mirrors/citra
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

video_core: Abstract shader generators. (#6990)

Browse source 
* video_core: Abstract shader generators.

* shader: Extract common generator structures and move generators to specific namespaces.

* shader: Minor fixes and clean-up.
This commit is contained in:
Steveice10 2023-09-30 02:06:06 -07:00 committed by GitHub
parent 1492d73ccb
commit 50f22d1f59
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
35 changed files with 1374 additions and 3344 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

20
CMakeModules/GenerateSCMRev.cmake
View file

@ -4,24 +4,24 @@ include(GenerateBuildInfo)
# The variable SRC_DIR must be passed into the script (since it uses the current build directory for all values of CMAKE_*_DIR) # The variable SRC_DIR must be passed into the script (since it uses the current build directory for all values of CMAKE_*_DIR)
set(VIDEO_CORE "${SRC_DIR}/src/video_core") set(VIDEO_CORE "${SRC_DIR}/src/video_core")
set(HASH_FILES set(HASH_FILES
"${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.cpp"
"${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.h"
"${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.cpp" "${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.cpp"
"${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.h" "${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.h"
"${VIDEO_CORE}/renderer_opengl/gl_shader_gen.cpp"
"${VIDEO_CORE}/renderer_opengl/gl_shader_gen.h"
"${VIDEO_CORE}/renderer_opengl/gl_shader_util.cpp" "${VIDEO_CORE}/renderer_opengl/gl_shader_util.cpp"
"${VIDEO_CORE}/renderer_opengl/gl_shader_util.h" "${VIDEO_CORE}/renderer_opengl/gl_shader_util.h"
"${VIDEO_CORE}/renderer_vulkan/vk_shader_gen.cpp"
"${VIDEO_CORE}/renderer_vulkan/vk_shader_gen.h"
"${VIDEO_CORE}/renderer_vulkan/vk_shader_gen_spv.cpp"
"${VIDEO_CORE}/renderer_vulkan/vk_shader_gen_spv.h"
"${VIDEO_CORE}/renderer_vulkan/vk_shader_util.cpp" "${VIDEO_CORE}/renderer_vulkan/vk_shader_util.cpp"
"${VIDEO_CORE}/renderer_vulkan/vk_shader_util.h" "${VIDEO_CORE}/renderer_vulkan/vk_shader_util.h"
"${VIDEO_CORE}/shader/generator/glsl_shader_decompiler.cpp"
"${VIDEO_CORE}/shader/generator/glsl_shader_decompiler.h"
"${VIDEO_CORE}/shader/generator/glsl_shader_gen.cpp"
"${VIDEO_CORE}/shader/generator/glsl_shader_gen.h"
"${VIDEO_CORE}/shader/generator/shader_gen.cpp"
"${VIDEO_CORE}/shader/generator/shader_gen.h"
"${VIDEO_CORE}/shader/generator/shader_uniforms.cpp"
"${VIDEO_CORE}/shader/generator/shader_uniforms.h"
"${VIDEO_CORE}/shader/generator/spv_shader_gen.cpp"
"${VIDEO_CORE}/shader/generator/spv_shader_gen.h"
"${VIDEO_CORE}/shader/shader.cpp" "${VIDEO_CORE}/shader/shader.cpp"
"${VIDEO_CORE}/shader/shader.h" "${VIDEO_CORE}/shader/shader.h"
"${VIDEO_CORE}/shader/shader_uniforms.cpp"
"${VIDEO_CORE}/shader/shader_uniforms.h"
"${VIDEO_CORE}/pica.cpp" "${VIDEO_CORE}/pica.cpp"
"${VIDEO_CORE}/pica.h" "${VIDEO_CORE}/pica.h"
"${VIDEO_CORE}/regs_framebuffer.h" "${VIDEO_CORE}/regs_framebuffer.h"

16
src/common/CMakeLists.txt
View file

@ -15,14 +15,22 @@ add_custom_command(OUTPUT scm_rev.cpp
DEPENDS DEPENDS
# WARNING! It was too much work to try and make a common location for this list, # WARNING! It was too much work to try and make a common location for this list,
# so if you need to change it, please update CMakeModules/GenerateSCMRev.cmake as well # so if you need to change it, please update CMakeModules/GenerateSCMRev.cmake as well
"${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.cpp"
"${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.h"
"${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.cpp" "${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.cpp"
"${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.h" "${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.h"
"${VIDEO_CORE}/renderer_opengl/gl_shader_gen.cpp"
"${VIDEO_CORE}/renderer_opengl/gl_shader_gen.h"
"${VIDEO_CORE}/renderer_opengl/gl_shader_util.cpp" "${VIDEO_CORE}/renderer_opengl/gl_shader_util.cpp"
"${VIDEO_CORE}/renderer_opengl/gl_shader_util.h" "${VIDEO_CORE}/renderer_opengl/gl_shader_util.h"
"${VIDEO_CORE}/renderer_vulkan/vk_shader_util.cpp"
"${VIDEO_CORE}/renderer_vulkan/vk_shader_util.h"
"${VIDEO_CORE}/shader/generator/glsl_shader_decompiler.cpp"
"${VIDEO_CORE}/shader/generator/glsl_shader_decompiler.h"
"${VIDEO_CORE}/shader/generator/glsl_shader_gen.cpp"
"${VIDEO_CORE}/shader/generator/glsl_shader_gen.h"
"${VIDEO_CORE}/shader/generator/shader_gen.cpp"
"${VIDEO_CORE}/shader/generator/shader_gen.h"
"${VIDEO_CORE}/shader/generator/shader_uniforms.cpp"
"${VIDEO_CORE}/shader/generator/shader_uniforms.h"
"${VIDEO_CORE}/shader/generator/spv_shader_gen.cpp"
"${VIDEO_CORE}/shader/generator/spv_shader_gen.h"
"${VIDEO_CORE}/shader/shader.cpp" "${VIDEO_CORE}/shader/shader.cpp"
"${VIDEO_CORE}/shader/shader.h" "${VIDEO_CORE}/shader/shader.h"
"${VIDEO_CORE}/pica.cpp" "${VIDEO_CORE}/pica.cpp"

20
src/video_core/CMakeLists.txt
View file

@ -61,12 +61,8 @@ add_library(video_core STATIC
renderer_opengl/gl_rasterizer_cache.cpp renderer_opengl/gl_rasterizer_cache.cpp
renderer_opengl/gl_resource_manager.cpp renderer_opengl/gl_resource_manager.cpp
renderer_opengl/gl_resource_manager.h renderer_opengl/gl_resource_manager.h
renderer_opengl/gl_shader_decompiler.cpp
renderer_opengl/gl_shader_decompiler.h
renderer_opengl/gl_shader_disk_cache.cpp renderer_opengl/gl_shader_disk_cache.cpp
renderer_opengl/gl_shader_disk_cache.h renderer_opengl/gl_shader_disk_cache.h
renderer_opengl/gl_shader_gen.cpp
renderer_opengl/gl_shader_gen.h
renderer_opengl/gl_shader_manager.cpp renderer_opengl/gl_shader_manager.cpp
renderer_opengl/gl_shader_manager.h renderer_opengl/gl_shader_manager.h
renderer_opengl/gl_shader_util.cpp renderer_opengl/gl_shader_util.cpp
@ -130,10 +126,6 @@ add_library(video_core STATIC
renderer_vulkan/vk_present_window.h renderer_vulkan/vk_present_window.h
renderer_vulkan/vk_renderpass_cache.cpp renderer_vulkan/vk_renderpass_cache.cpp
renderer_vulkan/vk_renderpass_cache.h renderer_vulkan/vk_renderpass_cache.h
renderer_vulkan/vk_shader_gen.cpp
renderer_vulkan/vk_shader_gen.h
renderer_vulkan/vk_shader_gen_spv.cpp
renderer_vulkan/vk_shader_gen_spv.h
renderer_vulkan/vk_shader_util.cpp renderer_vulkan/vk_shader_util.cpp
renderer_vulkan/vk_shader_util.h renderer_vulkan/vk_shader_util.h
renderer_vulkan/vk_stream_buffer.cpp renderer_vulkan/vk_stream_buffer.cpp
@ -143,6 +135,16 @@ add_library(video_core STATIC
renderer_vulkan/vk_texture_runtime.cpp renderer_vulkan/vk_texture_runtime.cpp
renderer_vulkan/vk_texture_runtime.h renderer_vulkan/vk_texture_runtime.h
shader/debug_data.h shader/debug_data.h
shader/generator/glsl_shader_decompiler.cpp
shader/generator/glsl_shader_decompiler.h
shader/generator/glsl_shader_gen.cpp
shader/generator/glsl_shader_gen.h
shader/generator/shader_gen.cpp
shader/generator/shader_gen.h
shader/generator/shader_uniforms.cpp
shader/generator/shader_uniforms.h
shader/generator/spv_shader_gen.cpp
shader/generator/spv_shader_gen.h
shader/shader.cpp shader/shader.cpp
shader/shader.h shader/shader.h
shader/shader_interpreter.cpp shader/shader_interpreter.cpp
@ -151,8 +153,6 @@ add_library(video_core STATIC
shader/shader_jit_x64_compiler.cpp shader/shader_jit_x64_compiler.cpp
shader/shader_jit_x64.h shader/shader_jit_x64.h
shader/shader_jit_x64_compiler.h shader/shader_jit_x64_compiler.h
shader/shader_uniforms.cpp
shader/shader_uniforms.h
texture/etc1.cpp texture/etc1.cpp
texture/etc1.h texture/etc1.h
texture/texture_decode.cpp texture/texture_decode.cpp

172
src/video_core/rasterizer_accelerated.cpp
View file

@ -54,7 +54,7 @@ RasterizerAccelerated::HardwareVertex::HardwareVertex(const Pica::Shader::Output
RasterizerAccelerated::RasterizerAccelerated(Memory::MemorySystem& memory_) RasterizerAccelerated::RasterizerAccelerated(Memory::MemorySystem& memory_)
: memory{memory_}, regs{Pica::g_state.regs} { : memory{memory_}, regs{Pica::g_state.regs} {
uniform_block_data.lighting_lut_dirty.fill(true); fs_uniform_block_data.lighting_lut_dirty.fill(true);
} }
/** /**
@ -135,7 +135,7 @@ void RasterizerAccelerated::SyncEntireState() {
SyncFixedState(); SyncFixedState();
// Sync uniforms // Sync uniforms
SyncClipCoef(); SyncClipPlane();
SyncDepthScale(); SyncDepthScale();
SyncDepthOffset(); SyncDepthOffset();
SyncAlphaTest(); SyncAlphaTest();
@ -199,7 +199,7 @@ void RasterizerAccelerated::NotifyPicaRegisterChanged(u32 id) {
case PICA_REG_INDEX(texturing.fog_lut_data[5]): case PICA_REG_INDEX(texturing.fog_lut_data[5]):
case PICA_REG_INDEX(texturing.fog_lut_data[6]): case PICA_REG_INDEX(texturing.fog_lut_data[6]):
case PICA_REG_INDEX(texturing.fog_lut_data[7]): case PICA_REG_INDEX(texturing.fog_lut_data[7]):
uniform_block_data.fog_lut_dirty = true; fs_uniform_block_data.fog_lut_dirty = true;
break; break;
// ProcTex state // ProcTex state
@ -227,19 +227,19 @@ void RasterizerAccelerated::NotifyPicaRegisterChanged(u32 id) {
using Pica::TexturingRegs; using Pica::TexturingRegs;
switch (regs.texturing.proctex_lut_config.ref_table.Value()) { switch (regs.texturing.proctex_lut_config.ref_table.Value()) {
case TexturingRegs::ProcTexLutTable::Noise: case TexturingRegs::ProcTexLutTable::Noise:
uniform_block_data.proctex_noise_lut_dirty = true; fs_uniform_block_data.proctex_noise_lut_dirty = true;
break; break;
case TexturingRegs::ProcTexLutTable::ColorMap: case TexturingRegs::ProcTexLutTable::ColorMap:
uniform_block_data.proctex_color_map_dirty = true; fs_uniform_block_data.proctex_color_map_dirty = true;
break; break;
case TexturingRegs::ProcTexLutTable::AlphaMap: case TexturingRegs::ProcTexLutTable::AlphaMap:
uniform_block_data.proctex_alpha_map_dirty = true; fs_uniform_block_data.proctex_alpha_map_dirty = true;
break; break;
case TexturingRegs::ProcTexLutTable::Color: case TexturingRegs::ProcTexLutTable::Color:
uniform_block_data.proctex_lut_dirty = true; fs_uniform_block_data.proctex_lut_dirty = true;
break; break;
case TexturingRegs::ProcTexLutTable::ColorDiff: case TexturingRegs::ProcTexLutTable::ColorDiff:
uniform_block_data.proctex_diff_lut_dirty = true; fs_uniform_block_data.proctex_diff_lut_dirty = true;
break; break;
} }
break; break;
@ -588,8 +588,8 @@ void RasterizerAccelerated::NotifyPicaRegisterChanged(u32 id) {
case PICA_REG_INDEX(lighting.lut_data[6]): case PICA_REG_INDEX(lighting.lut_data[6]):
case PICA_REG_INDEX(lighting.lut_data[7]): { case PICA_REG_INDEX(lighting.lut_data[7]): {
const auto& lut_config = regs.lighting.lut_config; const auto& lut_config = regs.lighting.lut_config;
uniform_block_data.lighting_lut_dirty[lut_config.type] = true; fs_uniform_block_data.lighting_lut_dirty[lut_config.type] = true;
uniform_block_data.lighting_lut_dirty_any = true; fs_uniform_block_data.lighting_lut_dirty_any = true;
break; break;
} }
@ -616,11 +616,12 @@ void RasterizerAccelerated::NotifyPicaRegisterChanged(u32 id) {
break; break;
// Clipping plane // Clipping plane
case PICA_REG_INDEX(rasterizer.clip_enable):
case PICA_REG_INDEX(rasterizer.clip_coef[0]): case PICA_REG_INDEX(rasterizer.clip_coef[0]):
case PICA_REG_INDEX(rasterizer.clip_coef[1]): case PICA_REG_INDEX(rasterizer.clip_coef[1]):
case PICA_REG_INDEX(rasterizer.clip_coef[2]): case PICA_REG_INDEX(rasterizer.clip_coef[2]):
case PICA_REG_INDEX(rasterizer.clip_coef[3]): case PICA_REG_INDEX(rasterizer.clip_coef[3]):
SyncClipCoef(); SyncClipPlane();
break; break;
} }
@ -631,18 +632,18 @@ void RasterizerAccelerated::NotifyPicaRegisterChanged(u32 id) {
void RasterizerAccelerated::SyncDepthScale() { void RasterizerAccelerated::SyncDepthScale() {
const f32 depth_scale = f24::FromRaw(regs.rasterizer.viewport_depth_range).ToFloat32(); const f32 depth_scale = f24::FromRaw(regs.rasterizer.viewport_depth_range).ToFloat32();
if (depth_scale != uniform_block_data.data.depth_scale) { if (depth_scale != fs_uniform_block_data.data.depth_scale) {
uniform_block_data.data.depth_scale = depth_scale; fs_uniform_block_data.data.depth_scale = depth_scale;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncDepthOffset() { void RasterizerAccelerated::SyncDepthOffset() {
const f32 depth_offset = f24::FromRaw(regs.rasterizer.viewport_depth_near_plane).ToFloat32(); const f32 depth_offset = f24::FromRaw(regs.rasterizer.viewport_depth_near_plane).ToFloat32();
if (depth_offset != uniform_block_data.data.depth_offset) { if (depth_offset != fs_uniform_block_data.data.depth_offset) {
uniform_block_data.data.depth_offset = depth_offset; fs_uniform_block_data.data.depth_offset = depth_offset;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -654,9 +655,9 @@ void RasterizerAccelerated::SyncFogColor() {
fog_color_regs.b.Value() / 255.0f, fog_color_regs.b.Value() / 255.0f,
}; };
if (fog_color != uniform_block_data.data.fog_color) { if (fog_color != fs_uniform_block_data.data.fog_color) {
uniform_block_data.data.fog_color = fog_color; fs_uniform_block_data.data.fog_color = fog_color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -674,13 +675,13 @@ void RasterizerAccelerated::SyncProcTexNoise() {
Pica::f16::FromRaw(regs.texturing.proctex_noise_v.phase).ToFloat32(), Pica::f16::FromRaw(regs.texturing.proctex_noise_v.phase).ToFloat32(),
}; };
if (proctex_noise_f != uniform_block_data.data.proctex_noise_f || if (proctex_noise_f != fs_uniform_block_data.data.proctex_noise_f ||
proctex_noise_a != uniform_block_data.data.proctex_noise_a || proctex_noise_a != fs_uniform_block_data.data.proctex_noise_a ||
proctex_noise_p != uniform_block_data.data.proctex_noise_p) { proctex_noise_p != fs_uniform_block_data.data.proctex_noise_p) {
uniform_block_data.data.proctex_noise_f = proctex_noise_f; fs_uniform_block_data.data.proctex_noise_f = proctex_noise_f;
uniform_block_data.data.proctex_noise_a = proctex_noise_a; fs_uniform_block_data.data.proctex_noise_a = proctex_noise_a;
uniform_block_data.data.proctex_noise_p = proctex_noise_p; fs_uniform_block_data.data.proctex_noise_p = proctex_noise_p;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -688,25 +689,25 @@ void RasterizerAccelerated::SyncProcTexBias() {
const auto proctex_bias = Pica::f16::FromRaw(regs.texturing.proctex.bias_low | const auto proctex_bias = Pica::f16::FromRaw(regs.texturing.proctex.bias_low |
(regs.texturing.proctex_lut.bias_high << 8)) (regs.texturing.proctex_lut.bias_high << 8))
.ToFloat32(); .ToFloat32();
if (proctex_bias != uniform_block_data.data.proctex_bias) { if (proctex_bias != fs_uniform_block_data.data.proctex_bias) {
uniform_block_data.data.proctex_bias = proctex_bias; fs_uniform_block_data.data.proctex_bias = proctex_bias;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncAlphaTest() { void RasterizerAccelerated::SyncAlphaTest() {
if (regs.framebuffer.output_merger.alpha_test.ref != if (regs.framebuffer.output_merger.alpha_test.ref !=
static_cast<u32>(uniform_block_data.data.alphatest_ref)) { static_cast<u32>(fs_uniform_block_data.data.alphatest_ref)) {
uniform_block_data.data.alphatest_ref = regs.framebuffer.output_merger.alpha_test.ref; fs_uniform_block_data.data.alphatest_ref = regs.framebuffer.output_merger.alpha_test.ref;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncCombinerColor() { void RasterizerAccelerated::SyncCombinerColor() {
const auto combiner_color = ColorRGBA8(regs.texturing.tev_combiner_buffer_color.raw); const auto combiner_color = ColorRGBA8(regs.texturing.tev_combiner_buffer_color.raw);
if (combiner_color != uniform_block_data.data.tev_combiner_buffer_color) { if (combiner_color != fs_uniform_block_data.data.tev_combiner_buffer_color) {
uniform_block_data.data.tev_combiner_buffer_color = combiner_color; fs_uniform_block_data.data.tev_combiner_buffer_color = combiner_color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -714,51 +715,51 @@ void RasterizerAccelerated::SyncTevConstColor(
const size_t stage_index, const Pica::TexturingRegs::TevStageConfig& tev_stage) { const size_t stage_index, const Pica::TexturingRegs::TevStageConfig& tev_stage) {
const auto const_color = ColorRGBA8(tev_stage.const_color); const auto const_color = ColorRGBA8(tev_stage.const_color);
if (const_color == uniform_block_data.data.const_color[stage_index]) { if (const_color == fs_uniform_block_data.data.const_color[stage_index]) {
return; return;
} }
uniform_block_data.data.const_color[stage_index] = const_color; fs_uniform_block_data.data.const_color[stage_index] = const_color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
void RasterizerAccelerated::SyncGlobalAmbient() { void RasterizerAccelerated::SyncGlobalAmbient() {
const auto color = LightColor(regs.lighting.global_ambient); const auto color = LightColor(regs.lighting.global_ambient);
if (color != uniform_block_data.data.lighting_global_ambient) { if (color != fs_uniform_block_data.data.lighting_global_ambient) {
uniform_block_data.data.lighting_global_ambient = color; fs_uniform_block_data.data.lighting_global_ambient = color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncLightSpecular0(int light_index) { void RasterizerAccelerated::SyncLightSpecular0(int light_index) {
const auto color = LightColor(regs.lighting.light[light_index].specular_0); const auto color = LightColor(regs.lighting.light[light_index].specular_0);
if (color != uniform_block_data.data.light_src[light_index].specular_0) { if (color != fs_uniform_block_data.data.light_src[light_index].specular_0) {
uniform_block_data.data.light_src[light_index].specular_0 = color; fs_uniform_block_data.data.light_src[light_index].specular_0 = color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncLightSpecular1(int light_index) { void RasterizerAccelerated::SyncLightSpecular1(int light_index) {
const auto color = LightColor(regs.lighting.light[light_index].specular_1); const auto color = LightColor(regs.lighting.light[light_index].specular_1);
if (color != uniform_block_data.data.light_src[light_index].specular_1) { if (color != fs_uniform_block_data.data.light_src[light_index].specular_1) {
uniform_block_data.data.light_src[light_index].specular_1 = color; fs_uniform_block_data.data.light_src[light_index].specular_1 = color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncLightDiffuse(int light_index) { void RasterizerAccelerated::SyncLightDiffuse(int light_index) {
const auto color = LightColor(regs.lighting.light[light_index].diffuse); const auto color = LightColor(regs.lighting.light[light_index].diffuse);
if (color != uniform_block_data.data.light_src[light_index].diffuse) { if (color != fs_uniform_block_data.data.light_src[light_index].diffuse) {
uniform_block_data.data.light_src[light_index].diffuse = color; fs_uniform_block_data.data.light_src[light_index].diffuse = color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncLightAmbient(int light_index) { void RasterizerAccelerated::SyncLightAmbient(int light_index) {
const auto color = LightColor(regs.lighting.light[light_index].ambient); const auto color = LightColor(regs.lighting.light[light_index].ambient);
if (color != uniform_block_data.data.light_src[light_index].ambient) { if (color != fs_uniform_block_data.data.light_src[light_index].ambient) {
uniform_block_data.data.light_src[light_index].ambient = color; fs_uniform_block_data.data.light_src[light_index].ambient = color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -769,9 +770,9 @@ void RasterizerAccelerated::SyncLightPosition(int light_index) {
Pica::f16::FromRaw(regs.lighting.light[light_index].z).ToFloat32(), Pica::f16::FromRaw(regs.lighting.light[light_index].z).ToFloat32(),
}; };
if (position != uniform_block_data.data.light_src[light_index].position) { if (position != fs_uniform_block_data.data.light_src[light_index].position) {
uniform_block_data.data.light_src[light_index].position = position; fs_uniform_block_data.data.light_src[light_index].position = position;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -780,9 +781,9 @@ void RasterizerAccelerated::SyncLightSpotDirection(int light_index) {
const auto spot_direction = const auto spot_direction =
Common::Vec3f{light.spot_x / 2047.0f, light.spot_y / 2047.0f, light.spot_z / 2047.0f}; Common::Vec3f{light.spot_x / 2047.0f, light.spot_y / 2047.0f, light.spot_z / 2047.0f};
if (spot_direction != uniform_block_data.data.light_src[light_index].spot_direction) { if (spot_direction != fs_uniform_block_data.data.light_src[light_index].spot_direction) {
uniform_block_data.data.light_src[light_index].spot_direction = spot_direction; fs_uniform_block_data.data.light_src[light_index].spot_direction = spot_direction;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -790,9 +791,9 @@ void RasterizerAccelerated::SyncLightDistanceAttenuationBias(int light_index) {
const f32 dist_atten_bias = const f32 dist_atten_bias =
Pica::f20::FromRaw(regs.lighting.light[light_index].dist_atten_bias).ToFloat32(); Pica::f20::FromRaw(regs.lighting.light[light_index].dist_atten_bias).ToFloat32();
if (dist_atten_bias != uniform_block_data.data.light_src[light_index].dist_atten_bias) { if (dist_atten_bias != fs_uniform_block_data.data.light_src[light_index].dist_atten_bias) {
uniform_block_data.data.light_src[light_index].dist_atten_bias = dist_atten_bias; fs_uniform_block_data.data.light_src[light_index].dist_atten_bias = dist_atten_bias;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -800,9 +801,9 @@ void RasterizerAccelerated::SyncLightDistanceAttenuationScale(int light_index) {
const f32 dist_atten_scale = const f32 dist_atten_scale =
Pica::f20::FromRaw(regs.lighting.light[light_index].dist_atten_scale).ToFloat32(); Pica::f20::FromRaw(regs.lighting.light[light_index].dist_atten_scale).ToFloat32();
if (dist_atten_scale != uniform_block_data.data.light_src[light_index].dist_atten_scale) { if (dist_atten_scale != fs_uniform_block_data.data.light_src[light_index].dist_atten_scale) {
uniform_block_data.data.light_src[light_index].dist_atten_scale = dist_atten_scale; fs_uniform_block_data.data.light_src[light_index].dist_atten_scale = dist_atten_scale;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -811,28 +812,28 @@ void RasterizerAccelerated::SyncShadowBias() {
const f32 constant = Pica::f16::FromRaw(shadow.constant).ToFloat32(); const f32 constant = Pica::f16::FromRaw(shadow.constant).ToFloat32();
const f32 linear = Pica::f16::FromRaw(shadow.linear).ToFloat32(); const f32 linear = Pica::f16::FromRaw(shadow.linear).ToFloat32();
if (constant != uniform_block_data.data.shadow_bias_constant || if (constant != fs_uniform_block_data.data.shadow_bias_constant ||
linear != uniform_block_data.data.shadow_bias_linear) { linear != fs_uniform_block_data.data.shadow_bias_linear) {
uniform_block_data.data.shadow_bias_constant = constant; fs_uniform_block_data.data.shadow_bias_constant = constant;
uniform_block_data.data.shadow_bias_linear = linear; fs_uniform_block_data.data.shadow_bias_linear = linear;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncShadowTextureBias() { void RasterizerAccelerated::SyncShadowTextureBias() {
const s32 bias = regs.texturing.shadow.bias << 1; const s32 bias = regs.texturing.shadow.bias << 1;
if (bias != uniform_block_data.data.shadow_texture_bias) { if (bias != fs_uniform_block_data.data.shadow_texture_bias) {
uniform_block_data.data.shadow_texture_bias = bias; fs_uniform_block_data.data.shadow_texture_bias = bias;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncTextureLodBias(int tex_index) { void RasterizerAccelerated::SyncTextureLodBias(int tex_index) {
const auto pica_textures = regs.texturing.GetTextures(); const auto pica_textures = regs.texturing.GetTextures();
const f32 bias = pica_textures[tex_index].config.lod.bias / 256.0f; const f32 bias = pica_textures[tex_index].config.lod.bias / 256.0f;
if (bias != uniform_block_data.data.tex_lod_bias[tex_index]) { if (bias != fs_uniform_block_data.data.tex_lod_bias[tex_index]) {
uniform_block_data.data.tex_lod_bias[tex_index] = bias; fs_uniform_block_data.data.tex_lod_bias[tex_index] = bias;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -840,19 +841,22 @@ void RasterizerAccelerated::SyncTextureBorderColor(int tex_index) {
const auto pica_textures = regs.texturing.GetTextures(); const auto pica_textures = regs.texturing.GetTextures();
const auto params = pica_textures[tex_index].config; const auto params = pica_textures[tex_index].config;
const Common::Vec4f border_color = ColorRGBA8(params.border_color.raw); const Common::Vec4f border_color = ColorRGBA8(params.border_color.raw);
if (border_color != uniform_block_data.data.tex_border_color[tex_index]) { if (border_color != fs_uniform_block_data.data.tex_border_color[tex_index]) {
uniform_block_data.data.tex_border_color[tex_index] = border_color; fs_uniform_block_data.data.tex_border_color[tex_index] = border_color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
void RasterizerAccelerated::SyncClipCoef() { void RasterizerAccelerated::SyncClipPlane() {
const bool enable_clip1 = regs.rasterizer.clip_enable != 0;
const auto raw_clip_coef = regs.rasterizer.GetClipCoef(); const auto raw_clip_coef = regs.rasterizer.GetClipCoef();
const Common::Vec4f new_clip_coef = {raw_clip_coef.x.ToFloat32(), raw_clip_coef.y.ToFloat32(), const Common::Vec4f new_clip_coef = {raw_clip_coef.x.ToFloat32(), raw_clip_coef.y.ToFloat32(),
raw_clip_coef.z.ToFloat32(), raw_clip_coef.w.ToFloat32()}; raw_clip_coef.z.ToFloat32(), raw_clip_coef.w.ToFloat32()};
if (new_clip_coef != uniform_block_data.data.clip_coef) { if (enable_clip1 != vs_uniform_block_data.data.enable_clip1 ||
uniform_block_data.data.clip_coef = new_clip_coef; new_clip_coef != vs_uniform_block_data.data.clip_coef) {
uniform_block_data.dirty = true; vs_uniform_block_data.data.enable_clip1 = enable_clip1;
vs_uniform_block_data.data.clip_coef = new_clip_coef;
vs_uniform_block_data.dirty = true;
} }
} }

