godotengine
/
godot
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
godot/drivers/gles2/rasterizer_canvas_base_gles...

1064 lines
41 KiB
C++
Raw Normal View History

GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
/*************************************************************************/
/* rasterizer_canvas_base_gles2.cpp */
/*************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* https://godotengine.org */
/*************************************************************************/
Update copyright statements to 2022 Happy new year to the wonderful Godot community!
2022-01-13 08:45:09 +00:00
/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */
/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +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. */
/*************************************************************************/
#include "rasterizer_canvas_base_gles2.h"
#include "core/os/os.h"
#include "core/project_settings.h"
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
#include "drivers/gles_common/rasterizer_asserts.h"
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
#include "rasterizer_scene_gles2.h"
#include "servers/visual/visual_server_raster.h"
#ifndef GLES_OVER_GL
#define glClearDepth glClearDepthf
#endif
RID RasterizerCanvasBaseGLES2::light_internal_create() {
return RID();
}
void RasterizerCanvasBaseGLES2::light_internal_update(RID p_rid, Light *p_light) {
}
void RasterizerCanvasBaseGLES2::light_internal_free(RID p_rid) {
}
void RasterizerCanvasBaseGLES2::canvas_begin() {
state.using_transparent_rt = false;
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
// always start with light_angle unset
state.using_light_angle = false;
Unified GLES2 / GLES3 Batching Batching is mostly separated into a common template which can be used with multiple backends (GLES2 and GLES3 here). Only necessary specifics are in the backend files. Batching is extended to cover more primitives.
2020-10-14 07:32:05 +00:00
state.using_large_vertex = false;
state.using_modulate = false;
state.canvas_shader.set_conditional(CanvasShaderGLES2::USE_ATTRIB_LIGHT_ANGLE, false);
state.canvas_shader.set_conditional(CanvasShaderGLES2::USE_ATTRIB_MODULATE, false);
state.canvas_shader.set_conditional(CanvasShaderGLES2::USE_ATTRIB_LARGE_VERTEX, false);
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
state.canvas_shader.bind();
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
int viewport_x, viewport_y, viewport_width, viewport_height;
if (storage->frame.current_rt) {
glBindFramebuffer(GL_FRAMEBUFFER, storage->frame.current_rt->fbo);
state.using_transparent_rt = storage->frame.current_rt->flags[RasterizerStorage::RENDER_TARGET_TRANSPARENT];
if (storage->frame.current_rt->flags[RasterizerStorage::RENDER_TARGET_DIRECT_TO_SCREEN]) {
// set Viewport and Scissor when rendering directly to screen
viewport_width = storage->frame.current_rt->width;
viewport_height = storage->frame.current_rt->height;
viewport_x = storage->frame.current_rt->x;
viewport_y = OS::get_singleton()->get_window_size().height - viewport_height - storage->frame.current_rt->y;
glScissor(viewport_x, viewport_y, viewport_width, viewport_height);
glViewport(viewport_x, viewport_y, viewport_width, viewport_height);
glEnable(GL_SCISSOR_TEST);
}
}
if (storage->frame.clear_request) {
glClearColor(storage->frame.clear_request_color.r,
storage->frame.clear_request_color.g,
storage->frame.clear_request_color.b,
state.using_transparent_rt ? storage->frame.clear_request_color.a : 1.0);
glClear(GL_COLOR_BUFFER_BIT);
storage->frame.clear_request = false;
}
/*
if (storage->frame.current_rt) {
glBindFramebuffer(GL_FRAMEBUFFER, storage->frame.current_rt->fbo);
glColorMask(1, 1, 1, 1);
}
*/
reset_canvas();
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, storage->resources.white_tex);
glVertexAttrib4f(VS::ARRAY_COLOR, 1, 1, 1, 1);
glDisableVertexAttribArray(VS::ARRAY_COLOR);
// set up default uniforms
Transform canvas_transform;
if (storage->frame.current_rt) {
float csy = 1.0;
if (storage->frame.current_rt && storage->frame.current_rt->flags[RasterizerStorage::RENDER_TARGET_VFLIP]) {
csy = -1.0;
}
canvas_transform.translate(-(storage->frame.current_rt->width / 2.0f), -(storage->frame.current_rt->height / 2.0f), 0.0f);
canvas_transform.scale(Vector3(2.0f / storage->frame.current_rt->width, csy * -2.0f / storage->frame.current_rt->height, 1.0f));
} else {
Vector2 ssize = OS::get_singleton()->get_window_size();
canvas_transform.translate(-(ssize.width / 2.0f), -(ssize.height / 2.0f), 0.0f);
canvas_transform.scale(Vector3(2.0f / ssize.width, -2.0f / ssize.height, 1.0f));
}
state.uniforms.projection_matrix = canvas_transform;
state.uniforms.final_modulate = Color(1, 1, 1, 1);
state.uniforms.modelview_matrix = Transform2D();
state.uniforms.extra_matrix = Transform2D();
_set_uniforms();
_bind_quad_buffer();
}
void RasterizerCanvasBaseGLES2::canvas_end() {
glBindBuffer(GL_ARRAY_BUFFER, 0);
for (int i = 0; i < VS::ARRAY_MAX; i++) {
glDisableVertexAttribArray(i);
}
if (storage->frame.current_rt && storage->frame.current_rt->flags[RasterizerStorage::RENDER_TARGET_DIRECT_TO_SCREEN]) {
//reset viewport to full window size
int viewport_width = OS::get_singleton()->get_window_size().width;
int viewport_height = OS::get_singleton()->get_window_size().height;
glViewport(0, 0, viewport_width, viewport_height);
glScissor(0, 0, viewport_width, viewport_height);
}
state.using_texture_rect = false;
state.using_skeleton = false;
state.using_ninepatch = false;
state.using_transparent_rt = false;
}
void RasterizerCanvasBaseGLES2::draw_generic_textured_rect(const Rect2 &p_rect, const Rect2 &p_src) {
state.canvas_shader.set_uniform(CanvasShaderGLES2::DST_RECT, Color(p_rect.position.x, p_rect.position.y, p_rect.size.x, p_rect.size.y));
state.canvas_shader.set_uniform(CanvasShaderGLES2::SRC_RECT, Color(p_src.position.x, p_src.position.y, p_src.size.x, p_src.size.y));
glDrawArrays(GL_TRIANGLE_FAN, 0, 4);
}
Unified GLES2 / GLES3 Batching Batching is mostly separated into a common template which can be used with multiple backends (GLES2 and GLES3 here). Only necessary specifics are in the backend files. Batching is extended to cover more primitives.
2020-10-14 07:32:05 +00:00
void RasterizerCanvasBaseGLES2::_set_texture_rect_mode(bool p_texture_rect, bool p_light_angle, bool p_modulate, bool p_large_vertex) {
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
// always set this directly (this could be state checked)
state.canvas_shader.set_conditional(CanvasShaderGLES2::USE_TEXTURE_RECT, p_texture_rect);
if (state.using_light_angle != p_light_angle) {
state.using_light_angle = p_light_angle;
Unified GLES2 / GLES3 Batching Batching is mostly separated into a common template which can be used with multiple backends (GLES2 and GLES3 here). Only necessary specifics are in the backend files. Batching is extended to cover more primitives.
