length = binding.array.dims[0];
} else {
info.length *= binding.array.dims[k];
}
}
}
} else if (need_block_size) {
info.length = binding.block.size;
} else {
info.length = 0;
}
info.binding = binding.binding;
uint32_t set = binding.set;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(set >= MAX_UNIFORM_SETS, Vector<uint8_t>(),
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
"On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' uses a set (" + itos(set) + ") index larger than what is supported (" + itos(MAX_UNIFORM_SETS) + ").");
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(set >= limits.maxBoundDescriptorSets, Vector<uint8_t>(),
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
"On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' uses a set (" + itos(set) + ") index larger than what is supported by the hardware (" + itos(limits.maxBoundDescriptorSets) + ").");
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
if (set < (uint32_t)uniform_info.size()) {
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
//check if this already exists
bool exists = false;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
for (int k = 0; k < uniform_info[set].size(); k++) {
if (uniform_info[set][k].binding == (uint32_t)info.binding) {
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
//already exists, verify that it's the same type
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(uniform_info[set][k].type != info.type, Vector<uint8_t>(),
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
"On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(info.binding) + " with different uniform type.");
//also, verify that it's the same size
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(uniform_info[set][k].length != info.length, Vector<uint8_t>(),
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
"On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(info.binding) + " with different uniform size.");
//just append stage mask and return
uniform_info.write[set].write[k].stages |= 1 << stage;
exists = true;
}
}
if (exists) {
continue; //merged
}
}
info.stages = 1 << stage;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
if (set >= (uint32_t)uniform_info.size()) {
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
uniform_info.resize(set + 1);
}
uniform_info.write[set].push_back(info);
}
}
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
{
//specialization constants
uint32_t sc_count = 0;
result = spvReflectEnumerateSpecializationConstants(&module, &sc_count, nullptr);
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating specialization constants.");
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
if (sc_count) {
Vector<SpvReflectSpecializationConstant *> spec_constants;
spec_constants.resize(sc_count);
result = spvReflectEnumerateSpecializationConstants(&module, &sc_count, spec_constants.ptrw());
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining specialization constants.");
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
for (uint32_t j = 0; j < sc_count; j++) {
int32_t existing = -1;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
RenderingDeviceVulkanShaderBinarySpecializationConstant sconst;
Fixing retrospective code for specialisation constants
2022-02-28 13:10:33 +00:00
SpvReflectSpecializationConstant *spc = spec_constants[j];
sconst.constant_id = spc->constant_id;
sconst.int_value = 0.0; // clear previous value JIC
switch (spc->constant_type) {
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
case SPV_REFLECT_SPECIALIZATION_CONSTANT_BOOL: {
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
sconst.type = PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL;
Fixing retrospective code for specialisation constants
2022-02-28 13:10:33 +00:00
sconst.bool_value = spc->default_value.int_bool_value != 0;
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
} break;
case SPV_REFLECT_SPECIALIZATION_CONSTANT_INT: {
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
sconst.type = PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT;
Fixing retrospective code for specialisation constants
2022-02-28 13:10:33 +00:00
sconst.int_value = spc->default_value.int_bool_value;
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
} break;
case SPV_REFLECT_SPECIALIZATION_CONSTANT_FLOAT: {
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
sconst.type = PIPELINE_SPECIALIZATION_CONSTANT_TYPE_FLOAT;
Fixing retrospective code for specialisation constants
2022-02-28 13:10:33 +00:00
sconst.float_value = spc->default_value.float_value;
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
} break;
}
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
sconst.stage_flags = 1 << p_spirv[i].shader_stage;
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
for (int k = 0; k < specialization_constants.size(); k++) {
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
if (specialization_constants[k].constant_id == sconst.constant_id) {
ERR_FAIL_COND_V_MSG(specialization_constants[k].type != sconst.type, Vector<uint8_t>(), "More than one specialization constant used for id (" + itos(sconst.constant_id) + "), but their types differ.");
ERR_FAIL_COND_V_MSG(specialization_constants[k].int_value != sconst.int_value, Vector<uint8_t>(), "More than one specialization constant used for id (" + itos(sconst.constant_id) + "), but their default values differ.");
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
existing = k;
break;
}
}
if (existing > 0) {
specialization_constants.write[existing].stage_flags |= sconst.stage_flags;
} else {
specialization_constants.push_back(sconst);
}
}
}
}
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
if (stage == SHADER_STAGE_VERTEX) {
uint32_t iv_count = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
result = spvReflectEnumerateInputVariables(&module, &iv_count, nullptr);
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating input variables.");
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
if (iv_count) {
Vector<SpvReflectInterfaceVariable *> input_vars;
input_vars.resize(iv_count);
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
result = spvReflectEnumerateInputVariables(&module, &iv_count, input_vars.ptrw());
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining input variables.");
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
for (uint32_t j = 0; j < iv_count; j++) {
Yay very basic 3D (only white) finally shows.
2019-08-20 20:54:03 +00:00
if (input_vars[j] && input_vars[j]->decoration_flags == 0) { //regular input
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
binary_data.vertex_input_mask |= (1 << uint32_t(input_vars[j]->location));
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
}
}
}
}
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
if (stage == SHADER_STAGE_FRAGMENT) {
uint32_t ov_count = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
result = spvReflectEnumerateOutputVariables(&module, &ov_count, nullptr);
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating output variables.");
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
if (ov_count) {
Vector<SpvReflectInterfaceVariable *> output_vars;
output_vars.resize(ov_count);
result = spvReflectEnumerateOutputVariables(&module, &ov_count, output_vars.ptrw());
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining output variables.");
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
for (uint32_t j = 0; j < ov_count; j++) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
const SpvReflectInterfaceVariable *refvar = output_vars[j];
if (refvar != nullptr && refvar->built_in != SpvBuiltInFragDepth) {
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
binary_data.fragment_outputs |= 1 << refvar->location;
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
}
}
}
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
uint32_t pc_count = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
result = spvReflectEnumeratePushConstantBlocks(&module, &pc_count, nullptr);
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating push constants.");
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
if (pc_count) {
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(pc_count > 1, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "': Only one push constant is supported, which should be the same across shader stages.");
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
Vector<SpvReflectBlockVariable *> pconstants;
pconstants.resize(pc_count);
result = spvReflectEnumeratePushConstantBlocks(&module, &pc_count, pconstants.ptrw());
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining push constants.");
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
#if 0
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
if (pconstants[0] == nullptr) {
Make FileAccess and DirAccess classes reference counted.
2022-03-23 09:08:58 +00:00
Ref<FileAccess> f = FileAccess::open("res://popo.spv", FileAccess::WRITE);
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
f->store_buffer((const uint8_t *)&SpirV[0], SpirV.size() * sizeof(uint32_t));
}
#endif
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
ERR_FAIL_COND_V_MSG(binary_data.push_constant_size && binary_data.push_constant_size != pconstants[0]->size, Vector<uint8_t>(),
"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "': Push constant block must be the same across shader stages.");
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
binary_data.push_constant_size = pconstants[0]->size;
binary_data.push_constants_vk_stage |= shader_stage_masks[stage];
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
//print_line("Stage: " + String(shader_stage_names[stage]) + " push constant of size=" + itos(push_constant.push_constant_size));
}
// Destroy the reflection data when no longer required.
spvReflectDestroyShaderModule(&module);
}
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
stages_processed |= (1 << p_spirv[i].shader_stage);
}
Vector<Vector<uint8_t>> compressed_stages;
Vector<uint32_t> smolv_size;
Vector<uint32_t> zstd_size; //if 0, stdno t used
uint32_t stages_binary_size = 0;
bool strip_debug = false;
for (int i = 0; i < p_spirv.size(); i++) {
smolv::ByteArray smolv;
if (!smolv::Encode(p_spirv[i].spir_v.ptr(), p_spirv[i].spir_v.size(), smolv, strip_debug ? smolv::kEncodeFlagStripDebugInfo : 0)) {
ERR_FAIL_V_MSG(Vector<uint8_t>(), "Error compressing shader stage :" + String(shader_stage_names[p_spirv[i].shader_stage]));
} else {
smolv_size.push_back(smolv.size());
{ //zstd
Vector<uint8_t> zstd;
zstd.resize(Compression::get_max_compressed_buffer_size(smolv.size(), Compression::MODE_ZSTD));
int dst_size = Compression::compress(zstd.ptrw(), &smolv[0], smolv.size(), Compression::MODE_ZSTD);
if (dst_size > 0 && (uint32_t)dst_size < smolv.size()) {
zstd_size.push_back(dst_size);
zstd.resize(dst_size);
compressed_stages.push_back(zstd);
} else {
Vector<uint8_t> smv;
smv.resize(smolv.size());
memcpy(smv.ptrw(), &smolv[0], smolv.size());
zstd_size.push_back(0); //not using zstd
compressed_stages.push_back(smv);
}
}
}
uint32_t s = compressed_stages[i].size();
if (s % 4 != 0) {
s += 4 - (s % 4);
}
stages_binary_size += s;
}
binary_data.specialization_constant_count = specialization_constants.size();
binary_data.set_count = uniform_info.size();
binary_data.stage_count = p_spirv.size();
Fixes to mobile renderer * Make sure shaders are named, to aid in debug in case of failure * SceneRenderRD was being wrongly initialized (virtual functions being called when derivative class not initialized). * Fixed some bugs resulting on the above being corrected.
2021-08-16 17:51:29 +00:00
CharString shader_name_utf = p_shader_name.utf8();
binary_data.shader_name_len = shader_name_utf.length();
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
uint32_t total_size = sizeof(uint32_t) * 3; //header + version + main datasize;
total_size += sizeof(RenderingDeviceVulkanShaderBinaryData);
Fixes to mobile renderer * Make sure shaders are named, to aid in debug in case of failure * SceneRenderRD was being wrongly initialized (virtual functions being called when derivative class not initialized). * Fixed some bugs resulting on the above being corrected.
2021-08-16 17:51:29 +00:00
total_size += binary_data.shader_name_len;
if ((binary_data.shader_name_len % 4) != 0) { //alignment rules are really strange
total_size += 4 - (binary_data.shader_name_len % 4);
}
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
for (int i = 0; i < uniform_info.size(); i++) {
total_size += sizeof(uint32_t);
total_size += uniform_info[i].size() * sizeof(RenderingDeviceVulkanShaderBinaryDataBinding);
}
total_size += sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) * specialization_constants.size();
total_size += compressed_stages.size() * sizeof(uint32_t) * 3; //sizes
total_size += stages_binary_size;
Vector<uint8_t> ret;
ret.resize(total_size);
{
Fix some issues found by cppcheck.
2022-04-05 10:40:26 +00:00
uint32_t offset = 0;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
uint8_t *binptr = ret.ptrw();
binptr[0] = 'G';
binptr[1] = 'V';
binptr[2] = 'B';
binptr[3] = 'D'; //godot vulkan binary data
offset += 4;
encode_uint32(SHADER_BINARY_VERSION, binptr + offset);
offset += sizeof(uint32_t);
encode_uint32(sizeof(RenderingDeviceVulkanShaderBinaryData), binptr + offset);
offset += sizeof(uint32_t);
memcpy(binptr + offset, &binary_data, sizeof(RenderingDeviceVulkanShaderBinaryData));
offset += sizeof(RenderingDeviceVulkanShaderBinaryData);
Fixes to mobile renderer * Make sure shaders are named, to aid in debug in case of failure * SceneRenderRD was being wrongly initialized (virtual functions being called when derivative class not initialized). * Fixed some bugs resulting on the above being corrected.
2021-08-16 17:51:29 +00:00
memcpy(binptr + offset, shader_name_utf.ptr(), binary_data.shader_name_len);
offset += binary_data.shader_name_len;
if ((binary_data.shader_name_len % 4) != 0) { //alignment rules are really strange
offset += 4 - (binary_data.shader_name_len % 4);
}
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
for (int i = 0; i < uniform_info.size(); i++) {
int count = uniform_info[i].size();
encode_uint32(count, binptr + offset);
offset += sizeof(uint32_t);
if (count > 0) {
memcpy(binptr + offset, uniform_info[i].ptr(), sizeof(RenderingDeviceVulkanShaderBinaryDataBinding) * count);
offset += sizeof(RenderingDeviceVulkanShaderBinaryDataBinding) * count;
}
}
if (specialization_constants.size()) {
memcpy(binptr + offset, specialization_constants.ptr(), sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) * specialization_constants.size());
offset += sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) * specialization_constants.size();
}
for (int i = 0; i < compressed_stages.size(); i++) {
encode_uint32(p_spirv[i].shader_stage, binptr + offset);
offset += sizeof(uint32_t);
encode_uint32(smolv_size[i], binptr + offset);
offset += sizeof(uint32_t);
encode_uint32(zstd_size[i], binptr + offset);
offset += sizeof(uint32_t);
memcpy(binptr + offset, compressed_stages[i].ptr(), compressed_stages[i].size());
uint32_t s = compressed_stages[i].size();
if (s % 4 != 0) {
s += 4 - (s % 4);
}
offset += s;
}
ERR_FAIL_COND_V(offset != (uint32_t)ret.size(), Vector<uint8_t>());
}
return ret;
}
RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_shader_binary) {
const uint8_t *binptr = p_shader_binary.ptr();
uint32_t binsize = p_shader_binary.size();
uint32_t read_offset = 0;
//consistency check
ERR_FAIL_COND_V(binsize < sizeof(uint32_t) * 3 + sizeof(RenderingDeviceVulkanShaderBinaryData), RID());
ERR_FAIL_COND_V(binptr[0] != 'G' || binptr[1] != 'V' || binptr[2] != 'B' || binptr[3] != 'D', RID());
uint32_t bin_version = decode_uint32(binptr + 4);
ERR_FAIL_COND_V(bin_version > SHADER_BINARY_VERSION, RID());
uint32_t bin_data_size = decode_uint32(binptr + 8);
Fix some issues found by cppcheck.
2022-04-05 10:40:26 +00:00
const RenderingDeviceVulkanShaderBinaryData &binary_data = *(reinterpret_cast<const RenderingDeviceVulkanShaderBinaryData *>(binptr + 12));
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
Shader::PushConstant push_constant;
push_constant.push_constant_size = binary_data.push_constant_size;
push_constant.push_constants_vk_stage = binary_data.push_constants_vk_stage;
uint32_t vertex_input_mask = binary_data.vertex_input_mask;
uint32_t fragment_outputs = binary_data.fragment_outputs;
bool is_compute = binary_data.is_compute;
Fix some issues found by cppcheck.
2022-04-05 10:40:26 +00:00
const uint32_t compute_local_size[3] = { binary_data.compute_local_size[0], binary_data.compute_local_size[1], binary_data.compute_local_size[2] };
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
read_offset += sizeof(uint32_t) * 3 + bin_data_size;
Fixes to mobile renderer * Make sure shaders are named, to aid in debug in case of failure * SceneRenderRD was being wrongly initialized (virtual functions being called when derivative class not initialized). * Fixed some bugs resulting on the above being corrected.
2021-08-16 17:51:29 +00:00
String name;
if (binary_data.shader_name_len) {
name.parse_utf8((const char *)(binptr + read_offset), binary_data.shader_name_len);
read_offset += binary_data.shader_name_len;
if ((binary_data.shader_name_len % 4) != 0) { //alignment rules are really strange
read_offset += 4 - (binary_data.shader_name_len % 4);
}
}
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
Vector<Vector<VkDescriptorSetLayoutBinding>> set_bindings;
Vector<Vector<UniformInfo>> uniform_info;
set_bindings.resize(binary_data.set_count);
uniform_info.resize(binary_data.set_count);
for (uint32_t i = 0; i < binary_data.set_count; i++) {
ERR_FAIL_COND_V(read_offset + sizeof(uint32_t) >= binsize, RID());
uint32_t set_count = decode_uint32(binptr + read_offset);
read_offset += sizeof(uint32_t);
Fix some issues found by cppcheck.
2022-04-05 10:40:26 +00:00
const RenderingDeviceVulkanShaderBinaryDataBinding *set_ptr = reinterpret_cast<const RenderingDeviceVulkanShaderBinaryDataBinding *>(binptr + read_offset);
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
uint32_t set_size = set_count * sizeof(RenderingDeviceVulkanShaderBinaryDataBinding);
ERR_FAIL_COND_V(read_offset + set_size >= binsize, RID());
for (uint32_t j = 0; j < set_count; j++) {
UniformInfo info;
info.type = UniformType(set_ptr[j].type);
info.length = set_ptr[j].length;
info.binding = set_ptr[j].binding;
info.stages = set_ptr[j].stages;
VkDescriptorSetLayoutBinding layout_binding;
layout_binding.pImmutableSamplers = nullptr;
layout_binding.binding = set_ptr[j].binding;
layout_binding.descriptorCount = 1;
layout_binding.stageFlags = 0;
for (uint32_t k = 0; k < SHADER_STAGE_MAX; k++) {
if (set_ptr[j].stages & (1 << k)) {
layout_binding.stageFlags |= shader_stage_masks[k];
}
}
switch (info.type) {
case UNIFORM_TYPE_SAMPLER: {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLER;
layout_binding.descriptorCount = set_ptr[j].length;
} break;
case UNIFORM_TYPE_SAMPLER_WITH_TEXTURE: {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
layout_binding.descriptorCount = set_ptr[j].length;
} break;
case UNIFORM_TYPE_TEXTURE: {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE;
layout_binding.descriptorCount = set_ptr[j].length;
} break;
case UNIFORM_TYPE_IMAGE: {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
layout_binding.descriptorCount = set_ptr[j].length;
} break;
case UNIFORM_TYPE_TEXTURE_BUFFER: {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
layout_binding.descriptorCount = set_ptr[j].length;
} break;
case UNIFORM_TYPE_IMAGE_BUFFER: {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER;
} break;
case UNIFORM_TYPE_UNIFORM_BUFFER: {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
} break;
case UNIFORM_TYPE_STORAGE_BUFFER: {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
} break;
case UNIFORM_TYPE_INPUT_ATTACHMENT: {
layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT;
} break;
default: {
ERR_FAIL_V(RID());
}
}
set_bindings.write[i].push_back(layout_binding);
uniform_info.write[i].push_back(info);
}
read_offset += set_size;
}
ERR_FAIL_COND_V(read_offset + binary_data.specialization_constant_count * sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) >= binsize, RID());
Vector<Shader::SpecializationConstant> specialization_constants;
for (uint32_t i = 0; i < binary_data.specialization_constant_count; i++) {
Fix some issues found by cppcheck.
2022-04-05 10:40:26 +00:00
const RenderingDeviceVulkanShaderBinarySpecializationConstant &src_sc = *(reinterpret_cast<const RenderingDeviceVulkanShaderBinarySpecializationConstant *>(binptr + read_offset));
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
Shader::SpecializationConstant sc;
sc.constant.int_value = src_sc.int_value;
sc.constant.type = PipelineSpecializationConstantType(src_sc.type);
sc.constant.constant_id = src_sc.constant_id;
sc.stage_flags = src_sc.stage_flags;
specialization_constants.push_back(sc);
read_offset += sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
Vector<Vector<uint8_t>> stage_spirv_data;
Vector<ShaderStage> stage_type;
for (uint32_t i = 0; i < binary_data.stage_count; i++) {
ERR_FAIL_COND_V(read_offset + sizeof(uint32_t) * 3 >= binsize, RID());
uint32_t stage = decode_uint32(binptr + read_offset);
read_offset += sizeof(uint32_t);
uint32_t smolv_size = decode_uint32(binptr + read_offset);
read_offset += sizeof(uint32_t);
uint32_t zstd_size = decode_uint32(binptr + read_offset);
read_offset += sizeof(uint32_t);
uint32_t buf_size = (zstd_size > 0) ? zstd_size : smolv_size;
Vector<uint8_t> smolv;
const uint8_t *src_smolv = nullptr;
if (zstd_size > 0) {
//decompress to smolv
smolv.resize(smolv_size);
int dec_smolv_size = Compression::decompress(smolv.ptrw(), smolv.size(), binptr + read_offset, zstd_size, Compression::MODE_ZSTD);
ERR_FAIL_COND_V(dec_smolv_size != (int32_t)smolv_size, RID());
src_smolv = smolv.ptr();
} else {
src_smolv = binptr + read_offset;
}
Vector<uint8_t> spirv;
uint32_t spirv_size = smolv::GetDecodedBufferSize(src_smolv, smolv_size);
spirv.resize(spirv_size);
if (!smolv::Decode(src_smolv, smolv_size, spirv.ptrw(), spirv_size)) {
ERR_FAIL_V_MSG(RID(), "Malformed smolv input uncompressing shader stage:" + String(shader_stage_names[stage]));
}
stage_spirv_data.push_back(spirv);
stage_type.push_back(ShaderStage(stage));
if (buf_size % 4 != 0) {
buf_size += 4 - (buf_size % 4);
}
ERR_FAIL_COND_V(read_offset + buf_size > binsize, RID());
read_offset += buf_size;
}
ERR_FAIL_COND_V(read_offset != binsize, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//all good, let's create modules
Added a spinlock template as well as a thread work pool class. Also, optimized shader compilation to happen on threads.
2019-07-29 15:59:18 +00:00
_THREAD_SAFE_METHOD_
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Shader shader;
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
shader.vertex_input_mask = vertex_input_mask;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
shader.fragment_output_mask = fragment_outputs;
Added support for push constants
2019-06-08 20:10:52 +00:00
shader.push_constant = push_constant;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
shader.is_compute = is_compute;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
shader.compute_local_size[0] = compute_local_size[0];
shader.compute_local_size[1] = compute_local_size[1];
shader.compute_local_size[2] = compute_local_size[2];
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
shader.specialization_constants = specialization_constants;
Fixes to mobile renderer * Make sure shaders are named, to aid in debug in case of failure * SceneRenderRD was being wrongly initialized (virtual functions being called when derivative class not initialized). * Fixed some bugs resulting on the above being corrected.
2021-08-16 17:51:29 +00:00
shader.name = name;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
String error_text;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
bool success = true;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
for (int i = 0; i < stage_spirv_data.size(); i++) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkShaderModuleCreateInfo shader_module_create_info;
shader_module_create_info.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
shader_module_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
shader_module_create_info.flags = 0;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
shader_module_create_info.codeSize = stage_spirv_data[i].size();
const uint8_t *r = stage_spirv_data[i].ptr();
Moved the shader source compilation code outside RenderingDevice and Vulkan
2019-07-28 22:58:32 +00:00
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
shader_module_create_info.pCode = (const uint32_t *)r;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkShaderModule module;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult res = vkCreateShaderModule(device, &shader_module_create_info, nullptr, &module);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (res) {
success = false;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
error_text = "Error (" + itos(res) + ") creating shader module for stage: " + String(shader_stage_names[stage_type[i]]);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
break;
}
const VkShaderStageFlagBits shader_stage_bits[SHADER_STAGE_MAX] = {
VK_SHADER_STAGE_VERTEX_BIT,
VK_SHADER_STAGE_FRAGMENT_BIT,
VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT,
VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT,
VK_SHADER_STAGE_COMPUTE_BIT,
};
VkPipelineShaderStageCreateInfo shader_stage;
shader_stage.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
shader_stage.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
shader_stage.flags = 0;
Implement Binary Shader Compilation * Added an extra stage before compiling shader, which is generating a binary blob. * On Vulkan, this allows caching the SPIRV reflection information, which is expensive to parse. * On other (future) RenderingDevices, it allows caching converted binary data, such as DXIL or MSL. This PR makes the shader cache include the reflection information, hence editor startup times are significantly improved. I tested this well and it appears to work, and I added a lot of consistency checks, but because it includes writing and reading binary information, rare bugs may pop up, so be aware. There was not much of a choice for storing the reflection information, given shaders can be a lot, take a lot of space and take time to parse.
2021-07-25 14:22:55 +00:00
shader_stage.stage = shader_stage_bits[stage_type[i]];
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
shader_stage.module = module;
shader_stage.pName = "main";
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
shader_stage.pSpecializationInfo = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
shader.pipeline_stages.push_back(shader_stage);
}
//proceed to create descriptor sets
if (success) {
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
for (int i = 0; i < set_bindings.size(); i++) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//empty ones are fine if they were not used according to spec (binding count will be 0)
VkDescriptorSetLayoutCreateInfo layout_create_info;
layout_create_info.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
layout_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
layout_create_info.flags = 0;
Replaced GLSLang reflection by SPIRV-Reflect, eventually allowing to move GLSLang out.
2019-07-28 16:42:15 +00:00
layout_create_info.bindingCount = set_bindings[i].size();
layout_create_info.pBindings = set_bindings[i].ptr();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkDescriptorSetLayout layout;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult res = vkCreateDescriptorSetLayout(device, &layout_create_info, nullptr, &layout);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (res) {
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
error_text = "Error (" + itos(res) + ") creating descriptor set layout for set " + itos(i);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
success = false;
break;
}
Shader::Set set;
set.descriptor_set_layout = layout;
set.uniform_info = uniform_info[i];
//sort and hash
set.uniform_info.sort();
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
uint32_t format = 0; //no format, default
if (set.uniform_info.size()) {
//has data, needs an actual format;
UniformSetFormat usformat;
usformat.uniform_info = set.uniform_info;
Map<UniformSetFormat, uint32_t>::Element *E = uniform_set_format_cache.find(usformat);
if (E) {
format = E->get();
} else {
format = uniform_set_format_cache.size() + 1;
uniform_set_format_cache.insert(usformat, format);
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
shader.sets.push_back(set);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
shader.set_formats.push_back(format);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
if (success) {
//create pipeline layout
VkPipelineLayoutCreateInfo pipeline_layout_create_info;
pipeline_layout_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
pipeline_layout_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
pipeline_layout_create_info.flags = 0;
pipeline_layout_create_info.setLayoutCount = shader.sets.size();
Vector<VkDescriptorSetLayout> layouts;
layouts.resize(shader.sets.size());
for (int i = 0; i < layouts.size(); i++) {
layouts.write[i] = shader.sets[i].descriptor_set_layout;
}
pipeline_layout_create_info.pSetLayouts = layouts.ptr();
Fix crash in RenderingDeviceVulkan::shader_create This commit moves the declaration of a local variable to ensure its scope survives long enough; at least in some versions of GCC and LLVM the associated memory was freed too early and thus caused issues ranging from black screens to crashes.
2020-10-24 22:56:54 +00:00
// Needs to be declared in this outer scope, otherwise it may not outlive its assignment
// to pipeline_layout_create_info.