21
src/video_core/rasterizer_accelerated.h
View file

@ -7,7 +7,7 @@
#include "common/vector_math.h" #include "common/vector_math.h"
#include "video_core/rasterizer_interface.h" #include "video_core/rasterizer_interface.h"
#include "video_core/regs_texturing.h" #include "video_core/regs_texturing.h"
#include "video_core/shader/shader_uniforms.h" #include "video_core/shader/generator/shader_uniforms.h"
namespace Memory { namespace Memory {
class MemorySystem; class MemorySystem;
@ -100,13 +100,19 @@ protected:
/// Syncs the texture border color to match the PICA registers /// Syncs the texture border color to match the PICA registers
void SyncTextureBorderColor(int tex_index); void SyncTextureBorderColor(int tex_index);
/// Syncs the clip coefficients to match the PICA register /// Syncs the clip plane state to match the PICA register
void SyncClipCoef(); void SyncClipPlane();
protected: protected:
/// Structure that keeps tracks of the uniform state /// Structure that keeps tracks of the vertex shader uniform state
struct UniformBlockData { struct VSUniformBlockData {
Pica::Shader::UniformData data{}; Pica::Shader::Generator::VSUniformData data{};
bool dirty = true;
};
/// Structure that keeps tracks of the fragment shader uniform state
struct FSUniformBlockData {
Pica::Shader::Generator::FSUniformData data{};
std::array<bool, Pica::LightingRegs::NumLightingSampler> lighting_lut_dirty{}; std::array<bool, Pica::LightingRegs::NumLightingSampler> lighting_lut_dirty{};
bool lighting_lut_dirty_any = true; bool lighting_lut_dirty_any = true;
bool fog_lut_dirty = true; bool fog_lut_dirty = true;
@ -149,7 +155,8 @@ protected:
std::vector<HardwareVertex> vertex_batch; std::vector<HardwareVertex> vertex_batch;
bool shader_dirty = true; bool shader_dirty = true;
UniformBlockData uniform_block_data{}; VSUniformBlockData vs_uniform_block_data{};
FSUniformBlockData fs_uniform_block_data{};
std::array<std::array<Common::Vec2f, 256>, Pica::LightingRegs::NumLightingSampler> std::array<std::array<Common::Vec2f, 256>, Pica::LightingRegs::NumLightingSampler>
lighting_lut_data{}; lighting_lut_data{};
std::array<Common::Vec2f, 128> fog_lut_data{}; std::array<Common::Vec2f, 128> fog_lut_data{};

2
src/video_core/renderer_opengl/gl_driver.cpp
View file

@ -168,7 +168,7 @@ void Driver::CheckExtensionSupport() {
arb_clear_texture = GLAD_GL_ARB_clear_texture; arb_clear_texture = GLAD_GL_ARB_clear_texture;
arb_get_texture_sub_image = GLAD_GL_ARB_get_texture_sub_image; arb_get_texture_sub_image = GLAD_GL_ARB_get_texture_sub_image;
arb_texture_compression_bptc = GLAD_GL_ARB_texture_compression_bptc; arb_texture_compression_bptc = GLAD_GL_ARB_texture_compression_bptc;
ext_clip_cull_distance = GLAD_GL_EXT_clip_cull_distance; clip_cull_distance = !is_gles || GLAD_GL_EXT_clip_cull_distance;
ext_texture_compression_s3tc = GLAD_GL_EXT_texture_compression_s3tc; ext_texture_compression_s3tc = GLAD_GL_EXT_texture_compression_s3tc;
shader_framebuffer_fetch = shader_framebuffer_fetch =
GLAD_GL_EXT_shader_framebuffer_fetch || GLAD_GL_ARM_shader_framebuffer_fetch; GLAD_GL_EXT_shader_framebuffer_fetch || GLAD_GL_ARM_shader_framebuffer_fetch;

8
src/video_core/renderer_opengl/gl_driver.h
View file

@ -100,9 +100,9 @@ public:
return arb_get_texture_sub_image; return arb_get_texture_sub_image;
} }
/// Returns true if the implementation supports EXT_clip_cull_distance /// Returns true if the implementation supports shader-defined clipping planes
bool HasExtClipCullDistance() const { bool HasClipCullDistance() const {
return ext_clip_cull_distance; return clip_cull_distance;
} }
/// Returns true if the implementation supports (EXT/ARM)_shader_framebuffer_fetch /// Returns true if the implementation supports (EXT/ARM)_shader_framebuffer_fetch
@ -132,7 +132,7 @@ private:
bool arb_buffer_storage{}; bool arb_buffer_storage{};
bool arb_clear_texture{}; bool arb_clear_texture{};
bool arb_get_texture_sub_image{}; bool arb_get_texture_sub_image{};
bool ext_clip_cull_distance{}; bool clip_cull_distance{};
bool ext_texture_compression_s3tc{}; bool ext_texture_compression_s3tc{};
bool arb_texture_compression_bptc{}; bool arb_texture_compression_bptc{};
bool shader_framebuffer_fetch{}; bool shader_framebuffer_fetch{};