2020-10-14 07:32:05 +00:00
state.canvas_shader.set_conditional(CanvasShaderGLES2::USE_ATTRIB_LIGHT_ANGLE, p_light_angle);
}
if (state.using_modulate != p_modulate) {
state.using_modulate = p_modulate;
state.canvas_shader.set_conditional(CanvasShaderGLES2::USE_ATTRIB_MODULATE, p_modulate);
}
if (state.using_large_vertex != p_large_vertex) {
state.using_large_vertex = p_large_vertex;
state.canvas_shader.set_conditional(CanvasShaderGLES2::USE_ATTRIB_LARGE_VERTEX, p_large_vertex);
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
}
}
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
RasterizerStorageGLES2::Texture *RasterizerCanvasBaseGLES2::_bind_canvas_texture(const RID &p_texture, const RID &p_normal_map) {
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
RasterizerStorageGLES2::Texture *tex_return = nullptr;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
if (p_texture.is_valid()) {
RasterizerStorageGLES2::Texture *texture = storage->texture_owner.getornull(p_texture);
if (!texture) {
state.current_tex = RID();
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
state.current_tex_ptr = nullptr;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glActiveTexture(GL_TEXTURE0 + storage->config.max_texture_image_units - 1);
glBindTexture(GL_TEXTURE_2D, storage->resources.white_tex);
} else {
if (texture->redraw_if_visible) {
VisualServerRaster::redraw_request();
}
texture = texture->get_ptr();
if (texture->render_target) {
texture->render_target->used_in_frame = true;
}
glActiveTexture(GL_TEXTURE0 + storage->config.max_texture_image_units - 1);
glBindTexture(GL_TEXTURE_2D, texture->tex_id);
state.current_tex = p_texture;
state.current_tex_ptr = texture;
tex_return = texture;
}
} else {
state.current_tex = RID();
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
state.current_tex_ptr = nullptr;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glActiveTexture(GL_TEXTURE0 + storage->config.max_texture_image_units - 1);
glBindTexture(GL_TEXTURE_2D, storage->resources.white_tex);
}
if (p_normal_map == state.current_normal) {
//do none
state.canvas_shader.set_uniform(CanvasShaderGLES2::USE_DEFAULT_NORMAL, state.current_normal.is_valid());
} else if (p_normal_map.is_valid()) {
RasterizerStorageGLES2::Texture *normal_map = storage->texture_owner.getornull(p_normal_map);
if (!normal_map) {
state.current_normal = RID();
glActiveTexture(GL_TEXTURE0 + storage->config.max_texture_image_units - 2);
glBindTexture(GL_TEXTURE_2D, storage->resources.normal_tex);
state.canvas_shader.set_uniform(CanvasShaderGLES2::USE_DEFAULT_NORMAL, false);
} else {
if (normal_map->redraw_if_visible) { //check before proxy, because this is usually used with proxies
VisualServerRaster::redraw_request();
}
normal_map = normal_map->get_ptr();
glActiveTexture(GL_TEXTURE0 + storage->config.max_texture_image_units - 2);
glBindTexture(GL_TEXTURE_2D, normal_map->tex_id);
state.current_normal = p_normal_map;
state.canvas_shader.set_uniform(CanvasShaderGLES2::USE_DEFAULT_NORMAL, true);
}
} else {
state.current_normal = RID();
glActiveTexture(GL_TEXTURE0 + storage->config.max_texture_image_units - 2);
glBindTexture(GL_TEXTURE_2D, storage->resources.normal_tex);
state.canvas_shader.set_uniform(CanvasShaderGLES2::USE_DEFAULT_NORMAL, false);
}
return tex_return;
}
void RasterizerCanvasBaseGLES2::draw_window_margins(int *black_margin, RID *black_image) {
Vector2 window_size = OS::get_singleton()->get_window_size();
int window_h = window_size.height;
int window_w = window_size.width;
glBindFramebuffer(GL_FRAMEBUFFER, storage->system_fbo);
glViewport(0, 0, window_size.width, window_size.height);
canvas_begin();
if (black_image[MARGIN_LEFT].is_valid()) {
_bind_canvas_texture(black_image[MARGIN_LEFT], RID());
Size2 sz(storage->texture_get_width(black_image[MARGIN_LEFT]), storage->texture_get_height(black_image[MARGIN_LEFT]));
Fixed images in black margins
2020-06-03 21:33:33 +00:00
draw_generic_textured_rect(Rect2(0, 0, black_margin[MARGIN_LEFT], window_h),
Rect2(0, 0, (float)black_margin[MARGIN_LEFT] / sz.x, (float)(window_h) / sz.y));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else if (black_margin[MARGIN_LEFT]) {
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, storage->resources.black_tex);
draw_generic_textured_rect(Rect2(0, 0, black_margin[MARGIN_LEFT], window_h), Rect2(0, 0, 1, 1));
}
if (black_image[MARGIN_RIGHT].is_valid()) {
_bind_canvas_texture(black_image[MARGIN_RIGHT], RID());
Size2 sz(storage->texture_get_width(black_image[MARGIN_RIGHT]), storage->texture_get_height(black_image[MARGIN_RIGHT]));
Fixed images in black margins
2020-06-03 21:33:33 +00:00
draw_generic_textured_rect(Rect2(window_w - black_margin[MARGIN_RIGHT], 0, black_margin[MARGIN_RIGHT], window_h),
Rect2(0, 0, (float)black_margin[MARGIN_RIGHT] / sz.x, (float)window_h / sz.y));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else if (black_margin[MARGIN_RIGHT]) {
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, storage->resources.black_tex);
draw_generic_textured_rect(Rect2(window_w - black_margin[MARGIN_RIGHT], 0, black_margin[MARGIN_RIGHT], window_h), Rect2(0, 0, 1, 1));
}
if (black_image[MARGIN_TOP].is_valid()) {
_bind_canvas_texture(black_image[MARGIN_TOP], RID());
Size2 sz(storage->texture_get_width(black_image[MARGIN_TOP]), storage->texture_get_height(black_image[MARGIN_TOP]));
Fixed images in black margins
2020-06-03 21:33:33 +00:00
draw_generic_textured_rect(Rect2(0, 0, window_w, black_margin[MARGIN_TOP]),
Rect2(0, 0, (float)window_w / sz.x, (float)black_margin[MARGIN_TOP] / sz.y));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else if (black_margin[MARGIN_TOP]) {
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, storage->resources.black_tex);
draw_generic_textured_rect(Rect2(0, 0, window_w, black_margin[MARGIN_TOP]), Rect2(0, 0, 1, 1));
}
if (black_image[MARGIN_BOTTOM].is_valid()) {
_bind_canvas_texture(black_image[MARGIN_BOTTOM], RID());
Size2 sz(storage->texture_get_width(black_image[MARGIN_BOTTOM]), storage->texture_get_height(black_image[MARGIN_BOTTOM]));
Fixed images in black margins
2020-06-03 21:33:33 +00:00
draw_generic_textured_rect(Rect2(0, window_h - black_margin[MARGIN_BOTTOM], window_w, black_margin[MARGIN_BOTTOM]),