VkPushConstantRange push_constant_range;
Added support for push constants
2019-06-08 20:10:52 +00:00
if (push_constant.push_constant_size) {
push_constant_range.stageFlags = push_constant.push_constants_vk_stage;
push_constant_range.offset = 0;
push_constant_range.size = push_constant.push_constant_size;
pipeline_layout_create_info.pushConstantRangeCount = 1;
pipeline_layout_create_info.pPushConstantRanges = &push_constant_range;
} else {
pipeline_layout_create_info.pushConstantRangeCount = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
pipeline_layout_create_info.pPushConstantRanges = nullptr;
Added support for push constants
2019-06-08 20:10:52 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult err = vkCreatePipelineLayout(device, &pipeline_layout_create_info, nullptr, &shader.pipeline_layout);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (err) {
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
error_text = "Error (" + itos(err) + ") creating pipeline layout.";
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
success = false;
}
}
if (!success) {
//clean up if failed
for (int i = 0; i < shader.pipeline_stages.size(); i++) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyShaderModule(device, shader.pipeline_stages[i].module, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
for (int i = 0; i < shader.sets.size(); i++) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyDescriptorSetLayout(device, shader.sets[i].descriptor_set_layout, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Moved the shader source compilation code outside RenderingDevice and Vulkan
2019-07-28 22:58:32 +00:00
ERR_FAIL_V_MSG(RID(), error_text);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
return shader_owner.make_rid(shader);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
uint32_t RenderingDeviceVulkan::shader_get_vertex_input_attribute_mask(RID p_shader) {
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
_THREAD_SAFE_METHOD_
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
const Shader *shader = shader_owner.get_or_null(p_shader);
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
ERR_FAIL_COND_V(!shader, 0);
return shader->vertex_input_mask;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
/******************/
/**** UNIFORMS ****/
/******************/
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
RID RenderingDeviceVulkan::uniform_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(p_data.size() && (uint32_t)p_data.size() != p_size_bytes, RID());
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Buffer buffer;
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
Error err = _buffer_allocate(&buffer, p_size_bytes, VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(err != OK, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_data.size()) {
uint64_t data_size = p_data.size();
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint8_t *r = p_data.ptr();
_buffer_update(&buffer, 0, r, data_size);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
_buffer_memory_barrier(buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_UNIFORM_READ_BIT, false);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
return uniform_buffer_owner.make_rid(buffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RID RenderingDeviceVulkan::storage_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data, uint32_t p_usage) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(p_data.size() && (uint32_t)p_data.size() != p_size_bytes, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Buffer buffer;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
buffer.usage = p_usage;
uint32_t flags = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
if (p_usage & STORAGE_BUFFER_USAGE_DISPATCH_INDIRECT) {
flags |= VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT;
}
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
Error err = _buffer_allocate(&buffer, p_size_bytes, flags, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(err != OK, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_data.size()) {
uint64_t data_size = p_data.size();
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint8_t *r = p_data.ptr();
_buffer_update(&buffer, 0, r, data_size);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
_buffer_memory_barrier(buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT, false);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
return storage_buffer_owner.make_rid(buffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
RID RenderingDeviceVulkan::texture_buffer_create(uint32_t p_size_elements, DataFormat p_format, const Vector<uint8_t> &p_data) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
"Creating buffers with data is forbidden during creation of a draw list");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
uint32_t element_size = get_format_vertex_size(p_format);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(element_size == 0, RID(), "Format requested is not supported for texture buffers");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
uint64_t size_bytes = uint64_t(element_size) * p_size_elements;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(p_data.size() && (uint32_t)p_data.size() != size_bytes, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
TextureBuffer texture_buffer;
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
Error err = _buffer_allocate(&texture_buffer.buffer, size_bytes, VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(err != OK, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_data.size()) {
uint64_t data_size = p_data.size();
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
const uint8_t *r = p_data.ptr();
_buffer_update(&texture_buffer.buffer, 0, r, data_size);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
_buffer_memory_barrier(texture_buffer.buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_SHADER_READ_BIT, false);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
VkBufferViewCreateInfo view_create_info;
view_create_info.sType = VK_STRUCTURE_TYPE_BUFFER_VIEW_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
view_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
view_create_info.flags = 0;
view_create_info.buffer = texture_buffer.buffer.buffer;
view_create_info.format = vulkan_formats[p_format];
view_create_info.offset = 0;
view_create_info.range = size_bytes;
texture_buffer.view = VK_NULL_HANDLE;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult res = vkCreateBufferView(device, &view_create_info, nullptr, &texture_buffer.view);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (res) {
_buffer_free(&texture_buffer.buffer);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_V_MSG(RID(), "Unable to create buffer view, error " + itos(res) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//allocate the view
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
return texture_buffer_owner.make_rid(texture_buffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
RenderingDeviceVulkan::DescriptorPool *RenderingDeviceVulkan::_descriptor_pool_allocate(const DescriptorPoolKey &p_key) {
if (!descriptor_pools.has(p_key)) {
descriptor_pools[p_key] = Set<DescriptorPool *>();
}
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
DescriptorPool *pool = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (Set<DescriptorPool *>::Element *E = descriptor_pools[p_key].front(); E; E = E->next()) {
if (E->get()->usage < max_descriptors_per_pool) {
pool = E->get();
break;
}
}
if (!pool) {
//create a new one
pool = memnew(DescriptorPool);
pool->usage = 0;
VkDescriptorPoolCreateInfo descriptor_pool_create_info;
descriptor_pool_create_info.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
descriptor_pool_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
descriptor_pool_create_info.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT; // can't think how somebody may NOT need this flag..
descriptor_pool_create_info.maxSets = max_descriptors_per_pool;
Vector<VkDescriptorPoolSize> sizes;
//here comes more vulkan API strangeness
if (p_key.uniform_type[UNIFORM_TYPE_SAMPLER]) {
VkDescriptorPoolSize s;
s.type = VK_DESCRIPTOR_TYPE_SAMPLER;
s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_SAMPLER] * max_descriptors_per_pool;
sizes.push_back(s);
}
if (p_key.uniform_type[UNIFORM_TYPE_SAMPLER_WITH_TEXTURE]) {
VkDescriptorPoolSize s;
s.type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_SAMPLER_WITH_TEXTURE] * max_descriptors_per_pool;
sizes.push_back(s);
}
if (p_key.uniform_type[UNIFORM_TYPE_TEXTURE]) {
VkDescriptorPoolSize s;
s.type = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE;
s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_TEXTURE] * max_descriptors_per_pool;
sizes.push_back(s);
}
if (p_key.uniform_type[UNIFORM_TYPE_IMAGE]) {
VkDescriptorPoolSize s;
s.type = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_IMAGE] * max_descriptors_per_pool;
sizes.push_back(s);
}
if (p_key.uniform_type[UNIFORM_TYPE_TEXTURE_BUFFER] || p_key.uniform_type[UNIFORM_TYPE_SAMPLER_WITH_TEXTURE_BUFFER]) {
VkDescriptorPoolSize s;
s.type = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
s.descriptorCount = (p_key.uniform_type[UNIFORM_TYPE_TEXTURE_BUFFER] + p_key.uniform_type[UNIFORM_TYPE_SAMPLER_WITH_TEXTURE_BUFFER]) * max_descriptors_per_pool;
sizes.push_back(s);
}
if (p_key.uniform_type[UNIFORM_TYPE_IMAGE_BUFFER]) {
VkDescriptorPoolSize s;
s.type = VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER;
s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_IMAGE_BUFFER] * max_descriptors_per_pool;
sizes.push_back(s);
}
if (p_key.uniform_type[UNIFORM_TYPE_UNIFORM_BUFFER]) {
VkDescriptorPoolSize s;
s.type = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_UNIFORM_BUFFER] * max_descriptors_per_pool;
sizes.push_back(s);
}
if (p_key.uniform_type[UNIFORM_TYPE_STORAGE_BUFFER]) {
VkDescriptorPoolSize s;
s.type = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_STORAGE_BUFFER] * max_descriptors_per_pool;
sizes.push_back(s);
}
if (p_key.uniform_type[UNIFORM_TYPE_INPUT_ATTACHMENT]) {
VkDescriptorPoolSize s;
s.type = VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT;
s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_INPUT_ATTACHMENT] * max_descriptors_per_pool;
sizes.push_back(s);
}
descriptor_pool_create_info.poolSizeCount = sizes.size();
descriptor_pool_create_info.pPoolSizes = sizes.ptr();
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult res = vkCreateDescriptorPool(device, &descriptor_pool_create_info, nullptr, &pool->pool);
Fixes bugs found by Sonarcloud and Coverity
2020-03-02 18:17:20 +00:00
if (res) {
memdelete(pool);
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
ERR_FAIL_COND_V_MSG(res, nullptr, "vkCreateDescriptorPool failed with error " + itos(res) + ".");
Fixes bugs found by Sonarcloud and Coverity
2020-03-02 18:17:20 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
descriptor_pools[p_key].insert(pool);
}
pool->usage++;
return pool;
}
void RenderingDeviceVulkan::_descriptor_pool_free(const DescriptorPoolKey &p_key, DescriptorPool *p_pool) {
#ifdef DEBUG_ENABLED
ERR_FAIL_COND(!descriptor_pools[p_key].has(p_pool));
#endif
ERR_FAIL_COND(p_pool->usage == 0);
p_pool->usage--;
if (p_pool->usage == 0) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyDescriptorPool(device, p_pool->pool, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
descriptor_pools[p_key].erase(p_pool);
memdelete(p_pool);
Rename empty() to is_empty()
2020-12-15 12:04:21 +00:00
if (descriptor_pools[p_key].is_empty()) {
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
descriptor_pools.erase(p_key);
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, RID p_shader, uint32_t p_shader_set) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(p_uniforms.size() == 0, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Shader *shader = shader_owner.get_or_null(p_shader);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!shader, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(p_shader_set >= (uint32_t)shader->sets.size() || shader->sets[p_shader_set].uniform_info.size() == 0, RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Desired set (" + itos(p_shader_set) + ") not used by shader.");
//see that all sets in shader are satisfied
const Shader::Set &set = shader->sets[p_shader_set];
uint32_t uniform_count = p_uniforms.size();
const Uniform *uniforms = p_uniforms.ptr();
uint32_t set_uniform_count = set.uniform_info.size();
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
const UniformInfo *set_uniforms = set.uniform_info.ptr();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Vector<VkWriteDescriptorSet> writes;
DescriptorPoolKey pool_key;
//to keep them alive until update call
Style: Set clang-format Standard to Cpp11 For us, it practically only changes the fact that `A<A<int>>` is now used instead of the C++03 compatible `A<A<int> >`. Note: clang-format 10+ changed the `Standard` arguments to fully specified `c++11`, `c++14`, etc. versions, but we can't use `c++17` now if we want to preserve compatibility with clang-format 8 and 9. `Cpp11` is still supported as deprecated alias for `Latest`.
2020-03-17 06:33:00 +00:00
List<Vector<VkDescriptorBufferInfo>> buffer_infos;
List<Vector<VkBufferView>> buffer_views;
List<Vector<VkDescriptorImageInfo>> image_infos;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//used for verification to make sure a uniform set does not use a framebuffer bound texture
Fix error spam on wrong attachment -For now, disable reading from depth this was always broken, needs to be fixed later -Give better error showing binding and set when this happens.
2020-12-19 13:18:08 +00:00
LocalVector<UniformSet::AttachableTexture> attachable_textures;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Vector<Texture *> mutable_sampled_textures;
Vector<Texture *> mutable_storage_textures;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (uint32_t i = 0; i < set_uniform_count; i++) {
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
const UniformInfo &set_uniform = set_uniforms[i];
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
int uniform_idx = -1;
for (int j = 0; j < (int)uniform_count; j++) {
if (uniforms[j].binding == set_uniform.binding) {
uniform_idx = j;
}
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(uniform_idx == -1, RID(),
Refactored 2D shader and lighting system -Removed normal/specular properties from nodes -Create CanvasTexture, which can contain normal/specular channels -Refactored, optimized and simplified 2D shaders -Use atlas for light textures. -Use a shadow atlas for shadow textures. -Use both items aboves to make light rendering stateless (faster). -Reorganized uniform sets for more efficiency.
2020-10-24 15:15:43 +00:00
"All the shader bindings for the given set must be covered by the uniforms provided. Binding (" + itos(set_uniform.binding) + "), set (" + itos(p_shader_set) + ") was not provided.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
const Uniform &uniform = uniforms[uniform_idx];
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
ERR_FAIL_COND_V_MSG(uniform.uniform_type != set_uniform.type, RID(),
"Mismatch uniform type for binding (" + itos(set_uniform.binding) + "), set (" + itos(p_shader_set) + "). Expected '" + shader_uniform_names[set_uniform.type] + "', supplied: '" + shader_uniform_names[uniform.uniform_type] + "'.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkWriteDescriptorSet write; //common header
write.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pNext = nullptr;
Vulkan: Use VK_NULL_HANDLE for handles, fixes 32-bit builds Some Vulkan types are defined as "non dispatchable handles" and use a different typedef on 32-bit and 64-bit systems (struct pointer on 64-bit, `uint64_t` otherwise). https://github.com/KhronosGroup/Vulkan-Headers/blob/0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5c/include/vulkan/vulkan_core.h#L59-L65 Contrarily to `NULL`, `nullptr` can't be converted to `uint64_t` so build was now failing on 32-bit after converting the codebase from using `NULL` to `nullptr`. Fixes #37620.
2020-04-06 08:17:42 +00:00
write.dstSet = VK_NULL_HANDLE; //will assign afterwards when everything is valid
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.dstBinding = set_uniform.binding;
Silence mulitple may be used uninitialized warnings in RenderingDeviceVulkan::uniform_set_create()
2020-05-15 10:48:00 +00:00
write.dstArrayElement = 0;
write.descriptorCount = 0;
write.descriptorType = VK_DESCRIPTOR_TYPE_MAX_ENUM; //Invalid value.
write.pImageInfo = nullptr;
write.pBufferInfo = nullptr;
write.pTexelBufferView = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
uint32_t type_size = 1;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
switch (uniform.uniform_type) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
case UNIFORM_TYPE_SAMPLER: {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (set_uniform.length > 1) {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "Sampler (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") sampler elements, so it should be provided equal number of sampler IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "Sampler (binding: " + itos(uniform.binding) + ") should provide one ID referencing a sampler (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Vector<VkDescriptorImageInfo> image_info;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
VkSampler *sampler = sampler_owner.get_or_null(uniform.get_id(j));
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!sampler, RID(), "Sampler (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid sampler.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkDescriptorImageInfo img_info;
img_info.sampler = *sampler;
img_info.imageView = VK_NULL_HANDLE;
img_info.imageLayout = VK_IMAGE_LAYOUT_UNDEFINED;
image_info.push_back(img_info);
}
write.dstArrayElement = 0;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
write.descriptorCount = uniform.get_id_count();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLER;
write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pBufferInfo = nullptr;
write.pTexelBufferView = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
type_size = uniform.get_id_count();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
case UNIFORM_TYPE_SAMPLER_WITH_TEXTURE: {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
if (uniform.get_id_count() != (uint32_t)set_uniform.length * 2) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (set_uniform.length > 1) {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "SamplerTexture (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") sampler&texture elements, so it should provided twice the amount of IDs (sampler,texture pairs) to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "SamplerTexture (binding: " + itos(uniform.binding) + ") should provide two IDs referencing a sampler and then a texture (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Vector<VkDescriptorImageInfo> image_info;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
for (uint32_t j = 0; j < uniform.get_id_count(); j += 2) {
VkSampler *sampler = sampler_owner.get_or_null(uniform.get_id(j + 0));
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!sampler, RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ", index " + itos(j + 1) + ") is not a valid sampler.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
Texture *texture = texture_owner.get_or_null(uniform.get_id(j + 1));
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!texture, RID(), "Texture (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT), RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Texture (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") needs the TEXTURE_USAGE_SAMPLING_BIT usage flag set in order to be used as uniform.");
VkDescriptorImageInfo img_info;
img_info.sampler = *sampler;
img_info.imageView = texture->view;
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
if (texture->usage_flags & (TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | TEXTURE_USAGE_INPUT_ATTACHMENT_BIT)) {
Fix error spam on wrong attachment -For now, disable reading from depth this was always broken, needs to be fixed later -Give better error showing binding and set when this happens.
2020-12-19 13:18:08 +00:00
UniformSet::AttachableTexture attachable_texture;
attachable_texture.bind = set_uniform.binding;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
attachable_texture.texture = texture->owner.is_valid() ? texture->owner : uniform.get_id(j + 1);
Fix error spam on wrong attachment -For now, disable reading from depth this was always broken, needs to be fixed later -Give better error showing binding and set when this happens.
2020-12-19 13:18:08 +00:00
attachable_textures.push_back(attachable_texture);
Improved validation of attachments
2019-06-07 18:20:01 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
if (texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT) {
//can also be used as storage, add to mutable sampled
mutable_sampled_textures.push_back(texture);
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
if (texture->owner.is_valid()) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
texture = texture_owner.get_or_null(texture->owner);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!texture, RID()); //bug, should never happen
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
img_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_info.push_back(img_info);
}
write.dstArrayElement = 0;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
write.descriptorCount = uniform.get_id_count() / 2;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pBufferInfo = nullptr;
write.pTexelBufferView = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
type_size = uniform.get_id_count() / 2;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
case UNIFORM_TYPE_TEXTURE: {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (set_uniform.length > 1) {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "Texture (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") textures, so it should be provided equal number of texture IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "Texture (binding: " + itos(uniform.binding) + ") should provide one ID referencing a texture (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Vector<VkDescriptorImageInfo> image_info;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
Texture *texture = texture_owner.get_or_null(uniform.get_id(j));
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!texture, RID(), "Texture (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT), RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Texture (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") needs the TEXTURE_USAGE_SAMPLING_BIT usage flag set in order to be used as uniform.");
VkDescriptorImageInfo img_info;
Vulkan: Use VK_NULL_HANDLE for handles, fixes 32-bit builds Some Vulkan types are defined as "non dispatchable handles" and use a different typedef on 32-bit and 64-bit systems (struct pointer on 64-bit, `uint64_t` otherwise). https://github.com/KhronosGroup/Vulkan-Headers/blob/0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5c/include/vulkan/vulkan_core.h#L59-L65 Contrarily to `NULL`, `nullptr` can't be converted to `uint64_t` so build was now failing on 32-bit after converting the codebase from using `NULL` to `nullptr`. Fixes #37620.
2020-04-06 08:17:42 +00:00
img_info.sampler = VK_NULL_HANDLE;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
img_info.imageView = texture->view;
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
if (texture->usage_flags & (TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | TEXTURE_USAGE_INPUT_ATTACHMENT_BIT)) {
Fix error spam on wrong attachment -For now, disable reading from depth this was always broken, needs to be fixed later -Give better error showing binding and set when this happens.
2020-12-19 13:18:08 +00:00
UniformSet::AttachableTexture attachable_texture;
attachable_texture.bind = set_uniform.binding;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
attachable_texture.texture = texture->owner.is_valid() ? texture->owner : uniform.get_id(j);
Fix error spam on wrong attachment -For now, disable reading from depth this was always broken, needs to be fixed later -Give better error showing binding and set when this happens.
2020-12-19 13:18:08 +00:00
attachable_textures.push_back(attachable_texture);
Improved validation of attachments
2019-06-07 18:20:01 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
if (texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT) {
//can also be used as storage, add to mutable sampled
mutable_sampled_textures.push_back(texture);
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
if (texture->owner.is_valid()) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
texture = texture_owner.get_or_null(texture->owner);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!texture, RID()); //bug, should never happen
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
img_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
image_info.push_back(img_info);
}
write.dstArrayElement = 0;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
write.descriptorCount = uniform.get_id_count();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE;
write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pBufferInfo = nullptr;
write.pTexelBufferView = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
type_size = uniform.get_id_count();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
case UNIFORM_TYPE_IMAGE: {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
if (set_uniform.length > 1) {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "Image (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") textures, so it should be provided equal number of texture IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
} else {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "Image (binding: " + itos(uniform.binding) + ") should provide one ID referencing a texture (IDs provided: " + itos(uniform.get_id_count()) + ").");
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
}
}
Vector<VkDescriptorImageInfo> image_info;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
Texture *texture = texture_owner.get_or_null(uniform.get_id(j));
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_COND_V_MSG(!texture, RID(),
"Image (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture.");
ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT), RID(),
"Image (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") needs the TEXTURE_USAGE_STORAGE_BIT usage flag set in order to be used as uniform.");
VkDescriptorImageInfo img_info;
Vulkan: Use VK_NULL_HANDLE for handles, fixes 32-bit builds Some Vulkan types are defined as "non dispatchable handles" and use a different typedef on 32-bit and 64-bit systems (struct pointer on 64-bit, `uint64_t` otherwise). https://github.com/KhronosGroup/Vulkan-Headers/blob/0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5c/include/vulkan/vulkan_core.h#L59-L65 Contrarily to `NULL`, `nullptr` can't be converted to `uint64_t` so build was now failing on 32-bit after converting the codebase from using `NULL` to `nullptr`. Fixes #37620.
2020-04-06 08:17:42 +00:00
img_info.sampler = VK_NULL_HANDLE;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
img_info.imageView = texture->view;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
if (texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT) {
//can also be used as storage, add to mutable sampled
mutable_storage_textures.push_back(texture);
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
if (texture->owner.is_valid()) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
texture = texture_owner.get_or_null(texture->owner);
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_COND_V(!texture, RID()); //bug, should never happen
}
img_info.imageLayout = VK_IMAGE_LAYOUT_GENERAL;
image_info.push_back(img_info);
}
write.dstArrayElement = 0;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
write.descriptorCount = uniform.get_id_count();
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
write.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pBufferInfo = nullptr;
write.pTexelBufferView = nullptr;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
type_size = uniform.get_id_count();
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
case UNIFORM_TYPE_TEXTURE_BUFFER: {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (set_uniform.length > 1) {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "Buffer (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") texture buffer elements, so it should be provided equal number of texture buffer IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "Buffer (binding: " + itos(uniform.binding) + ") should provide one ID referencing a texture buffer (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Vector<VkDescriptorBufferInfo> buffer_info;
Vector<VkBufferView> buffer_view;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
TextureBuffer *buffer = texture_buffer_owner.get_or_null(uniform.get_id(j));
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!buffer, RID(), "Texture Buffer (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture buffer.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
buffer_info.push_back(buffer->buffer.buffer_info);
buffer_view.push_back(buffer->view);
}
write.dstArrayElement = 0;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
write.descriptorCount = uniform.get_id_count();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pImageInfo = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.pBufferInfo = buffer_infos.push_back(buffer_info)->get().ptr();
write.pTexelBufferView = buffer_views.push_back(buffer_view)->get().ptr();
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
type_size = uniform.get_id_count();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
case UNIFORM_TYPE_SAMPLER_WITH_TEXTURE_BUFFER: {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
if (uniform.get_id_count() != (uint32_t)set_uniform.length * 2) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (set_uniform.length > 1) {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") sampler buffer elements, so it should provided twice the amount of IDs (sampler,buffer pairs) to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ") should provide two IDs referencing a sampler and then a texture buffer (IDs provided: " + itos(uniform.get_id_count()) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Vector<VkDescriptorImageInfo> image_info;
Vector<VkDescriptorBufferInfo> buffer_info;
Vector<VkBufferView> buffer_view;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
for (uint32_t j = 0; j < uniform.get_id_count(); j += 2) {
VkSampler *sampler = sampler_owner.get_or_null(uniform.get_id(j + 0));
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!sampler, RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ", index " + itos(j + 1) + ") is not a valid sampler.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
TextureBuffer *buffer = texture_buffer_owner.get_or_null(uniform.get_id(j + 1));
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkDescriptorImageInfo img_info;
img_info.sampler = *sampler;
img_info.imageView = VK_NULL_HANDLE;
img_info.imageLayout = VK_IMAGE_LAYOUT_UNDEFINED;
image_info.push_back(img_info);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!buffer, RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ", index " + itos(j + 1) + ") is not a valid texture buffer.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
buffer_info.push_back(buffer->buffer.buffer_info);
buffer_view.push_back(buffer->view);
}
write.dstArrayElement = 0;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
write.descriptorCount = uniform.get_id_count() / 2;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
write.pBufferInfo = buffer_infos.push_back(buffer_info)->get().ptr();
write.pTexelBufferView = buffer_views.push_back(buffer_view)->get().ptr();
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
type_size = uniform.get_id_count() / 2;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
case UNIFORM_TYPE_IMAGE_BUFFER: {
//todo
} break;
case UNIFORM_TYPE_UNIFORM_BUFFER: {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_COND_V_MSG(uniform.get_id_count() != 1, RID(),
"Uniform buffer supplied (binding: " + itos(uniform.binding) + ") must provide one ID (" + itos(uniform.get_id_count()) + " provided).");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
Buffer *buffer = uniform_buffer_owner.get_or_null(uniform.get_id(0));
Completed material/2D shader support (missing SCREEN_TEXTURE)
2019-07-21 14:31:30 +00:00
ERR_FAIL_COND_V_MSG(!buffer, RID(), "Uniform buffer supplied (binding: " + itos(uniform.binding) + ") is invalid.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(buffer->size != (uint32_t)set_uniform.length, RID(),
Completed material/2D shader support (missing SCREEN_TEXTURE)
2019-07-21 14:31:30 +00:00
"Uniform buffer supplied (binding: " + itos(uniform.binding) + ") size (" + itos(buffer->size) + " does not match size of shader uniform: (" + itos(set_uniform.length) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.dstArrayElement = 0;
write.descriptorCount = 1;
write.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pImageInfo = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.pBufferInfo = &buffer->buffer_info;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pTexelBufferView = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
case UNIFORM_TYPE_STORAGE_BUFFER: {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_COND_V_MSG(uniform.get_id_count() != 1, RID(),
"Storage buffer supplied (binding: " + itos(uniform.binding) + ") must provide one ID (" + itos(uniform.get_id_count()) + " provided).");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Reimplement skeletons and blend shapes Uses compute shaders, which only once, on demand, and all in parallel.
2020-12-16 14:07:08 +00:00
Buffer *buffer = nullptr;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
if (storage_buffer_owner.owns(uniform.get_id(0))) {
buffer = storage_buffer_owner.get_or_null(uniform.get_id(0));
} else if (vertex_buffer_owner.owns(uniform.get_id(0))) {
buffer = vertex_buffer_owner.get_or_null(uniform.get_id(0));
Reimplement skeletons and blend shapes Uses compute shaders, which only once, on demand, and all in parallel.
2020-12-16 14:07:08 +00:00
ERR_FAIL_COND_V_MSG(!(buffer->usage & VK_BUFFER_USAGE_STORAGE_BUFFER_BIT), RID(), "Vertex buffer supplied (binding: " + itos(uniform.binding) + ") was not created with storage flag.");
}
Completed material/2D shader support (missing SCREEN_TEXTURE)
2019-07-21 14:31:30 +00:00
ERR_FAIL_COND_V_MSG(!buffer, RID(), "Storage buffer supplied (binding: " + itos(uniform.binding) + ") is invalid.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
//if 0, then it's sized on link time
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ERR_FAIL_COND_V_MSG(set_uniform.length > 0 && buffer->size != (uint32_t)set_uniform.length, RID(),
Completed material/2D shader support (missing SCREEN_TEXTURE)
2019-07-21 14:31:30 +00:00
"Storage buffer supplied (binding: " + itos(uniform.binding) + ") size (" + itos(buffer->size) + " does not match size of shader uniform: (" + itos(set_uniform.length) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.dstArrayElement = 0;
write.descriptorCount = 1;
write.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pImageInfo = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
write.pBufferInfo = &buffer->buffer_info;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
write.pTexelBufferView = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
case UNIFORM_TYPE_INPUT_ATTACHMENT: {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND_V_MSG(shader->is_compute, RID(), "InputAttachment (binding: " + itos(uniform.binding) + ") supplied for compute shader (this is not allowed).");
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (set_uniform.length > 1) {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "InputAttachment (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") textures, so it should be provided equal number of texture IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
} else {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
ERR_FAIL_V_MSG(RID(), "InputAttachment (binding: " + itos(uniform.binding) + ") should provide one ID referencing a texture (IDs provided: " + itos(uniform.get_id_count()) + ").");
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
}
}
Vector<VkDescriptorImageInfo> image_info;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
Texture *texture = texture_owner.get_or_null(uniform.get_id(j));
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND_V_MSG(!texture, RID(),
"InputAttachment (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture.");
ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT), RID(),
"InputAttachment (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") needs the TEXTURE_USAGE_SAMPLING_BIT usage flag set in order to be used as uniform.");
VkDescriptorImageInfo img_info;
img_info.sampler = VK_NULL_HANDLE;
img_info.imageView = texture->view;
if (texture->owner.is_valid()) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
texture = texture_owner.get_or_null(texture->owner);
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND_V(!texture, RID()); //bug, should never happen
}
img_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
image_info.push_back(img_info);
}
write.dstArrayElement = 0;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
write.descriptorCount = uniform.get_id_count();
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
write.descriptorType = VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT;
write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
write.pBufferInfo = nullptr;
write.pTexelBufferView = nullptr;
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
type_size = uniform.get_id_count();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} break;
default: {
}
}
writes.push_back(write);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(pool_key.uniform_type[set_uniform.type] == MAX_DESCRIPTOR_POOL_ELEMENT, RID(),
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Uniform set reached the limit of bindings for the same type (" + itos(MAX_DESCRIPTOR_POOL_ELEMENT) + ").");
pool_key.uniform_type[set_uniform.type] += type_size;
}
//need a descriptor pool
DescriptorPool *pool = _descriptor_pool_allocate(pool_key);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!pool, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkDescriptorSetAllocateInfo descriptor_set_allocate_info;
descriptor_set_allocate_info.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
descriptor_set_allocate_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
descriptor_set_allocate_info.descriptorPool = pool->pool;
descriptor_set_allocate_info.descriptorSetCount = 1;
descriptor_set_allocate_info.pSetLayouts = &shader->sets[p_shader_set].descriptor_set_layout;
VkDescriptorSet descriptor_set;
VkResult res = vkAllocateDescriptorSets(device, &descriptor_set_allocate_info, &descriptor_set);
if (res) {
_descriptor_pool_free(pool_key, pool); // meh
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_V_MSG(RID(), "Cannot allocate descriptor sets, error " + itos(res) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
UniformSet uniform_set;
uniform_set.pool = pool;
uniform_set.pool_key = pool_key;
uniform_set.descriptor_set = descriptor_set;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
uniform_set.format = shader->set_formats[p_shader_set];
Improved validation of attachments
2019-06-07 18:20:01 +00:00
uniform_set.attachable_textures = attachable_textures;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
uniform_set.mutable_sampled_textures = mutable_sampled_textures;
uniform_set.mutable_storage_textures = mutable_storage_textures;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
uniform_set.shader_set = p_shader_set;
uniform_set.shader_id = p_shader;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RID id = uniform_set_owner.make_rid(uniform_set);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//add dependencies
_add_dependency(id, p_shader);
for (uint32_t i = 0; i < uniform_count; i++) {
const Uniform &uniform = uniforms[i];
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
int id_count = uniform.get_id_count();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (int j = 0; j < id_count; j++) {
Add a UniformSet cache * Changed syntax usage for RD::Uniform to create faster with a single RID * Converted render pass setup to use this in clustered renderer to test. This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
2022-03-06 11:57:09 +00:00
_add_dependency(id, uniform.get_id(j));
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
//write the contents
if (writes.size()) {
for (int i = 0; i < writes.size(); i++) {
writes.write[i].dstSet = descriptor_set;
}
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkUpdateDescriptorSets(device, writes.size(), writes.ptr(), 0, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return id;
}
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
bool RenderingDeviceVulkan::uniform_set_is_valid(RID p_uniform_set) {
return uniform_set_owner.owns(p_uniform_set);
}
Fix material invalidation on reimport. * IF a texture was reimported (calling replace as an example), it would invalidate all materials using it, causing plenty of errors. * Added the possibility to get a notification when a uniform set is erased. * With this notification, materials can be queued for update properly.