144
src/video_core/renderer_opengl/gl_rasterizer.cpp
View file

@ -11,9 +11,9 @@
#include "video_core/regs_framebuffer.h" #include "video_core/regs_framebuffer.h"
#include "video_core/regs_rasterizer.h" #include "video_core/regs_rasterizer.h"
#include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_rasterizer.h"
#include "video_core/renderer_opengl/gl_shader_gen.h"
#include "video_core/renderer_opengl/pica_to_gl.h" #include "video_core/renderer_opengl/pica_to_gl.h"
#include "video_core/renderer_opengl/renderer_opengl.h" #include "video_core/renderer_opengl/renderer_opengl.h"
#include "video_core/shader/generator/glsl_shader_gen.h"
#include "video_core/texture/texture_decode.h" #include "video_core/texture/texture_decode.h"
#include "video_core/video_core.h" #include "video_core/video_core.h"
@ -28,6 +28,7 @@ MICROPROFILE_DEFINE(OpenGL_Drawing, "OpenGL", "Drawing", MP_RGB(128, 128, 192));
MICROPROFILE_DEFINE(OpenGL_Display, "OpenGL", "Display", MP_RGB(128, 128, 192)); MICROPROFILE_DEFINE(OpenGL_Display, "OpenGL", "Display", MP_RGB(128, 128, 192));
using VideoCore::SurfaceType; using VideoCore::SurfaceType;
using namespace Pica::Shader::Generator;
constexpr std::size_t VERTEX_BUFFER_SIZE = 16 * 1024 * 1024; constexpr std::size_t VERTEX_BUFFER_SIZE = 16 * 1024 * 1024;
constexpr std::size_t INDEX_BUFFER_SIZE = 2 * 1024 * 1024; constexpr std::size_t INDEX_BUFFER_SIZE = 2 * 1024 * 1024;
@ -95,10 +96,12 @@ RasterizerOpenGL::RasterizerOpenGL(Memory::MemorySystem& memory,
hw_vao.Create(); hw_vao.Create();
glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &uniform_buffer_alignment); glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &uniform_buffer_alignment);
uniform_size_aligned_vs_pica =
Common::AlignUp<std::size_t>(sizeof(VSPicaUniformData), uniform_buffer_alignment);
uniform_size_aligned_vs = uniform_size_aligned_vs =
Common::AlignUp<std::size_t>(sizeof(Pica::Shader::VSUniformData), uniform_buffer_alignment); Common::AlignUp<std::size_t>(sizeof(VSUniformData), uniform_buffer_alignment);
uniform_size_aligned_fs = uniform_size_aligned_fs =
Common::AlignUp<std::size_t>(sizeof(Pica::Shader::UniformData), uniform_buffer_alignment); Common::AlignUp<std::size_t>(sizeof(FSUniformData), uniform_buffer_alignment);
// Set vertex attributes for software shader path // Set vertex attributes for software shader path
state.draw.vertex_array = sw_vao.handle; state.draw.vertex_array = sw_vao.handle;
@ -405,16 +408,16 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
// Update scissor uniforms // Update scissor uniforms
const auto [scissor_x1, scissor_y2, scissor_x2, scissor_y1] = fb_helper.Scissor(); const auto [scissor_x1, scissor_y2, scissor_x2, scissor_y1] = fb_helper.Scissor();
if (uniform_block_data.data.scissor_x1 != scissor_x1 || if (fs_uniform_block_data.data.scissor_x1 != scissor_x1 ||
uniform_block_data.data.scissor_x2 != scissor_x2 || fs_uniform_block_data.data.scissor_x2 != scissor_x2 ||
uniform_block_data.data.scissor_y1 != scissor_y1 || fs_uniform_block_data.data.scissor_y1 != scissor_y1 ||
uniform_block_data.data.scissor_y2 != scissor_y2) { fs_uniform_block_data.data.scissor_y2 != scissor_y2) {
uniform_block_data.data.scissor_x1 = scissor_x1; fs_uniform_block_data.data.scissor_x1 = scissor_x1;
uniform_block_data.data.scissor_x2 = scissor_x2; fs_uniform_block_data.data.scissor_x2 = scissor_x2;
uniform_block_data.data.scissor_y1 = scissor_y1; fs_uniform_block_data.data.scissor_y1 = scissor_y1;
uniform_block_data.data.scissor_y2 = scissor_y2; fs_uniform_block_data.data.scissor_y2 = scissor_y2;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
// Sync and bind the texture surfaces // Sync and bind the texture surfaces
@ -831,9 +834,9 @@ void RasterizerOpenGL::SyncBlendColor() {
state.blend.color.blue = blend_color[2]; state.blend.color.blue = blend_color[2];
state.blend.color.alpha = blend_color[3]; state.blend.color.alpha = blend_color[3];
if (blend_color != uniform_block_data.data.blend_color) { if (blend_color != fs_uniform_block_data.data.blend_color) {
uniform_block_data.data.blend_color = blend_color; fs_uniform_block_data.data.blend_color = blend_color;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
} }
@ -921,7 +924,7 @@ void RasterizerOpenGL::SyncAndUploadLUTsLF() {
sizeof(Common::Vec2f) * 256 * Pica::LightingRegs::NumLightingSampler + sizeof(Common::Vec2f) * 256 * Pica::LightingRegs::NumLightingSampler +
sizeof(Common::Vec2f) * 128; // fog sizeof(Common::Vec2f) * 128; // fog
if (!uniform_block_data.lighting_lut_dirty_any && !uniform_block_data.fog_lut_dirty) { if (!fs_uniform_block_data.lighting_lut_dirty_any && !fs_uniform_block_data.fog_lut_dirty) {
return; return;
} }
@ -931,9 +934,9 @@ void RasterizerOpenGL::SyncAndUploadLUTsLF() {
texture_lf_buffer.Map(max_size, sizeof(Common::Vec4f)); texture_lf_buffer.Map(max_size, sizeof(Common::Vec4f));
// Sync the lighting luts // Sync the lighting luts
if (uniform_block_data.lighting_lut_dirty_any || invalidate) { if (fs_uniform_block_data.lighting_lut_dirty_any || invalidate) {
for (unsigned index = 0; index < uniform_block_data.lighting_lut_dirty.size(); index++) { for (unsigned index = 0; index < fs_uniform_block_data.lighting_lut_dirty.size(); index++) {
if (uniform_block_data.lighting_lut_dirty[index] || invalidate) { if (fs_uniform_block_data.lighting_lut_dirty[index] || invalidate) {
std::array<Common::Vec2f, 256> new_data; std::array<Common::Vec2f, 256> new_data;
const auto& source_lut = Pica::g_state.lighting.luts[index]; const auto& source_lut = Pica::g_state.lighting.luts[index];
std::transform(source_lut.begin(), source_lut.end(), new_data.begin(), std::transform(source_lut.begin(), source_lut.end(), new_data.begin(),
@ -945,19 +948,19 @@ void RasterizerOpenGL::SyncAndUploadLUTsLF() {
lighting_lut_data[index] = new_data; lighting_lut_data[index] = new_data;
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f)); new_data.size() * sizeof(Common::Vec2f));
uniform_block_data.data.lighting_lut_offset[index / 4][index % 4] = fs_uniform_block_data.data.lighting_lut_offset[index / 4][index % 4] =
static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec2f)); static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec2f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f); bytes_used += new_data.size() * sizeof(Common::Vec2f);
} }
uniform_block_data.lighting_lut_dirty[index] = false; fs_uniform_block_data.lighting_lut_dirty[index] = false;
} }
} }
uniform_block_data.lighting_lut_dirty_any = false; fs_uniform_block_data.lighting_lut_dirty_any = false;
} }
// Sync the fog lut // Sync the fog lut
if (uniform_block_data.fog_lut_dirty || invalidate) { if (fs_uniform_block_data.fog_lut_dirty || invalidate) {
std::array<Common::Vec2f, 128> new_data; std::array<Common::Vec2f, 128> new_data;
std::transform(Pica::g_state.fog.lut.begin(), Pica::g_state.fog.lut.end(), new_data.begin(), std::transform(Pica::g_state.fog.lut.begin(), Pica::g_state.fog.lut.end(), new_data.begin(),
@ -969,12 +972,12 @@ void RasterizerOpenGL::SyncAndUploadLUTsLF() {
fog_lut_data = new_data; fog_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f)); new_data.size() * sizeof(Common::Vec2f));
uniform_block_data.data.fog_lut_offset = fs_uniform_block_data.data.fog_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f)); static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f); bytes_used += new_data.size() * sizeof(Common::Vec2f);
} }
uniform_block_data.fog_lut_dirty = false; fs_uniform_block_data.fog_lut_dirty = false;
} }
texture_lf_buffer.Unmap(bytes_used); texture_lf_buffer.Unmap(bytes_used);
@ -986,10 +989,10 @@ void RasterizerOpenGL::SyncAndUploadLUTs() {
sizeof(Common::Vec4f) * 256 + // proctex sizeof(Common::Vec4f) * 256 + // proctex
sizeof(Common::Vec4f) * 256; // proctex diff sizeof(Common::Vec4f) * 256; // proctex diff
if (!uniform_block_data.proctex_noise_lut_dirty && if (!fs_uniform_block_data.proctex_noise_lut_dirty &&
!uniform_block_data.proctex_color_map_dirty && !fs_uniform_block_data.proctex_color_map_dirty &&
!uniform_block_data.proctex_alpha_map_dirty && !uniform_block_data.proctex_lut_dirty && !fs_uniform_block_data.proctex_alpha_map_dirty &&
!uniform_block_data.proctex_diff_lut_dirty) { !fs_uniform_block_data.proctex_lut_dirty && !fs_uniform_block_data.proctex_diff_lut_dirty) {
return; return;
} }
@ -1012,34 +1015,34 @@ void RasterizerOpenGL::SyncAndUploadLUTs() {
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f)); new_data.size() * sizeof(Common::Vec2f));
lut_offset = static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec2f)); lut_offset = static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec2f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f); bytes_used += new_data.size() * sizeof(Common::Vec2f);
} }
}; };
// Sync the proctex noise lut // Sync the proctex noise lut
if (uniform_block_data.proctex_noise_lut_dirty || invalidate) { if (fs_uniform_block_data.proctex_noise_lut_dirty || invalidate) {
sync_proc_tex_value_lut(Pica::g_state.proctex.noise_table, proctex_noise_lut_data, sync_proc_tex_value_lut(Pica::g_state.proctex.noise_table, proctex_noise_lut_data,
uniform_block_data.data.proctex_noise_lut_offset); fs_uniform_block_data.data.proctex_noise_lut_offset);
uniform_block_data.proctex_noise_lut_dirty = false; fs_uniform_block_data.proctex_noise_lut_dirty = false;
} }
// Sync the proctex color map // Sync the proctex color map
if (uniform_block_data.proctex_color_map_dirty || invalidate) { if (fs_uniform_block_data.proctex_color_map_dirty || invalidate) {
sync_proc_tex_value_lut(Pica::g_state.proctex.color_map_table, proctex_color_map_data, sync_proc_tex_value_lut(Pica::g_state.proctex.color_map_table, proctex_color_map_data,
uniform_block_data.data.proctex_color_map_offset); fs_uniform_block_data.data.proctex_color_map_offset);
uniform_block_data.proctex_color_map_dirty = false; fs_uniform_block_data.proctex_color_map_dirty = false;
} }
// Sync the proctex alpha map // Sync the proctex alpha map
if (uniform_block_data.proctex_alpha_map_dirty || invalidate) { if (fs_uniform_block_data.proctex_alpha_map_dirty || invalidate) {
sync_proc_tex_value_lut(Pica::g_state.proctex.alpha_map_table, proctex_alpha_map_data, sync_proc_tex_value_lut(Pica::g_state.proctex.alpha_map_table, proctex_alpha_map_data,
uniform_block_data.data.proctex_alpha_map_offset); fs_uniform_block_data.data.proctex_alpha_map_offset);
uniform_block_data.proctex_alpha_map_dirty = false; fs_uniform_block_data.proctex_alpha_map_dirty = false;
} }
// Sync the proctex lut // Sync the proctex lut
if (uniform_block_data.proctex_lut_dirty || invalidate) { if (fs_uniform_block_data.proctex_lut_dirty || invalidate) {
std::array<Common::Vec4f, 256> new_data; std::array<Common::Vec4f, 256> new_data;
std::transform(Pica::g_state.proctex.color_table.begin(), std::transform(Pica::g_state.proctex.color_table.begin(),
@ -1053,16 +1056,16 @@ void RasterizerOpenGL::SyncAndUploadLUTs() {
proctex_lut_data = new_data; proctex_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec4f)); new_data.size() * sizeof(Common::Vec4f));
uniform_block_data.data.proctex_lut_offset = fs_uniform_block_data.data.proctex_lut_offset =
static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec4f)); static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec4f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec4f); bytes_used += new_data.size() * sizeof(Common::Vec4f);
} }
uniform_block_data.proctex_lut_dirty = false; fs_uniform_block_data.proctex_lut_dirty = false;
} }
// Sync the proctex difference lut // Sync the proctex difference lut
if (uniform_block_data.proctex_diff_lut_dirty || invalidate) { if (fs_uniform_block_data.proctex_diff_lut_dirty || invalidate) {
std::array<Common::Vec4f, 256> new_data; std::array<Common::Vec4f, 256> new_data;
std::transform(Pica::g_state.proctex.color_diff_table.begin(), std::transform(Pica::g_state.proctex.color_diff_table.begin(),
@ -1076,12 +1079,12 @@ void RasterizerOpenGL::SyncAndUploadLUTs() {
proctex_diff_lut_data = new_data; proctex_diff_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec4f)); new_data.size() * sizeof(Common::Vec4f));
uniform_block_data.data.proctex_diff_lut_offset = fs_uniform_block_data.data.proctex_diff_lut_offset =
static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec4f)); static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec4f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec4f); bytes_used += new_data.size() * sizeof(Common::Vec4f);
} }
uniform_block_data.proctex_diff_lut_dirty = false; fs_uniform_block_data.proctex_diff_lut_dirty = false;
} }
texture_buffer.Unmap(bytes_used); texture_buffer.Unmap(bytes_used);
@ -1092,38 +1095,47 @@ void RasterizerOpenGL::UploadUniforms(bool accelerate_draw) {
state.draw.uniform_buffer = uniform_buffer.GetHandle(); state.draw.uniform_buffer = uniform_buffer.GetHandle();
state.Apply(); state.Apply();
const bool sync_vs = accelerate_draw; const bool sync_vs_pica = accelerate_draw;
const bool sync_fs = uniform_block_data.dirty; const bool sync_vs = vs_uniform_block_data.dirty;
if (!sync_vs && !sync_fs) { const bool sync_fs = fs_uniform_block_data.dirty;
if (!sync_vs_pica && !sync_vs && !sync_fs) {
return; return;
} }
std::size_t uniform_size = uniform_size_aligned_vs + uniform_size_aligned_fs; std::size_t uniform_size =
uniform_size_aligned_vs_pica + uniform_size_aligned_vs + uniform_size_aligned_fs;
std::size_t used_bytes = 0; std::size_t used_bytes = 0;
const auto [uniforms, offset, invalidate] = const auto [uniforms, offset, invalidate] =
uniform_buffer.Map(uniform_size, uniform_buffer_alignment); uniform_buffer.Map(uniform_size, uniform_buffer_alignment);
if (sync_vs) { if (sync_vs || invalidate) {
Pica::Shader::VSUniformData vs_uniforms; std::memcpy(uniforms + used_bytes, &vs_uniform_block_data.data,
vs_uniforms.uniforms.SetFromRegs(regs.vs, Pica::g_state.vs); sizeof(vs_uniform_block_data.data));
std::memcpy(uniforms + used_bytes, &vs_uniforms, sizeof(vs_uniforms)); glBindBufferRange(GL_UNIFORM_BUFFER, UniformBindings::VSData, uniform_buffer.GetHandle(),
glBindBufferRange(GL_UNIFORM_BUFFER, static_cast<GLuint>(Pica::Shader::UniformBindings::VS), offset + used_bytes, sizeof(vs_uniform_block_data.data));
uniform_buffer.GetHandle(), offset + used_bytes, vs_uniform_block_data.dirty = false;
sizeof(Pica::Shader::VSUniformData));
used_bytes += uniform_size_aligned_vs; used_bytes += uniform_size_aligned_vs;
} }
if (sync_fs || invalidate) { if (sync_fs || invalidate) {
std::memcpy(uniforms + used_bytes, &uniform_block_data.data, std::memcpy(uniforms + used_bytes, &fs_uniform_block_data.data,
sizeof(Pica::Shader::UniformData)); sizeof(fs_uniform_block_data.data));
glBindBufferRange( glBindBufferRange(GL_UNIFORM_BUFFER, UniformBindings::FSData, uniform_buffer.GetHandle(),
GL_UNIFORM_BUFFER, static_cast<GLuint>(Pica::Shader::UniformBindings::Common), offset + used_bytes, sizeof(fs_uniform_block_data.data));
uniform_buffer.GetHandle(), offset + used_bytes, sizeof(Pica::Shader::UniformData)); fs_uniform_block_data.dirty = false;
uniform_block_data.dirty = false;
used_bytes += uniform_size_aligned_fs; used_bytes += uniform_size_aligned_fs;
} }
if (sync_vs_pica) {
VSPicaUniformData vs_uniforms;
vs_uniforms.uniforms.SetFromRegs(regs.vs, Pica::g_state.vs);
std::memcpy(uniforms + used_bytes, &vs_uniforms, sizeof(vs_uniforms));
glBindBufferRange(GL_UNIFORM_BUFFER, UniformBindings::VSPicaData,
uniform_buffer.GetHandle(), offset + used_bytes, sizeof(vs_uniforms));
used_bytes += uniform_size_aligned_vs_pica;
}
uniform_buffer.Unmap(used_bytes); uniform_buffer.Unmap(used_bytes);
} }

1
src/video_core/renderer_opengl/gl_rasterizer.h
View file

@ -148,6 +148,7 @@ private:
OGLStreamBuffer texture_buffer; OGLStreamBuffer texture_buffer;
OGLStreamBuffer texture_lf_buffer; OGLStreamBuffer texture_lf_buffer;
GLint uniform_buffer_alignment; GLint uniform_buffer_alignment;
std::size_t uniform_size_aligned_vs_pica;
std::size_t uniform_size_aligned_vs; std::size_t uniform_size_aligned_vs;
std::size_t uniform_size_aligned_fs; std::size_t uniform_size_aligned_fs;

29
src/video_core/renderer_opengl/gl_shader_decompiler.h
View file

@ -1,29 +0,0 @@
// Copyright 2017 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <array>
#include <functional>
#include <optional>
#include <string>
#include "common/common_types.h"
#include "video_core/shader/shader.h"
namespace OpenGL::ShaderDecompiler {
using RegGetter = std::function<std::string(u32)>;
struct ProgramResult {
std::string code;
};
std::string GetCommonDeclarations();
std::optional<ProgramResult> DecompileProgram(const Pica::Shader::ProgramCode& program_code,
const Pica::Shader::SwizzleData& swizzle_data,
u32 main_offset, const RegGetter& inputreg_getter,
const RegGetter& outputreg_getter, bool sanitize_mul);
} // namespace OpenGL::ShaderDecompiler

19
src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
View file

@ -297,35 +297,33 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCache::LoadDecompiledEntry()
} }
ShaderDiskCacheDecompiled entry; ShaderDiskCacheDecompiled entry;
entry.result.code = std::move(code); entry.code = std::move(code);
entry.sanitize_mul = sanitize_mul; entry.sanitize_mul = sanitize_mul;
return entry; return entry;
} }
void ShaderDiskCache::SaveDecompiledToFile(FileUtil::IOFile& file, u64 unique_identifier, void ShaderDiskCache::SaveDecompiledToFile(FileUtil::IOFile& file, u64 unique_identifier,
const ShaderDecompiler::ProgramResult& result, const std::string& code, bool sanitize_mul) {
bool sanitize_mul) {
if (!IsUsable()) if (!IsUsable())
return; return;
if (file.WriteObject(static_cast<u32>(PrecompiledEntryKind::Decompiled)) != 1 || if (file.WriteObject(static_cast<u32>(PrecompiledEntryKind::Decompiled)) != 1 ||
file.WriteObject(unique_identifier) != 1 || file.WriteObject(sanitize_mul) != 1 || file.WriteObject(unique_identifier) != 1 || file.WriteObject(sanitize_mul) != 1 ||
file.WriteObject(static_cast<u32>(result.code.size())) != 1 || file.WriteObject(static_cast<u32>(code.size())) != 1 ||
file.WriteArray(result.code.data(), result.code.size()) != result.code.size()) { file.WriteArray(code.data(), code.size()) != code.size()) {
LOG_ERROR(Render_OpenGL, "Failed to save decompiled cache entry - removing"); LOG_ERROR(Render_OpenGL, "Failed to save decompiled cache entry - removing");
file.Close(); file.Close();
InvalidatePrecompiled(); InvalidatePrecompiled();
} }
} }
bool ShaderDiskCache::SaveDecompiledToCache(u64 unique_identifier, bool ShaderDiskCache::SaveDecompiledToCache(u64 unique_identifier, const std::string& code,
const ShaderDecompiler::ProgramResult& result,
bool sanitize_mul) { bool sanitize_mul) {
if (!SaveObjectToPrecompiled(static_cast<u32>(PrecompiledEntryKind::Decompiled)) || if (!SaveObjectToPrecompiled(static_cast<u32>(PrecompiledEntryKind::Decompiled)) ||
!SaveObjectToPrecompiled(unique_identifier) || !SaveObjectToPrecompiled(sanitize_mul) || !SaveObjectToPrecompiled(unique_identifier) || !SaveObjectToPrecompiled(sanitize_mul) ||
!SaveObjectToPrecompiled(static_cast<u32>(result.code.size())) || !SaveObjectToPrecompiled(static_cast<u32>(code.size())) ||
!SaveArrayToPrecompiled(result.code.data(), result.code.size())) { !SaveArrayToPrecompiled(code.data(), code.size())) {
return false; return false;
} }
@ -374,8 +372,7 @@ void ShaderDiskCache::SaveRaw(const ShaderDiskCacheRaw& entry) {
transferable_file.Flush(); transferable_file.Flush();
} }
void ShaderDiskCache::SaveDecompiled(u64 unique_identifier, void ShaderDiskCache::SaveDecompiled(u64 unique_identifier, const std::string& code,
const ShaderDecompiler::ProgramResult& code,
bool sanitize_mul) { bool sanitize_mul) {
if (!IsUsable()) if (!IsUsable())
return; return;

14
src/video_core/renderer_opengl/gl_shader_disk_cache.h
View file

@ -20,8 +20,7 @@
#include "common/common_types.h" #include "common/common_types.h"
#include "common/file_util.h" #include "common/file_util.h"
#include "video_core/regs.h" #include "video_core/regs.h"
#include "video_core/renderer_opengl/gl_shader_decompiler.h" #include "video_core/shader/generator/glsl_shader_gen.h"
#include "video_core/renderer_opengl/gl_shader_gen.h"
namespace Core { namespace Core {
class System; class System;
@ -38,6 +37,7 @@ struct ShaderDiskCacheDump;
using RawShaderConfig = Pica::Regs; using RawShaderConfig = Pica::Regs;
using ProgramCode = std::vector<u32>; using ProgramCode = std::vector<u32>;
using ProgramType = Pica::Shader::Generator::ProgramType;
using ShaderDecompiledMap = std::unordered_map<u64, ShaderDiskCacheDecompiled>; using ShaderDecompiledMap = std::unordered_map<u64, ShaderDiskCacheDecompiled>;
using ShaderDumpsMap = std::unordered_map<u64, ShaderDiskCacheDump>; using ShaderDumpsMap = std::unordered_map<u64, ShaderDiskCacheDump>;
@ -78,7 +78,7 @@ private:
/// Contains decompiled data from a shader /// Contains decompiled data from a shader
struct ShaderDiskCacheDecompiled { struct ShaderDiskCacheDecompiled {
ShaderDecompiler::ProgramResult result; std::string code;
bool sanitize_mul; bool sanitize_mul;
}; };
@ -109,8 +109,7 @@ public:
void SaveRaw(const ShaderDiskCacheRaw& entry); void SaveRaw(const ShaderDiskCacheRaw& entry);
/// Saves a decompiled entry to the precompiled file. Does not check for collisions. /// Saves a decompiled entry to the precompiled file. Does not check for collisions.
void SaveDecompiled(u64 unique_identifier, const ShaderDecompiler::ProgramResult& code, void SaveDecompiled(u64 unique_identifier, const std::string& code, bool sanitize_mul);
bool sanitize_mul);
/// Saves a dump entry to the precompiled file. Does not check for collisions. /// Saves a dump entry to the precompiled file. Does not check for collisions.
void SaveDump(u64 unique_identifier, GLuint program); void SaveDump(u64 unique_identifier, GLuint program);
@ -132,11 +131,10 @@ private:
/// Saves a decompiled entry to the passed file. Does not check for collisions. /// Saves a decompiled entry to the passed file. Does not check for collisions.
void SaveDecompiledToFile(FileUtil::IOFile& file, u64 unique_identifier, void SaveDecompiledToFile(FileUtil::IOFile& file, u64 unique_identifier,
const ShaderDecompiler::ProgramResult& code, bool sanitize_mul); const std::string& code, bool sanitize_mul);
/// Saves a decompiled entry to the virtual precompiled cache. Does not check for collisions. /// Saves a decompiled entry to the virtual precompiled cache. Does not check for collisions.
bool SaveDecompiledToCache(u64 unique_identifier, const ShaderDecompiler::ProgramResult& code, bool SaveDecompiledToCache(u64 unique_identifier, const std::string& code, bool sanitize_mul);
bool sanitize_mul);
/// Returns if the cache can be used /// Returns if the cache can be used
bool IsUsable() const; bool IsUsable() const;