Rect2(0, 0, (float)window_w / sz.x, (float)black_margin[MARGIN_BOTTOM] / sz.y));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else if (black_margin[MARGIN_BOTTOM]) {
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, storage->resources.black_tex);
draw_generic_textured_rect(Rect2(0, window_h - black_margin[MARGIN_BOTTOM], window_w, black_margin[MARGIN_BOTTOM]), Rect2(0, 0, 1, 1));
}
canvas_end();
}
void RasterizerCanvasBaseGLES2::_bind_quad_buffer() {
glBindBuffer(GL_ARRAY_BUFFER, data.canvas_quad_vertices);
glEnableVertexAttribArray(VS::ARRAY_VERTEX);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glVertexAttribPointer(VS::ARRAY_VERTEX, 2, GL_FLOAT, GL_FALSE, 0, nullptr);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
}
void RasterizerCanvasBaseGLES2::_set_uniforms() {
state.canvas_shader.set_uniform(CanvasShaderGLES2::PROJECTION_MATRIX, state.uniforms.projection_matrix);
state.canvas_shader.set_uniform(CanvasShaderGLES2::MODELVIEW_MATRIX, state.uniforms.modelview_matrix);
state.canvas_shader.set_uniform(CanvasShaderGLES2::EXTRA_MATRIX, state.uniforms.extra_matrix);
state.canvas_shader.set_uniform(CanvasShaderGLES2::FINAL_MODULATE, state.uniforms.final_modulate);
state.canvas_shader.set_uniform(CanvasShaderGLES2::TIME, storage->frame.time[0]);
if (storage->frame.current_rt) {
Vector2 screen_pixel_size;
screen_pixel_size.x = 1.0 / storage->frame.current_rt->width;
screen_pixel_size.y = 1.0 / storage->frame.current_rt->height;
state.canvas_shader.set_uniform(CanvasShaderGLES2::SCREEN_PIXEL_SIZE, screen_pixel_size);
}
if (state.using_skeleton) {
state.canvas_shader.set_uniform(CanvasShaderGLES2::SKELETON_TRANSFORM, state.skeleton_transform);
state.canvas_shader.set_uniform(CanvasShaderGLES2::SKELETON_TRANSFORM_INVERSE, state.skeleton_transform_inverse);
state.canvas_shader.set_uniform(CanvasShaderGLES2::SKELETON_TEXTURE_SIZE, state.skeleton_texture_size);
}
if (state.using_light) {
Light *light = state.using_light;
state.canvas_shader.set_uniform(CanvasShaderGLES2::LIGHT_MATRIX, light->light_shader_xform);
Transform2D basis_inverse = light->light_shader_xform.affine_inverse().orthonormalized();
basis_inverse[2] = Vector2();
state.canvas_shader.set_uniform(CanvasShaderGLES2::LIGHT_MATRIX_INVERSE, basis_inverse);
state.canvas_shader.set_uniform(CanvasShaderGLES2::LIGHT_LOCAL_MATRIX, light->xform_cache.affine_inverse());
state.canvas_shader.set_uniform(CanvasShaderGLES2::LIGHT_COLOR, light->color * light->energy);
state.canvas_shader.set_uniform(CanvasShaderGLES2::LIGHT_POS, light->light_shader_pos);
state.canvas_shader.set_uniform(CanvasShaderGLES2::LIGHT_HEIGHT, light->height);
state.canvas_shader.set_uniform(CanvasShaderGLES2::LIGHT_OUTSIDE_ALPHA, light->mode == VS::CANVAS_LIGHT_MODE_MASK ? 1.0 : 0.0);
if (state.using_shadow) {
RasterizerStorageGLES2::CanvasLightShadow *cls = storage->canvas_light_shadow_owner.get(light->shadow_buffer);
glActiveTexture(GL_TEXTURE0 + storage->config.max_texture_image_units - 5);
glBindTexture(GL_TEXTURE_2D, cls->distance);
state.canvas_shader.set_uniform(CanvasShaderGLES2::SHADOW_MATRIX, light->shadow_matrix_cache);
state.canvas_shader.set_uniform(CanvasShaderGLES2::LIGHT_SHADOW_COLOR, light->shadow_color);
state.canvas_shader.set_uniform(CanvasShaderGLES2::SHADOWPIXEL_SIZE, (1.0 / light->shadow_buffer_size) * (1.0 + light->shadow_smooth));
if (light->radius_cache == 0) {
state.canvas_shader.set_uniform(CanvasShaderGLES2::SHADOW_GRADIENT, 0.0);
} else {
state.canvas_shader.set_uniform(CanvasShaderGLES2::SHADOW_GRADIENT, light->shadow_gradient_length / (light->radius_cache * 1.1));
}
state.canvas_shader.set_uniform(CanvasShaderGLES2::SHADOW_DISTANCE_MULT, light->radius_cache * 1.1);
/*canvas_shader.set_uniform(CanvasShaderGLES2::SHADOW_MATRIX,light->shadow_matrix_cache);
canvas_shader.set_uniform(CanvasShaderGLES2::SHADOW_ESM_MULTIPLIER,light->shadow_esm_mult);
canvas_shader.set_uniform(CanvasShaderGLES2::LIGHT_SHADOW_COLOR,light->shadow_color);*/
}
}
}
void RasterizerCanvasBaseGLES2::reset_canvas() {
glDisable(GL_CULL_FACE);
glDisable(GL_DEPTH_TEST);
glDisable(GL_SCISSOR_TEST);
glDisable(GL_DITHER);
glEnable(GL_BLEND);
if (storage->frame.current_rt && storage->frame.current_rt->flags[RasterizerStorage::RENDER_TARGET_TRANSPARENT]) {
glBlendFuncSeparate(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA, GL_ONE, GL_ONE_MINUS_SRC_ALPHA);
} else {
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
}
// bind the back buffer to a texture so shaders can use it.
// It should probably use texture unit -3 (as GLES2 does as well) but currently that's buggy.
// keeping this for now as there's nothing else that uses texture unit 2
// TODO ^
if (storage->frame.current_rt) {
// glActiveTexture(GL_TEXTURE0 + 2);
// glBindTexture(GL_TEXTURE_2D, storage->frame.current_rt->copy_screen_effect.color);
}
glBindBuffer(GL_ARRAY_BUFFER, 0);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0);
}
void RasterizerCanvasBaseGLES2::canvas_debug_viewport_shadows(Light *p_lights_with_shadow) {
}
void RasterizerCanvasBaseGLES2::_copy_texscreen(const Rect2 &p_rect) {
state.canvas_texscreen_used = true;
_copy_screen(p_rect);
// back to canvas, force rebind
state.using_texture_rect = false;
state.canvas_shader.bind();
_bind_canvas_texture(state.current_tex, state.current_normal);
_set_uniforms();
}
void RasterizerCanvasBaseGLES2::_draw_polygon(const int *p_indices, int p_index_count, int p_vertex_count, const Vector2 *p_vertices, const Vector2 *p_uvs, const Color *p_colors, bool p_singlecolor, const float *p_weights, const int *p_bones) {
glBindBuffer(GL_ARRAY_BUFFER, data.polygon_buffer);
uint32_t buffer_ofs = 0;
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
uint32_t buffer_ofs_after = buffer_ofs + (sizeof(Vector2) * p_vertex_count);
#ifdef DEBUG_ENABLED
ERR_FAIL_COND(buffer_ofs_after > data.polygon_buffer_size);
#endif
Add project settings to manually specify API usage As a result of the GLES specifications being vague about best practice for how buffers should be used dynamically, different GPUs / platforms appear to have different preferences. Mac in particular seems to have a number of problems in this area, and none of the rendering team uses Macs. So far we have relied on guesswork to choose the best usage, but in an attempt to pin this down, this PR begins to introduce manual selection of options for users to test their configurations.
