godotengine
/
godot
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Make shader binary alignment handling simpler and more robust 

Bonus:
Also simplified the rounding to block size in image size calculations.
This commit is contained in:
Pedro J. Estébanez 2024-03-06 11:06:17 +01:00
parent 7d2ca2d8ac
commit f77b4d155b
3 changed files with 37 additions and 67 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

47
drivers/d3d12/rendering_device_driver_d3d12.cpp
View File

@ -3246,10 +3246,7 @@ Vector<uint8_t> RenderingDeviceDriverD3D12::shader_compile_binary_from_spirv(Vec
compressed_stages.push_back(zstd);
uint32_t s = compressed_stages[i].size();
if (s % 4 != 0) {
s += 4 - (s % 4);
}
stages_binary_size += s;
stages_binary_size += STEPIFY(s, 4);
}
CharString shader_name_utf = p_shader_name.utf8();
@ -3259,10 +3256,7 @@ Vector<uint8_t> RenderingDeviceDriverD3D12::shader_compile_binary_from_spirv(Vec
uint32_t total_size = sizeof(uint32_t) * 3; // Header + version + main datasize;.
total_size += sizeof(ShaderBinary::Data);
total_size += binary_data.shader_name_len;
if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange.
total_size += 4 - (binary_data.shader_name_len % 4);
}
total_size += STEPIFY(binary_data.shader_name_len, 4);
for (int i = 0; i < sets_bindings.size(); i++) {
total_size += sizeof(uint32_t);
@ -3294,13 +3288,17 @@ Vector<uint8_t> RenderingDeviceDriverD3D12::shader_compile_binary_from_spirv(Vec
memcpy(binptr + offset, &binary_data, sizeof(ShaderBinary::Data));
offset += sizeof(ShaderBinary::Data);
#define ADVANCE_OFFSET_WITH_ALIGNMENT(m_bytes) \
{ \
offset += m_bytes; \
uint32_t padding = STEPIFY(m_bytes, 4) - m_bytes; \
memset(binptr + offset, 0, padding); /* Avoid garbage data. */ \
offset += padding; \
}
if (binary_data.shader_name_len > 0) {
memcpy(binptr + offset, shader_name_utf.ptr(), binary_data.shader_name_len);
offset += binary_data.shader_name_len;
if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange.
offset += 4 - (binary_data.shader_name_len % 4);
}
ADVANCE_OFFSET_WITH_ALIGNMENT(binary_data.shader_name_len);
}
for (int i = 0; i < sets_bindings.size(); i++) {
@ -3326,14 +3324,7 @@ Vector<uint8_t> RenderingDeviceDriverD3D12::shader_compile_binary_from_spirv(Vec
encode_uint32(zstd_size[i], binptr + offset);
offset += sizeof(uint32_t);
memcpy(binptr + offset, compressed_stages[i].ptr(), compressed_stages[i].size());
uint32_t s = compressed_stages[i].size();
if (s % 4 != 0) {
s += 4 - (s % 4);
}
offset += s;
ADVANCE_OFFSET_WITH_ALIGNMENT(compressed_stages[i].size());
}
memcpy(binptr + offset, root_sig_blob->GetBufferPointer(), root_sig_blob->GetBufferSize());
@ -3382,10 +3373,7 @@ RDD::ShaderID RenderingDeviceDriverD3D12::shader_create_from_bytecode(const Vect
if (binary_data.shader_name_len) {
r_name.parse_utf8((const char *)(binptr + read_offset), binary_data.shader_name_len);
read_offset += binary_data.shader_name_len;
if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange.
read_offset += 4 - (binary_data.shader_name_len % 4);
}
read_offset += STEPIFY(binary_data.shader_name_len, 4);
}
r_shader_desc.uniform_sets.resize(binary_data.set_count);
@ -3458,6 +3446,7 @@ RDD::ShaderID RenderingDeviceDriverD3D12::shader_create_from_bytecode(const Vect
for (uint32_t i = 0; i < binary_data.stage_count; i++) {
ERR_FAIL_COND_V(read_offset + sizeof(uint32_t) * 3 >= binsize, ShaderID());
uint32_t stage = decode_uint32(binptr + read_offset);
read_offset += sizeof(uint32_t);
uint32_t dxil_size = decode_uint32(binptr + read_offset);
@ -3472,13 +3461,9 @@ RDD::ShaderID RenderingDeviceDriverD3D12::shader_create_from_bytecode(const Vect
ERR_FAIL_COND_V(dec_dxil_size != (int32_t)dxil_size, ShaderID());
shader_info_in.stages_bytecode[ShaderStage(stage)] = dxil;
if (zstd_size % 4 != 0) {
zstd_size += 4 - (zstd_size % 4);
}
ERR_FAIL_COND_V(read_offset + zstd_size > binsize, ShaderID());
zstd_size = STEPIFY(zstd_size, 4);
read_offset += zstd_size;
ERR_FAIL_COND_V(read_offset > binsize, ShaderID());
}
const uint8_t *root_sig_data_ptr = binptr + read_offset;