1941
src/video_core/renderer_opengl/gl_shader_gen.cpp
View file

File diff suppressed because it is too large Load diff

266
src/video_core/renderer_opengl/gl_shader_gen.h
View file

@ -1,266 +0,0 @@
// Copyright 2015 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <functional>
#include <optional>
#include "common/hash.h"
#include "video_core/regs.h"
#include "video_core/shader/shader.h"
namespace OpenGL {
class Driver;
namespace ShaderDecompiler {
struct ProgramResult;
}
enum class ProgramType : u32 { VS, GS, FS };
enum Attributes {
ATTRIBUTE_POSITION,
ATTRIBUTE_COLOR,
ATTRIBUTE_TEXCOORD0,
ATTRIBUTE_TEXCOORD1,
ATTRIBUTE_TEXCOORD2,
ATTRIBUTE_TEXCOORD0_W,
ATTRIBUTE_NORMQUAT,
ATTRIBUTE_VIEW,
};
// Doesn't include const_color because we don't sync it, see comment in BuildFromRegs()
struct TevStageConfigRaw {
u32 sources_raw;
u32 modifiers_raw;
u32 ops_raw;
u32 scales_raw;
explicit operator Pica::TexturingRegs::TevStageConfig() const noexcept {
Pica::TexturingRegs::TevStageConfig stage;
stage.sources_raw = sources_raw;
stage.modifiers_raw = modifiers_raw;
stage.ops_raw = ops_raw;
stage.const_color = 0;
stage.scales_raw = scales_raw;
return stage;
}
};
struct PicaFSConfigState {
Pica::FramebufferRegs::CompareFunc alpha_test_func;
Pica::RasterizerRegs::ScissorMode scissor_test_mode;
Pica::TexturingRegs::TextureConfig::TextureType texture0_type;
bool texture2_use_coord1;
std::array<TevStageConfigRaw, 6> tev_stages;
u8 combiner_buffer_input;
Pica::RasterizerRegs::DepthBuffering depthmap_enable;
Pica::TexturingRegs::FogMode fog_mode;
bool fog_flip;
bool alphablend_enable;
Pica::FramebufferRegs::LogicOp logic_op;
struct {
struct {
unsigned num;
bool directional;
bool two_sided_diffuse;
bool dist_atten_enable;
bool spot_atten_enable;
bool geometric_factor_0;
bool geometric_factor_1;
bool shadow_enable;
} light[8];
bool enable;
unsigned src_num;
Pica::LightingRegs::LightingBumpMode bump_mode;
unsigned bump_selector;
bool bump_renorm;
bool clamp_highlights;
Pica::LightingRegs::LightingConfig config;
bool enable_primary_alpha;
bool enable_secondary_alpha;
bool enable_shadow;
bool shadow_primary;
bool shadow_secondary;
bool shadow_invert;
bool shadow_alpha;
unsigned shadow_selector;
struct {
bool enable;
bool abs_input;
Pica::LightingRegs::LightingLutInput type;
float scale;
} lut_d0, lut_d1, lut_sp, lut_fr, lut_rr, lut_rg, lut_rb;
} lighting;
struct {
bool enable;
u32 coord;
Pica::TexturingRegs::ProcTexClamp u_clamp, v_clamp;
Pica::TexturingRegs::ProcTexCombiner color_combiner, alpha_combiner;
bool separate_alpha;
bool noise_enable;
Pica::TexturingRegs::ProcTexShift u_shift, v_shift;
u32 lut_width;
u32 lut_offset0;
u32 lut_offset1;
u32 lut_offset2;
u32 lut_offset3;
u32 lod_min;
u32 lod_max;
Pica::TexturingRegs::ProcTexFilter lut_filter;
} proctex;
struct {
bool emulate_blending;
Pica::FramebufferRegs::BlendEquation eq;
Pica::FramebufferRegs::BlendFactor src_factor;
Pica::FramebufferRegs::BlendFactor dst_factor;
} rgb_blend, alpha_blend;
bool shadow_rendering;
bool shadow_texture_orthographic;
bool use_custom_normal_map;
};
/**
* This struct contains all state used to generate the GLSL fragment shader that emulates the
* current Pica register configuration. This struct is used as a cache key for generated GLSL shader
* programs. The functions in gl_shader_gen.cpp should retrieve state from this struct only, not by
* directly accessing Pica registers. This should reduce the risk of bugs in shader generation where
* Pica state is not being captured in the shader cache key, thereby resulting in (what should be)
* two separate shaders sharing the same key.
*/
struct PicaFSConfig : Common::HashableStruct<PicaFSConfigState> {
/// Construct a PicaFSConfig with the given Pica register configuration.
static PicaFSConfig BuildFromRegs(const Pica::Regs& regs, bool has_blend_minmax_factor,
bool use_normal = false);
bool TevStageUpdatesCombinerBufferColor(unsigned stage_index) const {
return (stage_index < 4) && (state.combiner_buffer_input & (1 << stage_index));
}
bool TevStageUpdatesCombinerBufferAlpha(unsigned stage_index) const {
return (stage_index < 4) && ((state.combiner_buffer_input >> 4) & (1 << stage_index));
}
};
/**
* This struct contains common information to identify a GL vertex/geometry shader generated from
* PICA vertex/geometry shader.
*/
struct PicaShaderConfigCommon {
void Init(const Pica::ShaderRegs& regs, Pica::Shader::ShaderSetup& setup);
u64 program_hash;
u64 swizzle_hash;
u32 main_offset;
bool sanitize_mul;
u32 num_outputs;
// output_map[output register index] -> output attribute index
std::array<u32, 16> output_map;
};
/**
* This struct contains information to identify a GL vertex shader generated from PICA vertex
* shader.
*/
struct PicaVSConfig : Common::HashableStruct<PicaShaderConfigCommon> {
explicit PicaVSConfig(const Pica::ShaderRegs& regs, Pica::Shader::ShaderSetup& setup) {
state.Init(regs, setup);
}
explicit PicaVSConfig(const PicaShaderConfigCommon& conf) {
state = conf;
}
};
struct PicaGSConfigCommonRaw {
void Init(const Pica::Regs& regs);
u32 vs_output_attributes;
u32 gs_output_attributes;
struct SemanticMap {
u32 attribute_index;
u32 component_index;
};
// semantic_maps[semantic name] -> GS output attribute index + component index
std::array<SemanticMap, 24> semantic_maps;
};
/**
* This struct contains information to identify a GL geometry shader generated from PICA no-geometry
* shader pipeline
*/
struct PicaFixedGSConfig : Common::HashableStruct<PicaGSConfigCommonRaw> {
explicit PicaFixedGSConfig(const Pica::Regs& regs) {
state.Init(regs);
}
};
/**
* Generates the GLSL vertex shader program source code that accepts vertices from software shader
* and directly passes them to the fragment shader.
* @param separable_shader generates shader that can be used for separate shader object
* @returns String of the shader source code
*/
ShaderDecompiler::ProgramResult GenerateTrivialVertexShader(bool separable_shader);
/**
* Generates the GLSL vertex shader program source code for the given VS program
* @returns String of the shader source code; boost::none on failure
*/
std::optional<ShaderDecompiler::ProgramResult> GenerateVertexShader(
const Pica::Shader::ShaderSetup& setup, const PicaVSConfig& config, bool separable_shader);
/*
* Generates the GLSL fixed geometry shader program source code for non-GS PICA pipeline
* @returns String of the shader source code
*/
ShaderDecompiler::ProgramResult GenerateFixedGeometryShader(const PicaFixedGSConfig& config,
bool separable_shader);
/**
* Generates the GLSL fragment shader program source code for the current Pica state
* @param config ShaderCacheKey object generated for the current Pica state, used for the shader
* configuration (NOTE: Use state in this struct only, not the Pica registers!)
* @param separable_shader generates shader that can be used for separate shader object
* @returns String of the shader source code
*/
ShaderDecompiler::ProgramResult GenerateFragmentShader(const PicaFSConfig& config,
bool separable_shader);
} // namespace OpenGL
namespace std {
template <>
struct hash<OpenGL::PicaFSConfig> {
std::size_t operator()(const OpenGL::PicaFSConfig& k) const noexcept {
return k.Hash();
}
};
template <>
struct hash<OpenGL::PicaVSConfig> {
std::size_t operator()(const OpenGL::PicaVSConfig& k) const noexcept {
return k.Hash();
}
};
template <>
struct hash<OpenGL::PicaFixedGSConfig> {
std::size_t operator()(const OpenGL::PicaFixedGSConfig& k) const noexcept {
return k.Hash();
}
};
} // namespace std

90
src/video_core/renderer_opengl/gl_shader_manager.cpp
View file

@ -14,9 +14,11 @@
#include "video_core/renderer_opengl/gl_shader_disk_cache.h" #include "video_core/renderer_opengl/gl_shader_disk_cache.h"
#include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/gl_shader_manager.h"
#include "video_core/renderer_opengl/gl_state.h" #include "video_core/renderer_opengl/gl_state.h"
#include "video_core/shader/shader_uniforms.h" #include "video_core/shader/generator/shader_uniforms.h"
#include "video_core/video_core.h" #include "video_core/video_core.h"
using namespace Pica::Shader::Generator;
namespace OpenGL { namespace OpenGL {
static u64 GetUniqueIdentifier(const Pica::Regs& regs, const ProgramCode& code) { static u64 GetUniqueIdentifier(const Pica::Regs& regs, const ProgramCode& code) {
@ -74,7 +76,7 @@ static std::set<GLenum> GetSupportedFormats() {
} }
static std::tuple<PicaVSConfig, Pica::Shader::ShaderSetup> BuildVSConfigFromRaw( static std::tuple<PicaVSConfig, Pica::Shader::ShaderSetup> BuildVSConfigFromRaw(
const ShaderDiskCacheRaw& raw) { const ShaderDiskCacheRaw& raw, const Driver& driver) {
Pica::Shader::ProgramCode program_code{}; Pica::Shader::ProgramCode program_code{};
Pica::Shader::SwizzleData swizzle_data{}; Pica::Shader::SwizzleData swizzle_data{};
std::copy_n(raw.GetProgramCode().begin(), Pica::Shader::MAX_PROGRAM_CODE_LENGTH, std::copy_n(raw.GetProgramCode().begin(), Pica::Shader::MAX_PROGRAM_CODE_LENGTH,
@ -84,7 +86,8 @@ static std::tuple<PicaVSConfig, Pica::Shader::ShaderSetup> BuildVSConfigFromRaw(
Pica::Shader::ShaderSetup setup; Pica::Shader::ShaderSetup setup;
setup.program_code = program_code; setup.program_code = program_code;
setup.swizzle_data = swizzle_data; setup.swizzle_data = swizzle_data;
return {PicaVSConfig{raw.GetRawShaderConfig().vs, setup}, setup}; return {PicaVSConfig{raw.GetRawShaderConfig(), setup, driver.HasClipCullDistance(), true},
setup};
} }
/** /**
@ -130,8 +133,10 @@ private:
class TrivialVertexShader { class TrivialVertexShader {
public: public:
explicit TrivialVertexShader(bool separable) : program(separable) { explicit TrivialVertexShader(const Driver& driver, bool separable) : program(separable) {
program.Create(GenerateTrivialVertexShader(separable).code.c_str(), GL_VERTEX_SHADER); const auto code =
GLSL::GenerateTrivialVertexShader(driver.HasClipCullDistance(), separable);
program.Create(code.c_str(), GL_VERTEX_SHADER);
} }
GLuint Get() const { GLuint Get() const {
return program.GetHandle(); return program.GetHandle();
@ -141,20 +146,18 @@ private:
OGLShaderStage program; OGLShaderStage program;
}; };
template <typename KeyConfigType, template <typename KeyConfigType, std::string (*CodeGenerator)(const KeyConfigType&, bool),
ShaderDecompiler::ProgramResult (*CodeGenerator)(const KeyConfigType&, bool),
GLenum ShaderType> GLenum ShaderType>
class ShaderCache { class ShaderCache {
public: public:
explicit ShaderCache(bool separable) : separable(separable) {} explicit ShaderCache(bool separable) : separable(separable) {}
std::tuple<GLuint, std::optional<ShaderDecompiler::ProgramResult>> Get( std::tuple<GLuint, std::optional<std::string>> Get(const KeyConfigType& config) {
const KeyConfigType& config) {
auto [iter, new_shader] = shaders.emplace(config, OGLShaderStage{separable}); auto [iter, new_shader] = shaders.emplace(config, OGLShaderStage{separable});
OGLShaderStage& cached_shader = iter->second; OGLShaderStage& cached_shader = iter->second;
std::optional<ShaderDecompiler::ProgramResult> result{}; std::optional<std::string> result{};
if (new_shader) { if (new_shader) {
result = CodeGenerator(config, separable); result = CodeGenerator(config, separable);
cached_shader.Create(result->code.c_str(), ShaderType); cached_shader.Create(result->c_str(), ShaderType);
} }
return {cached_shader.GetHandle(), std::move(result)}; return {cached_shader.GetHandle(), std::move(result)};
} }
@ -180,29 +183,27 @@ private:
// program buffer from the previous shader, which is hashed into the config, resulting several // program buffer from the previous shader, which is hashed into the config, resulting several
// different config values from the same shader program. // different config values from the same shader program.
template <typename KeyConfigType, template <typename KeyConfigType,
std::optional<ShaderDecompiler::ProgramResult> (*CodeGenerator)( std::string (*CodeGenerator)(const Pica::Shader::ShaderSetup&, const KeyConfigType&,
const Pica::Shader::ShaderSetup&, const KeyConfigType&, bool), bool),
GLenum ShaderType> GLenum ShaderType>
class ShaderDoubleCache { class ShaderDoubleCache {
public: public:
explicit ShaderDoubleCache(bool separable) : separable(separable) {} explicit ShaderDoubleCache(bool separable) : separable(separable) {}
std::tuple<GLuint, std::optional<ShaderDecompiler::ProgramResult>> Get( std::tuple<GLuint, std::optional<std::string>> Get(const KeyConfigType& key,
const KeyConfigType& key, const Pica::Shader::ShaderSetup& setup) { const Pica::Shader::ShaderSetup& setup) {
std::optional<ShaderDecompiler::ProgramResult> result{}; std::optional<std::string> result{};
auto map_it = shader_map.find(key); auto map_it = shader_map.find(key);
if (map_it == shader_map.end()) { if (map_it == shader_map.end()) {
auto program_opt = CodeGenerator(setup, key, separable); auto program = CodeGenerator(setup, key, separable);
if (!program_opt) { if (program.empty()) {
shader_map[key] = nullptr; shader_map[key] = nullptr;
return {0, std::nullopt}; return {0, std::nullopt};
} }
std::string& program = program_opt->code;
auto [iter, new_shader] = shader_cache.emplace(program, OGLShaderStage{separable}); auto [iter, new_shader] = shader_cache.emplace(program, OGLShaderStage{separable});
OGLShaderStage& cached_shader = iter->second; OGLShaderStage& cached_shader = iter->second;
if (new_shader) { if (new_shader) {
result.emplace(); result = program;
result->code = program;
cached_shader.Create(program.c_str(), ShaderType); cached_shader.Create(program.c_str(), ShaderType);
} }
shader_map[key] = &cached_shader; shader_map[key] = &cached_shader;
@ -237,18 +238,19 @@ private:
}; };
using ProgrammableVertexShaders = using ProgrammableVertexShaders =
ShaderDoubleCache<PicaVSConfig, &GenerateVertexShader, GL_VERTEX_SHADER>; ShaderDoubleCache<PicaVSConfig, &GLSL::GenerateVertexShader, GL_VERTEX_SHADER>;
using FixedGeometryShaders = using FixedGeometryShaders =
ShaderCache<PicaFixedGSConfig, &GenerateFixedGeometryShader, GL_GEOMETRY_SHADER>; ShaderCache<PicaFixedGSConfig, &GLSL::GenerateFixedGeometryShader, GL_GEOMETRY_SHADER>;
using FragmentShaders = ShaderCache<PicaFSConfig, &GenerateFragmentShader, GL_FRAGMENT_SHADER>; using FragmentShaders =
ShaderCache<PicaFSConfig, &GLSL::GenerateFragmentShader, GL_FRAGMENT_SHADER>;
class ShaderProgramManager::Impl { class ShaderProgramManager::Impl {
public: public:
explicit Impl(bool separable) explicit Impl(const Driver& driver, bool separable)
: separable(separable), programmable_vertex_shaders(separable), : separable(separable), programmable_vertex_shaders(separable),
trivial_vertex_shader(separable), fixed_geometry_shaders(separable), trivial_vertex_shader(driver, separable), fixed_geometry_shaders(separable),
fragment_shaders(separable), disk_cache(separable) { fragment_shaders(separable), disk_cache(separable) {
if (separable) if (separable)
pipeline.Create(); pipeline.Create();
@ -299,13 +301,13 @@ ShaderProgramManager::ShaderProgramManager(Frontend::EmuWindow& emu_window_, con
bool separable) bool separable)
: emu_window{emu_window_}, driver{driver_}, : emu_window{emu_window_}, driver{driver_},
strict_context_required{emu_window.StrictContextRequired()}, impl{std::make_unique<Impl>( strict_context_required{emu_window.StrictContextRequired()}, impl{std::make_unique<Impl>(
separable)} {} driver_, separable)} {}
ShaderProgramManager::~ShaderProgramManager() = default; ShaderProgramManager::~ShaderProgramManager() = default;
bool ShaderProgramManager::UseProgrammableVertexShader(const Pica::Regs& regs, bool ShaderProgramManager::UseProgrammableVertexShader(const Pica::Regs& regs,
Pica::Shader::ShaderSetup& setup) { Pica::Shader::ShaderSetup& setup) {
PicaVSConfig config{regs.vs, setup}; PicaVSConfig config{regs, setup, driver.HasClipCullDistance(), true};
auto [handle, result] = impl->programmable_vertex_shaders.Get(config, setup); auto [handle, result] = impl->programmable_vertex_shaders.Get(config, setup);
if (handle == 0) if (handle == 0)
return false; return false;
@ -333,7 +335,7 @@ void ShaderProgramManager::UseTrivialVertexShader() {
} }
void ShaderProgramManager::UseFixedGeometryShader(const Pica::Regs& regs) { void ShaderProgramManager::UseFixedGeometryShader(const Pica::Regs& regs) {
PicaFixedGSConfig gs_config(regs); PicaFixedGSConfig gs_config(regs, driver.HasClipCullDistance());
auto [handle, _] = impl->fixed_geometry_shaders.Get(gs_config); auto [handle, _] = impl->fixed_geometry_shaders.Get(gs_config);
impl->current.gs = handle; impl->current.gs = handle;
impl->current.gs_hash = gs_config.Hash(); impl->current.gs_hash = gs_config.Hash();
@ -345,8 +347,8 @@ void ShaderProgramManager::UseTrivialGeometryShader() {
} }
void ShaderProgramManager::UseFragmentShader(const Pica::Regs& regs, bool use_normal) { void ShaderProgramManager::UseFragmentShader(const Pica::Regs& regs, bool use_normal) {
PicaFSConfig config = PicaFSConfig config(regs, false, driver.IsOpenGLES(), false, driver.HasBlendMinMaxFactor(),
PicaFSConfig::BuildFromRegs(regs, driver.HasBlendMinMaxFactor(), use_normal); use_normal);
auto [handle, result] = impl->fragment_shaders.Get(config); auto [handle, result] = impl->fragment_shaders.Get(config);
impl->current.fs = handle; impl->current.fs = handle;
impl->current.fs_hash = config.Hash(); impl->current.fs_hash = config.Hash();
@ -463,12 +465,12 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
// we have both the binary shader and the decompiled, so inject it into the // we have both the binary shader and the decompiled, so inject it into the
// cache // cache
if (raw.GetProgramType() == ProgramType::VS) { if (raw.GetProgramType() == ProgramType::VS) {
auto [conf, setup] = BuildVSConfigFromRaw(raw); auto [conf, setup] = BuildVSConfigFromRaw(raw, driver);
std::scoped_lock lock(mutex); std::scoped_lock lock(mutex);
impl->programmable_vertex_shaders.Inject(conf, decomp->second.result.code, impl->programmable_vertex_shaders.Inject(conf, decomp->second.code,
std::move(shader)); std::move(shader));
} else if (raw.GetProgramType() == ProgramType::FS) { } else if (raw.GetProgramType() == ProgramType::FS) {
PicaFSConfig conf = PicaFSConfig::BuildFromRegs(raw.GetRawShaderConfig(), PicaFSConfig conf(raw.GetRawShaderConfig(), false, driver.IsOpenGLES(), false,
driver.HasBlendMinMaxFactor()); driver.HasBlendMinMaxFactor());
std::scoped_lock lock(mutex); std::scoped_lock lock(mutex);
impl->fragment_shaders.Inject(conf, std::move(shader)); impl->fragment_shaders.Inject(conf, std::move(shader));
@ -566,24 +568,24 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
bool sanitize_mul = false; bool sanitize_mul = false;
GLuint handle{0}; GLuint handle{0};
std::optional<ShaderDecompiler::ProgramResult> result; std::string code;
// Otherwise decompile and build the shader at boot and save the result to the // Otherwise decompile and build the shader at boot and save the result to the
// precompiled file // precompiled file
if (raw.GetProgramType() == ProgramType::VS) { if (raw.GetProgramType() == ProgramType::VS) {
auto [conf, setup] = BuildVSConfigFromRaw(raw); auto [conf, setup] = BuildVSConfigFromRaw(raw, driver);
result = GenerateVertexShader(setup, conf, impl->separable); code = GLSL::GenerateVertexShader(setup, conf, impl->separable);
OGLShaderStage stage{impl->separable}; OGLShaderStage stage{impl->separable};
stage.Create(result->code.c_str(), GL_VERTEX_SHADER); stage.Create(code.c_str(), GL_VERTEX_SHADER);
handle = stage.GetHandle(); handle = stage.GetHandle();
sanitize_mul = conf.state.sanitize_mul; sanitize_mul = conf.state.sanitize_mul;
std::scoped_lock lock(mutex); std::scoped_lock lock(mutex);
impl->programmable_vertex_shaders.Inject(conf, result->code, std::move(stage)); impl->programmable_vertex_shaders.Inject(conf, code, std::move(stage));
} else if (raw.GetProgramType() == ProgramType::FS) { } else if (raw.GetProgramType() == ProgramType::FS) {
PicaFSConfig conf = PicaFSConfig::BuildFromRegs(raw.GetRawShaderConfig(), PicaFSConfig conf(raw.GetRawShaderConfig(), false, driver.IsOpenGLES(), false,
driver.HasBlendMinMaxFactor()); driver.HasBlendMinMaxFactor());
result = GenerateFragmentShader(conf, impl->separable); code = GLSL::GenerateFragmentShader(conf, impl->separable);
OGLShaderStage stage{impl->separable}; OGLShaderStage stage{impl->separable};
stage.Create(result->code.c_str(), GL_FRAGMENT_SHADER); stage.Create(code.c_str(), GL_FRAGMENT_SHADER);
handle = stage.GetHandle(); handle = stage.GetHandle();
std::scoped_lock lock(mutex); std::scoped_lock lock(mutex);
impl->fragment_shaders.Inject(conf, std::move(stage)); impl->fragment_shaders.Inject(conf, std::move(stage));
@ -602,8 +604,8 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
std::scoped_lock lock(mutex); std::scoped_lock lock(mutex);
// If this is a new separable shader, add it the precompiled cache // If this is a new separable shader, add it the precompiled cache
if (result) { if (!code.empty()) {
disk_cache.SaveDecompiled(unique_identifier, *result, sanitize_mul); disk_cache.SaveDecompiled(unique_identifier, code, sanitize_mul);
disk_cache.SaveDump(unique_identifier, handle); disk_cache.SaveDump(unique_identifier, handle);
precompiled_cache_altered = true; precompiled_cache_altered = true;
} }