2020-10-25 15:38:54 +00:00
storage->buffer_orphan_and_upload(data.polygon_buffer_size, 0, sizeof(Vector2) * p_vertex_count, p_vertices, GL_ARRAY_BUFFER, _buffer_upload_usage_flag, true);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_VERTEX);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glVertexAttribPointer(VS::ARRAY_VERTEX, 2, GL_FLOAT, GL_FALSE, sizeof(Vector2), nullptr);
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
if (p_singlecolor) {
glDisableVertexAttribArray(VS::ARRAY_COLOR);
Color m = *p_colors;
glVertexAttrib4f(VS::ARRAY_COLOR, m.r, m.g, m.b, m.a);
} else if (!p_colors) {
glDisableVertexAttribArray(VS::ARRAY_COLOR);
glVertexAttrib4f(VS::ARRAY_COLOR, 1, 1, 1, 1);
} else {
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
RAST_FAIL_COND(!storage->safe_buffer_sub_data(data.polygon_buffer_size, GL_ARRAY_BUFFER, buffer_ofs, sizeof(Color) * p_vertex_count, p_colors, buffer_ofs_after));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_COLOR);
glVertexAttribPointer(VS::ARRAY_COLOR, 4, GL_FLOAT, GL_FALSE, sizeof(Color), CAST_INT_TO_UCHAR_PTR(buffer_ofs));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
}
if (p_uvs) {
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
RAST_FAIL_COND(!storage->safe_buffer_sub_data(data.polygon_buffer_size, GL_ARRAY_BUFFER, buffer_ofs, sizeof(Vector2) * p_vertex_count, p_uvs, buffer_ofs_after));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_TEX_UV);
glVertexAttribPointer(VS::ARRAY_TEX_UV, 2, GL_FLOAT, GL_FALSE, sizeof(Vector2), CAST_INT_TO_UCHAR_PTR(buffer_ofs));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else {
glDisableVertexAttribArray(VS::ARRAY_TEX_UV);
}
if (p_weights && p_bones) {
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
RAST_FAIL_COND(!storage->safe_buffer_sub_data(data.polygon_buffer_size, GL_ARRAY_BUFFER, buffer_ofs, sizeof(float) * 4 * p_vertex_count, p_weights, buffer_ofs_after));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_WEIGHTS);
glVertexAttribPointer(VS::ARRAY_WEIGHTS, 4, GL_FLOAT, GL_FALSE, sizeof(float) * 4, CAST_INT_TO_UCHAR_PTR(buffer_ofs));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
RAST_FAIL_COND(!storage->safe_buffer_sub_data(data.polygon_buffer_size, GL_ARRAY_BUFFER, buffer_ofs, sizeof(int) * 4 * p_vertex_count, p_bones, buffer_ofs_after));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_BONES);
glVertexAttribPointer(VS::ARRAY_BONES, 4, GL_UNSIGNED_INT, GL_FALSE, sizeof(int) * 4, CAST_INT_TO_UCHAR_PTR(buffer_ofs));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else {
glDisableVertexAttribArray(VS::ARRAY_WEIGHTS);
glDisableVertexAttribArray(VS::ARRAY_BONES);
}
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, data.polygon_index_buffer);
if (storage->config.support_32_bits_indices) { //should check for
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
#ifdef DEBUG_ENABLED
ERR_FAIL_COND((sizeof(int) * p_index_count) > data.polygon_index_buffer_size);
#endif
Add project settings to manually specify API usage As a result of the GLES specifications being vague about best practice for how buffers should be used dynamically, different GPUs / platforms appear to have different preferences. Mac in particular seems to have a number of problems in this area, and none of the rendering team uses Macs. So far we have relied on guesswork to choose the best usage, but in an attempt to pin this down, this PR begins to introduce manual selection of options for users to test their configurations.
2020-10-25 15:38:54 +00:00
storage->buffer_orphan_and_upload(data.polygon_index_buffer_size, 0, sizeof(int) * p_index_count, p_indices, GL_ELEMENT_ARRAY_BUFFER, _buffer_upload_usage_flag, true);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glDrawElements(GL_TRIANGLES, p_index_count, GL_UNSIGNED_INT, nullptr);
GLES3 - add counts for 2d drawcall and 2d items to performance monitor This had already been implemented for GLES2 but not GLES3.
2020-05-07 09:16:07 +00:00
storage->info.render._2d_draw_call_count++;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else {
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
#ifdef DEBUG_ENABLED
ERR_FAIL_COND((sizeof(uint16_t) * p_index_count) > data.polygon_index_buffer_size);
#endif
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
uint16_t *index16 = (uint16_t *)alloca(sizeof(uint16_t) * p_index_count);
for (int i = 0; i < p_index_count; i++) {
index16[i] = uint16_t(p_indices[i]);
}
Add project settings to manually specify API usage As a result of the GLES specifications being vague about best practice for how buffers should be used dynamically, different GPUs / platforms appear to have different preferences. Mac in particular seems to have a number of problems in this area, and none of the rendering team uses Macs. So far we have relied on guesswork to choose the best usage, but in an attempt to pin this down, this PR begins to introduce manual selection of options for users to test their configurations.
2020-10-25 15:38:54 +00:00
storage->buffer_orphan_and_upload(data.polygon_index_buffer_size, 0, sizeof(uint16_t) * p_index_count, index16, GL_ELEMENT_ARRAY_BUFFER, _buffer_upload_usage_flag, true);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glDrawElements(GL_TRIANGLES, p_index_count, GL_UNSIGNED_SHORT, nullptr);
GLES3 - add counts for 2d drawcall and 2d items to performance monitor This had already been implemented for GLES2 but not GLES3.
2020-05-07 09:16:07 +00:00
storage->info.render._2d_draw_call_count++;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
}
glBindBuffer(GL_ARRAY_BUFFER, 0);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0);
}
void RasterizerCanvasBaseGLES2::_draw_generic(GLuint p_primitive, int p_vertex_count, const Vector2 *p_vertices, const Vector2 *p_uvs, const Color *p_colors, bool p_singlecolor) {
glBindBuffer(GL_ARRAY_BUFFER, data.polygon_buffer);
uint32_t buffer_ofs = 0;
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
uint32_t buffer_ofs_after = buffer_ofs + (sizeof(Vector2) * p_vertex_count);
#ifdef DEBUG_ENABLED
ERR_FAIL_COND(buffer_ofs_after > data.polygon_buffer_size);
#endif
Add project settings to manually specify API usage As a result of the GLES specifications being vague about best practice for how buffers should be used dynamically, different GPUs / platforms appear to have different preferences. Mac in particular seems to have a number of problems in this area, and none of the rendering team uses Macs. So far we have relied on guesswork to choose the best usage, but in an attempt to pin this down, this PR begins to introduce manual selection of options for users to test their configurations.
2020-10-25 15:38:54 +00:00
storage->buffer_orphan_and_upload(data.polygon_buffer_size, 0, sizeof(Vector2) * p_vertex_count, p_vertices, GL_ARRAY_BUFFER, _buffer_upload_usage_flag, true);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_VERTEX);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glVertexAttribPointer(VS::ARRAY_VERTEX, 2, GL_FLOAT, GL_FALSE, sizeof(Vector2), nullptr);
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
if (p_singlecolor) {
glDisableVertexAttribArray(VS::ARRAY_COLOR);
Color m = *p_colors;
glVertexAttrib4f(VS::ARRAY_COLOR, m.r, m.g, m.b, m.a);
} else if (!p_colors) {
glDisableVertexAttribArray(VS::ARRAY_COLOR);
glVertexAttrib4f(VS::ARRAY_COLOR, 1, 1, 1, 1);
} else {
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
RAST_FAIL_COND(!storage->safe_buffer_sub_data(data.polygon_buffer_size, GL_ARRAY_BUFFER, buffer_ofs, sizeof(Color) * p_vertex_count, p_colors, buffer_ofs_after));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_COLOR);
glVertexAttribPointer(VS::ARRAY_COLOR, 4, GL_FLOAT, GL_FALSE, sizeof(Color), CAST_INT_TO_UCHAR_PTR(buffer_ofs));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
}
if (p_uvs) {
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
RAST_FAIL_COND(!storage->safe_buffer_sub_data(data.polygon_buffer_size, GL_ARRAY_BUFFER, buffer_ofs, sizeof(Vector2) * p_vertex_count, p_uvs, buffer_ofs_after));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_TEX_UV);
glVertexAttribPointer(VS::ARRAY_TEX_UV, 2, GL_FLOAT, GL_FALSE, sizeof(Vector2), CAST_INT_TO_UCHAR_PTR(buffer_ofs));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else {
glDisableVertexAttribArray(VS::ARRAY_TEX_UV);
}
glDrawArrays(p_primitive, 0, p_vertex_count);
GLES3 - add counts for 2d drawcall and 2d items to performance monitor This had already been implemented for GLES2 but not GLES3.