2021-07-07 22:55:20 +00:00
void RenderingDeviceVulkan::uniform_set_set_invalidation_callback(RID p_uniform_set, UniformSetInvalidatedCallback p_callback, void *p_userdata) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
UniformSet *us = uniform_set_owner.get_or_null(p_uniform_set);
Fix material invalidation on reimport. * IF a texture was reimported (calling replace as an example), it would invalidate all materials using it, causing plenty of errors. * Added the possibility to get a notification when a uniform set is erased. * With this notification, materials can be queued for update properly.
2021-07-07 22:55:20 +00:00
ERR_FAIL_COND(!us);
us->invalidated_callback = p_callback;
us->invalidated_callback_userdata = p_userdata;
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
Error RenderingDeviceVulkan::buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const void *p_data, uint32_t p_post_barrier) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
ERR_FAIL_COND_V_MSG(draw_list, ERR_INVALID_PARAMETER,
"Updating buffers is forbidden during creation of a draw list");
ERR_FAIL_COND_V_MSG(compute_list, ERR_INVALID_PARAMETER,
"Updating buffers is forbidden during creation of a compute list");
Bugfixes and ability to better specify filter and repeat modes everywhere. Removes antialiased flag for draw_* methods.
2019-06-25 01:24:07 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
VkPipelineStageFlags dst_stage_mask = 0;
VkAccessFlags dst_access = 0;
if (p_post_barrier & BARRIER_MASK_TRANSFER) {
// Protect subsequent updates...
dst_stage_mask = VK_PIPELINE_STAGE_TRANSFER_BIT;
dst_access = VK_ACCESS_TRANSFER_WRITE_BIT;
}
Buffer *buffer = _get_buffer_from_owner(p_buffer, dst_stage_mask, dst_access, p_post_barrier);
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
if (!buffer) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_V_MSG(ERR_INVALID_PARAMETER, "Buffer argument is not a valid buffer of any type.");
}
ERR_FAIL_COND_V_MSG(p_offset + p_size > buffer->size, ERR_INVALID_PARAMETER,
"Attempted to write buffer (" + itos((p_offset + p_size) - buffer->size) + " bytes) past the end.");
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
// no barrier should be needed here
// _buffer_memory_barrier(buffer->buffer, p_offset, p_size, dst_stage_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_access, VK_ACCESS_TRANSFER_WRITE_BIT, true);
Error err = _buffer_update(buffer, p_offset, (uint8_t *)p_data, p_size, p_post_barrier);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
if (err) {
return err;
}
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
#ifdef FORCE_FULL_BARRIER
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
_full_barrier(true);
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
#else
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (dst_stage_mask == 0) {
dst_stage_mask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
}
if (p_post_barrier != RD::BARRIER_MASK_NO_BARRIER) {
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
_buffer_memory_barrier(buffer->buffer, p_offset, p_size, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_stage_mask, VK_ACCESS_TRANSFER_WRITE_BIT, dst_access, true);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
#endif
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
return err;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
Error RenderingDeviceVulkan::buffer_clear(RID p_buffer, uint32_t p_offset, uint32_t p_size, uint32_t p_post_barrier) {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
_THREAD_SAFE_METHOD_
ERR_FAIL_COND_V_MSG((p_size % 4) != 0, ERR_INVALID_PARAMETER,
"Size must be a multiple of four");
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
ERR_FAIL_COND_V_MSG(draw_list, ERR_INVALID_PARAMETER,
"Updating buffers in is forbidden during creation of a draw list");
ERR_FAIL_COND_V_MSG(compute_list, ERR_INVALID_PARAMETER,
"Updating buffers is forbidden during creation of a compute list");
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
VkPipelineStageFlags dst_stage_mask = 0;
VkAccessFlags dst_access = 0;
if (p_post_barrier & BARRIER_MASK_TRANSFER) {
// Protect subsequent updates...
dst_stage_mask = VK_PIPELINE_STAGE_TRANSFER_BIT;
dst_access = VK_ACCESS_TRANSFER_WRITE_BIT;
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
Buffer *buffer = _get_buffer_from_owner(p_buffer, dst_stage_mask, dst_access, p_post_barrier);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (!buffer) {
ERR_FAIL_V_MSG(ERR_INVALID_PARAMETER, "Buffer argument is not a valid buffer of any type.");
}
ERR_FAIL_COND_V_MSG(p_offset + p_size > buffer->size, ERR_INVALID_PARAMETER,
"Attempted to write buffer (" + itos((p_offset + p_size) - buffer->size) + " bytes) past the end.");
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
// should not be needed
// _buffer_memory_barrier(buffer->buffer, p_offset, p_size, dst_stage_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_access, VK_ACCESS_TRANSFER_WRITE_BIT, p_post_barrier);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
vkCmdFillBuffer(frames[frame].draw_command_buffer, buffer->buffer, p_offset, p_size, 0);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
#ifdef FORCE_FULL_BARRIER
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
_full_barrier(true);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
#else
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (dst_stage_mask == 0) {
dst_stage_mask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
}
_buffer_memory_barrier(buffer->buffer, p_offset, p_size, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_stage_mask, VK_ACCESS_TRANSFER_WRITE_BIT, dst_access, dst_stage_mask);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
#endif
return OK;
}
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
Vector<uint8_t> RenderingDeviceVulkan::buffer_get_data(RID p_buffer) {
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
_THREAD_SAFE_METHOD_
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
// It could be this buffer was just created
VkPipelineShaderStageCreateFlags src_stage_mask = VK_PIPELINE_STAGE_TRANSFER_BIT;
VkAccessFlags src_access_mask = VK_ACCESS_TRANSFER_WRITE_BIT;
// Get the vulkan buffer and the potential stage/access possible
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
Buffer *buffer = _get_buffer_from_owner(p_buffer, src_stage_mask, src_access_mask, BARRIER_MASK_ALL);
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
if (!buffer) {
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
ERR_FAIL_V_MSG(Vector<uint8_t>(), "Buffer is either invalid or this type of buffer can't be retrieved. Only Index and Vertex buffers allow retrieving.");
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
}
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
// Make sure no one is using the buffer -- the "false" gets us to the same command buffer as below.
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
_buffer_memory_barrier(buffer->buffer, 0, buffer->size, src_stage_mask, src_access_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_READ_BIT, false);
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
VkCommandBuffer command_buffer = frames[frame].setup_command_buffer;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
Buffer tmp_buffer;
vk_mem_alloc: Update to upstream + Replace use of deprecated items
2022-03-29 09:28:09 +00:00
_buffer_allocate(&tmp_buffer, buffer->size, VK_BUFFER_USAGE_TRANSFER_DST_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_HOST, VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT);
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
VkBufferCopy region;
region.srcOffset = 0;
region.dstOffset = 0;
region.size = buffer->size;
vkCmdCopyBuffer(command_buffer, buffer->buffer, tmp_buffer.buffer, 1, &region); //dst buffer is in CPU, but I wonder if src buffer needs a barrier for this..
//flush everything so memory can be safely mapped
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
_flush(true);
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
void *buffer_mem;
VkResult vkerr = vmaMapMemory(allocator, tmp_buffer.allocation, &buffer_mem);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(vkerr, Vector<uint8_t>(), "vmaMapMemory failed with error " + itos(vkerr) + ".");
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
Vector<uint8_t> buffer_data;
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
{
buffer_data.resize(buffer->size);
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
uint8_t *w = buffer_data.ptrw();
Core: Drop custom `copymem`/`zeromem` defines We've been using standard C library functions `memcpy`/`memset` for these since 2016 with 67f65f66391327b2967a20a89c3627e1dd6e84eb. There was still the possibility for third-party platform ports to override the definitions with a custom header, but this doesn't seem useful anymore.
2021-04-27 14:19:21 +00:00
memcpy(w, buffer_mem, buffer->size);
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
}
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
vmaUnmapMemory(allocator, tmp_buffer.allocation);
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
_buffer_free(&tmp_buffer);
return buffer_data;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
/*************************/
/**** RENDER PIPELINE ****/
/*************************/
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferFormatID p_framebuffer_format, VertexFormatID p_vertex_format, RenderPrimitive p_render_primitive, const PipelineRasterizationState &p_rasterization_state, const PipelineMultisampleState &p_multisample_state, const PipelineDepthStencilState &p_depth_stencil_state, const PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags, uint32_t p_for_render_pass, const Vector<PipelineSpecializationConstant> &p_specialization_constants) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
//needs a shader
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Shader *shader = shader_owner.get_or_null(p_shader);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!shader, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_COND_V_MSG(shader->is_compute, RID(),
"Compute shaders can't be used in render pipelines");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_framebuffer_format == INVALID_ID) {
//if nothing provided, use an empty one (no attachments)
p_framebuffer_format = framebuffer_format_create(Vector<AttachmentFormat>());
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!framebuffer_formats.has(p_framebuffer_format), RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
const FramebufferFormat &fb_format = framebuffer_formats[p_framebuffer_format];
{ //validate shader vs framebuffer
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND_V_MSG(p_for_render_pass >= uint32_t(fb_format.E->key().passes.size()), RID(), "Render pass requested for pipeline creation (" + itos(p_for_render_pass) + ") is out of bounds");
const FramebufferPass &pass = fb_format.E->key().passes[p_for_render_pass];
uint32_t output_mask = 0;
for (int i = 0; i < pass.color_attachments.size(); i++) {
if (pass.color_attachments[i] != FramebufferPass::ATTACHMENT_UNUSED) {
output_mask |= 1 << i;
}
}
ERR_FAIL_COND_V_MSG(shader->fragment_output_mask != output_mask, RID(),
"Mismatch fragment shader output mask (" + itos(shader->fragment_output_mask) + ") and framebuffer color output mask (" + itos(output_mask) + ") when binding both in render pipeline.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//vertex
VkPipelineVertexInputStateCreateInfo pipeline_vertex_input_state_create_info;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
if (p_vertex_format != INVALID_ID) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//uses vertices, else it does not
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(!vertex_formats.has(p_vertex_format), RID());
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
const VertexDescriptionCache &vd = vertex_formats[p_vertex_format];
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
pipeline_vertex_input_state_create_info = vd.create_info;
//validate with inputs
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
for (uint32_t i = 0; i < 32; i++) {
Fix more issues found by cppcheck.
2022-04-07 10:23:40 +00:00
if (!(shader->vertex_input_mask & (1UL << i))) {
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
continue;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
bool found = false;
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
for (int j = 0; j < vd.vertex_formats.size(); j++) {
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
if (vd.vertex_formats[j].location == i) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
found = true;
}
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V_MSG(!found, RID(),
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
"Shader vertex input location (" + itos(i) + ") not provided in vertex input description for pipeline creation.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
} else {
//does not use vertices
pipeline_vertex_input_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
pipeline_vertex_input_state_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
pipeline_vertex_input_state_create_info.flags = 0;
pipeline_vertex_input_state_create_info.vertexBindingDescriptionCount = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
pipeline_vertex_input_state_create_info.pVertexBindingDescriptions = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
pipeline_vertex_input_state_create_info.vertexAttributeDescriptionCount = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
pipeline_vertex_input_state_create_info.pVertexAttributeDescriptions = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
ERR_FAIL_COND_V_MSG(shader->vertex_input_mask != 0, RID(),
"Shader contains vertex inputs, but no vertex input description was provided for pipeline creation.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//input assembly
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_render_primitive, RENDER_PRIMITIVE_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkPipelineInputAssemblyStateCreateInfo input_assembly_create_info;
input_assembly_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
input_assembly_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
input_assembly_create_info.flags = 0;
static const VkPrimitiveTopology topology_list[RENDER_PRIMITIVE_MAX] = {
VK_PRIMITIVE_TOPOLOGY_POINT_LIST,
VK_PRIMITIVE_TOPOLOGY_LINE_LIST,
VK_PRIMITIVE_TOPOLOGY_LINE_LIST_WITH_ADJACENCY,
VK_PRIMITIVE_TOPOLOGY_LINE_STRIP,
VK_PRIMITIVE_TOPOLOGY_LINE_STRIP_WITH_ADJACENCY,
VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST,
VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_WITH_ADJACENCY,
VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP,
VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP_WITH_ADJACENCY,
VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP,
VK_PRIMITIVE_TOPOLOGY_PATCH_LIST
};
input_assembly_create_info.topology = topology_list[p_render_primitive];
input_assembly_create_info.primitiveRestartEnable = (p_render_primitive == RENDER_PRIMITIVE_TRIANGLE_STRIPS_WITH_RESTART_INDEX);
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
//tessellation
VkPipelineTessellationStateCreateInfo tessellation_create_info;
tessellation_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO;
tessellation_create_info.pNext = nullptr;
tessellation_create_info.flags = 0;
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
ERR_FAIL_COND_V(limits.maxTessellationPatchSize > 0 && (p_rasterization_state.patch_control_points < 1 || p_rasterization_state.patch_control_points > limits.maxTessellationPatchSize), RID());
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
tessellation_create_info.patchControlPoints = p_rasterization_state.patch_control_points;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkPipelineViewportStateCreateInfo viewport_state_create_info;
viewport_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
viewport_state_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
viewport_state_create_info.flags = 0;
viewport_state_create_info.viewportCount = 1; //if VR extensions are supported at some point, this will have to be customizable in the framebuffer format
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
viewport_state_create_info.pViewports = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
viewport_state_create_info.scissorCount = 1;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
viewport_state_create_info.pScissors = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//rasterization
VkPipelineRasterizationStateCreateInfo rasterization_state_create_info;
rasterization_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
rasterization_state_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
rasterization_state_create_info.flags = 0;
rasterization_state_create_info.depthClampEnable = p_rasterization_state.enable_depth_clamp;
rasterization_state_create_info.rasterizerDiscardEnable = p_rasterization_state.discard_primitives;
rasterization_state_create_info.polygonMode = (p_rasterization_state.wireframe ? VK_POLYGON_MODE_LINE : VK_POLYGON_MODE_FILL);
Fix some issues found by cppcheck.
2022-04-05 10:40:26 +00:00
static const VkCullModeFlags cull_mode[3] = {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VK_CULL_MODE_NONE,
VK_CULL_MODE_FRONT_BIT,
VK_CULL_MODE_BACK_BIT
};
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_rasterization_state.cull_mode, 3, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
rasterization_state_create_info.cullMode = cull_mode[p_rasterization_state.cull_mode];
rasterization_state_create_info.frontFace = (p_rasterization_state.front_face == POLYGON_FRONT_FACE_CLOCKWISE ? VK_FRONT_FACE_CLOCKWISE : VK_FRONT_FACE_COUNTER_CLOCKWISE);
rasterization_state_create_info.depthBiasEnable = p_rasterization_state.depth_bias_enable;
rasterization_state_create_info.depthBiasConstantFactor = p_rasterization_state.depth_bias_constant_factor;
rasterization_state_create_info.depthBiasClamp = p_rasterization_state.depth_bias_clamp;
rasterization_state_create_info.depthBiasSlopeFactor = p_rasterization_state.depth_bias_slope_factor;
rasterization_state_create_info.lineWidth = p_rasterization_state.line_width;
//multisample
VkPipelineMultisampleStateCreateInfo multisample_state_create_info;
multisample_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
multisample_state_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
multisample_state_create_info.flags = 0;
multisample_state_create_info.rasterizationSamples = rasterization_sample_count[p_multisample_state.sample_count];
multisample_state_create_info.sampleShadingEnable = p_multisample_state.enable_sample_shading;
multisample_state_create_info.minSampleShading = p_multisample_state.min_sample_shading;
Vector<VkSampleMask> sample_mask;
if (p_multisample_state.sample_mask.size()) {
//use sample mask
Fix some issues found by cppcheck.
2022-04-05 10:40:26 +00:00
const int rasterization_sample_mask_expected_size[TEXTURE_SAMPLES_MAX] = {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
1, 2, 4, 8, 16, 32, 64
};
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_COND_V(rasterization_sample_mask_expected_size[p_multisample_state.sample_count] != p_multisample_state.sample_mask.size(), RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
sample_mask.resize(p_multisample_state.sample_mask.size());
for (int i = 0; i < p_multisample_state.sample_mask.size(); i++) {
VkSampleMask mask = p_multisample_state.sample_mask[i];
sample_mask.push_back(mask);
}
multisample_state_create_info.pSampleMask = sample_mask.ptr();
} else {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
multisample_state_create_info.pSampleMask = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
multisample_state_create_info.alphaToCoverageEnable = p_multisample_state.enable_alpha_to_coverage;
multisample_state_create_info.alphaToOneEnable = p_multisample_state.enable_alpha_to_one;
//depth stencil
VkPipelineDepthStencilStateCreateInfo depth_stencil_state_create_info;
depth_stencil_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
depth_stencil_state_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
depth_stencil_state_create_info.flags = 0;
depth_stencil_state_create_info.depthTestEnable = p_depth_stencil_state.enable_depth_test;
depth_stencil_state_create_info.depthWriteEnable = p_depth_stencil_state.enable_depth_write;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_depth_stencil_state.depth_compare_operator, COMPARE_OP_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
depth_stencil_state_create_info.depthCompareOp = compare_operators[p_depth_stencil_state.depth_compare_operator];
depth_stencil_state_create_info.depthBoundsTestEnable = p_depth_stencil_state.enable_depth_range;
depth_stencil_state_create_info.stencilTestEnable = p_depth_stencil_state.enable_stencil;
Exposed RenderingDevice to script API Also added an easier way to load native GLSL shaders. Extras: Had to fix no-cache for subresources in resource loader, it was not properly working, making shaders not properly reload. Note: The precommit hooks are broken because they don't seem to support enums from one class being used in another. Feel free to fix this after merging this PR.
2020-04-20 02:19:21 +00:00
ERR_FAIL_INDEX_V(p_depth_stencil_state.front_op.fail, STENCIL_OP_MAX, RID());
depth_stencil_state_create_info.front.failOp = stencil_operations[p_depth_stencil_state.front_op.fail];
ERR_FAIL_INDEX_V(p_depth_stencil_state.front_op.pass, STENCIL_OP_MAX, RID());
depth_stencil_state_create_info.front.passOp = stencil_operations[p_depth_stencil_state.front_op.pass];
ERR_FAIL_INDEX_V(p_depth_stencil_state.front_op.depth_fail, STENCIL_OP_MAX, RID());
depth_stencil_state_create_info.front.depthFailOp = stencil_operations[p_depth_stencil_state.front_op.depth_fail];
ERR_FAIL_INDEX_V(p_depth_stencil_state.front_op.compare, COMPARE_OP_MAX, RID());
depth_stencil_state_create_info.front.compareOp = compare_operators[p_depth_stencil_state.front_op.compare];
depth_stencil_state_create_info.front.compareMask = p_depth_stencil_state.front_op.compare_mask;
depth_stencil_state_create_info.front.writeMask = p_depth_stencil_state.front_op.write_mask;
depth_stencil_state_create_info.front.reference = p_depth_stencil_state.front_op.reference;
ERR_FAIL_INDEX_V(p_depth_stencil_state.back_op.fail, STENCIL_OP_MAX, RID());
depth_stencil_state_create_info.back.failOp = stencil_operations[p_depth_stencil_state.back_op.fail];
ERR_FAIL_INDEX_V(p_depth_stencil_state.back_op.pass, STENCIL_OP_MAX, RID());
depth_stencil_state_create_info.back.passOp = stencil_operations[p_depth_stencil_state.back_op.pass];
ERR_FAIL_INDEX_V(p_depth_stencil_state.back_op.depth_fail, STENCIL_OP_MAX, RID());
depth_stencil_state_create_info.back.depthFailOp = stencil_operations[p_depth_stencil_state.back_op.depth_fail];
ERR_FAIL_INDEX_V(p_depth_stencil_state.back_op.compare, COMPARE_OP_MAX, RID());
depth_stencil_state_create_info.back.compareOp = compare_operators[p_depth_stencil_state.back_op.compare];
depth_stencil_state_create_info.back.compareMask = p_depth_stencil_state.back_op.compare_mask;
depth_stencil_state_create_info.back.writeMask = p_depth_stencil_state.back_op.write_mask;
depth_stencil_state_create_info.back.reference = p_depth_stencil_state.back_op.reference;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
depth_stencil_state_create_info.minDepthBounds = p_depth_stencil_state.depth_range_min;
depth_stencil_state_create_info.maxDepthBounds = p_depth_stencil_state.depth_range_max;
//blend state
VkPipelineColorBlendStateCreateInfo color_blend_state_create_info;
color_blend_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
color_blend_state_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
color_blend_state_create_info.flags = 0;
color_blend_state_create_info.logicOpEnable = p_blend_state.enable_logic_op;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_FAIL_INDEX_V(p_blend_state.logic_op, LOGIC_OP_MAX, RID());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
color_blend_state_create_info.logicOp = logic_operations[p_blend_state.logic_op];
Vector<VkPipelineColorBlendAttachmentState> attachment_states;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
{
const FramebufferPass &pass = fb_format.E->key().passes[p_for_render_pass];
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
attachment_states.resize(pass.color_attachments.size());
ERR_FAIL_COND_V(p_blend_state.attachments.size() < pass.color_attachments.size(), RID());
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
for (int i = 0; i < pass.color_attachments.size(); i++) {
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
VkPipelineColorBlendAttachmentState state;
if (pass.color_attachments[i] == FramebufferPass::ATTACHMENT_UNUSED) {
state.blendEnable = false;
state.srcColorBlendFactor = VK_BLEND_FACTOR_ZERO;
state.dstColorBlendFactor = VK_BLEND_FACTOR_ZERO;
state.colorBlendOp = VK_BLEND_OP_ADD;
state.srcAlphaBlendFactor = VK_BLEND_FACTOR_ZERO;
state.dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO;
state.alphaBlendOp = VK_BLEND_OP_ADD;
state.colorWriteMask = 0;
} else {
state.blendEnable = p_blend_state.attachments[i].enable_blend;
ERR_FAIL_INDEX_V(p_blend_state.attachments[i].src_color_blend_factor, BLEND_FACTOR_MAX, RID());
state.srcColorBlendFactor = blend_factors[p_blend_state.attachments[i].src_color_blend_factor];
ERR_FAIL_INDEX_V(p_blend_state.attachments[i].dst_color_blend_factor, BLEND_FACTOR_MAX, RID());
state.dstColorBlendFactor = blend_factors[p_blend_state.attachments[i].dst_color_blend_factor];
ERR_FAIL_INDEX_V(p_blend_state.attachments[i].color_blend_op, BLEND_OP_MAX, RID());
state.colorBlendOp = blend_operations[p_blend_state.attachments[i].color_blend_op];
ERR_FAIL_INDEX_V(p_blend_state.attachments[i].src_alpha_blend_factor, BLEND_FACTOR_MAX, RID());
state.srcAlphaBlendFactor = blend_factors[p_blend_state.attachments[i].src_alpha_blend_factor];
ERR_FAIL_INDEX_V(p_blend_state.attachments[i].dst_alpha_blend_factor, BLEND_FACTOR_MAX, RID());
state.dstAlphaBlendFactor = blend_factors[p_blend_state.attachments[i].dst_alpha_blend_factor];
ERR_FAIL_INDEX_V(p_blend_state.attachments[i].alpha_blend_op, BLEND_OP_MAX, RID());
state.alphaBlendOp = blend_operations[p_blend_state.attachments[i].alpha_blend_op];
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
state.colorWriteMask = 0;
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (p_blend_state.attachments[i].write_r) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
state.colorWriteMask |= VK_COLOR_COMPONENT_R_BIT;
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (p_blend_state.attachments[i].write_g) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
state.colorWriteMask |= VK_COLOR_COMPONENT_G_BIT;
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (p_blend_state.attachments[i].write_b) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
state.colorWriteMask |= VK_COLOR_COMPONENT_B_BIT;
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (p_blend_state.attachments[i].write_a) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
state.colorWriteMask |= VK_COLOR_COMPONENT_A_BIT;
}
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
attachment_states.write[i] = state;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
color_blend_state_create_info.attachmentCount = attachment_states.size();
color_blend_state_create_info.pAttachments = attachment_states.ptr();
color_blend_state_create_info.blendConstants[0] = p_blend_state.blend_constant.r;
color_blend_state_create_info.blendConstants[1] = p_blend_state.blend_constant.g;
color_blend_state_create_info.blendConstants[2] = p_blend_state.blend_constant.b;
color_blend_state_create_info.blendConstants[3] = p_blend_state.blend_constant.a;
//dynamic state
VkPipelineDynamicStateCreateInfo dynamic_state_create_info;
dynamic_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
dynamic_state_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
dynamic_state_create_info.flags = 0;
Vector<VkDynamicState> dynamic_states; //vulkan is weird..
dynamic_states.push_back(VK_DYNAMIC_STATE_VIEWPORT); //viewport and scissor are always dynamic
dynamic_states.push_back(VK_DYNAMIC_STATE_SCISSOR);
if (p_dynamic_state_flags & DYNAMIC_STATE_LINE_WIDTH) {
dynamic_states.push_back(VK_DYNAMIC_STATE_LINE_WIDTH);
}
if (p_dynamic_state_flags & DYNAMIC_STATE_DEPTH_BIAS) {
dynamic_states.push_back(VK_DYNAMIC_STATE_DEPTH_BIAS);
}
if (p_dynamic_state_flags & DYNAMIC_STATE_BLEND_CONSTANTS) {
dynamic_states.push_back(VK_DYNAMIC_STATE_BLEND_CONSTANTS);
}
if (p_dynamic_state_flags & DYNAMIC_STATE_DEPTH_BOUNDS) {
dynamic_states.push_back(VK_DYNAMIC_STATE_DEPTH_BOUNDS);
}
if (p_dynamic_state_flags & DYNAMIC_STATE_STENCIL_COMPARE_MASK) {
dynamic_states.push_back(VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK);
}
if (p_dynamic_state_flags & DYNAMIC_STATE_STENCIL_WRITE_MASK) {
dynamic_states.push_back(VK_DYNAMIC_STATE_STENCIL_WRITE_MASK);
}
if (p_dynamic_state_flags & DYNAMIC_STATE_STENCIL_REFERENCE) {
dynamic_states.push_back(VK_DYNAMIC_STATE_STENCIL_REFERENCE);
}
dynamic_state_create_info.dynamicStateCount = dynamic_states.size();
dynamic_state_create_info.pDynamicStates = dynamic_states.ptr();
//finally, pipeline create info
VkGraphicsPipelineCreateInfo graphics_pipeline_create_info;
graphics_pipeline_create_info.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
graphics_pipeline_create_info.pNext = nullptr;
A lot of progress with canvas rendering, still far from working.