6
src/video_core/renderer_opengl/gl_shader_manager.h
View file

@ -24,6 +24,12 @@ namespace OpenGL {
class Driver; class Driver;
class OpenGLState; class OpenGLState;
enum UniformBindings {
VSPicaData = 0,
VSData = 1,
FSData = 2,
};
/// A class that manage different shader stages and configures them with given config data. /// A class that manage different shader stages and configures them with given config data.
class ShaderProgramManager { class ShaderProgramManager {
public: public:

16
src/video_core/renderer_opengl/gl_shader_util.cpp
View file

@ -14,9 +14,9 @@
namespace OpenGL { namespace OpenGL {
GLuint LoadShader(std::string_view source, GLenum type) { GLuint LoadShader(std::string_view source, GLenum type) {
const std::string version = GLES ? R"(#version 320 es std::string preamble;
if (GLES) {
#define CITRA_GLES preamble = R"(#version 320 es
#if defined(GL_ANDROID_extension_pack_es31a) #if defined(GL_ANDROID_extension_pack_es31a)
#extension GL_ANDROID_extension_pack_es31a : enable #extension GL_ANDROID_extension_pack_es31a : enable
@ -25,8 +25,10 @@ GLuint LoadShader(std::string_view source, GLenum type) {
#if defined(GL_EXT_clip_cull_distance) #if defined(GL_EXT_clip_cull_distance)
#extension GL_EXT_clip_cull_distance : enable #extension GL_EXT_clip_cull_distance : enable
#endif // defined(GL_EXT_clip_cull_distance) #endif // defined(GL_EXT_clip_cull_distance)
)" )";
: "#version 430 core\n"; } else {
preamble = "#version 430 core\n";
}
std::string_view debug_type; std::string_view debug_type;
switch (type) { switch (type) {
@ -43,8 +45,8 @@ GLuint LoadShader(std::string_view source, GLenum type) {
UNREACHABLE(); UNREACHABLE();
} }
std::array<const GLchar*, 2> src_arr{version.data(), source.data()}; std::array<const GLchar*, 2> src_arr{preamble.data(), source.data()};
std::array<GLint, 2> lengths{static_cast<GLint>(version.size()), std::array<GLint, 2> lengths{static_cast<GLint>(preamble.size()),
static_cast<GLint>(source.size())}; static_cast<GLint>(source.size())};
GLuint shader_id = glCreateShader(type); GLuint shader_id = glCreateShader(type);
glShaderSource(shader_id, static_cast<GLsizei>(src_arr.size()), src_arr.data(), lengths.data()); glShaderSource(shader_id, static_cast<GLsizei>(src_arr.size()), src_arr.data(), lengths.data());

15
src/video_core/renderer_opengl/gl_shader_util.h
View file

@ -9,21 +9,6 @@
namespace OpenGL { namespace OpenGL {
// High precision may or may not supported in GLES3. If it isn't, use medium precision instead.
static constexpr char fragment_shader_precision_OES[] = R"(
#ifdef GL_FRAGMENT_PRECISION_HIGH
precision highp int;
precision highp float;
precision highp samplerBuffer;
precision highp uimage2D;
#else
precision mediump int;
precision mediump float;
precision mediump samplerBuffer;
precision mediump uimage2D;
#endif // GL_FRAGMENT_PRECISION_HIGH
)";
/** /**
* Utility function to create and compile an OpenGL GLSL shader * Utility function to create and compile an OpenGL GLSL shader
* @param source String of the GLSL shader program * @param source String of the GLSL shader program

8
src/video_core/renderer_opengl/renderer_opengl.cpp
View file

@ -11,12 +11,12 @@
#include "core/hw/hw.h" #include "core/hw/hw.h"
#include "core/hw/lcd.h" #include "core/hw/lcd.h"
#include "core/memory.h" #include "core/memory.h"
#include "video_core/renderer_opengl/gl_shader_util.h"
#include "video_core/renderer_opengl/gl_state.h" #include "video_core/renderer_opengl/gl_state.h"
#include "video_core/renderer_opengl/gl_texture_mailbox.h" #include "video_core/renderer_opengl/gl_texture_mailbox.h"
#include "video_core/renderer_opengl/gl_vars.h" #include "video_core/renderer_opengl/gl_vars.h"
#include "video_core/renderer_opengl/post_processing_opengl.h" #include "video_core/renderer_opengl/post_processing_opengl.h"
#include "video_core/renderer_opengl/renderer_opengl.h" #include "video_core/renderer_opengl/renderer_opengl.h"
#include "video_core/shader/generator/glsl_shader_gen.h"
#include "video_core/video_core.h" #include "video_core/video_core.h"
#include "video_core/host_shaders/opengl_present_anaglyph_frag.h" #include "video_core/host_shaders/opengl_present_anaglyph_frag.h"
@ -387,11 +387,7 @@ void RendererOpenGL::InitOpenGLObjects() {
void RendererOpenGL::ReloadShader() { void RendererOpenGL::ReloadShader() {
// Link shaders and get variable locations // Link shaders and get variable locations
std::string shader_data; std::string shader_data = fragment_shader_precision_OES;
if (GLES) {
shader_data += fragment_shader_precision_OES;
}
if (Settings::values.render_3d.GetValue() == Settings::StereoRenderOption::Anaglyph) { if (Settings::values.render_3d.GetValue() == Settings::StereoRenderOption::Anaglyph) {
if (Settings::values.anaglyph_shader_name.GetValue() == "dubois (builtin)") { if (Settings::values.anaglyph_shader_name.GetValue() == "dubois (builtin)") {
shader_data += HostShaders::OPENGL_PRESENT_ANAGLYPH_FRAG; shader_data += HostShaders::OPENGL_PRESENT_ANAGLYPH_FRAG;

3
src/video_core/renderer_vulkan/vk_graphics_pipeline.h
View file

@ -5,7 +5,8 @@
#include "common/thread_worker.h" #include "common/thread_worker.h"
#include "video_core/rasterizer_cache/pixel_format.h" #include "video_core/rasterizer_cache/pixel_format.h"
#include "video_core/renderer_vulkan/vk_common.h" #include "video_core/renderer_vulkan/vk_common.h"
#include "video_core/renderer_vulkan/vk_shader_gen.h" #include "video_core/shader/generator/glsl_shader_gen.h"
#include "video_core/shader/generator/spv_shader_gen.h"
namespace Common { namespace Common {

43
src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
View file

@ -14,19 +14,14 @@
#include "video_core/renderer_vulkan/vk_pipeline_cache.h" #include "video_core/renderer_vulkan/vk_pipeline_cache.h"
#include "video_core/renderer_vulkan/vk_renderpass_cache.h" #include "video_core/renderer_vulkan/vk_renderpass_cache.h"
#include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_shader_gen_spv.h"
#include "video_core/renderer_vulkan/vk_shader_util.h" #include "video_core/renderer_vulkan/vk_shader_util.h"
using namespace Pica::Shader::Generator;
MICROPROFILE_DEFINE(Vulkan_Bind, "Vulkan", "Pipeline Bind", MP_RGB(192, 32, 32)); MICROPROFILE_DEFINE(Vulkan_Bind, "Vulkan", "Pipeline Bind", MP_RGB(192, 32, 32));
namespace Vulkan { namespace Vulkan {
enum ProgramType : u32 {
VS = 0,
GS = 2,
FS = 1,
};
u32 AttribBytes(Pica::PipelineRegs::VertexAttributeFormat format, u32 size) { u32 AttribBytes(Pica::PipelineRegs::VertexAttributeFormat format, u32 size) {
switch (format) { switch (format) {
case Pica::PipelineRegs::VertexAttributeFormat::FLOAT: case Pica::PipelineRegs::VertexAttributeFormat::FLOAT:
@ -52,14 +47,14 @@ AttribLoadFlags MakeAttribLoadFlag(Pica::PipelineRegs::VertexAttributeFormat for
} }
} }
constexpr std::array<vk::DescriptorSetLayoutBinding, 5> BUFFER_BINDINGS = {{ constexpr std::array<vk::DescriptorSetLayoutBinding, 6> BUFFER_BINDINGS = {{
{0, vk::DescriptorType::eUniformBufferDynamic, 1, vk::ShaderStageFlagBits::eVertex}, {0, vk::DescriptorType::eUniformBufferDynamic, 1, vk::ShaderStageFlagBits::eVertex},
{1, vk::DescriptorType::eUniformBufferDynamic, 1, {1, vk::DescriptorType::eUniformBufferDynamic, 1,
vk::ShaderStageFlagBits::eVertex | vk::ShaderStageFlagBits::eGeometry | vk::ShaderStageFlagBits::eVertex | vk::ShaderStageFlagBits::eGeometry},
vk::ShaderStageFlagBits::eFragment}, {2, vk::DescriptorType::eUniformBufferDynamic, 1, vk::ShaderStageFlagBits::eFragment},
{2, vk::DescriptorType::eUniformTexelBuffer, 1, vk::ShaderStageFlagBits::eFragment},
{3, vk::DescriptorType::eUniformTexelBuffer, 1, vk::ShaderStageFlagBits::eFragment}, {3, vk::DescriptorType::eUniformTexelBuffer, 1, vk::ShaderStageFlagBits::eFragment},
{4, vk::DescriptorType::eUniformTexelBuffer, 1, vk::ShaderStageFlagBits::eFragment}, {4, vk::DescriptorType::eUniformTexelBuffer, 1, vk::ShaderStageFlagBits::eFragment},
{5, vk::DescriptorType::eUniformTexelBuffer, 1, vk::ShaderStageFlagBits::eFragment},
}}; }};
constexpr std::array<vk::DescriptorSetLayoutBinding, 4> TEXTURE_BINDINGS = {{ constexpr std::array<vk::DescriptorSetLayoutBinding, 4> TEXTURE_BINDINGS = {{
@ -88,8 +83,9 @@ PipelineCache::PipelineCache(const Instance& instance_, Scheduler& scheduler_,
descriptor_set_providers{DescriptorSetProvider{instance, pool, BUFFER_BINDINGS}, descriptor_set_providers{DescriptorSetProvider{instance, pool, BUFFER_BINDINGS},
DescriptorSetProvider{instance, pool, TEXTURE_BINDINGS}, DescriptorSetProvider{instance, pool, TEXTURE_BINDINGS},
DescriptorSetProvider{instance, pool, SHADOW_BINDINGS}}, DescriptorSetProvider{instance, pool, SHADOW_BINDINGS}},
trivial_vertex_shader{instance, vk::ShaderStageFlagBits::eVertex, trivial_vertex_shader{
GenerateTrivialVertexShader(instance.IsShaderClipDistanceSupported())} { instance, vk::ShaderStageFlagBits::eVertex,
GLSL::GenerateTrivialVertexShader(instance.IsShaderClipDistanceSupported(), true)} {
BuildLayout(); BuildLayout();
} }
@ -294,8 +290,8 @@ bool PipelineCache::BindPipeline(const PipelineInfo& info, bool wait_built) {
bool PipelineCache::UseProgrammableVertexShader(const Pica::Regs& regs, bool PipelineCache::UseProgrammableVertexShader(const Pica::Regs& regs,
Pica::Shader::ShaderSetup& setup, Pica::Shader::ShaderSetup& setup,
const VertexLayout& layout) { const VertexLayout& layout) {
PicaVSConfig config{regs.rasterizer, regs.vs, setup, instance}; PicaVSConfig config{regs, setup, instance.IsShaderClipDistanceSupported(),
config.state.use_geometry_shader = instance.UseGeometryShaders(); instance.UseGeometryShaders()};
for (u32 i = 0; i < layout.attribute_count; i++) { for (u32 i = 0; i < layout.attribute_count; i++) {
const VertexAttribute& attr = layout.attributes[i]; const VertexAttribute& attr = layout.attributes[i];
@ -313,14 +309,13 @@ bool PipelineCache::UseProgrammableVertexShader(const Pica::Regs& regs,
auto [it, new_config] = programmable_vertex_map.try_emplace(config); auto [it, new_config] = programmable_vertex_map.try_emplace(config);
if (new_config) { if (new_config) {
auto code = GenerateVertexShader(setup, config); auto program = GLSL::GenerateVertexShader(setup, config, true);
if (!code) { if (program.empty()) {
LOG_ERROR(Render_Vulkan, "Failed to retrieve programmable vertex shader"); LOG_ERROR(Render_Vulkan, "Failed to retrieve programmable vertex shader");
programmable_vertex_map[config] = nullptr; programmable_vertex_map[config] = nullptr;
return false; return false;
} }
std::string& program = code.value();
auto [iter, new_program] = programmable_vertex_cache.try_emplace(program, instance); auto [iter, new_program] = programmable_vertex_cache.try_emplace(program, instance);
auto& shader = iter->second; auto& shader = iter->second;
@ -359,13 +354,13 @@ bool PipelineCache::UseFixedGeometryShader(const Pica::Regs& regs) {
return true; return true;
} }
const PicaFixedGSConfig gs_config{regs, instance}; const PicaFixedGSConfig gs_config{regs, instance.IsShaderClipDistanceSupported()};
auto [it, new_shader] = fixed_geometry_shaders.try_emplace(gs_config, instance); auto [it, new_shader] = fixed_geometry_shaders.try_emplace(gs_config, instance);
auto& shader = it->second; auto& shader = it->second;
if (new_shader) { if (new_shader) {
workers.QueueWork([gs_config, device = instance.GetDevice(), &shader]() { workers.QueueWork([gs_config, device = instance.GetDevice(), &shader]() {
const std::string code = GenerateFixedGeometryShader(gs_config); const auto code = GLSL::GenerateFixedGeometryShader(gs_config, true);
shader.module = Compile(code, vk::ShaderStageFlagBits::eGeometry, device); shader.module = Compile(code, vk::ShaderStageFlagBits::eGeometry, device);
shader.MarkDone(); shader.MarkDone();
}); });
@ -383,7 +378,9 @@ void PipelineCache::UseTrivialGeometryShader() {
} }
void PipelineCache::UseFragmentShader(const Pica::Regs& regs) { void PipelineCache::UseFragmentShader(const Pica::Regs& regs) {
const PicaFSConfig config{regs, instance}; const PicaFSConfig config{regs, instance.IsFragmentShaderInterlockSupported(),
instance.NeedsLogicOpEmulation(),
!instance.IsCustomBorderColorSupported(), false};
const auto [it, new_shader] = fragment_shaders.try_emplace(config, instance); const auto [it, new_shader] = fragment_shaders.try_emplace(config, instance);
auto& shader = it->second; auto& shader = it->second;
@ -395,12 +392,12 @@ void PipelineCache::UseFragmentShader(const Pica::Regs& regs) {
texture0_type == Pica::TexturingRegs::TextureConfig::ShadowCube || texture0_type == Pica::TexturingRegs::TextureConfig::ShadowCube ||
config.state.shadow_rendering.Value(); config.state.shadow_rendering.Value();
if (use_spirv && !is_shadow) { if (use_spirv && !is_shadow) {
const std::vector code = GenerateFragmentShaderSPV(config); const std::vector code = SPIRV::GenerateFragmentShader(config);
shader.module = CompileSPV(code, instance.GetDevice()); shader.module = CompileSPV(code, instance.GetDevice());
shader.MarkDone(); shader.MarkDone();
} else { } else {
workers.QueueWork([config, device = instance.GetDevice(), &shader]() { workers.QueueWork([config, device = instance.GetDevice(), &shader]() {
const std::string code = GenerateFragmentShader(config); const std::string code = GLSL::GenerateFragmentShader(config, true);
shader.module = Compile(code, vk::ShaderStageFlagBits::eFragment, device); shader.module = Compile(code, vk::ShaderStageFlagBits::eFragment, device);
shader.MarkDone(); shader.MarkDone();
}); });