50
drivers/vulkan/rendering_device_driver_vulkan.cpp
View File

@ -2957,10 +2957,7 @@ Vector<uint8_t> RenderingDeviceDriverVulkan::shader_compile_binary_from_spirv(Ve
}
}
uint32_t s = compressed_stages[i].size();
if (s % 4 != 0) {
s += 4 - (s % 4);
}
stages_binary_size += s;
stages_binary_size += STEPIFY(s, 4);
}
binary_data.specialization_constants_count = specialization_constants.size();
@ -2974,11 +2971,7 @@ Vector<uint8_t> RenderingDeviceDriverVulkan::shader_compile_binary_from_spirv(Ve
uint32_t total_size = sizeof(uint32_t) * 3; // Header + version + main datasize;.
total_size += sizeof(ShaderBinary::Data);
total_size += binary_data.shader_name_len;
if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange.
total_size += 4 - (binary_data.shader_name_len % 4);
}
total_size += STEPIFY(binary_data.shader_name_len, 4);
for (int i = 0; i < uniforms.size(); i++) {
total_size += sizeof(uint32_t);
@ -3007,13 +3000,17 @@ Vector<uint8_t> RenderingDeviceDriverVulkan::shader_compile_binary_from_spirv(Ve
memcpy(binptr + offset, &binary_data, sizeof(ShaderBinary::Data));
offset += sizeof(ShaderBinary::Data);
#define ADVANCE_OFFSET_WITH_ALIGNMENT(m_bytes) \
{ \
offset += m_bytes; \
uint32_t padding = STEPIFY(m_bytes, 4) - m_bytes; \
memset(binptr + offset, 0, padding); /* Avoid garbage data. */ \
offset += padding; \
}
if (binary_data.shader_name_len > 0) {
memcpy(binptr + offset, shader_name_utf.ptr(), binary_data.shader_name_len);
offset += binary_data.shader_name_len;
if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange.
offset += 4 - (binary_data.shader_name_len % 4);
}
ADVANCE_OFFSET_WITH_ALIGNMENT(binary_data.shader_name_len);
}
for (int i = 0; i < uniforms.size(); i++) {
@ -3039,14 +3036,7 @@ Vector<uint8_t> RenderingDeviceDriverVulkan::shader_compile_binary_from_spirv(Ve
encode_uint32(zstd_size[i], binptr + offset);
offset += sizeof(uint32_t);
memcpy(binptr + offset, compressed_stages[i].ptr(), compressed_stages[i].size());
uint32_t s = compressed_stages[i].size();
if (s % 4 != 0) {
s += 4 - (s % 4);
}
offset += s;
ADVANCE_OFFSET_WITH_ALIGNMENT(compressed_stages[i].size());
}
DEV_ASSERT(offset == (uint32_t)ret.size());
@ -3090,10 +3080,7 @@ RDD::ShaderID RenderingDeviceDriverVulkan::shader_create_from_bytecode(const Vec
if (binary_data.shader_name_len) {
r_name.parse_utf8((const char *)(binptr + read_offset), binary_data.shader_name_len);
read_offset += binary_data.shader_name_len;
if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange.
read_offset += 4 - (binary_data.shader_name_len % 4);
}
read_offset += STEPIFY(binary_data.shader_name_len, 4);
}
Vector<Vector<VkDescriptorSetLayoutBinding>> vk_set_bindings;
@ -3192,6 +3179,7 @@ RDD::ShaderID RenderingDeviceDriverVulkan::shader_create_from_bytecode(const Vec
for (uint32_t i = 0; i < binary_data.stage_count; i++) {
ERR_FAIL_COND_V(read_offset + sizeof(uint32_t) * 3 >= binsize, ShaderID());
uint32_t stage = decode_uint32(binptr + read_offset);
read_offset += sizeof(uint32_t);
uint32_t smolv_size = decode_uint32(binptr + read_offset);
@ -3223,16 +3211,12 @@ RDD::ShaderID RenderingDeviceDriverVulkan::shader_create_from_bytecode(const Vec
r_shader_desc.stages.set(i, ShaderStage(stage));
if (buf_size % 4 != 0) {
buf_size += 4 - (buf_size % 4);
}
DEV_ASSERT(read_offset + buf_size <= binsize);
buf_size = STEPIFY(buf_size, 4);
read_offset += buf_size;
ERR_FAIL_COND_V(read_offset > binsize, ShaderID());
}
DEV_ASSERT(read_offset == binsize);
ERR_FAIL_COND_V(read_offset != binsize, ShaderID());
// Modules.

7
servers/rendering/rendering_device_commons.cpp
View File

@ -711,12 +711,13 @@ uint32_t RenderingDeviceCommons::get_image_format_required_size(DataFormat p_for
uint32_t pixel_size = get_image_format_pixel_size(p_format);
uint32_t pixel_rshift = get_compressed_image_format_pixel_rshift(p_format);
uint32_t blockw, blockh;
uint32_t blockw = 0;
uint32_t blockh = 0;
get_compressed_image_format_block_dimensions(p_format, blockw, blockh);
for (uint32_t i = 0; i < p_mipmaps; i++) {
uint32_t bw = w % blockw != 0 ? w + (blockw - w % blockw) : w;
uint32_t bh = h % blockh != 0 ? h + (blockh - h % blockh) : h;
uint32_t bw = STEPIFY(w, blockw);
uint32_t bh = STEPIFY(h, blockh);
uint32_t s = bw * bh;