2019-06-16 02:45:24 +00:00
graphics_pipeline_create_info.flags = 0;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
Vector<VkPipelineShaderStageCreateInfo> pipeline_stages = shader->pipeline_stages;
Vector<VkSpecializationInfo> specialization_info;
Vector<Vector<VkSpecializationMapEntry>> specialization_map_entries;
Vector<uint32_t> specialization_constant_data;
if (shader->specialization_constants.size()) {
specialization_constant_data.resize(shader->specialization_constants.size());
uint32_t *data_ptr = specialization_constant_data.ptrw();
specialization_info.resize(pipeline_stages.size());
specialization_map_entries.resize(pipeline_stages.size());
for (int i = 0; i < shader->specialization_constants.size(); i++) {
Fix various typos Follow-up typos found via `codespell -q 3 -S ./thirdparty,*.po,./DONORS.md -L ackward,ang,ans,ba,beng,cas,childs,childrens,dof,doubleclick,fave,findn,hist,inout,leapyear,lod,nd,numer,ois,ony,paket,seeked,sinc,switchs,te,uint`
2021-07-25 11:18:25 +00:00
//see if overridden
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
const Shader::SpecializationConstant &sc = shader->specialization_constants[i];
data_ptr[i] = sc.constant.int_value; //just copy the 32 bits
for (int j = 0; j < p_specialization_constants.size(); j++) {
const PipelineSpecializationConstant &psc = p_specialization_constants[j];
if (psc.constant_id == sc.constant.constant_id) {
ERR_FAIL_COND_V_MSG(psc.type != sc.constant.type, RID(), "Specialization constant provided for id (" + itos(sc.constant.constant_id) + ") is of the wrong type.");
Use specialization constants in clustered renderer * Keep track of when projector, softshadow or directional sofshadow were enabled. * Enable them via specializaton constant where it makes sense. * Re-implements soft shadows. * Re-implements light projectors.
2021-07-12 23:32:05 +00:00
data_ptr[i] = psc.int_value;
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
break;
}
}
VkSpecializationMapEntry entry;
entry.constantID = sc.constant.constant_id;
entry.offset = i * sizeof(uint32_t);
entry.size = sizeof(uint32_t);
for (int j = 0; j < SHADER_STAGE_MAX; j++) {
if (sc.stage_flags & (1 << j)) {
VkShaderStageFlagBits stage = shader_stage_masks[j];
for (int k = 0; k < pipeline_stages.size(); k++) {
if (pipeline_stages[k].stage == stage) {
specialization_map_entries.write[k].push_back(entry);
}
}
}
}
}
Use specialization constants in clustered renderer * Keep track of when projector, softshadow or directional sofshadow were enabled. * Enable them via specializaton constant where it makes sense. * Re-implements soft shadows. * Re-implements light projectors.
2021-07-12 23:32:05 +00:00
for (int i = 0; i < pipeline_stages.size(); i++) {
if (specialization_map_entries[i].size()) {
specialization_info.write[i].dataSize = specialization_constant_data.size() * sizeof(uint32_t);
specialization_info.write[i].pData = data_ptr;
specialization_info.write[i].mapEntryCount = specialization_map_entries[i].size();
specialization_info.write[i].pMapEntries = specialization_map_entries[i].ptr();
pipeline_stages.write[i].pSpecializationInfo = specialization_info.ptr() + i;
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
}
}
}
graphics_pipeline_create_info.stageCount = pipeline_stages.size();
graphics_pipeline_create_info.pStages = pipeline_stages.ptr();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
graphics_pipeline_create_info.pVertexInputState = &pipeline_vertex_input_state_create_info;
graphics_pipeline_create_info.pInputAssemblyState = &input_assembly_create_info;
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
graphics_pipeline_create_info.pTessellationState = &tessellation_create_info;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
graphics_pipeline_create_info.pViewportState = &viewport_state_create_info;
graphics_pipeline_create_info.pRasterizationState = &rasterization_state_create_info;
graphics_pipeline_create_info.pMultisampleState = &multisample_state_create_info;
graphics_pipeline_create_info.pDepthStencilState = &depth_stencil_state_create_info;
graphics_pipeline_create_info.pColorBlendState = &color_blend_state_create_info;
graphics_pipeline_create_info.pDynamicState = &dynamic_state_create_info;
graphics_pipeline_create_info.layout = shader->pipeline_layout;
graphics_pipeline_create_info.renderPass = fb_format.render_pass;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
graphics_pipeline_create_info.subpass = p_for_render_pass;
Vulkan: Use VK_NULL_HANDLE for handles, fixes 32-bit builds Some Vulkan types are defined as "non dispatchable handles" and use a different typedef on 32-bit and 64-bit systems (struct pointer on 64-bit, `uint64_t` otherwise). https://github.com/KhronosGroup/Vulkan-Headers/blob/0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5c/include/vulkan/vulkan_core.h#L59-L65 Contrarily to `NULL`, `nullptr` can't be converted to `uint64_t` so build was now failing on 32-bit after converting the codebase from using `NULL` to `nullptr`. Fixes #37620.
2020-04-06 08:17:42 +00:00
graphics_pipeline_create_info.basePipelineHandle = VK_NULL_HANDLE;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
graphics_pipeline_create_info.basePipelineIndex = 0;
RenderPipeline pipeline;
Vulkan: Use VK_NULL_HANDLE for handles, fixes 32-bit builds Some Vulkan types are defined as "non dispatchable handles" and use a different typedef on 32-bit and 64-bit systems (struct pointer on 64-bit, `uint64_t` otherwise). https://github.com/KhronosGroup/Vulkan-Headers/blob/0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5c/include/vulkan/vulkan_core.h#L59-L65 Contrarily to `NULL`, `nullptr` can't be converted to `uint64_t` so build was now failing on 32-bit after converting the codebase from using `NULL` to `nullptr`. Fixes #37620.
2020-04-06 08:17:42 +00:00
VkResult err = vkCreateGraphicsPipelines(device, VK_NULL_HANDLE, 1, &graphics_pipeline_create_info, nullptr, &pipeline.pipeline);
Fixes to mobile renderer * Make sure shaders are named, to aid in debug in case of failure * SceneRenderRD was being wrongly initialized (virtual functions being called when derivative class not initialized). * Fixed some bugs resulting on the above being corrected.
2021-08-16 17:51:29 +00:00
ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateGraphicsPipelines failed with error " + itos(err) + " for shader '" + shader->name + "'.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
pipeline.set_formats = shader->set_formats;
Added support for push constants
2019-06-08 20:10:52 +00:00
pipeline.push_constant_stages = shader->push_constant.push_constants_vk_stage;
pipeline.pipeline_layout = shader->pipeline_layout;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
pipeline.shader = p_shader;
pipeline.push_constant_size = shader->push_constant.push_constant_size;
#ifdef DEBUG_ENABLED
pipeline.validation.dynamic_state = p_dynamic_state_flags;
pipeline.validation.framebuffer_format = p_framebuffer_format;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
pipeline.validation.render_pass = p_for_render_pass;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
pipeline.validation.vertex_format = p_vertex_format;
pipeline.validation.uses_restart_indices = input_assembly_create_info.primitiveRestartEnable;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
static const uint32_t primitive_divisor[RENDER_PRIMITIVE_MAX] = {
1, 2, 1, 1, 1, 3, 1, 1, 1, 1, 1
};
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
pipeline.validation.primitive_divisor = primitive_divisor[p_render_primitive];
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
static const uint32_t primitive_minimum[RENDER_PRIMITIVE_MAX] = {
1,
2,
2,
2,
2,
3,
3,
3,
3,
3,
1,
};
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
pipeline.validation.primitive_minimum = primitive_minimum[p_render_primitive];
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//create ID to associate with this pipeline
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
RID id = render_pipeline_owner.make_rid(pipeline);
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
//now add all the dependencies
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_add_dependency(id, p_shader);
return id;
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
bool RenderingDeviceVulkan::render_pipeline_is_valid(RID p_pipeline) {
_THREAD_SAFE_METHOD_
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
return render_pipeline_owner.owns(p_pipeline);
}
/**************************/
/**** COMPUTE PIPELINE ****/
/**************************/
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
RID RenderingDeviceVulkan::compute_pipeline_create(RID p_shader, const Vector<PipelineSpecializationConstant> &p_specialization_constants) {
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
_THREAD_SAFE_METHOD_
//needs a shader
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Shader *shader = shader_owner.get_or_null(p_shader);
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_COND_V(!shader, RID());
ERR_FAIL_COND_V_MSG(!shader->is_compute, RID(),
"Non-compute shaders can't be used in compute pipelines");
//finally, pipeline create info
VkComputePipelineCreateInfo compute_pipeline_create_info;
compute_pipeline_create_info.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
compute_pipeline_create_info.pNext = nullptr;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
compute_pipeline_create_info.flags = 0;
compute_pipeline_create_info.stage = shader->pipeline_stages[0];
compute_pipeline_create_info.layout = shader->pipeline_layout;
Vulkan: Use VK_NULL_HANDLE for handles, fixes 32-bit builds Some Vulkan types are defined as "non dispatchable handles" and use a different typedef on 32-bit and 64-bit systems (struct pointer on 64-bit, `uint64_t` otherwise). https://github.com/KhronosGroup/Vulkan-Headers/blob/0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5c/include/vulkan/vulkan_core.h#L59-L65 Contrarily to `NULL`, `nullptr` can't be converted to `uint64_t` so build was now failing on 32-bit after converting the codebase from using `NULL` to `nullptr`. Fixes #37620.
2020-04-06 08:17:42 +00:00
compute_pipeline_create_info.basePipelineHandle = VK_NULL_HANDLE;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
compute_pipeline_create_info.basePipelineIndex = 0;
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
VkSpecializationInfo specialization_info;
Vector<VkSpecializationMapEntry> specialization_map_entries;
Vector<uint32_t> specialization_constant_data;
if (shader->specialization_constants.size()) {
specialization_constant_data.resize(shader->specialization_constants.size());
uint32_t *data_ptr = specialization_constant_data.ptrw();
for (int i = 0; i < shader->specialization_constants.size(); i++) {
Fix various typos Follow-up typos found via `codespell -q 3 -S ./thirdparty,*.po,./DONORS.md -L ackward,ang,ans,ba,beng,cas,childs,childrens,dof,doubleclick,fave,findn,hist,inout,leapyear,lod,nd,numer,ois,ony,paket,seeked,sinc,switchs,te,uint`
2021-07-25 11:18:25 +00:00
//see if overridden
Implement Specialization Constants * Added support to our local copy of SpirV Reflect (which does not support it). * Pass them on render or compute pipeline creation. * Not implemented in our shaders yet.
2021-07-09 19:48:28 +00:00
const Shader::SpecializationConstant &sc = shader->specialization_constants[i];
data_ptr[i] = sc.constant.int_value; //just copy the 32 bits
for (int j = 0; j < p_specialization_constants.size(); j++) {
const PipelineSpecializationConstant &psc = p_specialization_constants[j];
if (psc.constant_id == sc.constant.constant_id) {
ERR_FAIL_COND_V_MSG(psc.type != sc.constant.type, RID(), "Specialization constant provided for id (" + itos(sc.constant.constant_id) + ") is of the wrong type.");
data_ptr[i] = sc.constant.int_value;
break;
}
}
VkSpecializationMapEntry entry;
entry.constantID = sc.constant.constant_id;
entry.offset = i * sizeof(uint32_t);
entry.size = sizeof(uint32_t);
specialization_map_entries.push_back(entry);
}
specialization_info.dataSize = specialization_constant_data.size() * sizeof(uint32_t);
specialization_info.pData = data_ptr;
specialization_info.mapEntryCount = specialization_map_entries.size();
specialization_info.pMapEntries = specialization_map_entries.ptr();
compute_pipeline_create_info.stage.pSpecializationInfo = &specialization_info;
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ComputePipeline pipeline;
Vulkan: Use VK_NULL_HANDLE for handles, fixes 32-bit builds Some Vulkan types are defined as "non dispatchable handles" and use a different typedef on 32-bit and 64-bit systems (struct pointer on 64-bit, `uint64_t` otherwise). https://github.com/KhronosGroup/Vulkan-Headers/blob/0e78ffd1dcfc3e9f14a966b9660dbc59bd967c5c/include/vulkan/vulkan_core.h#L59-L65 Contrarily to `NULL`, `nullptr` can't be converted to `uint64_t` so build was now failing on 32-bit after converting the codebase from using `NULL` to `nullptr`. Fixes #37620.
2020-04-06 08:17:42 +00:00
VkResult err = vkCreateComputePipelines(device, VK_NULL_HANDLE, 1, &compute_pipeline_create_info, nullptr, &pipeline.pipeline);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateComputePipelines failed with error " + itos(err) + ".");
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
pipeline.set_formats = shader->set_formats;
pipeline.push_constant_stages = shader->push_constant.push_constants_vk_stage;
pipeline.pipeline_layout = shader->pipeline_layout;
pipeline.shader = p_shader;
pipeline.push_constant_size = shader->push_constant.push_constant_size;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
pipeline.local_group_size[0] = shader->compute_local_size[0];
pipeline.local_group_size[1] = shader->compute_local_size[1];
pipeline.local_group_size[2] = shader->compute_local_size[2];
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
//create ID to associate with this pipeline
RID id = compute_pipeline_owner.make_rid(pipeline);
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
//now add all the dependencies
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
_add_dependency(id, p_shader);
return id;
}
bool RenderingDeviceVulkan::compute_pipeline_is_valid(RID p_pipeline) {
return compute_pipeline_owner.owns(p_pipeline);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
/****************/
/**** SCREEN ****/
/****************/
Added a Window node, and made it the scene root. Still a lot of work to do.
2020-03-04 01:51:12 +00:00
int RenderingDeviceVulkan::screen_get_width(DisplayServer::WindowID p_screen) const {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
ERR_FAIL_COND_V_MSG(local_device.is_valid(), -1, "Local devices have no screen");
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
return context->window_get_width(p_screen);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Added a Window node, and made it the scene root. Still a lot of work to do.
2020-03-04 01:51:12 +00:00
int RenderingDeviceVulkan::screen_get_height(DisplayServer::WindowID p_screen) const {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
ERR_FAIL_COND_V_MSG(local_device.is_valid(), -1, "Local devices have no screen");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
return context->window_get_height(p_screen);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::screen_get_framebuffer_format() const {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
ERR_FAIL_COND_V_MSG(local_device.is_valid(), INVALID_ID, "Local devices have no screen");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//very hacky, but not used often per frame so I guess ok
VkFormat vkformat = context->get_screen_format();
DataFormat format = DATA_FORMAT_MAX;
for (int i = 0; i < DATA_FORMAT_MAX; i++) {
if (vkformat == vulkan_formats[i]) {
format = DataFormat(i);
break;
}
}
ERR_FAIL_COND_V(format == DATA_FORMAT_MAX, INVALID_ID);
AttachmentFormat attachment;
attachment.format = format;
attachment.samples = TEXTURE_SAMPLES_1;
attachment.usage_flags = TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
Vector<AttachmentFormat> screen_attachment;
screen_attachment.push_back(attachment);
return const_cast<RenderingDeviceVulkan *>(this)->framebuffer_format_create(screen_attachment);
}
/*******************/
/**** DRAW LIST ****/
/*******************/
Added a Window node, and made it the scene root. Still a lot of work to do.
2020-03-04 01:51:12 +00:00
RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_begin_for_screen(DisplayServer::WindowID p_screen, const Color &p_clear_color) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
ERR_FAIL_COND_V_MSG(local_device.is_valid(), INVALID_ID, "Local devices have no screen");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
ERR_FAIL_COND_V_MSG(draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time.");
ERR_FAIL_COND_V_MSG(compute_list != nullptr, INVALID_ID, "Only one draw/compute list can be active at the same time.");
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
[Windows] Fix Vulkan driver crash on sub-window minimization.
2022-02-17 11:09:17 +00:00
if (!context->window_is_valid_swapchain(p_screen)) {
return INVALID_ID;
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
Size2i size = Size2i(context->window_get_width(p_screen), context->window_get_height(p_screen));
_draw_list_allocate(Rect2i(Vector2i(), size), 0, 0);
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#ifdef DEBUG_ENABLED
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
draw_list_framebuffer_format = screen_get_framebuffer_format();
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#endif
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
draw_list_subpass_count = 1;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkRenderPassBeginInfo render_pass_begin;
render_pass_begin.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
render_pass_begin.pNext = nullptr;
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
render_pass_begin.renderPass = context->window_get_render_pass(p_screen);
render_pass_begin.framebuffer = context->window_get_framebuffer(p_screen);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
render_pass_begin.renderArea.extent.width = size.width;
render_pass_begin.renderArea.extent.height = size.height;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
render_pass_begin.renderArea.offset.x = 0;
render_pass_begin.renderArea.offset.y = 0;
render_pass_begin.clearValueCount = 1;
VkClearValue clear_value;
clear_value.color.float32[0] = p_clear_color.r;
clear_value.color.float32[1] = p_clear_color.g;
clear_value.color.float32[2] = p_clear_color.b;
clear_value.color.float32[3] = p_clear_color.a;
render_pass_begin.pClearValues = &clear_value;
vkCmdBeginRenderPass(command_buffer, &render_pass_begin, VK_SUBPASS_CONTENTS_INLINE);
uint32_t size_x = screen_get_width(p_screen);
uint32_t size_y = screen_get_height(p_screen);
VkViewport viewport;
viewport.x = 0;
viewport.y = 0;
viewport.width = size_x;
viewport.height = size_y;
viewport.minDepth = 0;
viewport.maxDepth = 1.0;
vkCmdSetViewport(command_buffer, 0, 1, &viewport);
VkRect2D scissor;
scissor.offset.x = 0;
scissor.offset.y = 0;
scissor.extent.width = size_x;
Improve HiDPI handling, fix incorrect window viewport clipping.
2019-10-14 08:30:27 +00:00
scissor.extent.height = size_y;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vkCmdSetScissor(command_buffer, 0, 1, &scissor);
Threaded optimizations to cull and render -Reorganize thread work pool for rendering -Fixes to make secondary command buffers to work (disabled because they need more testing)
2021-01-04 20:00:44 +00:00
return int64_t(ID_TYPE_DRAW_LIST) << ID_BASE_SHIFT;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
Error RenderingDeviceVulkan::_draw_list_setup_framebuffer(Framebuffer *p_framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, VkFramebuffer *r_framebuffer, VkRenderPass *r_render_pass, uint32_t *r_subpass_count) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Framebuffer::VersionKey vk;
GIProbes working.
2019-10-03 20:39:08 +00:00
vk.initial_color_action = p_initial_color_action;
vk.final_color_action = p_final_color_action;
vk.initial_depth_action = p_initial_depth_action;
vk.final_depth_action = p_final_depth_action;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
vk.view_count = p_framebuffer->view_count;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (!p_framebuffer->framebuffers.has(vk)) {
//need to create this version
Framebuffer::Version version;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
version.render_pass = _render_pass_create(framebuffer_formats[p_framebuffer->format_id].E->key().attachments, framebuffer_formats[p_framebuffer->format_id].E->key().passes, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_framebuffer->view_count);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkFramebufferCreateInfo framebuffer_create_info;
framebuffer_create_info.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
framebuffer_create_info.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
framebuffer_create_info.flags = 0;
framebuffer_create_info.renderPass = version.render_pass;
Vector<VkImageView> attachments;
for (int i = 0; i < p_framebuffer->texture_ids.size(); i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(p_framebuffer->texture_ids[i]);
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (texture) {
attachments.push_back(texture->view);
ERR_FAIL_COND_V(texture->width != p_framebuffer->size.width, ERR_BUG);
ERR_FAIL_COND_V(texture->height != p_framebuffer->size.height, ERR_BUG);
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
framebuffer_create_info.attachmentCount = attachments.size();
framebuffer_create_info.pAttachments = attachments.ptr();
framebuffer_create_info.width = p_framebuffer->size.width;
framebuffer_create_info.height = p_framebuffer->size.height;
framebuffer_create_info.layers = 1;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult err = vkCreateFramebuffer(device, &framebuffer_create_info, nullptr, &version.framebuffer);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_V_MSG(err, ERR_CANT_CREATE, "vkCreateFramebuffer failed with error " + itos(err) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
version.subpass_count = framebuffer_formats[p_framebuffer->format_id].E->key().passes.size();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
p_framebuffer->framebuffers.insert(vk, version);
}
const Framebuffer::Version &version = p_framebuffer->framebuffers[vk];
*r_framebuffer = version.framebuffer;
*r_render_pass = version.render_pass;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
*r_subpass_count = version.subpass_count;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
return OK;
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Error RenderingDeviceVulkan::_draw_list_render_pass_begin(Framebuffer *framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_colors, float p_clear_depth, uint32_t p_clear_stencil, Point2i viewport_offset, Point2i viewport_size, VkFramebuffer vkframebuffer, VkRenderPass render_pass, VkCommandBuffer command_buffer, VkSubpassContents subpass_contents, const Vector<RID> &p_storage_textures) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkRenderPassBeginInfo render_pass_begin;
render_pass_begin.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
render_pass_begin.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
render_pass_begin.renderPass = render_pass;
render_pass_begin.framebuffer = vkframebuffer;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
/*
* Given how API works, it makes sense to always fully operate on the whole framebuffer.
* This allows better continue operations for operations like shadowmapping.
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
render_pass_begin.renderArea.extent.width = viewport_size.width;
render_pass_begin.renderArea.extent.height = viewport_size.height;
render_pass_begin.renderArea.offset.x = viewport_offset.x;
render_pass_begin.renderArea.offset.y = viewport_offset.y;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
*/
render_pass_begin.renderArea.extent.width = framebuffer->size.width;
render_pass_begin.renderArea.extent.height = framebuffer->size.height;
render_pass_begin.renderArea.offset.x = 0;
render_pass_begin.renderArea.offset.y = 0;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Vector<VkClearValue> clear_values;
GIProbes working.
2019-10-03 20:39:08 +00:00
clear_values.resize(framebuffer->texture_ids.size());
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
int clear_values_count = 0;
GIProbes working.
2019-10-03 20:39:08 +00:00
{
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
int color_index = 0;
for (int i = 0; i < framebuffer->texture_ids.size(); i++) {
VkClearValue clear_value;
GIProbes working.
2019-10-03 20:39:08 +00:00
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]);
if (!texture) {
color_index++;
continue;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
if (color_index < p_clear_colors.size() && texture->usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_INDEX_V(color_index, p_clear_colors.size(), ERR_BUG); //a bug
Color clear_color = p_clear_colors[color_index];
clear_value.color.float32[0] = clear_color.r;
clear_value.color.float32[1] = clear_color.g;
clear_value.color.float32[2] = clear_color.b;
clear_value.color.float32[3] = clear_color.a;
color_index++;
} else if (texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
GIProbes working.
2019-10-03 20:39:08 +00:00
clear_value.depthStencil.depth = p_clear_depth;
clear_value.depthStencil.stencil = p_clear_stencil;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
clear_value.color.float32[0] = 0;
clear_value.color.float32[1] = 0;
clear_value.color.float32[2] = 0;
clear_value.color.float32[3] = 0;
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
clear_values.write[clear_values_count++] = clear_value;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
render_pass_begin.clearValueCount = clear_values_count;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
render_pass_begin.pClearValues = clear_values.ptr();
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (int i = 0; i < p_storage_textures.size(); i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(p_storage_textures[i]);
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (!texture) {
continue;
}
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
ERR_CONTINUE_MSG(!(texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT), "Supplied storage texture " + itos(i) + " for draw list is not set to be used for storage.");
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT) {
//must change layout to general
VkImageMemoryBarrier image_memory_barrier;
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
image_memory_barrier.pNext = nullptr;
image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
image_memory_barrier.oldLayout = texture->layout;
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_GENERAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = texture->image;
image_memory_barrier.subresourceRange.aspectMask = texture->read_aspect_mask;
image_memory_barrier.subresourceRange.baseMipLevel = texture->base_mipmap;
image_memory_barrier.subresourceRange.levelCount = texture->mipmaps;
image_memory_barrier.subresourceRange.baseArrayLayer = texture->base_layer;
image_memory_barrier.subresourceRange.layerCount = texture->layers;
Revert "Revert "Synchronization validation fix patch set 3 (inclusive of all previous patches)""
2020-10-19 20:32:19 +00:00
vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
texture->layout = VK_IMAGE_LAYOUT_GENERAL;
draw_list_storage_textures.push_back(p_storage_textures[i]);
}
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vkCmdBeginRenderPass(command_buffer, &render_pass_begin, subpass_contents);
//mark textures as bound
draw_list_bound_textures.clear();
GIProbes working.
2019-10-03 20:39:08 +00:00
draw_list_unbind_color_textures = p_final_color_action != FINAL_ACTION_CONTINUE;
draw_list_unbind_depth_textures = p_final_depth_action != FINAL_ACTION_CONTINUE;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
for (int i = 0; i < framebuffer->texture_ids.size(); i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]);
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (!texture) {
continue;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
texture->bound = true;
draw_list_bound_textures.push_back(framebuffer->texture_ids[i]);
}
return OK;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
void RenderingDeviceVulkan::_draw_list_insert_clear_region(DrawList *draw_list, Framebuffer *framebuffer, Point2i viewport_offset, Point2i viewport_size, bool p_clear_color, const Vector<Color> &p_clear_colors, bool p_clear_depth, float p_depth, uint32_t p_stencil) {
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
ERR_FAIL_COND_MSG(p_clear_color && p_clear_colors.size() != framebuffer->texture_ids.size(), "Clear color values supplied (" + itos(p_clear_colors.size()) + ") differ from the amount required for framebuffer color attachments (" + itos(framebuffer->texture_ids.size()) + ").");
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Vector<VkClearAttachment> clear_attachments;
int color_index = 0;
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
int texture_index = 0;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
for (int i = 0; i < framebuffer->texture_ids.size(); i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]);
Fix may be used unitialized warnings in _render_pass_create() and _draw_list_insert_clear_region()
2020-05-18 14:49:52 +00:00
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (!texture) {
texture_index++;
continue;
}
VkClearAttachment clear_at = {};
GIProbes working.
2019-10-03 20:39:08 +00:00
if (p_clear_color && texture->usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
Color clear_color = p_clear_colors[texture_index++];
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
clear_at.clearValue.color.float32[0] = clear_color.r;
clear_at.clearValue.color.float32[1] = clear_color.g;
clear_at.clearValue.color.float32[2] = clear_color.b;
clear_at.clearValue.color.float32[3] = clear_color.a;
clear_at.colorAttachment = color_index++;
clear_at.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
GIProbes working.