10
src/video_core/renderer_vulkan/vk_pipeline_cache.h
View file

@ -9,6 +9,8 @@
#include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
#include "video_core/renderer_vulkan/vk_graphics_pipeline.h" #include "video_core/renderer_vulkan/vk_graphics_pipeline.h"
#include "video_core/shader/generator/glsl_shader_gen.h"
#include "video_core/shader/generator/spv_shader_gen.h"
namespace Pica { namespace Pica {
struct Regs; struct Regs;
@ -22,7 +24,7 @@ class RenderpassCache;
class DescriptorPool; class DescriptorPool;
constexpr u32 NUM_RASTERIZER_SETS = 3; constexpr u32 NUM_RASTERIZER_SETS = 3;
constexpr u32 NUM_DYNAMIC_OFFSETS = 2; constexpr u32 NUM_DYNAMIC_OFFSETS = 3;
/** /**
* Stores a collection of rasterizer pipelines used during rendering. * Stores a collection of rasterizer pipelines used during rendering.
@ -113,10 +115,10 @@ private:
std::array<u64, MAX_SHADER_STAGES> shader_hashes; std::array<u64, MAX_SHADER_STAGES> shader_hashes;
std::array<Shader*, MAX_SHADER_STAGES> current_shaders; std::array<Shader*, MAX_SHADER_STAGES> current_shaders;
std::unordered_map<PicaVSConfig, Shader*> programmable_vertex_map; std::unordered_map<Pica::Shader::Generator::PicaVSConfig, Shader*> programmable_vertex_map;
std::unordered_map<std::string, Shader> programmable_vertex_cache; std::unordered_map<std::string, Shader> programmable_vertex_cache;
std::unordered_map<PicaFixedGSConfig, Shader> fixed_geometry_shaders; std::unordered_map<Pica::Shader::Generator::PicaFixedGSConfig, Shader> fixed_geometry_shaders;
std::unordered_map<PicaFSConfig, Shader> fragment_shaders; std::unordered_map<Pica::Shader::Generator::PicaFSConfig, Shader> fragment_shaders;
Shader trivial_vertex_shader; Shader trivial_vertex_shader;
}; };

160
src/video_core/renderer_vulkan/vk_rasterizer.cpp
View file

@ -28,6 +28,8 @@ MICROPROFILE_DEFINE(Vulkan_Drawing, "Vulkan", "Drawing", MP_RGB(128, 128, 192));
using TriangleTopology = Pica::PipelineRegs::TriangleTopology; using TriangleTopology = Pica::PipelineRegs::TriangleTopology;
using VideoCore::SurfaceType; using VideoCore::SurfaceType;
using namespace Pica::Shader::Generator;
constexpr u64 STREAM_BUFFER_SIZE = 64 * 1024 * 1024; constexpr u64 STREAM_BUFFER_SIZE = 64 * 1024 * 1024;
constexpr u64 UNIFORM_BUFFER_SIZE = 4 * 1024 * 1024; constexpr u64 UNIFORM_BUFFER_SIZE = 4 * 1024 * 1024;
constexpr u64 TEXTURE_BUFFER_SIZE = 2 * 1024 * 1024; constexpr u64 TEXTURE_BUFFER_SIZE = 2 * 1024 * 1024;
@ -76,10 +78,10 @@ RasterizerVulkan::RasterizerVulkan(Memory::MemorySystem& memory,
vertex_buffers.fill(stream_buffer.Handle()); vertex_buffers.fill(stream_buffer.Handle());
uniform_buffer_alignment = instance.UniformMinAlignment(); uniform_buffer_alignment = instance.UniformMinAlignment();
uniform_size_aligned_vs = uniform_size_aligned_vs_pica =
Common::AlignUp(sizeof(Pica::Shader::VSUniformData), uniform_buffer_alignment); Common::AlignUp(sizeof(VSPicaUniformData), uniform_buffer_alignment);
uniform_size_aligned_fs = uniform_size_aligned_vs = Common::AlignUp(sizeof(VSUniformData), uniform_buffer_alignment);
Common::AlignUp(sizeof(Pica::Shader::UniformData), uniform_buffer_alignment); uniform_size_aligned_fs = Common::AlignUp(sizeof(FSUniformData), uniform_buffer_alignment);
// Define vertex layout for software shaders // Define vertex layout for software shaders
MakeSoftwareVertexLayout(); MakeSoftwareVertexLayout();
@ -107,11 +109,12 @@ RasterizerVulkan::RasterizerVulkan(Memory::MemorySystem& memory,
// Since we don't have access to VK_EXT_descriptor_indexing we need to intiallize // Since we don't have access to VK_EXT_descriptor_indexing we need to intiallize
// all descriptor sets even the ones we don't use. // all descriptor sets even the ones we don't use.
pipeline_cache.BindBuffer(0, uniform_buffer.Handle(), 0, sizeof(Pica::Shader::VSUniformData)); pipeline_cache.BindBuffer(0, uniform_buffer.Handle(), 0, sizeof(VSPicaUniformData));
pipeline_cache.BindBuffer(1, uniform_buffer.Handle(), 0, sizeof(Pica::Shader::UniformData)); pipeline_cache.BindBuffer(1, uniform_buffer.Handle(), 0, sizeof(VSUniformData));
pipeline_cache.BindTexelBuffer(2, *texture_lf_view); pipeline_cache.BindBuffer(2, uniform_buffer.Handle(), 0, sizeof(FSUniformData));
pipeline_cache.BindTexelBuffer(3, *texture_rg_view); pipeline_cache.BindTexelBuffer(3, *texture_lf_view);
pipeline_cache.BindTexelBuffer(4, *texture_rgba_view); pipeline_cache.BindTexelBuffer(4, *texture_rg_view);
pipeline_cache.BindTexelBuffer(5, *texture_rgba_view);
Surface& null_surface = res_cache.GetSurface(VideoCore::NULL_SURFACE_ID); Surface& null_surface = res_cache.GetSurface(VideoCore::NULL_SURFACE_ID);
Surface& null_cube_surface = res_cache.GetSurface(VideoCore::NULL_SURFACE_CUBE_ID); Surface& null_cube_surface = res_cache.GetSurface(VideoCore::NULL_SURFACE_CUBE_ID);
@ -140,7 +143,6 @@ void RasterizerVulkan::LoadDiskResources(const std::atomic_bool& stop_loading,
} }
void RasterizerVulkan::SyncFixedState() { void RasterizerVulkan::SyncFixedState() {
SyncClipEnabled();
SyncCullMode(); SyncCullMode();
SyncBlendEnabled(); SyncBlendEnabled();
SyncBlendFuncs(); SyncBlendFuncs();
@ -478,16 +480,16 @@ bool RasterizerVulkan::Draw(bool accelerate, bool is_indexed) {
// Update scissor uniforms // Update scissor uniforms
const auto [scissor_x1, scissor_y2, scissor_x2, scissor_y1] = fb_helper.Scissor(); const auto [scissor_x1, scissor_y2, scissor_x2, scissor_y1] = fb_helper.Scissor();
if (uniform_block_data.data.scissor_x1 != scissor_x1 || if (fs_uniform_block_data.data.scissor_x1 != scissor_x1 ||
uniform_block_data.data.scissor_x2 != scissor_x2 || fs_uniform_block_data.data.scissor_x2 != scissor_x2 ||
uniform_block_data.data.scissor_y1 != scissor_y1 || fs_uniform_block_data.data.scissor_y1 != scissor_y1 ||
uniform_block_data.data.scissor_y2 != scissor_y2) { fs_uniform_block_data.data.scissor_y2 != scissor_y2) {
uniform_block_data.data.scissor_x1 = scissor_x1; fs_uniform_block_data.data.scissor_x1 = scissor_x1;
uniform_block_data.data.scissor_x2 = scissor_x2; fs_uniform_block_data.data.scissor_x2 = scissor_x2;
uniform_block_data.data.scissor_y1 = scissor_y1; fs_uniform_block_data.data.scissor_y1 = scissor_y1;
uniform_block_data.data.scissor_y2 = scissor_y2; fs_uniform_block_data.data.scissor_y2 = scissor_y2;
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
} }
// Sync and bind the texture surfaces // Sync and bind the texture surfaces
@ -670,11 +672,6 @@ void RasterizerVulkan::UnbindSpecial() {
void RasterizerVulkan::NotifyFixedFunctionPicaRegisterChanged(u32 id) { void RasterizerVulkan::NotifyFixedFunctionPicaRegisterChanged(u32 id) {
switch (id) { switch (id) {
// Clipping plane
case PICA_REG_INDEX(rasterizer.clip_enable):
SyncClipEnabled();
break;
// Culling // Culling
case PICA_REG_INDEX(rasterizer.cull_mode): case PICA_REG_INDEX(rasterizer.cull_mode):
SyncCullMode(); SyncCullMode();
@ -831,14 +828,6 @@ void RasterizerVulkan::MakeSoftwareVertexLayout() {
} }
} }
void RasterizerVulkan::SyncClipEnabled() {
bool clip_enabled = regs.rasterizer.clip_enable != 0;
if (clip_enabled != uniform_block_data.data.enable_clip1) {
uniform_block_data.data.enable_clip1 = clip_enabled;
uniform_block_data.dirty = true;
}
}
void RasterizerVulkan::SyncCullMode() { void RasterizerVulkan::SyncCullMode() {
pipeline_info.rasterization.cull_mode.Assign(regs.rasterizer.cull_mode); pipeline_info.rasterization.cull_mode.Assign(regs.rasterizer.cull_mode);
} }
@ -946,7 +935,7 @@ void RasterizerVulkan::SyncAndUploadLUTsLF() {
sizeof(Common::Vec2f) * 256 * Pica::LightingRegs::NumLightingSampler + sizeof(Common::Vec2f) * 256 * Pica::LightingRegs::NumLightingSampler +
sizeof(Common::Vec2f) * 128; // fog sizeof(Common::Vec2f) * 128; // fog
if (!uniform_block_data.lighting_lut_dirty_any && !uniform_block_data.fog_lut_dirty) { if (!fs_uniform_block_data.lighting_lut_dirty_any && !fs_uniform_block_data.fog_lut_dirty) {
return; return;
} }
@ -954,9 +943,9 @@ void RasterizerVulkan::SyncAndUploadLUTsLF() {
auto [buffer, offset, invalidate] = texture_lf_buffer.Map(max_size, sizeof(Common::Vec4f)); auto [buffer, offset, invalidate] = texture_lf_buffer.Map(max_size, sizeof(Common::Vec4f));
// Sync the lighting luts // Sync the lighting luts
if (uniform_block_data.lighting_lut_dirty_any || invalidate) { if (fs_uniform_block_data.lighting_lut_dirty_any || invalidate) {
for (unsigned index = 0; index < uniform_block_data.lighting_lut_dirty.size(); index++) { for (unsigned index = 0; index < fs_uniform_block_data.lighting_lut_dirty.size(); index++) {
if (uniform_block_data.lighting_lut_dirty[index] || invalidate) { if (fs_uniform_block_data.lighting_lut_dirty[index] || invalidate) {
std::array<Common::Vec2f, 256> new_data; std::array<Common::Vec2f, 256> new_data;
const auto& source_lut = Pica::g_state.lighting.luts[index]; const auto& source_lut = Pica::g_state.lighting.luts[index];
std::transform(source_lut.begin(), source_lut.end(), new_data.begin(), std::transform(source_lut.begin(), source_lut.end(), new_data.begin(),
@ -968,19 +957,19 @@ void RasterizerVulkan::SyncAndUploadLUTsLF() {
lighting_lut_data[index] = new_data; lighting_lut_data[index] = new_data;
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f)); new_data.size() * sizeof(Common::Vec2f));
uniform_block_data.data.lighting_lut_offset[index / 4][index % 4] = fs_uniform_block_data.data.lighting_lut_offset[index / 4][index % 4] =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f)); static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f); bytes_used += new_data.size() * sizeof(Common::Vec2f);
} }
uniform_block_data.lighting_lut_dirty[index] = false; fs_uniform_block_data.lighting_lut_dirty[index] = false;
} }
} }
uniform_block_data.lighting_lut_dirty_any = false; fs_uniform_block_data.lighting_lut_dirty_any = false;
} }
// Sync the fog lut // Sync the fog lut
if (uniform_block_data.fog_lut_dirty || invalidate) { if (fs_uniform_block_data.fog_lut_dirty || invalidate) {
std::array<Common::Vec2f, 128> new_data; std::array<Common::Vec2f, 128> new_data;
std::transform(Pica::g_state.fog.lut.begin(), Pica::g_state.fog.lut.end(), new_data.begin(), std::transform(Pica::g_state.fog.lut.begin(), Pica::g_state.fog.lut.end(), new_data.begin(),
@ -992,12 +981,12 @@ void RasterizerVulkan::SyncAndUploadLUTsLF() {
fog_lut_data = new_data; fog_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f)); new_data.size() * sizeof(Common::Vec2f));
uniform_block_data.data.fog_lut_offset = fs_uniform_block_data.data.fog_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f)); static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f); bytes_used += new_data.size() * sizeof(Common::Vec2f);
} }
uniform_block_data.fog_lut_dirty = false; fs_uniform_block_data.fog_lut_dirty = false;
} }
texture_lf_buffer.Commit(static_cast<u32>(bytes_used)); texture_lf_buffer.Commit(static_cast<u32>(bytes_used));
@ -1010,10 +999,10 @@ void RasterizerVulkan::SyncAndUploadLUTs() {
sizeof(Common::Vec4f) * 256 + // proctex sizeof(Common::Vec4f) * 256 + // proctex
sizeof(Common::Vec4f) * 256; // proctex diff sizeof(Common::Vec4f) * 256; // proctex diff
if (!uniform_block_data.proctex_noise_lut_dirty && if (!fs_uniform_block_data.proctex_noise_lut_dirty &&
!uniform_block_data.proctex_color_map_dirty && !fs_uniform_block_data.proctex_color_map_dirty &&
!uniform_block_data.proctex_alpha_map_dirty && !uniform_block_data.proctex_lut_dirty && !fs_uniform_block_data.proctex_alpha_map_dirty &&
!uniform_block_data.proctex_diff_lut_dirty) { !fs_uniform_block_data.proctex_lut_dirty && !fs_uniform_block_data.proctex_diff_lut_dirty) {
return; return;
} }
@ -1035,34 +1024,34 @@ void RasterizerVulkan::SyncAndUploadLUTs() {
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f)); new_data.size() * sizeof(Common::Vec2f));
lut_offset = static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f)); lut_offset = static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f); bytes_used += new_data.size() * sizeof(Common::Vec2f);
} }
}; };
// Sync the proctex noise lut // Sync the proctex noise lut
if (uniform_block_data.proctex_noise_lut_dirty || invalidate) { if (fs_uniform_block_data.proctex_noise_lut_dirty || invalidate) {
sync_proctex_value_lut(proctex.noise_table, proctex_noise_lut_data, sync_proctex_value_lut(proctex.noise_table, proctex_noise_lut_data,
uniform_block_data.data.proctex_noise_lut_offset); fs_uniform_block_data.data.proctex_noise_lut_offset);
uniform_block_data.proctex_noise_lut_dirty = false; fs_uniform_block_data.proctex_noise_lut_dirty = false;
} }
// Sync the proctex color map // Sync the proctex color map
if (uniform_block_data.proctex_color_map_dirty || invalidate) { if (fs_uniform_block_data.proctex_color_map_dirty || invalidate) {
sync_proctex_value_lut(proctex.color_map_table, proctex_color_map_data, sync_proctex_value_lut(proctex.color_map_table, proctex_color_map_data,
uniform_block_data.data.proctex_color_map_offset); fs_uniform_block_data.data.proctex_color_map_offset);
uniform_block_data.proctex_color_map_dirty = false; fs_uniform_block_data.proctex_color_map_dirty = false;
} }
// Sync the proctex alpha map // Sync the proctex alpha map
if (uniform_block_data.proctex_alpha_map_dirty || invalidate) { if (fs_uniform_block_data.proctex_alpha_map_dirty || invalidate) {
sync_proctex_value_lut(proctex.alpha_map_table, proctex_alpha_map_data, sync_proctex_value_lut(proctex.alpha_map_table, proctex_alpha_map_data,
uniform_block_data.data.proctex_alpha_map_offset); fs_uniform_block_data.data.proctex_alpha_map_offset);
uniform_block_data.proctex_alpha_map_dirty = false; fs_uniform_block_data.proctex_alpha_map_dirty = false;
} }
// Sync the proctex lut // Sync the proctex lut
if (uniform_block_data.proctex_lut_dirty || invalidate) { if (fs_uniform_block_data.proctex_lut_dirty || invalidate) {
std::array<Common::Vec4f, 256> new_data; std::array<Common::Vec4f, 256> new_data;
std::transform(proctex.color_table.begin(), proctex.color_table.end(), new_data.begin(), std::transform(proctex.color_table.begin(), proctex.color_table.end(), new_data.begin(),
@ -1075,16 +1064,16 @@ void RasterizerVulkan::SyncAndUploadLUTs() {
proctex_lut_data = new_data; proctex_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec4f)); new_data.size() * sizeof(Common::Vec4f));
uniform_block_data.data.proctex_lut_offset = fs_uniform_block_data.data.proctex_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f)); static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec4f); bytes_used += new_data.size() * sizeof(Common::Vec4f);
} }
uniform_block_data.proctex_lut_dirty = false; fs_uniform_block_data.proctex_lut_dirty = false;
} }
// Sync the proctex difference lut // Sync the proctex difference lut
if (uniform_block_data.proctex_diff_lut_dirty || invalidate) { if (fs_uniform_block_data.proctex_diff_lut_dirty || invalidate) {
std::array<Common::Vec4f, 256> new_data; std::array<Common::Vec4f, 256> new_data;
std::transform(proctex.color_diff_table.begin(), proctex.color_diff_table.end(), std::transform(proctex.color_diff_table.begin(), proctex.color_diff_table.end(),
@ -1097,48 +1086,59 @@ void RasterizerVulkan::SyncAndUploadLUTs() {
proctex_diff_lut_data = new_data; proctex_diff_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(), std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec4f)); new_data.size() * sizeof(Common::Vec4f));
uniform_block_data.data.proctex_diff_lut_offset = fs_uniform_block_data.data.proctex_diff_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f)); static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f));
uniform_block_data.dirty = true; fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec4f); bytes_used += new_data.size() * sizeof(Common::Vec4f);
} }
uniform_block_data.proctex_diff_lut_dirty = false; fs_uniform_block_data.proctex_diff_lut_dirty = false;
} }
texture_buffer.Commit(static_cast<u32>(bytes_used)); texture_buffer.Commit(static_cast<u32>(bytes_used));
} }
void RasterizerVulkan::UploadUniforms(bool accelerate_draw) { void RasterizerVulkan::UploadUniforms(bool accelerate_draw) {
const bool sync_vs = accelerate_draw; const bool sync_vs_pica = accelerate_draw;
const bool sync_fs = uniform_block_data.dirty; const bool sync_vs = vs_uniform_block_data.dirty;
const bool sync_fs = fs_uniform_block_data.dirty;
if (!sync_vs && !sync_fs) { if (!sync_vs_pica && !sync_vs && !sync_fs) {
return; return;
} }
const u64 uniform_size = uniform_size_aligned_vs + uniform_size_aligned_fs; const u64 uniform_size =
uniform_size_aligned_vs_pica + uniform_size_aligned_vs + uniform_size_aligned_fs;
auto [uniforms, offset, invalidate] = auto [uniforms, offset, invalidate] =
uniform_buffer.Map(uniform_size, uniform_buffer_alignment); uniform_buffer.Map(uniform_size, uniform_buffer_alignment);
u32 used_bytes = 0; u32 used_bytes = 0;
if (sync_vs) {
Pica::Shader::VSUniformData vs_uniforms;
vs_uniforms.uniforms.SetFromRegs(regs.vs, Pica::g_state.vs);
std::memcpy(uniforms, &vs_uniforms, sizeof(vs_uniforms));
pipeline_cache.SetBufferOffset(0, offset); if (sync_vs || invalidate) {
std::memcpy(uniforms + used_bytes, &vs_uniform_block_data.data,
sizeof(vs_uniform_block_data.data));
pipeline_cache.SetBufferOffset(1, offset + used_bytes);
vs_uniform_block_data.dirty = false;
used_bytes += static_cast<u32>(uniform_size_aligned_vs); used_bytes += static_cast<u32>(uniform_size_aligned_vs);
} }
if (sync_fs || invalidate) { if (sync_fs || invalidate) {
std::memcpy(uniforms + used_bytes, &uniform_block_data.data, std::memcpy(uniforms + used_bytes, &fs_uniform_block_data.data,
sizeof(Pica::Shader::UniformData)); sizeof(fs_uniform_block_data.data));
pipeline_cache.SetBufferOffset(1, offset + used_bytes); pipeline_cache.SetBufferOffset(2, offset + used_bytes);
uniform_block_data.dirty = false; fs_uniform_block_data.dirty = false;
used_bytes += static_cast<u32>(uniform_size_aligned_fs); used_bytes += static_cast<u32>(uniform_size_aligned_fs);
} }
if (sync_vs_pica) {
VSPicaUniformData vs_uniforms;
vs_uniforms.uniforms.SetFromRegs(regs.vs, Pica::g_state.vs);
std::memcpy(uniforms + used_bytes, &vs_uniforms, sizeof(vs_uniforms));
pipeline_cache.SetBufferOffset(0, offset + used_bytes);
used_bytes += static_cast<u32>(uniform_size_aligned_vs_pica);
}
uniform_buffer.Commit(used_bytes); uniform_buffer.Commit(used_bytes);
} }