2020-05-07 09:16:07 +00:00
storage->info.render._2d_draw_call_count++;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glBindBuffer(GL_ARRAY_BUFFER, 0);
}
void RasterizerCanvasBaseGLES2::_draw_generic_indices(GLuint p_primitive, const int *p_indices, int p_index_count, int p_vertex_count, const Vector2 *p_vertices, const Vector2 *p_uvs, const Color *p_colors, bool p_singlecolor) {
glBindBuffer(GL_ARRAY_BUFFER, data.polygon_buffer);
uint32_t buffer_ofs = 0;
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
uint32_t buffer_ofs_after = buffer_ofs + (sizeof(Vector2) * p_vertex_count);
#ifdef DEBUG_ENABLED
ERR_FAIL_COND(buffer_ofs_after > data.polygon_buffer_size);
#endif
Add project settings to manually specify API usage As a result of the GLES specifications being vague about best practice for how buffers should be used dynamically, different GPUs / platforms appear to have different preferences. Mac in particular seems to have a number of problems in this area, and none of the rendering team uses Macs. So far we have relied on guesswork to choose the best usage, but in an attempt to pin this down, this PR begins to introduce manual selection of options for users to test their configurations.
2020-10-25 15:38:54 +00:00
storage->buffer_orphan_and_upload(data.polygon_buffer_size, 0, sizeof(Vector2) * p_vertex_count, p_vertices, GL_ARRAY_BUFFER, _buffer_upload_usage_flag, true);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_VERTEX);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glVertexAttribPointer(VS::ARRAY_VERTEX, 2, GL_FLOAT, GL_FALSE, sizeof(Vector2), nullptr);
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
if (p_singlecolor) {
glDisableVertexAttribArray(VS::ARRAY_COLOR);
Color m = *p_colors;
glVertexAttrib4f(VS::ARRAY_COLOR, m.r, m.g, m.b, m.a);
} else if (!p_colors) {
glDisableVertexAttribArray(VS::ARRAY_COLOR);
glVertexAttrib4f(VS::ARRAY_COLOR, 1, 1, 1, 1);
} else {
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
RAST_FAIL_COND(!storage->safe_buffer_sub_data(data.polygon_buffer_size, GL_ARRAY_BUFFER, buffer_ofs, sizeof(Color) * p_vertex_count, p_colors, buffer_ofs_after));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_COLOR);
glVertexAttribPointer(VS::ARRAY_COLOR, 4, GL_FLOAT, GL_FALSE, sizeof(Color), CAST_INT_TO_UCHAR_PTR(buffer_ofs));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
}
if (p_uvs) {
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
RAST_FAIL_COND(!storage->safe_buffer_sub_data(data.polygon_buffer_size, GL_ARRAY_BUFFER, buffer_ofs, sizeof(Vector2) * p_vertex_count, p_uvs, buffer_ofs_after));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glEnableVertexAttribArray(VS::ARRAY_TEX_UV);
glVertexAttribPointer(VS::ARRAY_TEX_UV, 2, GL_FLOAT, GL_FALSE, sizeof(Vector2), CAST_INT_TO_UCHAR_PTR(buffer_ofs));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
buffer_ofs = buffer_ofs_after;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else {
glDisableVertexAttribArray(VS::ARRAY_TEX_UV);
}
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
#ifdef RASTERIZER_EXTRA_CHECKS
// very slow, do not enable in normal use
for (int n = 0; n < p_index_count; n++) {
RAST_DEV_DEBUG_ASSERT(p_indices[n] < p_vertex_count);
}
#endif
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, data.polygon_index_buffer);
if (storage->config.support_32_bits_indices) { //should check for
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
#ifdef DEBUG_ENABLED
ERR_FAIL_COND((sizeof(int) * p_index_count) > data.polygon_index_buffer_size);
#endif
Add project settings to manually specify API usage As a result of the GLES specifications being vague about best practice for how buffers should be used dynamically, different GPUs / platforms appear to have different preferences. Mac in particular seems to have a number of problems in this area, and none of the rendering team uses Macs. So far we have relied on guesswork to choose the best usage, but in an attempt to pin this down, this PR begins to introduce manual selection of options for users to test their configurations.
2020-10-25 15:38:54 +00:00
storage->buffer_orphan_and_upload(data.polygon_index_buffer_size, 0, sizeof(int) * p_index_count, p_indices, GL_ELEMENT_ARRAY_BUFFER, _buffer_upload_usage_flag, true);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glDrawElements(p_primitive, p_index_count, GL_UNSIGNED_INT, nullptr);
GLES3 - add counts for 2d drawcall and 2d items to performance monitor This had already been implemented for GLES2 but not GLES3.
2020-05-07 09:16:07 +00:00
storage->info.render._2d_draw_call_count++;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
} else {
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
#ifdef DEBUG_ENABLED
ERR_FAIL_COND((sizeof(uint16_t) * p_index_count) > data.polygon_index_buffer_size);
#endif
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
uint16_t *index16 = (uint16_t *)alloca(sizeof(uint16_t) * p_index_count);
for (int i = 0; i < p_index_count; i++) {
index16[i] = uint16_t(p_indices[i]);
}
Add project settings to manually specify API usage As a result of the GLES specifications being vague about best practice for how buffers should be used dynamically, different GPUs / platforms appear to have different preferences. Mac in particular seems to have a number of problems in this area, and none of the rendering team uses Macs. So far we have relied on guesswork to choose the best usage, but in an attempt to pin this down, this PR begins to introduce manual selection of options for users to test their configurations.
2020-10-25 15:38:54 +00:00
storage->buffer_orphan_and_upload(data.polygon_index_buffer_size, 0, sizeof(uint16_t) * p_index_count, index16, GL_ELEMENT_ARRAY_BUFFER, _buffer_upload_usage_flag, true);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glDrawElements(p_primitive, p_index_count, GL_UNSIGNED_SHORT, nullptr);
GLES3 - add counts for 2d drawcall and 2d items to performance monitor This had already been implemented for GLES2 but not GLES3.
2020-05-07 09:16:07 +00:00
storage->info.render._2d_draw_call_count++;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
}
glBindBuffer(GL_ARRAY_BUFFER, 0);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0);
}
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
void RasterizerCanvasBaseGLES2::_draw_gui_primitive(int p_points, const Vector2 *p_vertices, const Color *p_colors, const Vector2 *p_uvs, const float *p_light_angles) {
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
static const GLenum prim[5] = { GL_POINTS, GL_POINTS, GL_LINES, GL_TRIANGLES, GL_TRIANGLE_FAN };
int color_offset = 0;
int uv_offset = 0;
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
int light_angle_offset = 0;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
int stride = 2;
if (p_colors) {
color_offset = stride;
stride += 4;
}
if (p_uvs) {
uv_offset = stride;
stride += 2;
}
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
if (p_light_angles) { //light_angles
light_angle_offset = stride;
stride += 1;
}
Add DEV_ASSERT and DEV_CHECK macros Change the existing DEV_ASSERT function to be switched on and off by the DEV_ENABLED define. DEV_ASSERT breaks into the debugger as soon as hit. Add error macros DEV_CHECK and DEV_CHECK_ONCE to add an alternative check that ERR_PRINT when a condition fails, again only enabled in DEV_ENABLED builds.