2019-10-03 20:39:08 +00:00
} else if (p_clear_depth && texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
clear_at.clearValue.depthStencil.depth = p_depth;
clear_at.clearValue.depthStencil.stencil = p_stencil;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
clear_at.colorAttachment = 0;
clear_at.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
if (format_has_stencil(texture->format)) {
clear_at.aspectMask |= VK_IMAGE_ASPECT_STENCIL_BIT;
}
} else {
ERR_CONTINUE(true);
}
clear_attachments.push_back(clear_at);
}
VkClearRect cr;
cr.baseArrayLayer = 0;
cr.layerCount = 1;
cr.rect.offset.x = viewport_offset.x;
cr.rect.offset.y = viewport_offset.y;
cr.rect.extent.width = viewport_size.width;
cr.rect.extent.height = viewport_size.height;
vkCmdClearAttachments(draw_list->command_buffer, clear_attachments.size(), clear_attachments.ptr(), 1, &cr);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_begin(RID p_framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values, float p_clear_depth, uint32_t p_clear_stencil, const Rect2 &p_region, const Vector<RID> &p_storage_textures) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
ERR_FAIL_COND_V_MSG(draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time.");
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
ERR_FAIL_COND_V_MSG(compute_list != nullptr && !compute_list->state.allow_draw_overlap, INVALID_ID, "Only one draw/compute list can be active at the same time.");
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_framebuffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_V(!framebuffer, INVALID_ID);
Point2i viewport_offset;
Point2i viewport_size = framebuffer->size;
GIProbes working.
2019-10-03 20:39:08 +00:00
bool needs_clear_color = false;
bool needs_clear_depth = false;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
if (p_region != Rect2() && p_region != Rect2(Vector2(), viewport_size)) { //check custom region
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rect2i viewport(viewport_offset, viewport_size);
Rect2i regioni = p_region;
if (!(regioni.position.x >= viewport.position.x) && (regioni.position.y >= viewport.position.y) &&
((regioni.position.x + regioni.size.x) <= (viewport.position.x + viewport.size.x)) &&
((regioni.position.y + regioni.size.y) <= (viewport.position.y + viewport.size.y))) {
ERR_FAIL_V_MSG(INVALID_ID, "When supplying a custom region, it must be contained within the framebuffer rectangle");
}
viewport_offset = regioni.position;
viewport_size = regioni.size;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (p_initial_color_action == INITIAL_ACTION_CLEAR_REGION_CONTINUE) {
needs_clear_color = true;
p_initial_color_action = INITIAL_ACTION_CONTINUE;
}
if (p_initial_depth_action == INITIAL_ACTION_CLEAR_REGION_CONTINUE) {
needs_clear_depth = true;
p_initial_depth_action = INITIAL_ACTION_CONTINUE;
}
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
if (p_initial_color_action == INITIAL_ACTION_CLEAR_REGION) {
GIProbes working.
2019-10-03 20:39:08 +00:00
needs_clear_color = true;
p_initial_color_action = INITIAL_ACTION_KEEP;
}
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
if (p_initial_depth_action == INITIAL_ACTION_CLEAR_REGION) {
GIProbes working.
2019-10-03 20:39:08 +00:00
needs_clear_depth = true;
p_initial_depth_action = INITIAL_ACTION_KEEP;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
GIProbes working.
2019-10-03 20:39:08 +00:00
if (p_initial_color_action == INITIAL_ACTION_CLEAR) { //check clear values
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
int color_count = 0;
for (int i = 0; i < framebuffer->texture_ids.size(); i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]);
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (!texture || !(texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
color_count++;
}
}
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
ERR_FAIL_COND_V_MSG(p_clear_color_values.size() != color_count, INVALID_ID, "Clear color values supplied (" + itos(p_clear_color_values.size()) + ") differ from the amount required for framebuffer color attachments (" + itos(color_count) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
VkFramebuffer vkframebuffer;
VkRenderPass render_pass;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
Error err = _draw_list_setup_framebuffer(framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, &vkframebuffer, &render_pass, &draw_list_subpass_count);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_V(err != OK, INVALID_ID);
VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
err = _draw_list_render_pass_begin(framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, viewport_offset, viewport_size, vkframebuffer, render_pass, command_buffer, VK_SUBPASS_CONTENTS_INLINE, p_storage_textures);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (err != OK) {
return INVALID_ID;
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
draw_list_render_pass = render_pass;
draw_list_vkframebuffer = vkframebuffer;
_draw_list_allocate(Rect2i(viewport_offset, viewport_size), 0, 0);
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#ifdef DEBUG_ENABLED
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
draw_list_framebuffer_format = framebuffer->format_id;
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#endif
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
draw_list_current_subpass = 0;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
GIProbes working.
2019-10-03 20:39:08 +00:00
if (needs_clear_color || needs_clear_depth) {
_draw_list_insert_clear_region(draw_list, framebuffer, viewport_offset, viewport_size, needs_clear_color, p_clear_color_values, needs_clear_depth, p_clear_depth, p_clear_stencil);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkViewport viewport;
viewport.x = viewport_offset.x;
viewport.y = viewport_offset.y;
viewport.width = viewport_size.width;
viewport.height = viewport_size.height;
viewport.minDepth = 0;
viewport.maxDepth = 1.0;
vkCmdSetViewport(command_buffer, 0, 1, &viewport);
VkRect2D scissor;
scissor.offset.x = viewport_offset.x;
scissor.offset.y = viewport_offset.y;
scissor.extent.width = viewport_size.width;
scissor.extent.height = viewport_size.height;
vkCmdSetScissor(command_buffer, 0, 1, &scissor);
Threaded optimizations to cull and render -Reorganize thread work pool for rendering -Fixes to make secondary command buffers to work (disabled because they need more testing)
2021-01-04 20:00:44 +00:00
return int64_t(ID_TYPE_DRAW_LIST) << ID_BASE_SHIFT;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Error RenderingDeviceVulkan::draw_list_begin_split(RID p_framebuffer, uint32_t p_splits, DrawListID *r_split_ids, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values, float p_clear_depth, uint32_t p_clear_stencil, const Rect2 &p_region, const Vector<RID> &p_storage_textures) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
ERR_FAIL_COND_V(p_splits < 1, ERR_INVALID_DECLARATION);
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_framebuffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_V(!framebuffer, ERR_INVALID_DECLARATION);
Point2i viewport_offset;
Point2i viewport_size = framebuffer->size;
GIProbes working.
2019-10-03 20:39:08 +00:00
bool needs_clear_color = false;
bool needs_clear_depth = false;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
if (p_region != Rect2() && p_region != Rect2(Vector2(), viewport_size)) { //check custom region
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rect2i viewport(viewport_offset, viewport_size);
Rect2i regioni = p_region;
if (!(regioni.position.x >= viewport.position.x) && (regioni.position.y >= viewport.position.y) &&
((regioni.position.x + regioni.size.x) <= (viewport.position.x + viewport.size.x)) &&
((regioni.position.y + regioni.size.y) <= (viewport.position.y + viewport.size.y))) {
ERR_FAIL_V_MSG(ERR_INVALID_PARAMETER, "When supplying a custom region, it must be contained within the framebuffer rectangle");
}
viewport_offset = regioni.position;
viewport_size = regioni.size;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
if (p_initial_color_action == INITIAL_ACTION_CLEAR_REGION) {
GIProbes working.
2019-10-03 20:39:08 +00:00
needs_clear_color = true;
p_initial_color_action = INITIAL_ACTION_KEEP;
}
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
if (p_initial_depth_action == INITIAL_ACTION_CLEAR_REGION) {
GIProbes working.
2019-10-03 20:39:08 +00:00
needs_clear_depth = true;
p_initial_depth_action = INITIAL_ACTION_KEEP;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
GIProbes working.
2019-10-03 20:39:08 +00:00
if (p_initial_color_action == INITIAL_ACTION_CLEAR) { //check clear values
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
int color_count = 0;
for (int i = 0; i < framebuffer->texture_ids.size(); i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Add color pass flags to Forward Clustered renderer This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask. The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass. This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used. Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
2022-02-17 09:56:22 +00:00
if (!texture || !(texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
color_count++;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND_V_MSG(p_clear_color_values.size() != color_count, ERR_INVALID_PARAMETER,
"Clear color values supplied (" + itos(p_clear_color_values.size()) + ") differ from the amount required for framebuffer (" + itos(color_count) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
VkFramebuffer vkframebuffer;
VkRenderPass render_pass;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
Error err = _draw_list_setup_framebuffer(framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, &vkframebuffer, &render_pass, &draw_list_subpass_count);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_V(err != OK, ERR_CANT_CREATE);
VkCommandBuffer frame_command_buffer = frames[frame].draw_command_buffer;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
err = _draw_list_render_pass_begin(framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, viewport_offset, viewport_size, vkframebuffer, render_pass, frame_command_buffer, VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS, p_storage_textures);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (err != OK) {
return ERR_CANT_CREATE;
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
draw_list_current_subpass = 0;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#ifdef DEBUG_ENABLED
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
draw_list_framebuffer_format = framebuffer->format_id;
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#endif
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
draw_list_render_pass = render_pass;
draw_list_vkframebuffer = vkframebuffer;
GIProbes working.
2019-10-03 20:39:08 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
err = _draw_list_allocate(Rect2i(viewport_offset, viewport_size), p_splits, 0);
if (err != OK) {
return err;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (needs_clear_color || needs_clear_depth) {
_draw_list_insert_clear_region(&draw_list[0], framebuffer, viewport_offset, viewport_size, needs_clear_color, p_clear_color_values, needs_clear_depth, p_clear_depth, p_clear_stencil);
}
for (uint32_t i = 0; i < p_splits; i++) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkViewport viewport;
viewport.x = viewport_offset.x;
viewport.y = viewport_offset.y;
viewport.width = viewport_size.width;
viewport.height = viewport_size.height;
viewport.minDepth = 0;
viewport.maxDepth = 1.0;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
vkCmdSetViewport(draw_list[i].command_buffer, 0, 1, &viewport);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkRect2D scissor;
scissor.offset.x = viewport_offset.x;
scissor.offset.y = viewport_offset.y;
scissor.extent.width = viewport_size.width;
scissor.extent.height = viewport_size.height;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
vkCmdSetScissor(draw_list[i].command_buffer, 0, 1, &scissor);
Threaded optimizations to cull and render -Reorganize thread work pool for rendering -Fixes to make secondary command buffers to work (disabled because they need more testing)
2021-01-04 20:00:44 +00:00
r_split_ids[i] = (int64_t(ID_TYPE_SPLIT_DRAW_LIST) << ID_BASE_SHIFT) + i;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return OK;
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
RenderingDeviceVulkan::DrawList *RenderingDeviceVulkan::_get_draw_list_ptr(DrawListID p_id) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (p_id < 0) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
return nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
if (!draw_list) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
return nullptr;
Threaded optimizations to cull and render -Reorganize thread work pool for rendering -Fixes to make secondary command buffers to work (disabled because they need more testing)
2021-01-04 20:00:44 +00:00
} else if (p_id == (int64_t(ID_TYPE_DRAW_LIST) << ID_BASE_SHIFT)) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (draw_list_split) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
return nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return draw_list;
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
} else if (p_id >> DrawListID(ID_BASE_SHIFT) == ID_TYPE_SPLIT_DRAW_LIST) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (!draw_list_split) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
return nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
uint64_t index = p_id & ((DrawListID(1) << DrawListID(ID_BASE_SHIFT)) - 1); //mask
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (index >= draw_list_count) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
return nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
return &draw_list[index];
} else {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
return nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::draw_list_bind_render_pipeline(DrawListID p_list, RID p_render_pipeline) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
DrawList *dl = _get_draw_list_ptr(p_list);
ERR_FAIL_COND(!dl);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
const RenderPipeline *pipeline = render_pipeline_owner.get_or_null(p_render_pipeline);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND(!pipeline);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
ERR_FAIL_COND(pipeline->validation.framebuffer_format != draw_list_framebuffer_format && pipeline->validation.render_pass != draw_list_current_subpass);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
if (p_render_pipeline == dl->state.pipeline) {
return; //redundant state, return.
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
dl->state.pipeline = p_render_pipeline;
dl->state.pipeline_layout = pipeline->pipeline_layout;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vkCmdBindPipeline(dl->command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline->pipeline);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
if (dl->state.pipeline_shader != pipeline->shader) {
// shader changed, so descriptor sets may become incompatible.
//go through ALL sets, and unbind them (and all those above) if the format is different
uint32_t pcount = pipeline->set_formats.size(); //formats count in this pipeline
dl->state.set_count = MAX(dl->state.set_count, pcount);
const uint32_t *pformats = pipeline->set_formats.ptr(); //pipeline set formats
bool sets_valid = true; //once invalid, all above become invalid
for (uint32_t i = 0; i < pcount; i++) {
//if a part of the format is different, invalidate it (and the rest)
if (!sets_valid || dl->state.sets[i].pipeline_expected_format != pformats[i]) {
dl->state.sets[i].bound = false;
dl->state.sets[i].pipeline_expected_format = pformats[i];
sets_valid = false;
}
}
for (uint32_t i = pcount; i < dl->state.set_count; i++) {
//unbind the ones above (not used) if exist
dl->state.sets[i].bound = false;
}
dl->state.set_count = pcount; //update set count
if (pipeline->push_constant_size) {
dl->state.pipeline_push_constant_stages = pipeline->push_constant_stages;
#ifdef DEBUG_ENABLED
Vulkan: Initialize struct members on declaration Supersedes and closes #38945.
2020-06-03 08:45:44 +00:00
dl->validation.pipeline_push_constant_supplied = false;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
}
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
dl->state.pipeline_shader = pipeline->shader;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
}
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//update render pass pipeline info
dl->validation.pipeline_active = true;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
dl->validation.pipeline_dynamic_state = pipeline->validation.dynamic_state;
dl->validation.pipeline_vertex_format = pipeline->validation.vertex_format;
dl->validation.pipeline_uses_restart_indices = pipeline->validation.uses_restart_indices;
dl->validation.pipeline_primitive_divisor = pipeline->validation.primitive_divisor;
dl->validation.pipeline_primitive_minimum = pipeline->validation.primitive_minimum;
Added support for push constants
2019-06-08 20:10:52 +00:00
dl->validation.pipeline_push_constant_size = pipeline->push_constant_size;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::draw_list_bind_uniform_set(DrawListID p_list, RID p_uniform_set, uint32_t p_index) {
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Static analyzer fixes: Removes unused code in OS. Fixes return types. Fixes few typos.
2020-12-08 18:58:49 +00:00
ERR_FAIL_COND_MSG(p_index >= limits.maxBoundDescriptorSets || p_index >= MAX_UNIFORM_SETS,
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
"Attempting to bind a descriptor set (" + itos(p_index) + ") greater than what the hardware supports (" + itos(limits.maxBoundDescriptorSets) + ").");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
DrawList *dl = _get_draw_list_ptr(p_list);
ERR_FAIL_COND(!dl);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
const UniformSet *uniform_set = uniform_set_owner.get_or_null(p_uniform_set);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND(!uniform_set);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
if (p_index > dl->state.set_count) {
dl->state.set_count = p_index;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Improved validation of attachments
2019-06-07 18:20:01 +00:00
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
dl->state.sets[p_index].descriptor_set = uniform_set->descriptor_set; //update set pointer
dl->state.sets[p_index].bound = false; //needs rebind
dl->state.sets[p_index].uniform_set_format = uniform_set->format;
dl->state.sets[p_index].uniform_set = p_uniform_set;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
uint32_t mst_count = uniform_set->mutable_storage_textures.size();
if (mst_count) {
Texture **mst_textures = const_cast<UniformSet *>(uniform_set)->mutable_storage_textures.ptrw();
for (uint32_t i = 0; i < mst_count; i++) {
if (mst_textures[i]->used_in_frame != frames_drawn) {
mst_textures[i]->used_in_frame = frames_drawn;
mst_textures[i]->used_in_transfer = false;
mst_textures[i]->used_in_compute = false;
}
mst_textures[i]->used_in_raster = true;
}
}
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Improved validation of attachments
2019-06-07 18:20:01 +00:00
{ //validate that textures bound are not attached as framebuffer bindings
uint32_t attachable_count = uniform_set->attachable_textures.size();
Fix error spam on wrong attachment -For now, disable reading from depth this was always broken, needs to be fixed later -Give better error showing binding and set when this happens.
2020-12-19 13:18:08 +00:00
const UniformSet::AttachableTexture *attachable_ptr = uniform_set->attachable_textures.ptr();
Improved validation of attachments
2019-06-07 18:20:01 +00:00
uint32_t bound_count = draw_list_bound_textures.size();
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
const RID *bound_ptr = draw_list_bound_textures.ptr();
Improved validation of attachments
2019-06-07 18:20:01 +00:00
for (uint32_t i = 0; i < attachable_count; i++) {
for (uint32_t j = 0; j < bound_count; j++) {
Fix error spam on wrong attachment -For now, disable reading from depth this was always broken, needs to be fixed later -Give better error showing binding and set when this happens.
2020-12-19 13:18:08 +00:00
ERR_FAIL_COND_MSG(attachable_ptr[i].texture == bound_ptr[j],
"Attempted to use the same texture in framebuffer attachment and a uniform (set: " + itos(p_index) + ", binding: " + itos(attachable_ptr[i].bind) + "), this is not allowed.");
Improved validation of attachments
2019-06-07 18:20:01 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::draw_list_bind_vertex_array(DrawListID p_list, RID p_vertex_array) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
DrawList *dl = _get_draw_list_ptr(p_list);
ERR_FAIL_COND(!dl);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
const VertexArray *vertex_array = vertex_array_owner.get_or_null(p_vertex_array);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND(!vertex_array);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
if (dl->state.vertex_array == p_vertex_array) {
return; //already set
}
dl->state.vertex_array = p_vertex_array;
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
dl->validation.vertex_format = vertex_array->description;
dl->validation.vertex_max_instances_allowed = vertex_array->max_instances_allowed;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
dl->validation.vertex_array_size = vertex_array->vertex_count;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vkCmdBindVertexBuffers(dl->command_buffer, 0, vertex_array->buffers.size(), vertex_array->buffers.ptr(), vertex_array->offsets.ptr());
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::draw_list_bind_index_array(DrawListID p_list, RID p_index_array) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
DrawList *dl = _get_draw_list_ptr(p_list);
ERR_FAIL_COND(!dl);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
const IndexArray *index_array = index_array_owner.get_or_null(p_index_array);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND(!index_array);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
if (dl->state.index_array == p_index_array) {
return; //already set
}
dl->state.index_array = p_index_array;
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
dl->validation.index_array_max_index = index_array->max_index;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
dl->validation.index_array_size = index_array->indices;
dl->validation.index_array_offset = index_array->offset;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vkCmdBindIndexBuffer(dl->command_buffer, index_array->buffer, index_array->offset, index_array->index_type);
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
void RenderingDeviceVulkan::draw_list_set_line_width(DrawListID p_list, float p_width) {
DrawList *dl = _get_draw_list_ptr(p_list);
ERR_FAIL_COND(!dl);
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#ifdef DEBUG_ENABLED
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#endif
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
vkCmdSetLineWidth(dl->command_buffer, p_width);
}
Exposed RenderingDevice to script API Also added an easier way to load native GLSL shaders. Extras: Had to fix no-cache for subresources in resource loader, it was not properly working, making shaders not properly reload. Note: The precommit hooks are broken because they don't seem to support enums from one class being used in another. Feel free to fix this after merging this PR.
2020-04-20 02:19:21 +00:00
void RenderingDeviceVulkan::draw_list_set_push_constant(DrawListID p_list, const void *p_data, uint32_t p_data_size) {
Added support for push constants
2019-06-08 20:10:52 +00:00
DrawList *dl = _get_draw_list_ptr(p_list);
ERR_FAIL_COND(!dl);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif
#ifdef DEBUG_ENABLED
Added support for push constants
2019-06-08 20:10:52 +00:00
ERR_FAIL_COND_MSG(p_data_size != dl->validation.pipeline_push_constant_size,
"This render pipeline requires (" + itos(dl->validation.pipeline_push_constant_size) + ") bytes of push constant data, supplied: (" + itos(p_data_size) + ")");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
vkCmdPushConstants(dl->command_buffer, dl->state.pipeline_layout, dl->state.pipeline_push_constant_stages, 0, p_data_size, p_data);
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#ifdef DEBUG_ENABLED
Vulkan: Initialize struct members on declaration Supersedes and closes #38945.
2020-06-03 08:45:44 +00:00
dl->validation.pipeline_push_constant_supplied = true;
Add runtime GLES2 / Vulkan context selection.
2019-07-12 13:18:30 +00:00
#endif
Added support for push constants
2019-06-08 20:10:52 +00:00
}
GIProbes working.
2019-10-03 20:39:08 +00:00
void RenderingDeviceVulkan::draw_list_draw(DrawListID p_list, bool p_use_indices, uint32_t p_instances, uint32_t p_procedural_vertices) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
DrawList *dl = _get_draw_list_ptr(p_list);
ERR_FAIL_COND(!dl);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(!dl->validation.pipeline_active,
"No render pipeline was set before attempting to draw.");
if (dl->validation.pipeline_vertex_format != INVALID_ID) {
//pipeline uses vertices, validate format
ERR_FAIL_COND_MSG(dl->validation.vertex_format == INVALID_ID,
"No vertex array was bound, and render pipeline expects vertices.");
//make sure format is right
ERR_FAIL_COND_MSG(dl->validation.pipeline_vertex_format != dl->validation.vertex_format,
"The vertex format used to create the pipeline does not match the vertex format bound.");
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
//make sure number of instances is valid
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(p_instances > dl->validation.vertex_max_instances_allowed,
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
"Number of instances requested (" + itos(p_instances) + " is larger than the maximum number supported by the bound vertex array (" + itos(dl->validation.vertex_max_instances_allowed) + ").");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Added support for push constants
2019-06-08 20:10:52 +00:00
if (dl->validation.pipeline_push_constant_size > 0) {
//using push constants, check that they were supplied
Vulkan: Initialize struct members on declaration Supersedes and closes #38945.
2020-06-03 08:45:44 +00:00
ERR_FAIL_COND_MSG(!dl->validation.pipeline_push_constant_supplied,
Added support for push constants
2019-06-08 20:10:52 +00:00
"The shader in this pipeline requires a push constant to be set before drawing, but it's not present.");
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
//Bind descriptor sets
for (uint32_t i = 0; i < dl->state.set_count; i++) {
if (dl->state.sets[i].pipeline_expected_format == 0) {
continue; //nothing expected by this pipeline
}
#ifdef DEBUG_ENABLED
if (dl->state.sets[i].pipeline_expected_format != dl->state.sets[i].uniform_set_format) {
if (dl->state.sets[i].uniform_set_format == 0) {
ERR_FAIL_MSG("Uniforms were never supplied for set (" + itos(i) + ") at the time of drawing, which are required by the pipeline");
} else if (uniform_set_owner.owns(dl->state.sets[i].uniform_set)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
UniformSet *us = uniform_set_owner.get_or_null(dl->state.sets[i].uniform_set);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + "):\n" + _shader_uniform_debug(us->shader_id, us->shader_set) + "\nare not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(dl->state.pipeline_shader));
} else {
ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + ", which was was just freed) are not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(dl->state.pipeline_shader));
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
if (!dl->state.sets[i].bound) {
//All good, see if this requires re-binding
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkCmdBindDescriptorSets(dl->command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS, dl->state.pipeline_layout, i, 1, &dl->state.sets[i].descriptor_set, 0, nullptr);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
dl->state.sets[i].bound = true;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
if (p_use_indices) {
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
GIProbes working.
2019-10-03 20:39:08 +00:00
ERR_FAIL_COND_MSG(p_procedural_vertices > 0,
"Procedural vertices can't be used together with indices.");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(!dl->validation.index_array_size,
"Draw command requested indices, but no index buffer was set.");
GIProbes working.
2019-10-03 20:39:08 +00:00
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (dl->validation.pipeline_vertex_format != INVALID_ID) {
//uses vertices, do some vertex validations
ERR_FAIL_COND_MSG(dl->validation.vertex_array_size < dl->validation.index_array_max_index,
"Index array references (max index: " + itos(dl->validation.index_array_max_index) + ") indices beyond the vertex array size (" + itos(dl->validation.vertex_array_size) + ").");
}
ERR_FAIL_COND_MSG(dl->validation.pipeline_uses_restart_indices != dl->validation.index_buffer_uses_restart_indices,
"The usage of restart indices in index buffer does not match the render primitive in the pipeline.");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
uint32_t to_draw = dl->validation.index_array_size;
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(to_draw < dl->validation.pipeline_primitive_minimum,
"Too few indices (" + itos(to_draw) + ") for the render primitive set in the render pipeline (" + itos(dl->validation.pipeline_primitive_minimum) + ").");
ERR_FAIL_COND_MSG((to_draw % dl->validation.pipeline_primitive_divisor) != 0,
"Index amount (" + itos(to_draw) + ") must be a multiple of the amount of indices required by the render primitive (" + itos(dl->validation.pipeline_primitive_divisor) + ").");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vkCmdDrawIndexed(dl->command_buffer, to_draw, p_instances, dl->validation.index_array_offset, 0, 0);
} else {
GIProbes working.
2019-10-03 20:39:08 +00:00
uint32_t to_draw;
if (p_procedural_vertices > 0) {
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Correct condition wrongly converted to ERR_FAIL_COND_MSG Fixes debugging of giprobes not working, likely other stuff
2020-02-22 14:36:26 +00:00
ERR_FAIL_COND_MSG(dl->validation.pipeline_vertex_format != INVALID_ID,
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
"Procedural vertices requested, but pipeline expects a vertex array.");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
GIProbes working.
2019-10-03 20:39:08 +00:00
to_draw = p_procedural_vertices;
} else {
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(dl->validation.pipeline_vertex_format == INVALID_ID,
"Draw command lacks indices, but pipeline format does not use vertices.");
#endif
to_draw = dl->validation.vertex_array_size;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
ERR_FAIL_COND_MSG(to_draw < dl->validation.pipeline_primitive_minimum,
"Too few vertices (" + itos(to_draw) + ") for the render primitive set in the render pipeline (" + itos(dl->validation.pipeline_primitive_minimum) + ").");
ERR_FAIL_COND_MSG((to_draw % dl->validation.pipeline_primitive_divisor) != 0,
"Vertex amount (" + itos(to_draw) + ") must be a multiple of the amount of vertices required by the render primitive (" + itos(dl->validation.pipeline_primitive_divisor) + ").");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
GIProbes working.
2019-10-03 20:39:08 +00:00
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vkCmdDraw(dl->command_buffer, to_draw, p_instances, 0, 0);
}
}
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::draw_list_enable_scissor(DrawListID p_list, const Rect2 &p_rect) {
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
DrawList *dl = _get_draw_list_ptr(p_list);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
ERR_FAIL_COND(!dl);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
Rect2i rect = p_rect;
rect.position += dl->viewport.position;
Rename Rect2 and Rect2i clip() to intersection()
2020-12-19 12:43:35 +00:00
rect = dl->viewport.intersection(rect);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
if (rect.get_area() == 0) {
return;
}
VkRect2D scissor;
scissor.offset.x = rect.position.x;
scissor.offset.y = rect.position.y;
scissor.extent.width = rect.size.width;
scissor.extent.height = rect.size.height;
vkCmdSetScissor(dl->command_buffer, 0, 1, &scissor);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::draw_list_disable_scissor(DrawListID p_list) {
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
DrawList *dl = _get_draw_list_ptr(p_list);
ERR_FAIL_COND(!dl);
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#ifdef DEBUG_ENABLED
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
Custom material support seems complete.