4
src/video_core/renderer_vulkan/vk_rasterizer.h
View file

@ -60,9 +60,6 @@ public:
private: private:
void NotifyFixedFunctionPicaRegisterChanged(u32 id) override; void NotifyFixedFunctionPicaRegisterChanged(u32 id) override;
/// Syncs the clip enabled status to match the PICA register
void SyncClipEnabled();
/// Syncs the cull mode to match the PICA register /// Syncs the cull mode to match the PICA register
void SyncCullMode(); void SyncCullMode();
@ -163,6 +160,7 @@ private:
vk::UniqueBufferView texture_rg_view; vk::UniqueBufferView texture_rg_view;
vk::UniqueBufferView texture_rgba_view; vk::UniqueBufferView texture_rgba_view;
u64 uniform_buffer_alignment; u64 uniform_buffer_alignment;
u64 uniform_size_aligned_vs_pica;
u64 uniform_size_aligned_vs; u64 uniform_size_aligned_vs;
u64 uniform_size_aligned_fs; u64 uniform_size_aligned_fs;
bool async_shaders{false}; bool async_shaders{false};

30
src/video_core/renderer_opengl/gl_shader_decompiler.cpp → src/video_core/shader/generator/glsl_shader_decompiler.cpp
View file

@ -12,9 +12,9 @@
#include <nihstro/shader_bytecode.h> #include <nihstro/shader_bytecode.h>
#include "common/assert.h" #include "common/assert.h"
#include "common/common_types.h" #include "common/common_types.h"
#include "video_core/renderer_opengl/gl_shader_decompiler.h" #include "video_core/shader/generator/glsl_shader_decompiler.h"
namespace OpenGL::ShaderDecompiler { namespace Pica::Shader::Generator::GLSL {
using nihstro::DestRegister; using nihstro::DestRegister;
using nihstro::Instruction; using nihstro::Instruction;
@ -939,34 +939,20 @@ private:
ShaderWriter shader; ShaderWriter shader;
}; };
std::string GetCommonDeclarations() { std::string DecompileProgram(const Pica::Shader::ProgramCode& program_code,
return R"( const Pica::Shader::SwizzleData& swizzle_data, u32 main_offset,
struct pica_uniforms { const RegGetter& inputreg_getter, const RegGetter& outputreg_getter,
bool b[16];
uvec4 i[4];
vec4 f[96];
};
bool exec_shader();
)";
}
std::optional<ProgramResult> DecompileProgram(const Pica::Shader::ProgramCode& program_code,
const Pica::Shader::SwizzleData& swizzle_data,
u32 main_offset, const RegGetter& inputreg_getter,
const RegGetter& outputreg_getter,
bool sanitize_mul) { bool sanitize_mul) {
try { try {
auto subroutines = ControlFlowAnalyzer(program_code, main_offset).MoveSubroutines(); auto subroutines = ControlFlowAnalyzer(program_code, main_offset).MoveSubroutines();
GLSLGenerator generator(subroutines, program_code, swizzle_data, main_offset, GLSLGenerator generator(subroutines, program_code, swizzle_data, main_offset,
inputreg_getter, outputreg_getter, sanitize_mul); inputreg_getter, outputreg_getter, sanitize_mul);
return {ProgramResult{generator.MoveShaderCode()}}; return generator.MoveShaderCode();
} catch (const DecompileFail& exception) { } catch (const DecompileFail& exception) {
LOG_INFO(HW_GPU, "Shader decompilation failed: {}", exception.what()); LOG_INFO(HW_GPU, "Shader decompilation failed: {}", exception.what());
return std::nullopt; return "";
} }
} }
} // namespace OpenGL::ShaderDecompiler } // namespace Pica::Shader::Generator::GLSL

21
src/video_core/shader/generator/glsl_shader_decompiler.h Normal file
View file

@ -0,0 +1,21 @@
// Copyright 2017 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <functional>
#include <string>
#include "common/common_types.h"
#include "video_core/shader/shader.h"
namespace Pica::Shader::Generator::GLSL {
using RegGetter = std::function<std::string(u32)>;
std::string DecompileProgram(const Pica::Shader::ProgramCode& program_code,
const Pica::Shader::SwizzleData& swizzle_data, u32 main_offset,
const RegGetter& inputreg_getter, const RegGetter& outputreg_getter,
bool sanitize_mul);
} // namespace Pica::Shader::Generator::GLSL

821
src/video_core/renderer_vulkan/vk_shader_gen.cpp → src/video_core/shader/generator/glsl_shader_gen.cpp
View file

File diff suppressed because it is too large Load diff

57
src/video_core/shader/generator/glsl_shader_gen.h Normal file
View file

@ -0,0 +1,57 @@
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include "video_core/shader/generator/shader_gen.h"
#include "video_core/shader/shader.h"
// High precision may or may not be supported in GLES3. If it isn't, use medium precision instead.
static constexpr char fragment_shader_precision_OES[] = R"(
#if GL_ES
#ifdef GL_FRAGMENT_PRECISION_HIGH
precision highp int;
precision highp float;
precision highp samplerBuffer;
precision highp uimage2D;
#else
precision mediump int;
precision mediump float;
precision mediump samplerBuffer;
precision mediump uimage2D;
#endif // GL_FRAGMENT_PRECISION_HIGH
#endif
)";
namespace Pica::Shader::Generator::GLSL {
/**
* Generates the GLSL vertex shader program source code that accepts vertices from software shader
* and directly passes them to the fragment shader.
* @returns String of the shader source code
*/
std::string GenerateTrivialVertexShader(bool use_clip_planes, bool separable_shader);
/**
* Generates the GLSL vertex shader program source code for the given VS program
* @returns String of the shader source code; empty on failure
*/
std::string GenerateVertexShader(const Pica::Shader::ShaderSetup& setup, const PicaVSConfig& config,
bool separable_shader);
/**
* Generates the GLSL fixed geometry shader program source code for non-GS PICA pipeline
* @returns String of the shader source code
*/
std::string GenerateFixedGeometryShader(const PicaFixedGSConfig& config, bool separable_shader);
/**
* Generates the GLSL fragment shader program source code for the current Pica state
* @param config ShaderCacheKey object generated for the current Pica state, used for the shader
* configuration (NOTE: Use state in this struct only, not the Pica registers!)
* @returns String of the shader source code
*/
std::string GenerateFragmentShader(const PicaFSConfig& config, bool separable_shader);
} // namespace Pica::Shader::Generator::GLSL

281
src/video_core/shader/generator/shader_gen.cpp Normal file
View file

@ -0,0 +1,281 @@
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "common/bit_set.h"
#include "common/logging/log.h"
#include "video_core/shader/generator/shader_gen.h"
#include "video_core/video_core.h"
namespace Pica::Shader::Generator {
PicaFSConfig::PicaFSConfig(const Pica::Regs& regs, bool has_fragment_shader_interlock,
bool emulate_logic_op, bool emulate_custom_border_color,
bool emulate_blend_minmax_factor, bool use_custom_normal_map) {
state.scissor_test_mode.Assign(regs.rasterizer.scissor_test.mode);
state.depthmap_enable.Assign(regs.rasterizer.depthmap_enable);
state.alpha_test_func.Assign(regs.framebuffer.output_merger.alpha_test.enable
? regs.framebuffer.output_merger.alpha_test.func.Value()
: Pica::FramebufferRegs::CompareFunc::Always);
state.texture0_type.Assign(regs.texturing.texture0.type);
state.texture2_use_coord1.Assign(regs.texturing.main_config.texture2_use_coord1 != 0);
const auto pica_textures = regs.texturing.GetTextures();
for (u32 tex_index = 0; tex_index < 3; tex_index++) {
const auto config = pica_textures[tex_index].config;
state.texture_border_color[tex_index].enable_s.Assign(
emulate_custom_border_color &&
config.wrap_s == Pica::TexturingRegs::TextureConfig::WrapMode::ClampToBorder);
state.texture_border_color[tex_index].enable_t.Assign(
emulate_custom_border_color &&
config.wrap_t == Pica::TexturingRegs::TextureConfig::WrapMode::ClampToBorder);
}
// Emulate logic op in the shader if not supported. This is mostly for mobile GPUs
const bool needs_emulate_logic_op =
emulate_logic_op && !regs.framebuffer.output_merger.alphablend_enable;
state.emulate_logic_op.Assign(needs_emulate_logic_op);
if (needs_emulate_logic_op) {
state.logic_op.Assign(regs.framebuffer.output_merger.logic_op);
} else {
state.logic_op.Assign(Pica::FramebufferRegs::LogicOp::NoOp);
}
// Copy relevant tev stages fields.
// We don't sync const_color here because of the high variance, it is a
// shader uniform instead.
const auto& tev_stages = regs.texturing.GetTevStages();
DEBUG_ASSERT(state.tev_stages.size() == tev_stages.size());
for (std::size_t i = 0; i < tev_stages.size(); i++) {
const auto& tev_stage = tev_stages[i];
state.tev_stages[i].sources_raw = tev_stage.sources_raw;
state.tev_stages[i].modifiers_raw = tev_stage.modifiers_raw;
state.tev_stages[i].ops_raw = tev_stage.ops_raw;
state.tev_stages[i].scales_raw = tev_stage.scales_raw;
if (tev_stage.color_op == Pica::TexturingRegs::TevStageConfig::Operation::Dot3_RGBA) {
state.tev_stages[i].sources_raw &= 0xFFF;
state.tev_stages[i].modifiers_raw &= 0xFFF;
state.tev_stages[i].ops_raw &= 0xF;
}
}
state.fog_mode.Assign(regs.texturing.fog_mode);
state.fog_flip.Assign(regs.texturing.fog_flip != 0);
state.combiner_buffer_input.Assign(
regs.texturing.tev_combiner_buffer_input.update_mask_rgb.Value() |
regs.texturing.tev_combiner_buffer_input.update_mask_a.Value() << 4);
// Fragment lighting
state.lighting.enable.Assign(!regs.lighting.disable);
if (state.lighting.enable) {
state.lighting.src_num.Assign(regs.lighting.max_light_index + 1);
for (u32 light_index = 0; light_index < state.lighting.src_num; ++light_index) {
const u32 num = regs.lighting.light_enable.GetNum(light_index);
const auto& light = regs.lighting.light[num];
state.lighting.light[light_index].num.Assign(num);
state.lighting.light[light_index].directional.Assign(light.config.directional != 0);
state.lighting.light[light_index].two_sided_diffuse.Assign(
light.config.two_sided_diffuse != 0);
state.lighting.light[light_index].geometric_factor_0.Assign(
light.config.geometric_factor_0 != 0);
state.lighting.light[light_index].geometric_factor_1.Assign(
light.config.geometric_factor_1 != 0);
state.lighting.light[light_index].dist_atten_enable.Assign(
!regs.lighting.IsDistAttenDisabled(num));
state.lighting.light[light_index].spot_atten_enable.Assign(
!regs.lighting.IsSpotAttenDisabled(num));
state.lighting.light[light_index].shadow_enable.Assign(
!regs.lighting.IsShadowDisabled(num));
}
state.lighting.lut_d0.enable.Assign(regs.lighting.config1.disable_lut_d0 == 0);
if (state.lighting.lut_d0.enable) {
state.lighting.lut_d0.abs_input.Assign(regs.lighting.abs_lut_input.disable_d0 == 0);
state.lighting.lut_d0.type.Assign(regs.lighting.lut_input.d0.Value());
state.lighting.lut_d0.scale =
regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.d0);
}
state.lighting.lut_d1.enable.Assign(regs.lighting.config1.disable_lut_d1 == 0);
if (state.lighting.lut_d1.enable) {
state.lighting.lut_d1.abs_input.Assign(regs.lighting.abs_lut_input.disable_d1 == 0);
state.lighting.lut_d1.type.Assign(regs.lighting.lut_input.d1.Value());
state.lighting.lut_d1.scale =
regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.d1);
}
// this is a dummy field due to lack of the corresponding register
state.lighting.lut_sp.enable.Assign(1);
state.lighting.lut_sp.abs_input.Assign(regs.lighting.abs_lut_input.disable_sp == 0);
state.lighting.lut_sp.type.Assign(regs.lighting.lut_input.sp.Value());
state.lighting.lut_sp.scale = regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.sp);
state.lighting.lut_fr.enable.Assign(regs.lighting.config1.disable_lut_fr == 0);
if (state.lighting.lut_fr.enable) {
state.lighting.lut_fr.abs_input.Assign(regs.lighting.abs_lut_input.disable_fr == 0);
state.lighting.lut_fr.type.Assign(regs.lighting.lut_input.fr.Value());
state.lighting.lut_fr.scale =
regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.fr);
}
state.lighting.lut_rr.enable.Assign(regs.lighting.config1.disable_lut_rr == 0);
if (state.lighting.lut_rr.enable) {
state.lighting.lut_rr.abs_input.Assign(regs.lighting.abs_lut_input.disable_rr == 0);
state.lighting.lut_rr.type.Assign(regs.lighting.lut_input.rr.Value());
state.lighting.lut_rr.scale =
regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.rr);
}
state.lighting.lut_rg.enable.Assign(regs.lighting.config1.disable_lut_rg == 0);
if (state.lighting.lut_rg.enable) {
state.lighting.lut_rg.abs_input.Assign(regs.lighting.abs_lut_input.disable_rg == 0);
state.lighting.lut_rg.type.Assign(regs.lighting.lut_input.rg.Value());
state.lighting.lut_rg.scale =
regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.rg);
}
state.lighting.lut_rb.enable.Assign(regs.lighting.config1.disable_lut_rb == 0);
if (state.lighting.lut_rb.enable) {
state.lighting.lut_rb.abs_input.Assign(regs.lighting.abs_lut_input.disable_rb == 0);
state.lighting.lut_rb.type.Assign(regs.lighting.lut_input.rb.Value());
state.lighting.lut_rb.scale =
regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.rb);
}
state.lighting.config.Assign(regs.lighting.config0.config);
state.lighting.enable_primary_alpha.Assign(regs.lighting.config0.enable_primary_alpha);
state.lighting.enable_secondary_alpha.Assign(regs.lighting.config0.enable_secondary_alpha);
state.lighting.bump_mode.Assign(regs.lighting.config0.bump_mode);
state.lighting.bump_selector.Assign(regs.lighting.config0.bump_selector);
state.lighting.bump_renorm.Assign(regs.lighting.config0.disable_bump_renorm == 0);
state.lighting.clamp_highlights.Assign(regs.lighting.config0.clamp_highlights != 0);
state.lighting.enable_shadow.Assign(regs.lighting.config0.enable_shadow != 0);
if (state.lighting.enable_shadow) {
state.lighting.shadow_primary.Assign(regs.lighting.config0.shadow_primary != 0);
state.lighting.shadow_secondary.Assign(regs.lighting.config0.shadow_secondary != 0);
state.lighting.shadow_invert.Assign(regs.lighting.config0.shadow_invert != 0);
state.lighting.shadow_alpha.Assign(regs.lighting.config0.shadow_alpha != 0);
state.lighting.shadow_selector.Assign(regs.lighting.config0.shadow_selector);
}
}
state.proctex.enable.Assign(regs.texturing.main_config.texture3_enable);
if (state.proctex.enable) {
state.proctex.coord.Assign(regs.texturing.main_config.texture3_coordinates);
state.proctex.u_clamp.Assign(regs.texturing.proctex.u_clamp);
state.proctex.v_clamp.Assign(regs.texturing.proctex.v_clamp);
state.proctex.color_combiner.Assign(regs.texturing.proctex.color_combiner);
state.proctex.alpha_combiner.Assign(regs.texturing.proctex.alpha_combiner);
state.proctex.separate_alpha.Assign(regs.texturing.proctex.separate_alpha);
state.proctex.noise_enable.Assign(regs.texturing.proctex.noise_enable);
state.proctex.u_shift.Assign(regs.texturing.proctex.u_shift);
state.proctex.v_shift.Assign(regs.texturing.proctex.v_shift);
state.proctex.lut_width = regs.texturing.proctex_lut.width;
state.proctex.lut_offset0 = regs.texturing.proctex_lut_offset.level0;
state.proctex.lut_offset1 = regs.texturing.proctex_lut_offset.level1;
state.proctex.lut_offset2 = regs.texturing.proctex_lut_offset.level2;
state.proctex.lut_offset3 = regs.texturing.proctex_lut_offset.level3;
state.proctex.lod_min = regs.texturing.proctex_lut.lod_min;
state.proctex.lod_max = regs.texturing.proctex_lut.lod_max;
state.proctex.lut_filter.Assign(regs.texturing.proctex_lut.filter);
}
const auto alpha_eq = regs.framebuffer.output_merger.alpha_blending.blend_equation_a.Value();
const auto rgb_eq = regs.framebuffer.output_merger.alpha_blending.blend_equation_rgb.Value();
if (emulate_blend_minmax_factor && regs.framebuffer.output_merger.alphablend_enable) {
if (rgb_eq == Pica::FramebufferRegs::BlendEquation::Max ||
rgb_eq == Pica::FramebufferRegs::BlendEquation::Min) {
state.rgb_blend.emulate_blending = true;
state.rgb_blend.eq = rgb_eq;
state.rgb_blend.src_factor =
regs.framebuffer.output_merger.alpha_blending.factor_source_rgb;
state.rgb_blend.dst_factor =
regs.framebuffer.output_merger.alpha_blending.factor_dest_rgb;
}
if (alpha_eq == Pica::FramebufferRegs::BlendEquation::Max ||
alpha_eq == Pica::FramebufferRegs::BlendEquation::Min) {
state.alpha_blend.emulate_blending = true;
state.alpha_blend.eq = alpha_eq;
state.alpha_blend.src_factor =
regs.framebuffer.output_merger.alpha_blending.factor_source_a;
state.alpha_blend.dst_factor =
regs.framebuffer.output_merger.alpha_blending.factor_dest_a;
}
}
state.shadow_rendering.Assign(regs.framebuffer.output_merger.fragment_operation_mode ==
Pica::FramebufferRegs::FragmentOperationMode::Shadow);
state.shadow_texture_orthographic.Assign(regs.texturing.shadow.orthographic != 0);
// We only need fragment shader interlock when shadow rendering.
state.use_fragment_shader_interlock.Assign(state.shadow_rendering &&
has_fragment_shader_interlock);
state.use_custom_normal_map.Assign(use_custom_normal_map);
}
void PicaGSConfigState::Init(const Pica::Regs& regs, bool use_clip_planes_) {
use_clip_planes = use_clip_planes_;
vs_output_attributes = Common::BitSet<u32>(regs.vs.output_mask).Count();
gs_output_attributes = vs_output_attributes;
semantic_maps.fill({16, 0});
for (u32 attrib = 0; attrib < regs.rasterizer.vs_output_total; ++attrib) {
const std::array semantics{
regs.rasterizer.vs_output_attributes[attrib].map_x.Value(),
regs.rasterizer.vs_output_attributes[attrib].map_y.Value(),
regs.rasterizer.vs_output_attributes[attrib].map_z.Value(),
regs.rasterizer.vs_output_attributes[attrib].map_w.Value(),
};
for (u32 comp = 0; comp < 4; ++comp) {
const auto semantic = semantics[comp];
if (static_cast<std::size_t>(semantic) < 24) {
semantic_maps[static_cast<std::size_t>(semantic)] = {attrib, comp};
} else if (semantic != Pica::RasterizerRegs::VSOutputAttributes::INVALID) {
LOG_ERROR(Render, "Invalid/unknown semantic id: {}", semantic);
}
}
}
}
void PicaVSConfigState::Init(const Pica::Regs& regs, Pica::Shader::ShaderSetup& setup,
bool use_clip_planes_, bool use_geometry_shader_) {
use_clip_planes = use_clip_planes_;
use_geometry_shader = use_geometry_shader_;
program_hash = setup.GetProgramCodeHash();
swizzle_hash = setup.GetSwizzleDataHash();
main_offset = regs.vs.main_offset;
sanitize_mul = VideoCore::g_hw_shader_accurate_mul;
num_outputs = 0;
load_flags.fill(AttribLoadFlags::Float);
output_map.fill(16);
for (int reg : Common::BitSet<u32>(regs.vs.output_mask)) {
output_map[reg] = num_outputs++;
}
if (!use_geometry_shader_) {
gs_state.Init(regs, use_clip_planes_);
}
}
PicaVSConfig::PicaVSConfig(const Pica::Regs& regs, Pica::Shader::ShaderSetup& setup,
bool use_clip_planes_, bool use_geometry_shader_) {
state.Init(regs, setup, use_clip_planes_, use_geometry_shader_);
}
PicaFixedGSConfig::PicaFixedGSConfig(const Pica::Regs& regs, bool use_clip_planes_) {
state.Init(regs, use_clip_planes_);
}
} // namespace Pica::Shader::Generator