2021-10-04 13:13:43 +00:00
DEV_ASSERT(p_points <= 4);
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
float buffer_data[(2 + 2 + 4 + 1) * 4];
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
for (int i = 0; i < p_points; i++) {
buffer_data[stride * i + 0] = p_vertices[i].x;
buffer_data[stride * i + 1] = p_vertices[i].y;
}
if (p_colors) {
for (int i = 0; i < p_points; i++) {
buffer_data[stride * i + color_offset + 0] = p_colors[i].r;
buffer_data[stride * i + color_offset + 1] = p_colors[i].g;
buffer_data[stride * i + color_offset + 2] = p_colors[i].b;
buffer_data[stride * i + color_offset + 3] = p_colors[i].a;
}
}
if (p_uvs) {
for (int i = 0; i < p_points; i++) {
buffer_data[stride * i + uv_offset + 0] = p_uvs[i].x;
buffer_data[stride * i + uv_offset + 1] = p_uvs[i].y;
}
}
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
if (p_light_angles) {
for (int i = 0; i < p_points; i++) {
buffer_data[stride * i + light_angle_offset + 0] = p_light_angles[i];
}
}
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glBindBuffer(GL_ARRAY_BUFFER, data.polygon_buffer);
Add project settings to manually specify API usage As a result of the GLES specifications being vague about best practice for how buffers should be used dynamically, different GPUs / platforms appear to have different preferences. Mac in particular seems to have a number of problems in this area, and none of the rendering team uses Macs. So far we have relied on guesswork to choose the best usage, but in an attempt to pin this down, this PR begins to introduce manual selection of options for users to test their configurations.
2020-10-25 15:38:54 +00:00
storage->buffer_orphan_and_upload(data.polygon_buffer_size, 0, p_points * stride * 4 * sizeof(float), buffer_data, GL_ARRAY_BUFFER, _buffer_upload_usage_flag, true);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glVertexAttribPointer(VS::ARRAY_VERTEX, 2, GL_FLOAT, GL_FALSE, stride * sizeof(float), nullptr);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
if (p_colors) {
glVertexAttribPointer(VS::ARRAY_COLOR, 4, GL_FLOAT, GL_FALSE, stride * sizeof(float), CAST_INT_TO_UCHAR_PTR(color_offset * sizeof(float)));
glEnableVertexAttribArray(VS::ARRAY_COLOR);
}
if (p_uvs) {
glVertexAttribPointer(VS::ARRAY_TEX_UV, 2, GL_FLOAT, GL_FALSE, stride * sizeof(float), CAST_INT_TO_UCHAR_PTR(uv_offset * sizeof(float)));
glEnableVertexAttribArray(VS::ARRAY_TEX_UV);
}
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
if (p_light_angles) {
glVertexAttribPointer(VS::ARRAY_TANGENT, 1, GL_FLOAT, GL_FALSE, stride * sizeof(float), CAST_INT_TO_UCHAR_PTR(light_angle_offset * sizeof(float)));
glEnableVertexAttribArray(VS::ARRAY_TANGENT);
}
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glDrawArrays(prim[p_points], 0, p_points);
GLES3 - add counts for 2d drawcall and 2d items to performance monitor This had already been implemented for GLES2 but not GLES3.
2020-05-07 09:16:07 +00:00
storage->info.render._2d_draw_call_count++;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
if (p_light_angles) {
// may not be needed
glDisableVertexAttribArray(VS::ARRAY_TANGENT);
}
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glBindBuffer(GL_ARRAY_BUFFER, 0);
}
void RasterizerCanvasBaseGLES2::_copy_screen(const Rect2 &p_rect) {
if (storage->frame.current_rt->flags[RasterizerStorage::RENDER_TARGET_DIRECT_TO_SCREEN]) {
ERR_PRINT_ONCE("Cannot use screen texture copying in render target set to render direct to screen.");
return;
}
Improve the error message about `SCREEN_TEXTURE` being unavailable This also tweaks the Viewport `usage` property hint to match the project setting used for the root viewport.
2021-09-11 21:17:07 +00:00
ERR_FAIL_COND_MSG(storage->frame.current_rt->copy_screen_effect.color == 0, "Can't use screen texture copying in a render target configured without copy buffers. To resolve this, change the viewport's Usage property to \"2D\" or \"3D\" instead of \"2D Without Sampling\" or \"3D Without Effects\" respectively.");
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glDisable(GL_BLEND);
Vector2 wh(storage->frame.current_rt->width, storage->frame.current_rt->height);
Color copy_section(p_rect.position.x / wh.x, p_rect.position.y / wh.y, p_rect.size.x / wh.x, p_rect.size.y / wh.y);
if (p_rect != Rect2()) {
storage->shaders.copy.set_conditional(CopyShaderGLES2::USE_COPY_SECTION, true);
}
storage->shaders.copy.set_conditional(CopyShaderGLES2::USE_NO_ALPHA, !state.using_transparent_rt);
glBindFramebuffer(GL_FRAMEBUFFER, storage->frame.current_rt->copy_screen_effect.fbo);
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, storage->frame.current_rt->color);
storage->shaders.copy.bind();
storage->shaders.copy.set_uniform(CopyShaderGLES2::COPY_SECTION, copy_section);
const Vector2 vertpos[4] = {
Vector2(-1, -1),
Vector2(-1, 1),
Vector2(1, 1),
Vector2(1, -1),
};
const Vector2 uvpos[4] = {
Vector2(0, 0),
Vector2(0, 1),
Vector2(1, 1),
Vector2(1, 0)
};
const int indexpos[6] = {
0, 1, 2,
2, 3, 0
};
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
_draw_polygon(indexpos, 6, 4, vertpos, uvpos, nullptr, false);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
storage->shaders.copy.set_conditional(CopyShaderGLES2::USE_COPY_SECTION, false);
storage->shaders.copy.set_conditional(CopyShaderGLES2::USE_NO_ALPHA, false);
glBindFramebuffer(GL_FRAMEBUFFER, storage->frame.current_rt->fbo); //back to front
glEnable(GL_BLEND);
}
void RasterizerCanvasBaseGLES2::canvas_light_shadow_buffer_update(RID p_buffer, const Transform2D &p_light_xform, int p_light_mask, float p_near, float p_far, LightOccluderInstance *p_occluders, CameraMatrix *p_xform_cache) {
RasterizerStorageGLES2::CanvasLightShadow *cls = storage->canvas_light_shadow_owner.get(p_buffer);
ERR_FAIL_COND(!cls);
glDisable(GL_BLEND);
glDisable(GL_SCISSOR_TEST);
glDisable(GL_DITHER);
glDisable(GL_CULL_FACE);
glDepthFunc(GL_LEQUAL);
glEnable(GL_DEPTH_TEST);
glDepthMask(true);
glBindFramebuffer(GL_FRAMEBUFFER, cls->fbo);
state.canvas_shadow_shader.set_conditional(CanvasShadowShaderGLES2::USE_RGBA_SHADOWS, storage->config.use_rgba_2d_shadows);
state.canvas_shadow_shader.bind();
glViewport(0, 0, cls->size, cls->height);
glClearDepth(1.0f);
glClearColor(1, 1, 1, 1);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
VS::CanvasOccluderPolygonCullMode cull = VS::CANVAS_OCCLUDER_POLYGON_CULL_DISABLED;
for (int i = 0; i < 4; i++) {
//make sure it remains orthogonal, makes easy to read angle later
Transform light;
light.origin[0] = p_light_xform[2][0];
light.origin[1] = p_light_xform[2][1];
light.basis[0][0] = p_light_xform[0][0];
light.basis[0][1] = p_light_xform[1][0];
light.basis[1][0] = p_light_xform[0][1];
light.basis[1][1] = p_light_xform[1][1];
//light.basis.scale(Vector3(to_light.elements[0].length(),to_light.elements[1].length(),1));
//p_near=1;
CameraMatrix projection;
{
real_t fov = 90;
real_t nearp = p_near;
real_t farp = p_far;
real_t aspect = 1.0;
real_t ymax = nearp * Math::tan(Math::deg2rad(fov * 0.5));
real_t ymin = -ymax;
real_t xmin = ymin * aspect;
real_t xmax = ymax * aspect;
projection.set_frustum(xmin, xmax, ymin, ymax, nearp, farp);
}
Vector3 cam_target = Basis(Vector3(0, 0, Math_PI * 2 * (i / 4.0))).xform(Vector3(0, 1, 0));
projection = projection * CameraMatrix(Transform().looking_at(cam_target, Vector3(0, 0, -1)).affine_inverse());
state.canvas_shadow_shader.set_uniform(CanvasShadowShaderGLES2::PROJECTION_MATRIX, projection);
state.canvas_shadow_shader.set_uniform(CanvasShadowShaderGLES2::LIGHT_MATRIX, light);
state.canvas_shadow_shader.set_uniform(CanvasShadowShaderGLES2::DISTANCE_NORM, 1.0 / p_far);
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2021-05-05 10:44:11 +00:00
if (i == 0) {
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
*p_xform_cache = projection;
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2021-05-05 10:44:11 +00:00
}
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glViewport(0, (cls->height / 4) * i, cls->size, cls->height / 4);
LightOccluderInstance *instance = p_occluders;
while (instance) {
RasterizerStorageGLES2::CanvasOccluder *cc = storage->canvas_occluder_owner.getornull(instance->polygon_buffer);
if (!cc || cc->len == 0 || !(p_light_mask & instance->light_mask)) {
instance = instance->next;
continue;
}
state.canvas_shadow_shader.set_uniform(CanvasShadowShaderGLES2::WORLD_MATRIX, instance->xform_cache);
VS::CanvasOccluderPolygonCullMode transformed_cull_cache = instance->cull_cache;
if (transformed_cull_cache != VS::CANVAS_OCCLUDER_POLYGON_CULL_DISABLED &&
(p_light_xform.basis_determinant() * instance->xform_cache.basis_determinant()) < 0) {
clang-format: Enable `BreakBeforeTernaryOperators` clang-format keeps breaking the way it handles break *after* ternary operators, so I give up and go with the only style they seem to actually test.