2019-07-27 13:23:24 +00:00
#endif
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
VkRect2D scissor;
scissor.offset.x = dl->viewport.position.x;
scissor.offset.y = dl->viewport.position.y;
scissor.extent.width = dl->viewport.size.width;
scissor.extent.height = dl->viewport.size.height;
vkCmdSetScissor(dl->command_buffer, 0, 1, &scissor);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
uint32_t RenderingDeviceVulkan::draw_list_get_current_pass() {
return draw_list_current_subpass;
}
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_switch_to_next_pass() {
ERR_FAIL_COND_V(draw_list == nullptr, INVALID_ID);
ERR_FAIL_COND_V(draw_list_current_subpass >= draw_list_subpass_count - 1, INVALID_FORMAT_ID);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
draw_list_current_subpass++;
Rect2i viewport;
_draw_list_free(&viewport);
vkCmdNextSubpass(frames[frame].draw_command_buffer, VK_SUBPASS_CONTENTS_INLINE);
_draw_list_allocate(viewport, 0, draw_list_current_subpass);
return int64_t(ID_TYPE_DRAW_LIST) << ID_BASE_SHIFT;
}
Error RenderingDeviceVulkan::draw_list_switch_to_next_pass_split(uint32_t p_splits, DrawListID *r_split_ids) {
ERR_FAIL_COND_V(draw_list == nullptr, ERR_INVALID_PARAMETER);
ERR_FAIL_COND_V(draw_list_current_subpass >= draw_list_subpass_count - 1, ERR_INVALID_PARAMETER);
draw_list_current_subpass++;
Rect2i viewport;
_draw_list_free(&viewport);
vkCmdNextSubpass(frames[frame].draw_command_buffer, VK_SUBPASS_CONTENTS_INLINE);
_draw_list_allocate(viewport, p_splits, draw_list_current_subpass);
for (uint32_t i = 0; i < p_splits; i++) {
r_split_ids[i] = (int64_t(ID_TYPE_SPLIT_DRAW_LIST) << ID_BASE_SHIFT) + i;
}
return OK;
}
Error RenderingDeviceVulkan::_draw_list_allocate(const Rect2i &p_viewport, uint32_t p_splits, uint32_t p_subpass) {
Make draw/compute lists threadsafe Lock the rendering device while command buffers are in use
2021-11-11 18:55:09 +00:00
// Lock while draw_list is active
_THREAD_SAFE_LOCK_
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (p_splits == 0) {
draw_list = memnew(DrawList);
draw_list->command_buffer = frames[frame].draw_command_buffer;
draw_list->viewport = p_viewport;
draw_list_count = 0;
draw_list_split = false;
} else {
if (p_splits > (uint32_t)split_draw_list_allocators.size()) {
uint32_t from = split_draw_list_allocators.size();
split_draw_list_allocators.resize(p_splits);
for (uint32_t i = from; i < p_splits; i++) {
VkCommandPoolCreateInfo cmd_pool_info;
cmd_pool_info.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO;
cmd_pool_info.pNext = nullptr;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
cmd_pool_info.queueFamilyIndex = context->get_graphics_queue_family_index();
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
cmd_pool_info.flags = VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT;
VkResult res = vkCreateCommandPool(device, &cmd_pool_info, nullptr, &split_draw_list_allocators.write[i].command_pool);
ERR_FAIL_COND_V_MSG(res, ERR_CANT_CREATE, "vkCreateCommandPool failed with error " + itos(res) + ".");
for (int j = 0; j < frame_count; j++) {
VkCommandBuffer command_buffer;
VkCommandBufferAllocateInfo cmdbuf;
//no command buffer exists, create it.
cmdbuf.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
cmdbuf.pNext = nullptr;
cmdbuf.commandPool = split_draw_list_allocators[i].command_pool;
cmdbuf.level = VK_COMMAND_BUFFER_LEVEL_SECONDARY;
cmdbuf.commandBufferCount = 1;
VkResult err = vkAllocateCommandBuffers(device, &cmdbuf, &command_buffer);
ERR_FAIL_COND_V_MSG(err, ERR_CANT_CREATE, "vkAllocateCommandBuffers failed with error " + itos(err) + ".");
split_draw_list_allocators.write[i].command_buffers.push_back(command_buffer);
}
}
}
draw_list = memnew_arr(DrawList, p_splits);
draw_list_count = p_splits;
draw_list_split = true;
for (uint32_t i = 0; i < p_splits; i++) {
//take a command buffer and initialize it
VkCommandBuffer command_buffer = split_draw_list_allocators[i].command_buffers[frame];
VkCommandBufferInheritanceInfo inheritance_info;
inheritance_info.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_INHERITANCE_INFO;
inheritance_info.pNext = nullptr;
inheritance_info.renderPass = draw_list_render_pass;
inheritance_info.subpass = p_subpass;
inheritance_info.framebuffer = draw_list_vkframebuffer;
inheritance_info.occlusionQueryEnable = false;
inheritance_info.queryFlags = 0; //?
inheritance_info.pipelineStatistics = 0;
VkCommandBufferBeginInfo cmdbuf_begin;
cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
cmdbuf_begin.pNext = nullptr;
cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT | VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT;
cmdbuf_begin.pInheritanceInfo = &inheritance_info;
VkResult res = vkResetCommandBuffer(command_buffer, 0);
if (res) {
memdelete_arr(draw_list);
draw_list = nullptr;
ERR_FAIL_V_MSG(ERR_CANT_CREATE, "vkResetCommandBuffer failed with error " + itos(res) + ".");
}
res = vkBeginCommandBuffer(command_buffer, &cmdbuf_begin);
if (res) {
memdelete_arr(draw_list);
draw_list = nullptr;
ERR_FAIL_V_MSG(ERR_CANT_CREATE, "vkBeginCommandBuffer failed with error " + itos(res) + ".");
}
draw_list[i].command_buffer = command_buffer;
draw_list[i].viewport = p_viewport;
}
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
return OK;
}
void RenderingDeviceVulkan::_draw_list_free(Rect2i *r_last_viewport) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (draw_list_split) {
//send all command buffers
VkCommandBuffer *command_buffers = (VkCommandBuffer *)alloca(sizeof(VkCommandBuffer) * draw_list_count);
for (uint32_t i = 0; i < draw_list_count; i++) {
Threaded optimizations to cull and render -Reorganize thread work pool for rendering -Fixes to make secondary command buffers to work (disabled because they need more testing)
2021-01-04 20:00:44 +00:00
vkEndCommandBuffer(draw_list[i].command_buffer);
command_buffers[i] = draw_list[i].command_buffer;
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (r_last_viewport) {
if (i == 0 || draw_list[i].viewport_set) {
*r_last_viewport = draw_list[i].viewport;
}
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
vkCmdExecuteCommands(frames[frame].draw_command_buffer, draw_list_count, command_buffers);
memdelete_arr(draw_list);
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
draw_list = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
if (r_last_viewport) {
*r_last_viewport = draw_list->viewport;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//just end the list
memdelete(draw_list);
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
draw_list = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Make draw/compute lists threadsafe Lock the rendering device while command buffers are in use
2021-11-11 18:55:09 +00:00
// draw_list is no longer active
_THREAD_SAFE_UNLOCK_
Implement Framebuffer Subpass support * Required for better optimizing mobile platforms * Will be used by the Vulkan mobile renderer.
2021-06-24 13:58:36 +00:00
}
void RenderingDeviceVulkan::draw_list_end(uint32_t p_post_barrier) {
_THREAD_SAFE_METHOD_
ERR_FAIL_COND_MSG(!draw_list, "Immediate draw list is already inactive.");
_draw_list_free();
vkCmdEndRenderPass(frames[frame].draw_command_buffer);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
GIProbes working.
2019-10-03 20:39:08 +00:00
for (int i = 0; i < draw_list_bound_textures.size(); i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(draw_list_bound_textures[i]);
GIProbes working.
2019-10-03 20:39:08 +00:00
ERR_CONTINUE(!texture); //wtf
if (draw_list_unbind_color_textures && (texture->usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT)) {
texture->bound = false;
}
if (draw_list_unbind_depth_textures && (texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
texture->bound = false;
}
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
uint32_t barrier_flags = 0;
uint32_t access_flags = 0;
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (p_post_barrier & BARRIER_MASK_RASTER) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT /*| VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT*/;
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT /*| VK_ACCESS_INDIRECT_COMMAND_READ_BIT*/;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
}
if (p_post_barrier & BARRIER_MASK_TRANSFER) {
barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT;
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (barrier_flags == 0) {
barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
draw_list_bound_textures.clear();
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
VkImageMemoryBarrier *image_barriers = nullptr;
uint32_t image_barrier_count = draw_list_storage_textures.size();
if (image_barrier_count) {
image_barriers = (VkImageMemoryBarrier *)alloca(sizeof(VkImageMemoryBarrier) * draw_list_storage_textures.size());
}
uint32_t src_stage = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
uint32_t src_access = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
if (image_barrier_count) {
src_stage |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
src_access |= VK_ACCESS_SHADER_WRITE_BIT;
}
for (uint32_t i = 0; i < image_barrier_count; i++) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(draw_list_storage_textures[i]);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
VkImageMemoryBarrier &image_memory_barrier = image_barriers[i];
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
image_memory_barrier.pNext = nullptr;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
image_memory_barrier.srcAccessMask = src_access;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
image_memory_barrier.dstAccessMask = access_flags;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
image_memory_barrier.oldLayout = texture->layout;
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = texture->image;
image_memory_barrier.subresourceRange.aspectMask = texture->read_aspect_mask;
image_memory_barrier.subresourceRange.baseMipLevel = texture->base_mipmap;
image_memory_barrier.subresourceRange.levelCount = texture->mipmaps;
image_memory_barrier.subresourceRange.baseArrayLayer = texture->base_layer;
image_memory_barrier.subresourceRange.layerCount = texture->layers;
texture->layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
}
draw_list_storage_textures.clear();
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
// To ensure proper synchronization, we must make sure rendering is done before:
// * Some buffer is copied
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
// * Another render pass happens (since we may be done)
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
#ifdef FORCE_FULL_BARRIER
_full_barrier(true);
#else
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
VkMemoryBarrier mem_barrier;
mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER;
mem_barrier.pNext = nullptr;
mem_barrier.srcAccessMask = src_access;
mem_barrier.dstAccessMask = access_flags;
if (image_barrier_count > 0 || p_post_barrier != BARRIER_MASK_NO_BARRIER) {
vkCmdPipelineBarrier(frames[frame].draw_command_buffer, src_stage, barrier_flags, 0, 1, &mem_barrier, 0, nullptr, image_barrier_count, image_barriers);
}
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
#endif
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
/***********************/
/**** COMPUTE LISTS ****/
/***********************/
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RenderingDevice::ComputeListID RenderingDeviceVulkan::compute_list_begin(bool p_allow_draw_overlap) {
ERR_FAIL_COND_V_MSG(!p_allow_draw_overlap && draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time.");
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
ERR_FAIL_COND_V_MSG(compute_list != nullptr, INVALID_ID, "Only one draw/compute list can be active at the same time.");
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Make draw/compute lists threadsafe Lock the rendering device while command buffers are in use
2021-11-11 18:55:09 +00:00
// Lock while compute_list is active
_THREAD_SAFE_LOCK_
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
compute_list = memnew(ComputeList);
compute_list->command_buffer = frames[frame].draw_command_buffer;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
compute_list->state.allow_draw_overlap = p_allow_draw_overlap;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
return ID_TYPE_COMPUTE_LIST;
}
void RenderingDeviceVulkan::compute_list_bind_compute_pipeline(ComputeListID p_list, RID p_compute_pipeline) {
ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
ERR_FAIL_COND(!compute_list);
ComputeList *cl = compute_list;
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
const ComputePipeline *pipeline = compute_pipeline_owner.get_or_null(p_compute_pipeline);
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_COND(!pipeline);
if (p_compute_pipeline == cl->state.pipeline) {
return; //redundant state, return.
}
cl->state.pipeline = p_compute_pipeline;
cl->state.pipeline_layout = pipeline->pipeline_layout;
vkCmdBindPipeline(cl->command_buffer, VK_PIPELINE_BIND_POINT_COMPUTE, pipeline->pipeline);
if (cl->state.pipeline_shader != pipeline->shader) {
// shader changed, so descriptor sets may become incompatible.
//go through ALL sets, and unbind them (and all those above) if the format is different
uint32_t pcount = pipeline->set_formats.size(); //formats count in this pipeline
cl->state.set_count = MAX(cl->state.set_count, pcount);
const uint32_t *pformats = pipeline->set_formats.ptr(); //pipeline set formats
bool sets_valid = true; //once invalid, all above become invalid
for (uint32_t i = 0; i < pcount; i++) {
//if a part of the format is different, invalidate it (and the rest)
if (!sets_valid || cl->state.sets[i].pipeline_expected_format != pformats[i]) {
cl->state.sets[i].bound = false;
cl->state.sets[i].pipeline_expected_format = pformats[i];
sets_valid = false;
}
}
for (uint32_t i = pcount; i < cl->state.set_count; i++) {
//unbind the ones above (not used) if exist
cl->state.sets[i].bound = false;
}
cl->state.set_count = pcount; //update set count
if (pipeline->push_constant_size) {
cl->state.pipeline_push_constant_stages = pipeline->push_constant_stages;
#ifdef DEBUG_ENABLED
Vulkan: Initialize struct members on declaration Supersedes and closes #38945.
2020-06-03 08:45:44 +00:00
cl->validation.pipeline_push_constant_supplied = false;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
#endif
}
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
cl->state.pipeline_shader = pipeline->shader;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
cl->state.local_group_size[0] = pipeline->local_group_size[0];
cl->state.local_group_size[1] = pipeline->local_group_size[1];
cl->state.local_group_size[2] = pipeline->local_group_size[2];
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
}
#ifdef DEBUG_ENABLED
//update compute pass pipeline info
cl->validation.pipeline_active = true;
cl->validation.pipeline_push_constant_size = pipeline->push_constant_size;
#endif
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
void RenderingDeviceVulkan::compute_list_bind_uniform_set(ComputeListID p_list, RID p_uniform_set, uint32_t p_index) {
ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
ERR_FAIL_COND(!compute_list);
ComputeList *cl = compute_list;
#ifdef DEBUG_ENABLED
Static analyzer fixes: Removes unused code in OS. Fixes return types. Fixes few typos.
2020-12-08 18:58:49 +00:00
ERR_FAIL_COND_MSG(p_index >= limits.maxBoundDescriptorSets || p_index >= MAX_UNIFORM_SETS,
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
"Attempting to bind a descriptor set (" + itos(p_index) + ") greater than what the hardware supports (" + itos(limits.maxBoundDescriptorSets) + ").");
#endif
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(!cl->validation.active, "Submitted Compute Lists can no longer be modified.");
#endif
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
UniformSet *uniform_set = uniform_set_owner.get_or_null(p_uniform_set);
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_COND(!uniform_set);
if (p_index > cl->state.set_count) {
cl->state.set_count = p_index;
}
cl->state.sets[p_index].descriptor_set = uniform_set->descriptor_set; //update set pointer
cl->state.sets[p_index].bound = false; //needs rebind
cl->state.sets[p_index].uniform_set_format = uniform_set->format;
cl->state.sets[p_index].uniform_set = p_uniform_set;
uint32_t textures_to_sampled_count = uniform_set->mutable_sampled_textures.size();
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
uint32_t textures_to_storage_count = uniform_set->mutable_storage_textures.size();
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Texture **textures_to_sampled = uniform_set->mutable_sampled_textures.ptrw();
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
VkImageMemoryBarrier *texture_barriers = nullptr;
if (textures_to_sampled_count + textures_to_storage_count) {
texture_barriers = (VkImageMemoryBarrier *)alloca(sizeof(VkImageMemoryBarrier) * (textures_to_sampled_count + textures_to_storage_count));
}
uint32_t texture_barrier_count = 0;
uint32_t src_stage_flags = 0;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
for (uint32_t i = 0; i < textures_to_sampled_count; i++) {
if (textures_to_sampled[i]->layout != VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
src_stage_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
VkImageMemoryBarrier &image_memory_barrier = texture_barriers[texture_barrier_count++];
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
image_memory_barrier.oldLayout = textures_to_sampled[i]->layout;
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = textures_to_sampled[i]->image;
image_memory_barrier.subresourceRange.aspectMask = textures_to_sampled[i]->read_aspect_mask;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = textures_to_sampled[i]->base_mipmap;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.subresourceRange.levelCount = textures_to_sampled[i]->mipmaps;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseArrayLayer = textures_to_sampled[i]->base_layer;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.subresourceRange.layerCount = textures_to_sampled[i]->layers;
textures_to_sampled[i]->layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
cl->state.textures_to_sampled_layout.erase(textures_to_sampled[i]);
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (textures_to_sampled[i]->used_in_frame != frames_drawn) {
textures_to_sampled[i]->used_in_frame = frames_drawn;
textures_to_sampled[i]->used_in_transfer = false;
textures_to_sampled[i]->used_in_raster = false;
}
textures_to_sampled[i]->used_in_compute = true;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
}
Texture **textures_to_storage = uniform_set->mutable_storage_textures.ptrw();
for (uint32_t i = 0; i < textures_to_storage_count; i++) {
if (textures_to_storage[i]->layout != VK_IMAGE_LAYOUT_GENERAL) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
uint32_t src_access_flags = 0;
if (textures_to_storage[i]->used_in_frame == frames_drawn) {
if (textures_to_storage[i]->used_in_compute) {
src_stage_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
src_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (textures_to_storage[i]->used_in_raster) {
src_stage_flags |= VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT;
src_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
}
if (textures_to_storage[i]->used_in_transfer) {
src_stage_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
src_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT;
}
textures_to_storage[i]->used_in_compute = false;
textures_to_storage[i]->used_in_raster = false;
Remove redundant assignments. Use used_in_transfer instead of used_in_compute twice.
2021-08-21 09:37:21 +00:00
textures_to_storage[i]->used_in_transfer = false;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
} else {
src_access_flags = 0;
textures_to_storage[i]->used_in_compute = false;
textures_to_storage[i]->used_in_raster = false;
Remove redundant assignments. Use used_in_transfer instead of used_in_compute twice.
2021-08-21 09:37:21 +00:00
textures_to_storage[i]->used_in_transfer = false;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
textures_to_storage[i]->used_in_frame = frames_drawn;
}
VkImageMemoryBarrier &image_memory_barrier = texture_barriers[texture_barrier_count++];
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
image_memory_barrier.srcAccessMask = src_access_flags;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
image_memory_barrier.oldLayout = textures_to_storage[i]->layout;
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_GENERAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = textures_to_storage[i]->image;
GIProbes working.
2019-10-03 20:39:08 +00:00
image_memory_barrier.subresourceRange.aspectMask = textures_to_storage[i]->read_aspect_mask;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = textures_to_storage[i]->base_mipmap;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.subresourceRange.levelCount = textures_to_storage[i]->mipmaps;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseArrayLayer = textures_to_storage[i]->base_layer;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.subresourceRange.layerCount = textures_to_storage[i]->layers;
textures_to_storage[i]->layout = VK_IMAGE_LAYOUT_GENERAL;
cl->state.textures_to_sampled_layout.insert(textures_to_storage[i]); //needs to go back to sampled layout afterwards
}
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (texture_barrier_count) {
if (src_stage_flags == 0) {
src_stage_flags = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
}
vkCmdPipelineBarrier(cl->command_buffer, src_stage_flags, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, texture_barrier_count, texture_barriers);
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
#if 0
{ //validate that textures bound are not attached as framebuffer bindings
uint32_t attachable_count = uniform_set->attachable_textures.size();
const RID *attachable_ptr = uniform_set->attachable_textures.ptr();
uint32_t bound_count = draw_list_bound_textures.size();
const RID *bound_ptr = draw_list_bound_textures.ptr();
for (uint32_t i = 0; i < attachable_count; i++) {
for (uint32_t j = 0; j < bound_count; j++) {
ERR_FAIL_COND_MSG(attachable_ptr[i] == bound_ptr[j],
"Attempted to use the same texture in framebuffer attachment and a uniform set, this is not allowed.");
}
}
}
#endif
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Exposed RenderingDevice to script API Also added an easier way to load native GLSL shaders. Extras: Had to fix no-cache for subresources in resource loader, it was not properly working, making shaders not properly reload. Note: The precommit hooks are broken because they don't seem to support enums from one class being used in another. Feel free to fix this after merging this PR.
2020-04-20 02:19:21 +00:00
void RenderingDeviceVulkan::compute_list_set_push_constant(ComputeListID p_list, const void *p_data, uint32_t p_data_size) {
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
ERR_FAIL_COND(!compute_list);
ComputeList *cl = compute_list;
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(!cl->validation.active, "Submitted Compute Lists can no longer be modified.");
#endif
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(p_data_size != cl->validation.pipeline_push_constant_size,
"This compute pipeline requires (" + itos(cl->validation.pipeline_push_constant_size) + ") bytes of push constant data, supplied: (" + itos(p_data_size) + ")");
#endif
vkCmdPushConstants(cl->command_buffer, cl->state.pipeline_layout, cl->state.pipeline_push_constant_stages, 0, p_data_size, p_data);
#ifdef DEBUG_ENABLED
Vulkan: Initialize struct members on declaration Supersedes and closes #38945.
2020-06-03 08:45:44 +00:00
cl->validation.pipeline_push_constant_supplied = true;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
#endif
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
void RenderingDeviceVulkan::compute_list_dispatch(ComputeListID p_list, uint32_t p_x_groups, uint32_t p_y_groups, uint32_t p_z_groups) {
ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
ERR_FAIL_COND(!compute_list);
ComputeList *cl = compute_list;
#ifdef DEBUG_ENABLED
Disallow compute dispatch with zero dimensions.
2021-10-29 21:10:48 +00:00
ERR_FAIL_COND_MSG(p_x_groups == 0, "Dispatch amount of X compute groups (" + itos(p_x_groups) + ") is zero.");
ERR_FAIL_COND_MSG(p_z_groups == 0, "Dispatch amount of Z compute groups (" + itos(p_z_groups) + ") is zero.");
ERR_FAIL_COND_MSG(p_y_groups == 0, "Dispatch amount of Y compute groups (" + itos(p_y_groups) + ") is zero.");
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
ERR_FAIL_COND_MSG(p_x_groups > limits.maxComputeWorkGroupCount[0],
"Dispatch amount of X compute groups (" + itos(p_x_groups) + ") is larger than device limit (" + itos(limits.maxComputeWorkGroupCount[0]) + ")");
ERR_FAIL_COND_MSG(p_y_groups > limits.maxComputeWorkGroupCount[1],
Addition of FogVolumes, FogShaders, FogMaterial, and overhaul of VolumetricFog Co-authored-by: Brian Semrau <brian.semrau@gmail.com>
2021-10-03 11:28:55 +00:00
"Dispatch amount of Y compute groups (" + itos(p_y_groups) + ") is larger than device limit (" + itos(limits.maxComputeWorkGroupCount[1]) + ")");
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
ERR_FAIL_COND_MSG(p_z_groups > limits.maxComputeWorkGroupCount[2],
Addition of FogVolumes, FogShaders, FogMaterial, and overhaul of VolumetricFog Co-authored-by: Brian Semrau <brian.semrau@gmail.com>
2021-10-03 11:28:55 +00:00
"Dispatch amount of Z compute groups (" + itos(p_z_groups) + ") is larger than device limit (" + itos(limits.maxComputeWorkGroupCount[2]) + ")");
GIProbes working.
2019-10-03 20:39:08 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_COND_MSG(!cl->validation.active, "Submitted Compute Lists can no longer be modified.");
#endif
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw.");
if (cl->validation.pipeline_push_constant_size > 0) {
//using push constants, check that they were supplied
Vulkan: Initialize struct members on declaration Supersedes and closes #38945.
2020-06-03 08:45:44 +00:00
ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied,
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
"The shader in this pipeline requires a push constant to be set before drawing, but it's not present.");
}
#endif
//Bind descriptor sets
for (uint32_t i = 0; i < cl->state.set_count; i++) {
if (cl->state.sets[i].pipeline_expected_format == 0) {
continue; //nothing expected by this pipeline
}
#ifdef DEBUG_ENABLED
if (cl->state.sets[i].pipeline_expected_format != cl->state.sets[i].uniform_set_format) {
if (cl->state.sets[i].uniform_set_format == 0) {
ERR_FAIL_MSG("Uniforms were never supplied for set (" + itos(i) + ") at the time of drawing, which are required by the pipeline");
} else if (uniform_set_owner.owns(cl->state.sets[i].uniform_set)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
UniformSet *us = uniform_set_owner.get_or_null(cl->state.sets[i].uniform_set);
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + "):\n" + _shader_uniform_debug(us->shader_id, us->shader_set) + "\nare not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(cl->state.pipeline_shader));
} else {
ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + ", which was was just freed) are not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(cl->state.pipeline_shader));
}
}
#endif
if (!cl->state.sets[i].bound) {
//All good, see if this requires re-binding
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkCmdBindDescriptorSets(cl->command_buffer, VK_PIPELINE_BIND_POINT_COMPUTE, cl->state.pipeline_layout, i, 1, &cl->state.sets[i].descriptor_set, 0, nullptr);
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
cl->state.sets[i].bound = true;
}
}
vkCmdDispatch(cl->command_buffer, p_x_groups, p_y_groups, p_z_groups);
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
void RenderingDeviceVulkan::compute_list_dispatch_threads(ComputeListID p_list, uint32_t p_x_threads, uint32_t p_y_threads, uint32_t p_z_threads) {
ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
ERR_FAIL_COND(!compute_list);
Disallow compute dispatch with zero dimensions.
2021-10-29 21:10:48 +00:00
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(p_x_threads == 0, "Dispatch amount of X compute threads (" + itos(p_x_threads) + ") is zero.");
ERR_FAIL_COND_MSG(p_y_threads == 0, "Dispatch amount of Y compute threads (" + itos(p_y_threads) + ") is zero.");
ERR_FAIL_COND_MSG(p_z_threads == 0, "Dispatch amount of Z compute threads (" + itos(p_z_threads) + ") is zero.");
#endif
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
ComputeList *cl = compute_list;
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw.");
if (cl->validation.pipeline_push_constant_size > 0) {
//using push constants, check that they were supplied
ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied,
"The shader in this pipeline requires a push constant to be set before drawing, but it's not present.");
}
#endif
compute_list_dispatch(p_list, (p_x_threads - 1) / cl->state.local_group_size[0] + 1, (p_y_threads - 1) / cl->state.local_group_size[1] + 1, (p_z_threads - 1) / cl->state.local_group_size[2] + 1);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
void RenderingDeviceVulkan::compute_list_dispatch_indirect(ComputeListID p_list, RID p_buffer, uint32_t p_offset) {
ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
ERR_FAIL_COND(!compute_list);
ComputeList *cl = compute_list;
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Buffer *buffer = storage_buffer_owner.get_or_null(p_buffer);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
ERR_FAIL_COND(!buffer);
ERR_FAIL_COND_MSG(!(buffer->usage & STORAGE_BUFFER_USAGE_DISPATCH_INDIRECT), "Buffer provided was not created to do indirect dispatch.");
ERR_FAIL_COND_MSG(p_offset + 12 > buffer->size, "Offset provided (+12) is past the end of buffer.");
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(!cl->validation.active, "Submitted Compute Lists can no longer be modified.");
#endif
#ifdef DEBUG_ENABLED
ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw.");
if (cl->validation.pipeline_push_constant_size > 0) {
//using push constants, check that they were supplied
ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied,
"The shader in this pipeline requires a push constant to be set before drawing, but it's not present.");
}
#endif
//Bind descriptor sets
for (uint32_t i = 0; i < cl->state.set_count; i++) {
if (cl->state.sets[i].pipeline_expected_format == 0) {
continue; //nothing expected by this pipeline
}
#ifdef DEBUG_ENABLED
if (cl->state.sets[i].pipeline_expected_format != cl->state.sets[i].uniform_set_format) {
if (cl->state.sets[i].uniform_set_format == 0) {
ERR_FAIL_MSG("Uniforms were never supplied for set (" + itos(i) + ") at the time of drawing, which are required by the pipeline");
} else if (uniform_set_owner.owns(cl->state.sets[i].uniform_set)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
UniformSet *us = uniform_set_owner.get_or_null(cl->state.sets[i].uniform_set);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + "):\n" + _shader_uniform_debug(us->shader_id, us->shader_set) + "\nare not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(cl->state.pipeline_shader));
} else {
ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + ", which was was just freed) are not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(cl->state.pipeline_shader));
}
}
#endif
if (!cl->state.sets[i].bound) {
//All good, see if this requires re-binding
vkCmdBindDescriptorSets(cl->command_buffer, VK_PIPELINE_BIND_POINT_COMPUTE, cl->state.pipeline_layout, i, 1, &cl->state.sets[i].descriptor_set, 0, nullptr);
cl->state.sets[i].bound = true;
}
}
vkCmdDispatchIndirect(cl->command_buffer, buffer->buffer, p_offset);
}
GIProbes working.
2019-10-03 20:39:08 +00:00
void RenderingDeviceVulkan::compute_list_add_barrier(ComputeListID p_list) {
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
#ifdef FORCE_FULL_BARRIER
_full_barrier(true);
#else
GIProbes working.
2019-10-03 20:39:08 +00:00
_memory_barrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_ACCESS_SHADER_WRITE_BIT, VK_ACCESS_SHADER_READ_BIT, true);
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
#endif
GIProbes working.
2019-10-03 20:39:08 +00:00
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
void RenderingDeviceVulkan::compute_list_end(uint32_t p_post_barrier) {
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
ERR_FAIL_COND(!compute_list);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
uint32_t barrier_flags = 0;
uint32_t access_flags = 0;
if (p_post_barrier & BARRIER_MASK_COMPUTE) {
barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
}
if (p_post_barrier & BARRIER_MASK_RASTER) {
barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT;
access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT;
}
if (p_post_barrier & BARRIER_MASK_TRANSFER) {
barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT;
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (barrier_flags == 0) {
barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
}
VkImageMemoryBarrier *image_barriers = nullptr;
uint32_t image_barrier_count = compute_list->state.textures_to_sampled_layout.size();
if (image_barrier_count) {
image_barriers = (VkImageMemoryBarrier *)alloca(sizeof(VkImageMemoryBarrier) * image_barrier_count);
}
uint32_t barrier_idx = 0;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
for (Set<Texture *>::Element *E = compute_list->state.textures_to_sampled_layout.front(); E; E = E->next()) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
VkImageMemoryBarrier &image_memory_barrier = image_barriers[barrier_idx++];
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
image_memory_barrier.pNext = nullptr;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
image_memory_barrier.dstAccessMask = access_flags;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.oldLayout = E->get()->layout;
image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
image_memory_barrier.image = E->get()->image;
image_memory_barrier.subresourceRange.aspectMask = E->get()->read_aspect_mask;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseMipLevel = E->get()->base_mipmap;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.subresourceRange.levelCount = E->get()->mipmaps;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
image_memory_barrier.subresourceRange.baseArrayLayer = E->get()->base_layer;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
image_memory_barrier.subresourceRange.layerCount = E->get()->layers;
E->get()->layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (E->get()->used_in_frame != frames_drawn) {
E->get()->used_in_transfer = false;
E->get()->used_in_raster = false;
E->get()->used_in_compute = false;
E->get()->used_in_frame = frames_drawn;
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
}
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
#ifdef FORCE_FULL_BARRIER
_full_barrier(true);
#else
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
VkMemoryBarrier mem_barrier;
mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER;
mem_barrier.pNext = nullptr;
mem_barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
mem_barrier.dstAccessMask = access_flags;
if (image_barrier_count > 0 || p_post_barrier != BARRIER_MASK_NO_BARRIER) {
vkCmdPipelineBarrier(compute_list->command_buffer, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, barrier_flags, 0, 1, &mem_barrier, 0, nullptr, image_barrier_count, image_barriers);
}
Several fixes to GIProbes
2019-10-31 22:54:21 +00:00
#endif
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
memdelete(compute_list);
compute_list = nullptr;
Make draw/compute lists threadsafe Lock the rendering device while command buffers are in use
2021-11-11 18:55:09 +00:00
// compute_list is no longer active
_THREAD_SAFE_UNLOCK_
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
void RenderingDeviceVulkan::barrier(uint32_t p_from, uint32_t p_to) {
uint32_t src_barrier_flags = 0;
uint32_t src_access_flags = 0;
if (p_from & BARRIER_MASK_COMPUTE) {
src_barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
src_access_flags |= VK_ACCESS_SHADER_WRITE_BIT;
}
if (p_from & BARRIER_MASK_RASTER) {
src_barrier_flags |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
src_access_flags |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
}
if (p_from & BARRIER_MASK_TRANSFER) {
src_barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
src_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
}
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
if (p_from == 0) {
src_barrier_flags = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
uint32_t dst_barrier_flags = 0;
uint32_t dst_access_flags = 0;
if (p_to & BARRIER_MASK_COMPUTE) {
dst_barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
}
if (p_to & BARRIER_MASK_RASTER) {
dst_barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT;
dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT;
}
if (p_to & BARRIER_MASK_TRANSFER) {
dst_barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
dst_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT;
}
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
if (p_to == 0) {
dst_barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
_memory_barrier(src_barrier_flags, dst_barrier_flags, src_access_flags, dst_access_flags, true);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
void RenderingDeviceVulkan::full_barrier() {
#ifndef DEBUG_ENABLED
ERR_PRINT("Full barrier is debug-only, should not be used in production");
#endif
_full_barrier(true);
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
#if 0
void RenderingDeviceVulkan::draw_list_render_secondary_to_framebuffer(ID p_framebuffer, ID *p_draw_lists, uint32_t p_draw_list_count, InitialAction p_initial_action, FinalAction p_final_action, const Vector<Variant> &p_clear_colors) {
VkCommandBuffer frame_cmdbuf = frames[frame].frame_buffer;
ERR_FAIL_COND(!frame_cmdbuf);
VkRenderPassBeginInfo render_pass_begin;
render_pass_begin.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
render_pass_begin.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
render_pass_begin.renderPass = context->get_render_pass();
render_pass_begin.framebuffer = context->get_frame_framebuffer(frame);
render_pass_begin.renderArea.extent.width = context->get_screen_width(p_screen);
render_pass_begin.renderArea.extent.height = context->get_screen_height(p_screen);
render_pass_begin.renderArea.offset.x = 0;
render_pass_begin.renderArea.offset.y = 0;
render_pass_begin.clearValueCount = 1;
VkClearValue clear_value;
clear_value.color.float32[0] = p_clear_color.r;
clear_value.color.float32[1] = p_clear_color.g;
clear_value.color.float32[2] = p_clear_color.b;
clear_value.color.float32[3] = p_clear_color.a;
render_pass_begin.pClearValues = &clear_value;
vkCmdBeginRenderPass(frame_cmdbuf, &render_pass_begin, VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS);
ID screen_format = screen_get_framebuffer_format();
{
VkCommandBuffer *command_buffers = (VkCommandBuffer *)alloca(sizeof(VkCommandBuffer) * p_draw_list_count);
uint32_t command_buffer_count = 0;
for (uint32_t i = 0; i < p_draw_list_count; i++) {
DrawList *dl = _get_draw_list_ptr(p_draw_lists[i]);
ERR_CONTINUE_MSG(!dl, "Draw list index (" + itos(i) + ") is not a valid draw list ID.");
ERR_CONTINUE_MSG(dl->validation.framebuffer_format != p_format_check,
"Draw list index (" + itos(i) + ") is created with a framebuffer format incompatible with this render pass.");
if (dl->validation.active) {
//needs to be closed, so close it.
vkEndCommandBuffer(dl->command_buffer);
dl->validation.active = false;
}
command_buffers[command_buffer_count++] = dl->command_buffer;
}
print_line("to draw: " + itos(command_buffer_count));
vkCmdExecuteCommands(p_primary, command_buffer_count, command_buffers);
}
vkCmdEndRenderPass(frame_cmdbuf);
}
#endif
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::_free_internal(RID p_id) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//push everything so it's disposed of next time this frame index is processed (means, it's safe to do it)
if (texture_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].textures_to_dispose_of.push_back(*texture);
texture_owner.free(p_id);
} else if (framebuffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].framebuffers_to_dispose_of.push_back(*framebuffer);
framebuffer_owner.free(p_id);
} else if (sampler_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
VkSampler *sampler = sampler_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].samplers_to_dispose_of.push_back(*sampler);
sampler_owner.free(p_id);
} else if (vertex_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Buffer *vertex_buffer = vertex_buffer_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].buffers_to_dispose_of.push_back(*vertex_buffer);
vertex_buffer_owner.free(p_id);
} else if (vertex_array_owner.owns(p_id)) {
vertex_array_owner.free(p_id);
} else if (index_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
IndexBuffer *index_buffer = index_buffer_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Buffer b;
b.allocation = index_buffer->allocation;
b.buffer = index_buffer->buffer;
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
b.size = index_buffer->size;
Fix buffer_info may be used uninitialized
2021-09-07 15:59:27 +00:00
b.buffer_info = {};
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].buffers_to_dispose_of.push_back(b);
index_buffer_owner.free(p_id);
} else if (index_array_owner.owns(p_id)) {
index_array_owner.free(p_id);
} else if (shader_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Shader *shader = shader_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].shaders_to_dispose_of.push_back(*shader);
shader_owner.free(p_id);
} else if (uniform_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Buffer *uniform_buffer = uniform_buffer_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].buffers_to_dispose_of.push_back(*uniform_buffer);
uniform_buffer_owner.free(p_id);
} else if (texture_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
TextureBuffer *texture_buffer = texture_buffer_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].buffers_to_dispose_of.push_back(texture_buffer->buffer);
frames[frame].buffer_views_to_dispose_of.push_back(texture_buffer->view);
texture_buffer_owner.free(p_id);
} else if (storage_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Buffer *storage_buffer = storage_buffer_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].buffers_to_dispose_of.push_back(*storage_buffer);
storage_buffer_owner.free(p_id);
} else if (uniform_set_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
UniformSet *uniform_set = uniform_set_owner.get_or_null(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
frames[frame].uniform_sets_to_dispose_of.push_back(*uniform_set);
Fix materials not updating when texture replaced/deleted
2021-11-01 19:33:59 +00:00
uniform_set_owner.free(p_id);
Fix material invalidation on reimport. * IF a texture was reimported (calling replace as an example), it would invalidate all materials using it, causing plenty of errors. * Added the possibility to get a notification when a uniform set is erased. * With this notification, materials can be queued for update properly.
2021-07-07 22:55:20 +00:00
if (uniform_set->invalidated_callback != nullptr) {
Fix materials not updating when texture replaced/deleted
2021-11-01 19:33:59 +00:00
uniform_set->invalidated_callback(uniform_set->invalidated_callback_userdata);
Fix material invalidation on reimport. * IF a texture was reimported (calling replace as an example), it would invalidate all materials using it, causing plenty of errors. * Added the possibility to get a notification when a uniform set is erased. * With this notification, materials can be queued for update properly.
2021-07-07 22:55:20 +00:00
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
} else if (render_pipeline_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
RenderPipeline *pipeline = render_pipeline_owner.get_or_null(p_id);
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
frames[frame].render_pipelines_to_dispose_of.push_back(*pipeline);
render_pipeline_owner.free(p_id);
} else if (compute_pipeline_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
ComputePipeline *pipeline = compute_pipeline_owner.get_or_null(p_id);
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
frames[frame].compute_pipelines_to_dispose_of.push_back(*pipeline);
compute_pipeline_owner.free(p_id);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
} else {
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
ERR_PRINT("Attempted to free invalid ID: " + itos(p_id.get_id()));
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
void RenderingDeviceVulkan::free(RID p_id) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
_THREAD_SAFE_METHOD_
_free_dependencies(p_id); //recursively erase dependencies first, to avoid potential API problems
_free_internal(p_id);
}
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
// The full list of resources that can be named is in the VkObjectType enum
// We just expose the resources that are owned and can be accessed easily.
void RenderingDeviceVulkan::set_resource_name(RID p_id, const String p_name) {
if (texture_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *texture = texture_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
if (texture->owner.is_null()) {
// Don't set the source texture's name when calling on a texture view
context->set_object_name(VK_OBJECT_TYPE_IMAGE, uint64_t(texture->image), p_name);
}
context->set_object_name(VK_OBJECT_TYPE_IMAGE_VIEW, uint64_t(texture->view), p_name + " View");
} else if (framebuffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
//Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
// Not implemented for now as the relationship between Framebuffer and RenderPass is very complex
} else if (sampler_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
VkSampler *sampler = sampler_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_SAMPLER, uint64_t(*sampler), p_name);
} else if (vertex_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Buffer *vertex_buffer = vertex_buffer_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(vertex_buffer->buffer), p_name);
} else if (index_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
IndexBuffer *index_buffer = index_buffer_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(index_buffer->buffer), p_name);
} else if (shader_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Shader *shader = shader_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_PIPELINE_LAYOUT, uint64_t(shader->pipeline_layout), p_name + " Pipeline Layout");
for (int i = 0; i < shader->sets.size(); i++) {
context->set_object_name(VK_OBJECT_TYPE_DESCRIPTOR_SET_LAYOUT, uint64_t(shader->sets[i].descriptor_set_layout), p_name);
}
} else if (uniform_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Buffer *uniform_buffer = uniform_buffer_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(uniform_buffer->buffer), p_name);
} else if (texture_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
TextureBuffer *texture_buffer = texture_buffer_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(texture_buffer->buffer.buffer), p_name);
context->set_object_name(VK_OBJECT_TYPE_BUFFER_VIEW, uint64_t(texture_buffer->view), p_name + " View");
} else if (storage_buffer_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Buffer *storage_buffer = storage_buffer_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(storage_buffer->buffer), p_name);
} else if (uniform_set_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
UniformSet *uniform_set = uniform_set_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_DESCRIPTOR_SET, uint64_t(uniform_set->descriptor_set), p_name);
} else if (render_pipeline_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
RenderPipeline *pipeline = render_pipeline_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_PIPELINE, uint64_t(pipeline->pipeline), p_name);
context->set_object_name(VK_OBJECT_TYPE_PIPELINE_LAYOUT, uint64_t(pipeline->pipeline_layout), p_name + " Layout");
} else if (compute_pipeline_owner.owns(p_id)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
ComputePipeline *pipeline = compute_pipeline_owner.get_or_null(p_id);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
context->set_object_name(VK_OBJECT_TYPE_PIPELINE, uint64_t(pipeline->pipeline), p_name);
context->set_object_name(VK_OBJECT_TYPE_PIPELINE_LAYOUT, uint64_t(pipeline->pipeline_layout), p_name + " Layout");
} else {
ERR_PRINT("Attempted to name invalid ID: " + itos(p_id.get_id()));
}
}
void RenderingDeviceVulkan::draw_command_begin_label(String p_label_name, const Color p_color) {
context->command_begin_label(frames[frame].draw_command_buffer, p_label_name, p_color);
}
void RenderingDeviceVulkan::draw_command_insert_label(String p_label_name, const Color p_color) {
context->command_insert_label(frames[frame].draw_command_buffer, p_label_name, p_color);
}
void RenderingDeviceVulkan::draw_command_end_label() {
context->command_end_label(frames[frame].draw_command_buffer);
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
String RenderingDeviceVulkan::get_device_vendor_name() const {
return context->get_device_vendor_name();
}
Add `get_video_adapter_api_version()` to RenderingServer This method can be used to get the graphics API version currently in use (such as Vulkan). It can be used by projects for troubleshooting or statistical purposes.
2021-07-31 13:39:46 +00:00
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
String RenderingDeviceVulkan::get_device_name() const {
return context->get_device_name();
}
Add `RenderingServer.get_video_adapter_type()` method This can be used to distinguish between integrated, dedicated, virtual and software-emulated GPUs. This in turn can be used to automatically adjust graphics settings, or warn users about features that may run slowly on their hardware.
2021-12-10 16:01:51 +00:00
RenderingDevice::DeviceType RenderingDeviceVulkan::get_device_type() const {
return context->get_device_type();
}
Add `get_video_adapter_api_version()` to RenderingServer This method can be used to get the graphics API version currently in use (such as Vulkan). It can be used by projects for troubleshooting or statistical purposes.
2021-07-31 13:39:46 +00:00
String RenderingDeviceVulkan::get_device_api_version() const {
return context->get_device_api_version();
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
String RenderingDeviceVulkan::get_device_pipeline_cache_uuid() const {
return context->get_device_pipeline_cache_uuid();
}
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
void RenderingDeviceVulkan::_finalize_command_bufers() {
if (draw_list) {
ERR_PRINT("Found open draw list at the end of the frame, this should never happen (further drawing will likely not work).");
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
if (compute_list) {
ERR_PRINT("Found open compute list at the end of the frame, this should never happen (further compute will likely not work).");
}
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
{ //complete the setup buffer (that needs to be processed before anything else)
vkEndCommandBuffer(frames[frame].setup_command_buffer);
vkEndCommandBuffer(frames[frame].draw_command_buffer);
}
}
void RenderingDeviceVulkan::_begin_frame() {
//erase pending resources
_free_pending_resources(frame);
//create setup command buffer and set as the setup buffer
{
VkCommandBufferBeginInfo cmdbuf_begin;
cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
cmdbuf_begin.pNext = nullptr;
cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
cmdbuf_begin.pInheritanceInfo = nullptr;
VkResult err = vkResetCommandBuffer(frames[frame].setup_command_buffer, 0);
ERR_FAIL_COND_MSG(err, "vkResetCommandBuffer failed with error " + itos(err) + ".");
err = vkBeginCommandBuffer(frames[frame].setup_command_buffer, &cmdbuf_begin);
ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
err = vkBeginCommandBuffer(frames[frame].draw_command_buffer, &cmdbuf_begin);
ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
if (local_device.is_null()) {
context->append_command_buffer(frames[frame].draw_command_buffer);
context->set_setup_buffer(frames[frame].setup_command_buffer); //append now so it's added before everything else
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
}
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
//advance current frame
frames_drawn++;
//advance staging buffer if used
if (staging_buffer_used) {
staging_buffer_current = (staging_buffer_current + 1) % staging_buffer_blocks.size();
staging_buffer_used = false;
}
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
if (frames[frame].timestamp_count) {
vkGetQueryPoolResults(device, frames[frame].timestamp_pool, 0, frames[frame].timestamp_count, sizeof(uint64_t) * max_timestamp_query_elements, frames[frame].timestamp_result_values, sizeof(uint64_t), VK_QUERY_RESULT_64_BIT);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
vkCmdResetQueryPool(frames[frame].setup_command_buffer, frames[frame].timestamp_pool, 0, frames[frame].timestamp_count);
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
SWAP(frames[frame].timestamp_names, frames[frame].timestamp_result_names);
SWAP(frames[frame].timestamp_cpu_values, frames[frame].timestamp_cpu_result_values);
}
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
frames[frame].timestamp_result_count = frames[frame].timestamp_count;
frames[frame].timestamp_count = 0;
frames[frame].index = Engine::get_singleton()->get_frames_drawn();
}
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
void RenderingDeviceVulkan::swap_buffers() {
ERR_FAIL_COND_MSG(local_device.is_valid(), "Local devices can't swap buffers.");
_THREAD_SAFE_METHOD_
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
_finalize_command_bufers();
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
screen_prepared = false;
//swap buffers
context->swap_buffers();
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
frame = (frame + 1) % frame_count;
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
_begin_frame();
}
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
void RenderingDeviceVulkan::submit() {
ERR_FAIL_COND_MSG(local_device.is_null(), "Only local devices can submit and sync.");
ERR_FAIL_COND_MSG(local_device_processing, "device already submitted, call sync to wait until done.");
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
_finalize_command_bufers();
VkCommandBuffer command_buffers[2] = { frames[frame].setup_command_buffer, frames[frame].draw_command_buffer };
context->local_device_push_command_buffers(local_device, command_buffers, 2);
local_device_processing = true;
}
void RenderingDeviceVulkan::sync() {
ERR_FAIL_COND_MSG(local_device.is_null(), "Only local devices can submit and sync.");
ERR_FAIL_COND_MSG(!local_device_processing, "sync can only be called after a submit");
context->local_device_sync(local_device);
_begin_frame();
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
local_device_processing = false;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
vk_mem_alloc: Update to upstream + Adapt approach to small objects pooling This updates VMA and instead of using the custom small pool approach from 4e6c9d3ae979f2eb0151cf581fe61d2f3194ea72, lazily creates pools for the relevant memory type indices, which doesn't require patching VMA. Also, patches already merged upstream or not needed any longer are removed.
2022-02-24 11:47:34 +00:00
VmaPool RenderingDeviceVulkan::_find_or_create_small_allocs_pool(uint32_t p_mem_type_index) {
if (small_allocs_pools.has(p_mem_type_index)) {
return small_allocs_pools[p_mem_type_index];
}
print_verbose("Creating VMA small objects pool for memory type index " + itos(p_mem_type_index));
VmaPoolCreateInfo pci;
pci.memoryTypeIndex = p_mem_type_index;
pci.flags = 0;
pci.blockSize = 0;
pci.minBlockCount = 0;
pci.maxBlockCount = SIZE_MAX;
pci.priority = 0.5f;
pci.minAllocationAlignment = 0;
pci.pMemoryAllocateNext = nullptr;
VmaPool pool = VK_NULL_HANDLE;
VkResult res = vmaCreatePool(allocator, &pci, &pool);
small_allocs_pools[p_mem_type_index] = pool; // Don't try to create it again if failed the first time
ERR_FAIL_COND_V_MSG(res, pool, "vmaCreatePool failed with error " + itos(res) + ".");
return pool;
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
void RenderingDeviceVulkan::_free_pending_resources(int p_frame) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//free in dependency usage order, so nothing weird happens
//pipelines
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
while (frames[p_frame].render_pipelines_to_dispose_of.front()) {
RenderPipeline *pipeline = &frames[p_frame].render_pipelines_to_dispose_of.front()->get();
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyPipeline(device, pipeline->pipeline, nullptr);
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
frames[p_frame].render_pipelines_to_dispose_of.pop_front();
}
while (frames[p_frame].compute_pipelines_to_dispose_of.front()) {
ComputePipeline *pipeline = &frames[p_frame].compute_pipelines_to_dispose_of.front()->get();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyPipeline(device, pipeline->pipeline, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
frames[p_frame].compute_pipelines_to_dispose_of.pop_front();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//uniform sets
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
while (frames[p_frame].uniform_sets_to_dispose_of.front()) {
UniformSet *uniform_set = &frames[p_frame].uniform_sets_to_dispose_of.front()->get();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vkFreeDescriptorSets(device, uniform_set->pool->pool, 1, &uniform_set->descriptor_set);
_descriptor_pool_free(uniform_set->pool_key, uniform_set->pool);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
frames[p_frame].uniform_sets_to_dispose_of.pop_front();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//buffer views
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
while (frames[p_frame].buffer_views_to_dispose_of.front()) {
VkBufferView buffer_view = frames[p_frame].buffer_views_to_dispose_of.front()->get();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyBufferView(device, buffer_view, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
frames[p_frame].buffer_views_to_dispose_of.pop_front();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//shaders
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
while (frames[p_frame].shaders_to_dispose_of.front()) {
Shader *shader = &frames[p_frame].shaders_to_dispose_of.front()->get();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//descriptor set layout for each set
for (int i = 0; i < shader->sets.size(); i++) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyDescriptorSetLayout(device, shader->sets[i].descriptor_set_layout, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//pipeline layout
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyPipelineLayout(device, shader->pipeline_layout, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//shaders themselves
for (int i = 0; i < shader->pipeline_stages.size(); i++) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyShaderModule(device, shader->pipeline_stages[i].module, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
frames[p_frame].shaders_to_dispose_of.pop_front();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//samplers
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
while (frames[p_frame].samplers_to_dispose_of.front()) {
VkSampler sampler = frames[p_frame].samplers_to_dispose_of.front()->get();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroySampler(device, sampler, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
frames[p_frame].samplers_to_dispose_of.pop_front();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//framebuffers
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
while (frames[p_frame].framebuffers_to_dispose_of.front()) {
Framebuffer *framebuffer = &frames[p_frame].framebuffers_to_dispose_of.front()->get();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Use range iterators for `Map`
2021-08-09 20:13:42 +00:00
for (const KeyValue<Framebuffer::VersionKey, Framebuffer::Version> &E : framebuffer->framebuffers) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//first framebuffer, then render pass because it depends on it
Use range iterators for `Map`
2021-08-09 20:13:42 +00:00
vkDestroyFramebuffer(device, E.value.framebuffer, nullptr);
vkDestroyRenderPass(device, E.value.render_pass, nullptr);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
frames[p_frame].framebuffers_to_dispose_of.pop_front();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//textures
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
while (frames[p_frame].textures_to_dispose_of.front()) {
Texture *texture = &frames[p_frame].textures_to_dispose_of.front()->get();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
if (texture->bound) {
WARN_PRINT("Deleted a texture while it was bound..");
}
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyImageView(device, texture->view, nullptr);
Modify RenderingDevice to use RIDs, now that they are O(1)
2019-06-10 17:12:24 +00:00
if (texture->owner.is_null()) {
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
//actually owns the image and the allocation too
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
image_memory -= texture->allocation_info.size;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vmaDestroyImage(allocator, texture->image, texture->allocation);
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
frames[p_frame].textures_to_dispose_of.pop_front();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
//buffers
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
while (frames[p_frame].buffers_to_dispose_of.front()) {
_buffer_free(&frames[p_frame].buffers_to_dispose_of.front()->get());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
frames[p_frame].buffers_to_dispose_of.pop_front();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
void RenderingDeviceVulkan::prepare_screen_for_drawing() {
_THREAD_SAFE_METHOD_
context->prepare_buffers();
screen_prepared = true;
}
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
uint32_t RenderingDeviceVulkan::get_frame_delay() const {
return frame_count;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
uint64_t RenderingDeviceVulkan::get_memory_usage(MemoryType p_type) const {
if (p_type == MEMORY_BUFFERS) {
return buffer_memory;
} else if (p_type == MEMORY_TEXTURES) {
return image_memory;
} else {
vk_mem_alloc: Update to upstream + Adapt approach to small objects pooling This updates VMA and instead of using the custom small pool approach from 4e6c9d3ae979f2eb0151cf581fe61d2f3194ea72, lazily creates pools for the relevant memory type indices, which doesn't require patching VMA. Also, patches already merged upstream or not needed any longer are removed.
2022-02-24 11:47:34 +00:00
VmaTotalStatistics stats;
vmaCalculateStatistics(allocator, &stats);
return stats.total.statistics.allocationBytes;
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
}
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
}
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
void RenderingDeviceVulkan::_flush(bool p_current_frame) {
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
if (local_device.is_valid() && !p_current_frame) {
Fixes small typos and grammar correction
2021-03-12 13:35:16 +00:00
return; //flushing previous frames has no effect with local device
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
}
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
//not doing this crashes RADV (undefined behavior)
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
if (p_current_frame) {
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
vkEndCommandBuffer(frames[frame].setup_command_buffer);
vkEndCommandBuffer(frames[frame].draw_command_buffer);
}
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
if (local_device.is_valid()) {
VkCommandBuffer command_buffers[2] = { frames[frame].setup_command_buffer, frames[frame].draw_command_buffer };
context->local_device_push_command_buffers(local_device, command_buffers, 2);
context->local_device_sync(local_device);
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
VkCommandBufferBeginInfo cmdbuf_begin;
cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
cmdbuf_begin.pNext = nullptr;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
cmdbuf_begin.pInheritanceInfo = nullptr;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
VkResult err = vkBeginCommandBuffer(frames[frame].setup_command_buffer, &cmdbuf_begin);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
err = vkBeginCommandBuffer(frames[frame].draw_command_buffer, &cmdbuf_begin);
ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
} else {
context->flush(p_current_frame, p_current_frame);
//re-create the setup command
if (p_current_frame) {
VkCommandBufferBeginInfo cmdbuf_begin;
cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
cmdbuf_begin.pNext = nullptr;
cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
cmdbuf_begin.pInheritanceInfo = nullptr;
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
VkResult err = vkBeginCommandBuffer(frames[frame].setup_command_buffer, &cmdbuf_begin);
ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
context->set_setup_buffer(frames[frame].setup_command_buffer); //append now so it's added before everything else
}
if (p_current_frame) {
VkCommandBufferBeginInfo cmdbuf_begin;
cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
cmdbuf_begin.pNext = nullptr;
cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
cmdbuf_begin.pInheritanceInfo = nullptr;
VkResult err = vkBeginCommandBuffer(frames[frame].draw_command_buffer, &cmdbuf_begin);
ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
context->append_command_buffer(frames[frame].draw_command_buffer);
}
Added ability to retrieve back textures stored on GPU
2019-06-25 22:49:52 +00:00
}
}
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
void RenderingDeviceVulkan::initialize(VulkanContext *p_context, bool p_local_device) {
Obtain supported Vulkan API
2021-03-22 10:04:55 +00:00
// get our device capabilities
{
device_capabilities.version_major = p_context->get_vulkan_major();
device_capabilities.version_minor = p_context->get_vulkan_minor();
// get info about subgroups
VulkanContext::SubgroupCapabilities subgroup_capabilities = p_context->get_subgroup_capabilities();
device_capabilities.subgroup_size = subgroup_capabilities.size;
device_capabilities.subgroup_in_shaders = subgroup_capabilities.supported_stages_flags_rd();
device_capabilities.subgroup_operations = subgroup_capabilities.supported_operations_flags_rd();
Cleanup vulkan capabilities check and add multiview check
2021-05-04 05:30:21 +00:00
// get info about further features
VulkanContext::MultiviewCapabilities multiview_capabilies = p_context->get_multiview_capabilities();
device_capabilities.supports_multiview = multiview_capabilies.is_supported && multiview_capabilies.max_view_count > 1;
Implemented AMD's FSR as a computer shader for upscaling 3D scenes
2021-11-23 21:16:03 +00:00
device_capabilities.supports_fsr_half_float = p_context->get_shader_capabilities().shader_float16_is_supported && p_context->get_storage_buffer_capabilities().storage_buffer_16_bit_access_is_supported;
Obtain supported Vulkan API
2021-03-22 10:04:55 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
context = p_context;
device = p_context->get_device();
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
if (p_local_device) {
frame_count = 1;
local_device = p_context->local_device_create();
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
device = p_context->local_device_get_vk_device(local_device);
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
} else {
frame_count = p_context->get_swapchain_image_count() + 1; //always need one extra to ensure it's unused at any time, without having to use a fence for this.