2021-10-28 12:35:34 +00:00
transformed_cull_cache = (transformed_cull_cache == VS::CANVAS_OCCLUDER_POLYGON_CULL_CLOCKWISE)
? VS::CANVAS_OCCLUDER_POLYGON_CULL_COUNTER_CLOCKWISE
: VS::CANVAS_OCCLUDER_POLYGON_CULL_CLOCKWISE;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
}
if (cull != transformed_cull_cache) {
cull = transformed_cull_cache;
switch (cull) {
case VS::CANVAS_OCCLUDER_POLYGON_CULL_DISABLED: {
glDisable(GL_CULL_FACE);
} break;
case VS::CANVAS_OCCLUDER_POLYGON_CULL_CLOCKWISE: {
glEnable(GL_CULL_FACE);
glCullFace(GL_FRONT);
} break;
case VS::CANVAS_OCCLUDER_POLYGON_CULL_COUNTER_CLOCKWISE: {
glEnable(GL_CULL_FACE);
glCullFace(GL_BACK);
} break;
}
}
glBindBuffer(GL_ARRAY_BUFFER, cc->vertex_id);
glEnableVertexAttribArray(VS::ARRAY_VERTEX);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glVertexAttribPointer(VS::ARRAY_VERTEX, 3, GL_FLOAT, false, 0, nullptr);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, cc->index_id);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glDrawElements(GL_TRIANGLES, cc->len * 3, GL_UNSIGNED_SHORT, nullptr);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
instance = instance->next;
}
}
glBindBuffer(GL_ARRAY_BUFFER, 0);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0);
}
void RasterizerCanvasBaseGLES2::draw_lens_distortion_rect(const Rect2 &p_rect, float p_k1, float p_k2, const Vector2 &p_eye_center, float p_oversample) {
Vector2 half_size;
if (storage->frame.current_rt) {
half_size = Vector2(storage->frame.current_rt->width, storage->frame.current_rt->height);
} else {
half_size = OS::get_singleton()->get_window_size();
}
half_size *= 0.5;
Vector2 offset((p_rect.position.x - half_size.x) / half_size.x, (p_rect.position.y - half_size.y) / half_size.y);
Vector2 scale(p_rect.size.x / half_size.x, p_rect.size.y / half_size.y);
float aspect_ratio = p_rect.size.x / p_rect.size.y;
// setup our lens shader
state.lens_shader.bind();
state.lens_shader.set_uniform(LensDistortedShaderGLES2::OFFSET, offset);
state.lens_shader.set_uniform(LensDistortedShaderGLES2::SCALE, scale);
state.lens_shader.set_uniform(LensDistortedShaderGLES2::K1, p_k1);
state.lens_shader.set_uniform(LensDistortedShaderGLES2::K2, p_k2);
state.lens_shader.set_uniform(LensDistortedShaderGLES2::EYE_CENTER, p_eye_center);
state.lens_shader.set_uniform(LensDistortedShaderGLES2::UPSCALE, p_oversample);
state.lens_shader.set_uniform(LensDistortedShaderGLES2::ASPECT_RATIO, aspect_ratio);
// bind our quad buffer
_bind_quad_buffer();
// and draw
glDrawArrays(GL_TRIANGLE_FAN, 0, 4);
// and cleanup
glBindBuffer(GL_ARRAY_BUFFER, 0);
for (int i = 0; i < VS::ARRAY_MAX; i++) {
glDisableVertexAttribArray(i);
}
}
void RasterizerCanvasBaseGLES2::initialize() {
Add buffer orphan / stream options Allows users to override default API usage, in order to get best performance on different platforms. Also changes the default legacy flags to use STREAM rather than DYNAMIC.
2021-04-13 16:34:27 +00:00
int flag_stream_mode = GLOBAL_GET("rendering/2d/opengl/legacy_stream");
switch (flag_stream_mode) {
default: {
_buffer_upload_usage_flag = GL_STREAM_DRAW;
} break;
case 1: {
_buffer_upload_usage_flag = GL_DYNAMIC_DRAW;
} break;
case 2: {
_buffer_upload_usage_flag = GL_STREAM_DRAW;
} break;
}
Add project settings to manually specify API usage As a result of the GLES specifications being vague about best practice for how buffers should be used dynamically, different GPUs / platforms appear to have different preferences. Mac in particular seems to have a number of problems in this area, and none of the rendering team uses Macs. So far we have relied on guesswork to choose the best usage, but in an attempt to pin this down, this PR begins to introduce manual selection of options for users to test their configurations.