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
limits = p_context->get_device_limits();
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
max_timestamp_query_elements = 256;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
{ //initialize allocator
VmaAllocatorCreateInfo allocatorInfo;
memset(&allocatorInfo, 0, sizeof(VmaAllocatorCreateInfo));
allocatorInfo.physicalDevice = p_context->get_physical_device();
allocatorInfo.device = device;
Upgrade Vulkan memory allocator
2021-08-12 22:02:20 +00:00
allocatorInfo.instance = p_context->get_instance();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
vmaCreateAllocator(&allocatorInfo, &allocator);
}
frames = memnew_arr(Frame, frame_count);
frame = 0;
//create setup and frame buffers
for (int i = 0; i < frame_count; i++) {
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
frames[i].index = 0;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
{ //create command pool, one per frame is recommended
VkCommandPoolCreateInfo cmd_pool_info;
cmd_pool_info.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
cmd_pool_info.pNext = nullptr;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
cmd_pool_info.queueFamilyIndex = p_context->get_graphics_queue_family_index();
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
cmd_pool_info.flags = VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
VkResult res = vkCreateCommandPool(device, &cmd_pool_info, nullptr, &frames[i].command_pool);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_MSG(res, "vkCreateCommandPool failed with error " + itos(res) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
{ //create command buffers
VkCommandBufferAllocateInfo cmdbuf;
//no command buffer exists, create it.
cmdbuf.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
cmdbuf.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
cmdbuf.commandPool = frames[i].command_pool;
cmdbuf.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
cmdbuf.commandBufferCount = 1;
VkResult err = vkAllocateCommandBuffers(device, &cmdbuf, &frames[i].setup_command_buffer);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_CONTINUE_MSG(err, "vkAllocateCommandBuffers failed with error " + itos(err) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
err = vkAllocateCommandBuffers(device, &cmdbuf, &frames[i].draw_command_buffer);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_CONTINUE_MSG(err, "vkAllocateCommandBuffers failed with error " + itos(err) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
{
//create query pool
VkQueryPoolCreateInfo query_pool_create_info;
query_pool_create_info.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO;
query_pool_create_info.flags = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
query_pool_create_info.pNext = nullptr;
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
query_pool_create_info.queryType = VK_QUERY_TYPE_TIMESTAMP;
query_pool_create_info.queryCount = max_timestamp_query_elements;
query_pool_create_info.pipelineStatistics = 0;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkCreateQueryPool(device, &query_pool_create_info, nullptr, &frames[i].timestamp_pool);
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
frames[i].timestamp_names = memnew_arr(String, max_timestamp_query_elements);
frames[i].timestamp_cpu_values = memnew_arr(uint64_t, max_timestamp_query_elements);
frames[i].timestamp_count = 0;
frames[i].timestamp_result_names = memnew_arr(String, max_timestamp_query_elements);
frames[i].timestamp_cpu_result_values = memnew_arr(uint64_t, max_timestamp_query_elements);
frames[i].timestamp_result_values = memnew_arr(uint64_t, max_timestamp_query_elements);
frames[i].timestamp_result_count = 0;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
{
//begin the first command buffer for the first frame, so
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
//setting up things can be done in the meantime until swap_buffers(), which is called before advance.
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkCommandBufferBeginInfo cmdbuf_begin;
cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
cmdbuf_begin.pNext = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
cmdbuf_begin.pInheritanceInfo = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
VkResult err = vkBeginCommandBuffer(frames[0].setup_command_buffer, &cmdbuf_begin);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
err = vkBeginCommandBuffer(frames[0].draw_command_buffer, &cmdbuf_begin);
More explicit error messages when vulkan calls return errors
2020-03-08 08:26:22 +00:00
ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
if (local_device.is_null()) {
context->set_setup_buffer(frames[0].setup_command_buffer); //append now so it's added before everything else
context->append_command_buffer(frames[0].draw_command_buffer);
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
doctool: Fix differences between headless and Vulkan rendering backends Fixes #53913.
2021-10-20 14:37:19 +00:00
// Note: If adding new project settings here, also duplicate their definition in
// rendering_server.cpp for headless doctool.
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
staging_buffer_block_size = GLOBAL_DEF("rendering/vulkan/staging_buffer/block_size_kb", 256);
Change some math macros to constexpr Changes `MAX`, `MIN`, `ABS`, `CLAMP` and `SIGN`.
2022-03-08 14:10:48 +00:00
staging_buffer_block_size = MAX(4u, staging_buffer_block_size);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
staging_buffer_block_size *= 1024; //kb -> bytes
staging_buffer_max_size = GLOBAL_DEF("rendering/vulkan/staging_buffer/max_size_mb", 128);
Change some math macros to constexpr Changes `MAX`, `MIN`, `ABS`, `CLAMP` and `SIGN`.
2022-03-08 14:10:48 +00:00
staging_buffer_max_size = MAX(1u, staging_buffer_max_size);
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
staging_buffer_max_size *= 1024 * 1024;
if (staging_buffer_max_size < staging_buffer_block_size * 4) {
//validate enough blocks
staging_buffer_max_size = staging_buffer_block_size * 4;
}
texture_upload_region_size_px = GLOBAL_DEF("rendering/vulkan/staging_buffer/texture_upload_region_size_px", 64);
texture_upload_region_size_px = nearest_power_of_2_templated(texture_upload_region_size_px);
frames_drawn = frame_count; //start from frame count, so everything else is immediately old
//ensure current staging block is valid and at least one per frame exists
staging_buffer_current = 0;
staging_buffer_used = false;
for (int i = 0; i < frame_count; i++) {
//staging was never used, create a block
Error err = _insert_staging_block();
ERR_CONTINUE(err != OK);
}
max_descriptors_per_pool = GLOBAL_DEF("rendering/vulkan/descriptor_pools/max_descriptors_per_pool", 64);
//check to make sure DescriptorPoolKey is good
Change when we can ERR_FAIL_COND and ERR_FAIL_CONV_V to static_assert
2020-02-20 14:13:08 +00:00
static_assert(sizeof(uint64_t) * 3 >= UNIFORM_TYPE_MAX * sizeof(uint16_t));
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
draw_list = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
draw_list_count = 0;
draw_list_split = false;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
compute_list = nullptr;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
template <class T>
void RenderingDeviceVulkan::_free_rids(T &p_owner, const char *p_type) {
List<RID> owned;
p_owner.get_owned_list(&owned);
if (owned.size()) {
Tweak warning messages related to leaked RIDs - Use hardcoded pluralization for messages. - Since these messages are English-only, it's fine to hardcode it. - Use double quotes for consistency with other messages.
2021-05-03 14:05:38 +00:00
if (owned.size() == 1) {
WARN_PRINT(vformat("1 RID of type \"%s\" was leaked.", p_type));
} else {
WARN_PRINT(vformat("%d RIDs of type \"%s\" were leaked.", owned.size(), p_type));
}
Use const references where possible for List range iterators
2021-07-24 13:46:25 +00:00
for (const RID &E : owned) {
Use C++ iterators for Lists in many situations
2021-07-16 03:45:57 +00:00
free(E);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
}
}
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
void RenderingDeviceVulkan::capture_timestamp(const String &p_name) {
Fix various typos Found via `codespell -q 3 -S ./thirdparty,*.po,./DONORS.md -L ackward,ang,ans,ba,beng,cas,childs,childrens,dof,doubleclick,expct,fave,findn,gird,hist,inh,inout,leapyear,lod,nd,numer,ois,ony,paket,ro,seeked,sinc,switchs,te,uint,varn,vew`
2022-01-02 18:47:52 +00:00
ERR_FAIL_COND_MSG(draw_list != nullptr, "Capturing timestamps during draw list creation is not allowed. Offending timestamp was: " + p_name);
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
ERR_FAIL_COND(frames[frame].timestamp_count >= max_timestamp_query_elements);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
//this should be optional for profiling, else it will slow things down
Better GIProbe quality settings.
2019-10-03 23:15:38 +00:00
{
VkMemoryBarrier memoryBarrier;
memoryBarrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER;
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
memoryBarrier.pNext = nullptr;
Better GIProbe quality settings.
2019-10-03 23:15:38 +00:00
memoryBarrier.srcAccessMask = VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
clang-format: Disable alignment of operands, too unreliable Sets `AlignOperands` to `DontAlign`. `clang-format` developers seem to mostly care about space-based indentation and every other version of clang-format breaks the bad mismatch of tabs and spaces that it seems to use for operand alignment. So it's better without, so that it respects our two-tabs `ContinuationIndentWidth`.
2021-10-28 13:19:35 +00:00
VK_ACCESS_INDEX_READ_BIT |
VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
VK_ACCESS_UNIFORM_READ_BIT |
VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
VK_ACCESS_SHADER_READ_BIT |
VK_ACCESS_SHADER_WRITE_BIT |
VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
VK_ACCESS_TRANSFER_READ_BIT |
VK_ACCESS_TRANSFER_WRITE_BIT |
VK_ACCESS_HOST_READ_BIT |
VK_ACCESS_HOST_WRITE_BIT;
Better GIProbe quality settings.
2019-10-03 23:15:38 +00:00
memoryBarrier.dstAccessMask = VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
clang-format: Disable alignment of operands, too unreliable Sets `AlignOperands` to `DontAlign`. `clang-format` developers seem to mostly care about space-based indentation and every other version of clang-format breaks the bad mismatch of tabs and spaces that it seems to use for operand alignment. So it's better without, so that it respects our two-tabs `ContinuationIndentWidth`.
2021-10-28 13:19:35 +00:00
VK_ACCESS_INDEX_READ_BIT |
VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
VK_ACCESS_UNIFORM_READ_BIT |
VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
VK_ACCESS_SHADER_READ_BIT |
VK_ACCESS_SHADER_WRITE_BIT |
VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
VK_ACCESS_TRANSFER_READ_BIT |
VK_ACCESS_TRANSFER_WRITE_BIT |
VK_ACCESS_HOST_READ_BIT |
VK_ACCESS_HOST_WRITE_BIT;
Better GIProbe quality settings.
2019-10-03 23:15:38 +00:00
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
vkCmdPipelineBarrier(frames[frame].draw_command_buffer, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, 1, &memoryBarrier, 0, nullptr, 0, nullptr);
Better GIProbe quality settings.
2019-10-03 23:15:38 +00:00
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
vkCmdWriteTimestamp(frames[frame].draw_command_buffer, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, frames[frame].timestamp_pool, frames[frame].timestamp_count);
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
frames[frame].timestamp_names[frames[frame].timestamp_count] = p_name;
frames[frame].timestamp_cpu_values[frames[frame].timestamp_count] = OS::get_singleton()->get_ticks_usec();
frames[frame].timestamp_count++;
}
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
uint64_t RenderingDeviceVulkan::get_driver_resource(DriverResource p_resource, RID p_rid, uint64_t p_index) {
_THREAD_SAFE_METHOD_
switch (p_resource) {
case DRIVER_RESOURCE_VULKAN_DEVICE: {
return (uint64_t)context->get_device();
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_PHYSICAL_DEVICE: {
return (uint64_t)context->get_physical_device();
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_INSTANCE: {
return (uint64_t)context->get_instance();
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_QUEUE: {
return (uint64_t)context->get_graphics_queue();
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_QUEUE_FAMILY_INDEX: {
return context->get_graphics_queue_family_index();
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_IMAGE: {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *tex = texture_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
ERR_FAIL_NULL_V(tex, 0);
return (uint64_t)tex->image;
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_IMAGE_VIEW: {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *tex = texture_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
ERR_FAIL_NULL_V(tex, 0);
return (uint64_t)tex->view;
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_IMAGE_NATIVE_TEXTURE_FORMAT: {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
Texture *tex = texture_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
ERR_FAIL_NULL_V(tex, 0);
return vulkan_formats[tex->format];
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_SAMPLER: {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
VkSampler *sampler = sampler_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
ERR_FAIL_NULL_V(sampler, 0);
return uint64_t(*sampler);
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_DESCRIPTOR_SET: {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
UniformSet *uniform_set = uniform_set_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
ERR_FAIL_NULL_V(uniform_set, 0);
return uint64_t(uniform_set->descriptor_set);
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_BUFFER: {
Buffer *buffer = nullptr;
if (vertex_buffer_owner.owns(p_rid)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = vertex_buffer_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
} else if (index_buffer_owner.owns(p_rid)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = index_buffer_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
} else if (uniform_buffer_owner.owns(p_rid)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = uniform_buffer_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
} else if (texture_buffer_owner.owns(p_rid)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = &texture_buffer_owner.get_or_null(p_rid)->buffer;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
} else if (storage_buffer_owner.owns(p_rid)) {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
buffer = storage_buffer_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
}
ERR_FAIL_NULL_V(buffer, 0);
return uint64_t(buffer->buffer);
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_COMPUTE_PIPELINE: {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
ComputePipeline *compute_pipeline = compute_pipeline_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
ERR_FAIL_NULL_V(compute_pipeline, 0);
return uint64_t(compute_pipeline->pipeline);
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
case DRIVER_RESOURCE_VULKAN_RENDER_PIPELINE: {
Rename RID's `getornull()` to `get_or_null()`
2021-09-29 17:08:41 +00:00
RenderPipeline *render_pipeline = render_pipeline_owner.get_or_null(p_rid);
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
ERR_FAIL_NULL_V(render_pipeline, 0);
return uint64_t(render_pipeline->pipeline);
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
default: {
// not supported for this driver
return 0;
Style: Cleanup single-line blocks, semicolons, dead code Remove currently unused implementation of TextureBasisU, could be re-added later on if needed and ported.
2022-02-16 12:56:32 +00:00
} break;
Expose Vulkan internal values for access from extensions
2021-08-29 02:52:19 +00:00
}
}
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
uint32_t RenderingDeviceVulkan::get_captured_timestamps_count() const {
return frames[frame].timestamp_result_count;
}
uint64_t RenderingDeviceVulkan::get_captured_timestamps_frame() const {
return frames[frame].index;
}
Shows proper scene render time in editor info Also fixed GPU profiler, which was not working on nvidia hardware.
2020-04-10 17:18:42 +00:00
static void mult64to128(uint64_t u, uint64_t v, uint64_t &h, uint64_t &l) {
uint64_t u1 = (u & 0xffffffff);
uint64_t v1 = (v & 0xffffffff);
uint64_t t = (u1 * v1);
uint64_t w3 = (t & 0xffffffff);
uint64_t k = (t >> 32);
u >>= 32;
t = (u * v1) + k;
k = (t & 0xffffffff);
uint64_t w1 = (t >> 32);
v >>= 32;
t = (u1 * v) + k;
k = (t >> 32);
h = (u * v) + w1 + k;
l = (t << 32) + w3;
}
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
uint64_t RenderingDeviceVulkan::get_captured_timestamp_gpu_time(uint32_t p_index) const {
Fix various GCC compilation warnings after Vulkan merge Part of #36132.
2020-02-13 14:53:32 +00:00
ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, 0);
Shows proper scene render time in editor info Also fixed GPU profiler, which was not working on nvidia hardware.
2020-04-10 17:18:42 +00:00
// this sucks because timestampPeriod multiplier is a float, while the timestamp is 64 bits nanosecs.
// so, in cases like nvidia which give you enormous numbers and 1 as multiplier, multiplying is next to impossible
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
// need to do 128 bits fixed point multiplication to get the right value
Shows proper scene render time in editor info Also fixed GPU profiler, which was not working on nvidia hardware.
2020-04-10 17:18:42 +00:00
uint64_t shift_bits = 16;
uint64_t h, l;
mult64to128(frames[frame].timestamp_result_values[p_index], uint64_t(double(limits.timestampPeriod) * double(1 << shift_bits)), h, l);
l >>= shift_bits;
l |= h << (64 - shift_bits);
return l;
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
uint64_t RenderingDeviceVulkan::get_captured_timestamp_cpu_time(uint32_t p_index) const {
Fix various GCC compilation warnings after Vulkan merge Part of #36132.
2020-02-13 14:53:32 +00:00
ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, 0);
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
return frames[frame].timestamp_cpu_result_values[p_index];
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
String RenderingDeviceVulkan::get_captured_timestamp_name(uint32_t p_index) const {
Fix various GCC compilation warnings after Vulkan merge Part of #36132.
2020-02-13 14:53:32 +00:00
ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, String());
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
return frames[frame].timestamp_result_names[p_index];
}
Fix device limit exceeding for uniform buffer
2022-03-10 17:43:27 +00:00
uint64_t RenderingDeviceVulkan::limit_get(Limit p_limit) {
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
switch (p_limit) {
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case LIMIT_MAX_BOUND_UNIFORM_SETS:
return limits.maxBoundDescriptorSets;
case LIMIT_MAX_FRAMEBUFFER_COLOR_ATTACHMENTS:
return limits.maxColorAttachments;
case LIMIT_MAX_TEXTURES_PER_UNIFORM_SET:
return limits.maxDescriptorSetSampledImages;
case LIMIT_MAX_SAMPLERS_PER_UNIFORM_SET:
return limits.maxDescriptorSetSamplers;
case LIMIT_MAX_STORAGE_BUFFERS_PER_UNIFORM_SET:
return limits.maxDescriptorSetStorageBuffers;
case LIMIT_MAX_STORAGE_IMAGES_PER_UNIFORM_SET:
return limits.maxDescriptorSetStorageImages;
case LIMIT_MAX_UNIFORM_BUFFERS_PER_UNIFORM_SET:
return limits.maxDescriptorSetUniformBuffers;
case LIMIT_MAX_DRAW_INDEXED_INDEX:
return limits.maxDrawIndexedIndexValue;
case LIMIT_MAX_FRAMEBUFFER_HEIGHT:
return limits.maxFramebufferHeight;
case LIMIT_MAX_FRAMEBUFFER_WIDTH:
return limits.maxFramebufferWidth;
case LIMIT_MAX_TEXTURE_ARRAY_LAYERS:
return limits.maxImageArrayLayers;
case LIMIT_MAX_TEXTURE_SIZE_1D:
return limits.maxImageDimension1D;
case LIMIT_MAX_TEXTURE_SIZE_2D:
return limits.maxImageDimension2D;
case LIMIT_MAX_TEXTURE_SIZE_3D:
return limits.maxImageDimension3D;
case LIMIT_MAX_TEXTURE_SIZE_CUBE:
return limits.maxImageDimensionCube;
case LIMIT_MAX_TEXTURES_PER_SHADER_STAGE:
return limits.maxPerStageDescriptorSampledImages;
case LIMIT_MAX_SAMPLERS_PER_SHADER_STAGE:
return limits.maxPerStageDescriptorSamplers;
case LIMIT_MAX_STORAGE_BUFFERS_PER_SHADER_STAGE:
return limits.maxPerStageDescriptorStorageBuffers;
case LIMIT_MAX_STORAGE_IMAGES_PER_SHADER_STAGE:
return limits.maxPerStageDescriptorStorageImages;
case LIMIT_MAX_UNIFORM_BUFFERS_PER_SHADER_STAGE:
return limits.maxPerStageDescriptorUniformBuffers;
case LIMIT_MAX_PUSH_CONSTANT_SIZE:
return limits.maxPushConstantsSize;
case LIMIT_MAX_UNIFORM_BUFFER_SIZE:
return limits.maxUniformBufferRange;
case LIMIT_MAX_VERTEX_INPUT_ATTRIBUTE_OFFSET:
return limits.maxVertexInputAttributeOffset;
case LIMIT_MAX_VERTEX_INPUT_ATTRIBUTES:
return limits.maxVertexInputAttributes;
case LIMIT_MAX_VERTEX_INPUT_BINDINGS:
return limits.maxVertexInputBindings;
case LIMIT_MAX_VERTEX_INPUT_BINDING_STRIDE:
return limits.maxVertexInputBindingStride;
case LIMIT_MIN_UNIFORM_BUFFER_OFFSET_ALIGNMENT:
return limits.minUniformBufferOffsetAlignment;
case LIMIT_MAX_COMPUTE_WORKGROUP_COUNT_X:
return limits.maxComputeWorkGroupCount[0];
case LIMIT_MAX_COMPUTE_WORKGROUP_COUNT_Y:
return limits.maxComputeWorkGroupCount[1];
case LIMIT_MAX_COMPUTE_WORKGROUP_COUNT_Z:
return limits.maxComputeWorkGroupCount[2];
case LIMIT_MAX_COMPUTE_WORKGROUP_INVOCATIONS:
return limits.maxComputeWorkGroupInvocations;
case LIMIT_MAX_COMPUTE_WORKGROUP_SIZE_X:
return limits.maxComputeWorkGroupSize[0];
case LIMIT_MAX_COMPUTE_WORKGROUP_SIZE_Y:
return limits.maxComputeWorkGroupSize[1];
case LIMIT_MAX_COMPUTE_WORKGROUP_SIZE_Z:
return limits.maxComputeWorkGroupSize[2];
default:
ERR_FAIL_V(0);
Modified light rendering to make it more compatible. Modified polygon management to make it more compatible with MoltenVK
2019-07-10 20:44:55 +00:00
}
return 0;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
void RenderingDeviceVulkan::finalize() {
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
//free all resources
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
_flush(false);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
_free_rids(render_pipeline_owner, "Pipeline");
_free_rids(compute_pipeline_owner, "Compute");
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
_free_rids(uniform_set_owner, "UniformSet");
_free_rids(texture_buffer_owner, "TextureBuffer");
_free_rids(storage_buffer_owner, "StorageBuffer");
_free_rids(uniform_buffer_owner, "UniformBuffer");
_free_rids(shader_owner, "Shader");
_free_rids(index_array_owner, "IndexArray");
_free_rids(index_buffer_owner, "IndexBuffer");
_free_rids(vertex_array_owner, "VertexArray");
_free_rids(vertex_buffer_owner, "VertexBuffer");
_free_rids(framebuffer_owner, "Framebuffer");
_free_rids(sampler_owner, "Sampler");
{
//for textures it's a bit more difficult because they may be shared
List<RID> owned;
texture_owner.get_owned_list(&owned);
if (owned.size()) {
Tweak warning messages related to leaked RIDs - Use hardcoded pluralization for messages. - Since these messages are English-only, it's fine to hardcode it. - Use double quotes for consistency with other messages.
2021-05-03 14:05:38 +00:00
if (owned.size() == 1) {
WARN_PRINT("1 RID of type \"Texture\" was leaked.");
} else {
WARN_PRINT(vformat("%d RIDs of type \"Texture\" were leaked.", owned.size()));
}
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
//free shared first
for (List<RID>::Element *E = owned.front(); E;) {
List<RID>::Element *N = E->next();
if (texture_is_shared(E->get())) {
free(E->get());
Use C++ iterators for Lists in many situations
2021-07-16 03:45:57 +00:00
owned.erase(E);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
}
E = N;
}
//free non shared second, this will avoid an error trying to free unexisting textures due to dependencies.
Use const references where possible for List range iterators
2021-07-24 13:46:25 +00:00
for (const RID &E : owned) {
Use C++ iterators for Lists in many situations
2021-07-16 03:45:57 +00:00
free(E);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
}
}
}
//free everything pending
for (int i = 0; i < frame_count; i++) {
int f = (frame + i) % frame_count;
_free_pending_resources(f);
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyCommandPool(device, frames[i].command_pool, nullptr);
vkDestroyQueryPool(device, frames[i].timestamp_pool, nullptr);
Visual GPU profiler and related profiling support in Vulkan.
2019-09-20 20:58:06 +00:00
memdelete_arr(frames[i].timestamp_names);
memdelete_arr(frames[i].timestamp_cpu_values);
memdelete_arr(frames[i].timestamp_result_names);
memdelete_arr(frames[i].timestamp_result_values);
memdelete_arr(frames[i].timestamp_cpu_result_values);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
}
for (int i = 0; i < split_draw_list_allocators.size(); i++) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
vkDestroyCommandPool(device, split_draw_list_allocators[i].command_pool, nullptr);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
memdelete_arr(frames);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
for (int i = 0; i < staging_buffer_blocks.size(); i++) {
vmaDestroyBuffer(allocator, staging_buffer_blocks[i].buffer, staging_buffer_blocks[i].allocation);
}
vk_mem_alloc: Update to upstream + Adapt approach to small objects pooling This updates VMA and instead of using the custom small pool approach from 4e6c9d3ae979f2eb0151cf581fe61d2f3194ea72, lazily creates pools for the relevant memory type indices, which doesn't require patching VMA. Also, patches already merged upstream or not needed any longer are removed.
2022-02-24 11:47:34 +00:00
while (small_allocs_pools.size()) {
Map<uint32_t, VmaPool>::Element *E = small_allocs_pools.front();
vmaDestroyPool(allocator, E->get());
small_allocs_pools.erase(E);
}
Fix leak with Vulkan allocator
2020-04-04 18:16:48 +00:00
vmaDestroyAllocator(allocator);
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
Fixed leak with vertex_formats
2020-04-06 16:07:31 +00:00
while (vertex_formats.size()) {
Map<VertexFormatID, VertexDescriptionCache>::Element *temp = vertex_formats.front();
memdelete_arr(temp->get().bindings);
memdelete_arr(temp->get().attributes);
vertex_formats.erase(temp);
}
Fixes leaks in ResourceCache, Vulkan and X11
2020-04-15 08:27:57 +00:00
for (int i = 0; i < framebuffer_formats.size(); i++) {
vkDestroyRenderPass(device, framebuffer_formats[i].render_pass, nullptr);
}
framebuffer_formats.clear();
basic 2D engine is more or less working with Vulkan, including editor. Still a lot to do
2019-06-24 19:13:06 +00:00
//all these should be clear at this point
ERR_FAIL_COND(descriptor_pools.size());
ERR_FAIL_COND(dependency_map.size());
ERR_FAIL_COND(reverse_dependency_map.size());
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
RenderingDevice *RenderingDeviceVulkan::create_local_device() {
RenderingDeviceVulkan *rd = memnew(RenderingDeviceVulkan);
rd->initialize(context, true);
return rd;
}
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
RenderingDeviceVulkan::RenderingDeviceVulkan() {
Obtain supported Vulkan API
2021-03-22 10:04:55 +00:00
device_capabilities.device_family = DEVICE_VULKAN;
Initial work on Vulkan: -Added VulkanContext -Added an X11 implementation -Added a rendering device abstraction -added a Vulkan rendering device abstraction -Engine does not work, only shows Godot logo (run it from bin/)
2019-06-07 16:07:57 +00:00
}
Ability to create local RenderingDevice instances.
2020-04-18 23:30:57 +00:00
RenderingDeviceVulkan::~RenderingDeviceVulkan() {
if (local_device.is_valid()) {
finalize();
context->local_device_free(local_device);
}
}