2020-10-25 15:38:54 +00:00
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
// quad buffer
{
glGenBuffers(1, &data.canvas_quad_vertices);
glBindBuffer(GL_ARRAY_BUFFER, data.canvas_quad_vertices);
const float qv[8] = {
0, 0,
0, 1,
1, 1,
1, 0
};
glBufferData(GL_ARRAY_BUFFER, sizeof(float) * 8, qv, GL_STATIC_DRAW);
glBindBuffer(GL_ARRAY_BUFFER, 0);
}
// polygon buffer
{
uint32_t poly_size = GLOBAL_DEF("rendering/limits/buffers/canvas_polygon_buffer_size_kb", 128);
ProjectSettings::get_singleton()->set_custom_property_info("rendering/limits/buffers/canvas_polygon_buffer_size_kb", PropertyInfo(Variant::INT, "rendering/limits/buffers/canvas_polygon_buffer_size_kb", PROPERTY_HINT_RANGE, "0,256,1,or_greater"));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
poly_size = MAX(poly_size, 2); // minimum 2k, may still see anomalies in editor
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
poly_size *= 1024;
glGenBuffers(1, &data.polygon_buffer);
glBindBuffer(GL_ARRAY_BUFFER, data.polygon_buffer);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glBufferData(GL_ARRAY_BUFFER, poly_size, nullptr, GL_DYNAMIC_DRAW);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
data.polygon_buffer_size = poly_size;
glBindBuffer(GL_ARRAY_BUFFER, 0);
uint32_t index_size = GLOBAL_DEF("rendering/limits/buffers/canvas_polygon_index_buffer_size_kb", 128);
ProjectSettings::get_singleton()->set_custom_property_info("rendering/limits/buffers/canvas_polygon_index_buffer_size_kb", PropertyInfo(Variant::INT, "rendering/limits/buffers/canvas_polygon_index_buffer_size_kb", PROPERTY_HINT_RANGE, "0,256,1,or_greater"));
2d Legacy - close vulnerabilities and more debug checks While adding more debug checks to legacy renderer, I closed 2 types of vulnerabilities: * TYPE_PRIMITIVE would previously read from uninitialized data if only specifying a single color * Other legacy draw operations would fail in debug AFTER accessing out of bounds memory rather than before Many calls to glBufferSubData are wrapped in a safe version which checks for out of bounds and exits the draw function if this is detected.
2020-11-22 11:24:29 +00:00
index_size = MAX(index_size, 2);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
index_size *= 1024; // kb
glGenBuffers(1, &data.polygon_index_buffer);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, data.polygon_index_buffer);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glBufferData(GL_ELEMENT_ARRAY_BUFFER, index_size, nullptr, GL_DYNAMIC_DRAW);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0);
data.polygon_index_buffer_size = index_size;
}
// ninepatch buffers
{
// array buffer
glGenBuffers(1, &data.ninepatch_vertices);
glBindBuffer(GL_ARRAY_BUFFER, data.ninepatch_vertices);
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
glBufferData(GL_ARRAY_BUFFER, sizeof(float) * (16 + 16) * 2, nullptr, GL_DYNAMIC_DRAW);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
glBindBuffer(GL_ARRAY_BUFFER, 0);
// element buffer
glGenBuffers(1, &data.ninepatch_elements);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, data.ninepatch_elements);
#define _EIDX(y, x) (y * 4 + x)
uint8_t elems[3 * 2 * 9] = {
// first row
_EIDX(0, 0), _EIDX(0, 1), _EIDX(1, 1),
_EIDX(1, 1), _EIDX(1, 0), _EIDX(0, 0),
_EIDX(0, 1), _EIDX(0, 2), _EIDX(1, 2),
_EIDX(1, 2), _EIDX(1, 1), _EIDX(0, 1),
_EIDX(0, 2), _EIDX(0, 3), _EIDX(1, 3),
_EIDX(1, 3), _EIDX(1, 2), _EIDX(0, 2),
// second row
_EIDX(1, 0), _EIDX(1, 1), _EIDX(2, 1),
_EIDX(2, 1), _EIDX(2, 0), _EIDX(1, 0),
// the center one would be here, but we'll put it at the end
// so it's easier to disable the center and be able to use
// one draw call for both
_EIDX(1, 2), _EIDX(1, 3), _EIDX(2, 3),
_EIDX(2, 3), _EIDX(2, 2), _EIDX(1, 2),
// third row
_EIDX(2, 0), _EIDX(2, 1), _EIDX(3, 1),
_EIDX(3, 1), _EIDX(3, 0), _EIDX(2, 0),
_EIDX(2, 1), _EIDX(2, 2), _EIDX(3, 2),
_EIDX(3, 2), _EIDX(3, 1), _EIDX(2, 1),
_EIDX(2, 2), _EIDX(2, 3), _EIDX(3, 3),
_EIDX(3, 3), _EIDX(3, 2), _EIDX(2, 2),
// center field
_EIDX(1, 1), _EIDX(1, 2), _EIDX(2, 2),
_EIDX(2, 2), _EIDX(2, 1), _EIDX(1, 1)
};
#undef _EIDX
glBufferData(GL_ELEMENT_ARRAY_BUFFER, sizeof(elems), elems, GL_STATIC_DRAW);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0);
}
state.canvas_shadow_shader.init();
state.canvas_shader.init();
Batching - fix a couple of minor valgrind bugs. Valgrind reported two instances of reading uninitialized memory in the batching. They are both pretty benign (as evidenced by no bug reports) but wise to close these. The first is that when changing batch from a default batch it reads the batch color which is not set (as it is not relevant for default batches). The segment of code is not necessary when it has already deemed a batch change necessary (which will occur from a default batch). In addition this means that the count of color changes will be more accurate, rather than having a possible random value in. The second is that on initialization _set_texture_rect_mode is called before the state has been properly initialized (it is initialized at the beginning of each canvas_begin, but this occurs outside of that).
2021-02-16 18:00:20 +00:00
state.using_light_angle = false;
state.using_large_vertex = false;
state.using_modulate = false;
GLES2 2D fix normal mapping - batching and nvidia workaround Normal mapping previously took no account of rotation or flips in any path except the TEXTURE_RECT (uniform draw) method. This passed flips to the shader in uniforms. In order to pass flips and rotations to the shader in batching and nvidia workaround, a per vertex attribute is required rather than a uniform. This introduces LIGHT_ANGLE which encodes both the rotation of a quad (vertex) and the horizontal and vertical flip. In order to optionally store light angles in batching, we switch to using a 'unit' sized array which can be reused for different FVF types, as there is no need for a separate array for each FVF, as it is a waste of memory.
2020-08-15 13:31:16 +00:00
_set_texture_rect_mode(true);
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
state.canvas_shader.set_conditional(CanvasShaderGLES2::USE_RGBA_SHADOWS, storage->config.use_rgba_2d_shadows);
state.canvas_shader.bind();
state.lens_shader.init();
Renaming rendering/2d project settings. The rendering/quality/2d section of project settings is becoming considerably expanded in 3.2.4, and arguably was not the correct place for settings that were not really to do with quality. 3.2.4 is the last sensible opportunity we will have to move these settings, as the only existing one likely to break compatibility in a small way is `pixel_snap`, and given that the whole snapping area is being overhauled we can draw attention to the fact it has changed in the release notes. Class reference is also updated and slightly improved. `pixel_snap` is renamed to `gpu_pixel_snap` in the project settings and code to help differentiate from CPU side transform snapping.
2021-03-01 11:26:11 +00:00
state.canvas_shader.set_conditional(CanvasShaderGLES2::USE_PIXEL_SNAP, GLOBAL_DEF("rendering/2d/snapping/use_gpu_pixel_snap", false));
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
Style: Replaces uses of 0/NULL by nullptr (C++11) Using clang-tidy's `modernize-use-nullptr`. https://clang.llvm.org/extra/clang-tidy/checks/modernize-use-nullptr.html
2021-05-04 14:00:45 +00:00
state.using_light = nullptr;
GLES2 2d Batch rendering (across items) 2d rendering is currently bottlenecked by drawing primitives one at a time, limiting OpenGL efficiency. This PR batches primitives and renders in fewer drawcalls, resulting in significant performance improvements. This also speeds up text rendering. This PR batches across canvas items as well as within items. The code dynamically chooses between a vertex format with and without color, depending on the input data for a frame, in order to optimize throughput and maximize batch size. It also adds an option to use glScissor to reduce fillrate in light passes.
2020-03-27 09:19:37 +00:00
state.using_transparent_rt = false;
state.using_skeleton = false;
}
void RasterizerCanvasBaseGLES2::finalize() {
}
RasterizerCanvasBaseGLES2::RasterizerCanvasBaseGLES2() {
}