164
src/video_core/renderer_vulkan/vk_shader_gen.h → src/video_core/shader/generator/shader_gen.h
View file

@ -4,14 +4,17 @@
#pragma once #pragma once
#include <optional>
#include "common/hash.h" #include "common/hash.h"
#include "video_core/regs.h" #include "video_core/regs.h"
#include "video_core/shader/shader.h" #include "video_core/shader/shader.h"
namespace Vulkan { namespace Pica::Shader::Generator {
class Instance; enum ProgramType : u32 {
VS = 0,
GS = 2,
FS = 1,
};
enum Attributes { enum Attributes {
ATTRIBUTE_POSITION, ATTRIBUTE_POSITION,
@ -31,13 +34,13 @@ struct TevStageConfigRaw {
u32 ops_raw; u32 ops_raw;
u32 scales_raw; u32 scales_raw;
explicit operator Pica::TexturingRegs::TevStageConfig() const noexcept { explicit operator Pica::TexturingRegs::TevStageConfig() const noexcept {
Pica::TexturingRegs::TevStageConfig stage; return {
stage.sources_raw = sources_raw; .sources_raw = sources_raw,
stage.modifiers_raw = modifiers_raw; .modifiers_raw = modifiers_raw,
stage.ops_raw = ops_raw; .ops_raw = ops_raw,
stage.const_color = 0; .const_color = 0,
stage.scales_raw = scales_raw; .scales_raw = scales_raw,
return stage; };
} }
}; };
@ -56,6 +59,7 @@ struct PicaFSConfigState {
BitField<27, 1, u32> shadow_rendering; BitField<27, 1, u32> shadow_rendering;
BitField<28, 1, u32> shadow_texture_orthographic; BitField<28, 1, u32> shadow_texture_orthographic;
BitField<29, 1, u32> use_fragment_shader_interlock; BitField<29, 1, u32> use_fragment_shader_interlock;
BitField<30, 1, u32> use_custom_normal_map;
}; };
union { union {
@ -127,24 +131,33 @@ struct PicaFSConfigState {
u8 lod_min; u8 lod_min;
u8 lod_max; u8 lod_max;
} proctex; } proctex;
struct {
bool emulate_blending;
Pica::FramebufferRegs::BlendEquation eq;
Pica::FramebufferRegs::BlendFactor src_factor;
Pica::FramebufferRegs::BlendFactor dst_factor;
} rgb_blend, alpha_blend;
}; };
/** /**
* This struct contains all state used to generate the GLSL fragment shader that emulates the * This struct contains all state used to generate the GLSL fragment shader that emulates the
* current Pica register configuration. This struct is used as a cache key for generated GLSL shader * current Pica register configuration. This struct is used as a cache key for generated GLSL shader
* programs. The functions in gl_shader_gen.cpp should retrieve state from this struct only, not by * programs. The functions in glsl_shader_gen.cpp should retrieve state from this struct only, not
* directly accessing Pica registers. This should reduce the risk of bugs in shader generation where * by directly accessing Pica registers. This should reduce the risk of bugs in shader generation
* Pica state is not being captured in the shader cache key, thereby resulting in (what should be) * where Pica state is not being captured in the shader cache key, thereby resulting in (what should
* two separate shaders sharing the same key. * be) two separate shaders sharing the same key.
*/ */
struct PicaFSConfig : Common::HashableStruct<PicaFSConfigState> { struct PicaFSConfig : Common::HashableStruct<PicaFSConfigState> {
PicaFSConfig(const Pica::Regs& regs, const Instance& instance); PicaFSConfig(const Pica::Regs& regs, bool has_fragment_shader_interlock, bool emulate_logic_op,
bool emulate_custom_border_color, bool emulate_blend_minmax_factor,
bool use_custom_normal_map = false);
bool TevStageUpdatesCombinerBufferColor(unsigned stage_index) const { [[nodiscard]] bool TevStageUpdatesCombinerBufferColor(unsigned stage_index) const {
return (stage_index < 4) && (state.combiner_buffer_input & (1 << stage_index)); return (stage_index < 4) && (state.combiner_buffer_input & (1 << stage_index));
} }
bool TevStageUpdatesCombinerBufferAlpha(unsigned stage_index) const { [[nodiscard]] bool TevStageUpdatesCombinerBufferAlpha(unsigned stage_index) const {
return (stage_index < 4) && ((state.combiner_buffer_input >> 4) & (1 << stage_index)); return (stage_index < 4) && ((state.combiner_buffer_input >> 4) & (1 << stage_index));
} }
}; };
@ -158,12 +171,36 @@ enum class AttribLoadFlags {
DECLARE_ENUM_FLAG_OPERATORS(AttribLoadFlags) DECLARE_ENUM_FLAG_OPERATORS(AttribLoadFlags)
/** /**
* This struct contains common information to identify a GL vertex/geometry shader generated from * This struct contains common information to identify a GLSL geometry shader generated from
* PICA vertex/geometry shader. * PICA geometry shader.
*/ */
struct PicaShaderConfigCommon { struct PicaGSConfigState {
void Init(const Pica::RasterizerRegs& rasterizer, const Pica::ShaderRegs& regs, void Init(const Pica::Regs& regs, bool use_clip_planes_);
Pica::Shader::ShaderSetup& setup);
bool use_clip_planes;
u32 vs_output_attributes;
u32 gs_output_attributes;
struct SemanticMap {
u32 attribute_index;
u32 component_index;
};
// semantic_maps[semantic name] -> GS output attribute index + component index
std::array<SemanticMap, 24> semantic_maps;
};
/**
* This struct contains common information to identify a GLSL vertex shader generated from
* PICA vertex shader.
*/
struct PicaVSConfigState {
void Init(const Pica::Regs& regs, Pica::Shader::ShaderSetup& setup, bool use_clip_planes_,
bool use_geometry_shader_);
bool use_clip_planes;
bool use_geometry_shader;
u64 program_hash; u64 program_hash;
u64 swizzle_hash; u64 swizzle_hash;
@ -177,103 +214,46 @@ struct PicaShaderConfigCommon {
// output_map[output register index] -> output attribute index // output_map[output register index] -> output attribute index
std::array<u32, 16> output_map; std::array<u32, 16> output_map;
bool use_geometry_shader; PicaGSConfigState gs_state;
u32 vs_output_attributes;
u32 gs_output_attributes;
struct SemanticMap {
u32 attribute_index;
u32 component_index;
};
// semantic_maps[semantic name] -> GS output attribute index + component index
std::array<SemanticMap, 24> semantic_maps;
}; };
/** /**
* This struct contains information to identify a GL vertex shader generated from PICA vertex * This struct contains information to identify a GL vertex shader generated from PICA vertex
* shader. * shader.
*/ */
struct PicaVSConfig : Common::HashableStruct<PicaShaderConfigCommon> { struct PicaVSConfig : Common::HashableStruct<PicaVSConfigState> {
explicit PicaVSConfig(const Pica::RasterizerRegs& rasterizer, const Pica::ShaderRegs& regs, explicit PicaVSConfig(const Pica::Regs& regs, Pica::Shader::ShaderSetup& setup,
Pica::Shader::ShaderSetup& setup, const Instance& instance); bool use_clip_planes_, bool use_geometry_shader_);
bool use_clip_planes;
};
struct PicaGSConfigCommonRaw {
void Init(const Pica::Regs& regs);
u32 vs_output_attributes;
u32 gs_output_attributes;
struct SemanticMap {
u32 attribute_index;
u32 component_index;
};
// semantic_maps[semantic name] -> GS output attribute index + component index
std::array<SemanticMap, 24> semantic_maps;
}; };
/** /**
* This struct contains information to identify a GL geometry shader generated from PICA no-geometry * This struct contains information to identify a GL geometry shader generated from PICA no-geometry
* shader pipeline * shader pipeline
*/ */
struct PicaFixedGSConfig : Common::HashableStruct<PicaGSConfigCommonRaw> { struct PicaFixedGSConfig : Common::HashableStruct<PicaGSConfigState> {
explicit PicaFixedGSConfig(const Pica::Regs& regs, const Instance& instance); explicit PicaFixedGSConfig(const Pica::Regs& regs, bool use_clip_planes_);
bool use_clip_planes;
}; };
/** } // namespace Pica::Shader::Generator
* Generates the GLSL vertex shader program source code that accepts vertices from software shader
* and directly passes them to the fragment shader.
* @param separable_shader generates shader that can be used for separate shader object
* @returns String of the shader source code
*/
std::string GenerateTrivialVertexShader(bool use_clip_planes);
/**
* Generates the GLSL vertex shader program source code for the given VS program
* @returns String of the shader source code; boost::none on failure
*/
std::optional<std::string> GenerateVertexShader(const Pica::Shader::ShaderSetup& setup,
const PicaVSConfig& config);
/**
* Generates the GLSL fixed geometry shader program source code for non-GS PICA pipeline
* @returns String of the shader source code
*/
std::string GenerateFixedGeometryShader(const PicaFixedGSConfig& config);
/**
* Generates the GLSL fragment shader program source code for the current Pica state
* @param config ShaderCacheKey object generated for the current Pica state, used for the shader
* configuration (NOTE: Use state in this struct only, not the Pica registers!)
* @param separable_shader generates shader that can be used for separate shader object
* @returns String of the shader source code
*/
std::string GenerateFragmentShader(const PicaFSConfig& config);
} // namespace Vulkan
namespace std { namespace std {
template <> template <>
struct hash<Vulkan::PicaFSConfig> { struct hash<Pica::Shader::Generator::PicaFSConfig> {
std::size_t operator()(const Vulkan::PicaFSConfig& k) const noexcept { std::size_t operator()(const Pica::Shader::Generator::PicaFSConfig& k) const noexcept {
return k.Hash(); return k.Hash();
} }
}; };
template <> template <>
struct hash<Vulkan::PicaVSConfig> { struct hash<Pica::Shader::Generator::PicaVSConfig> {
std::size_t operator()(const Vulkan::PicaVSConfig& k) const noexcept { std::size_t operator()(const Pica::Shader::Generator::PicaVSConfig& k) const noexcept {
return k.Hash(); return k.Hash();
} }
}; };
template <> template <>
struct hash<Vulkan::PicaFixedGSConfig> { struct hash<Pica::Shader::Generator::PicaFixedGSConfig> {
std::size_t operator()(const Vulkan::PicaFixedGSConfig& k) const noexcept { std::size_t operator()(const Pica::Shader::Generator::PicaFixedGSConfig& k) const noexcept {
return k.Hash(); return k.Hash();
} }
}; };

26
src/video_core/shader/generator/shader_uniforms.cpp Normal file
View file

@ -0,0 +1,26 @@
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <algorithm>
#include "video_core/shader/generator/shader_uniforms.h"
#include "video_core/shader/shader.h"
namespace Pica::Shader::Generator {
void PicaUniformsData::SetFromRegs(const Pica::ShaderRegs& regs,
const Pica::Shader::ShaderSetup& setup) {
std::transform(std::begin(setup.uniforms.b), std::end(setup.uniforms.b), std::begin(bools),
[](bool value) -> BoolAligned { return {value ? 1 : 0}; });
std::transform(std::begin(regs.int_uniforms), std::end(regs.int_uniforms), std::begin(i),
[](const auto& value) -> Common::Vec4u {
return {value.x.Value(), value.y.Value(), value.z.Value(), value.w.Value()};
});
std::transform(std::begin(setup.uniforms.f), std::end(setup.uniforms.f), std::begin(f),
[](const auto& value) -> Common::Vec4f {
return {value.x.ToFloat32(), value.y.ToFloat32(), value.z.ToFloat32(),
value.w.ToFloat32()};
});
}
} // namespace Pica::Shader::Generator

27
src/video_core/shader/shader_uniforms.h → src/video_core/shader/generator/shader_uniforms.h
View file

@ -12,10 +12,10 @@ struct ShaderRegs;
} }
namespace Pica::Shader { namespace Pica::Shader {
struct ShaderSetup; struct ShaderSetup;
}
enum class UniformBindings : u32 { Common, VS, GS }; namespace Pica::Shader::Generator {
struct LightSrc { struct LightSrc {
alignas(16) Common::Vec3f specular_0; alignas(16) Common::Vec3f specular_0;
@ -34,7 +34,7 @@ struct LightSrc {
* the end of a uniform block is included in UNIFORM_BLOCK_DATA_SIZE or not. * the end of a uniform block is included in UNIFORM_BLOCK_DATA_SIZE or not.
* Not following that rule will cause problems on some AMD drivers. * Not following that rule will cause problems on some AMD drivers.
*/ */
struct UniformData { struct FSUniformData {
int framebuffer_scale; int framebuffer_scale;
int alphatest_ref; int alphatest_ref;
float depth_scale; float depth_scale;
@ -53,7 +53,6 @@ struct UniformData {
int proctex_diff_lut_offset; int proctex_diff_lut_offset;
float proctex_bias; float proctex_bias;
int shadow_texture_bias; int shadow_texture_bias;
alignas(4) bool enable_clip1;
alignas(16) Common::Vec4i lighting_lut_offset[LightingRegs::NumLightingSampler / 4]; alignas(16) Common::Vec4i lighting_lut_offset[LightingRegs::NumLightingSampler / 4];
alignas(16) Common::Vec3f fog_color; alignas(16) Common::Vec3f fog_color;
alignas(8) Common::Vec2f proctex_noise_f; alignas(8) Common::Vec2f proctex_noise_f;
@ -65,13 +64,12 @@ struct UniformData {
alignas(16) Common::Vec4f tev_combiner_buffer_color; alignas(16) Common::Vec4f tev_combiner_buffer_color;
alignas(16) Common::Vec3f tex_lod_bias; alignas(16) Common::Vec3f tex_lod_bias;
alignas(16) Common::Vec4f tex_border_color[3]; alignas(16) Common::Vec4f tex_border_color[3];
alignas(16) Common::Vec4f clip_coef;
alignas(16) Common::Vec4f blend_color; alignas(16) Common::Vec4f blend_color;
}; };
static_assert(sizeof(UniformData) == 0x540, static_assert(sizeof(FSUniformData) == 0x530,
"The size of the UniformData does not match the structure in the shader"); "The size of the UniformData does not match the structure in the shader");
static_assert(sizeof(UniformData) < 16384, static_assert(sizeof(FSUniformData) < 16384,
"UniformData structure must be less than 16kb as per the OpenGL spec"); "UniformData structure must be less than 16kb as per the OpenGL spec");
/** /**
@ -91,13 +89,20 @@ struct PicaUniformsData {
}; };
struct VSUniformData { struct VSUniformData {
PicaUniformsData uniforms; bool enable_clip1;
alignas(16) Common::Vec4f clip_coef;
}; };
static_assert(sizeof(VSUniformData) == 1856, static_assert(sizeof(VSUniformData) == 32,
"The size of the VSUniformData does not match the structure in the shader"); "The size of the VSUniformData does not match the structure in the shader");
static_assert(sizeof(VSUniformData) < 16384, static_assert(sizeof(VSUniformData) < 16384,
"VSUniformData structure must be less than 16kb as per the OpenGL spec"); "VSUniformData structure must be less than 16kb as per the OpenGL spec");
std::string BuildShaderUniformDefinitions(const std::string& extra_layout_parameters = ""); struct VSPicaUniformData {
alignas(16) PicaUniformsData uniforms;
};
static_assert(sizeof(VSPicaUniformData) == 1856,
"The size of the VSPicaUniformData does not match the structure in the shader");
static_assert(sizeof(VSPicaUniformData) < 16384,
"VSPicaUniformData structure must be less than 16kb as per the OpenGL spec");
} // namespace Pica::Shader } // namespace Pica::Shader::Generator

71
src/video_core/renderer_vulkan/vk_shader_gen_spv.cpp → src/video_core/shader/generator/spv_shader_gen.cpp
View file

@ -4,7 +4,7 @@
#include "core/core.h" #include "core/core.h"
#include "core/telemetry_session.h" #include "core/telemetry_session.h"
#include "video_core/renderer_vulkan/vk_shader_gen_spv.h" #include "video_core/shader/generator/spv_shader_gen.h"
using Pica::FramebufferRegs; using Pica::FramebufferRegs;
using Pica::LightingRegs; using Pica::LightingRegs;
@ -12,7 +12,7 @@ using Pica::RasterizerRegs;
using Pica::TexturingRegs; using Pica::TexturingRegs;
using TevStageConfig = TexturingRegs::TevStageConfig; using TevStageConfig = TexturingRegs::TevStageConfig;
namespace Vulkan { namespace Pica::Shader::Generator::SPIRV {
constexpr u32 SPIRV_VERSION_1_3 = 0x00010300; constexpr u32 SPIRV_VERSION_1_3 = 0x00010300;
@ -52,7 +52,7 @@ void FragmentModule::Generate() {
} }
combiner_buffer = ConstF32(0.f, 0.f, 0.f, 0.f); combiner_buffer = ConstF32(0.f, 0.f, 0.f, 0.f);
next_combiner_buffer = GetShaderDataMember(vec_ids.Get(4), ConstS32(27)); next_combiner_buffer = GetShaderDataMember(vec_ids.Get(4), ConstS32(26));
last_tex_env_out = rounded_primary_color; last_tex_env_out = rounded_primary_color;
// Write shader bytecode to emulate PICA TEV stages // Write shader bytecode to emulate PICA TEV stages
@ -192,7 +192,7 @@ void FragmentModule::WriteFog() {
// Blend the fog // Blend the fog
const Id tex_env_rgb{ const Id tex_env_rgb{
OpVectorShuffle(vec_ids.Get(3), last_tex_env_out, last_tex_env_out, 0, 1, 2)}; OpVectorShuffle(vec_ids.Get(3), last_tex_env_out, last_tex_env_out, 0, 1, 2)};
const Id fog_color{GetShaderDataMember(vec_ids.Get(3), ConstS32(20))}; const Id fog_color{GetShaderDataMember(vec_ids.Get(3), ConstS32(19))};
const Id fog_factor_rgb{ const Id fog_factor_rgb{
OpCompositeConstruct(vec_ids.Get(3), fog_factor, fog_factor, fog_factor)}; OpCompositeConstruct(vec_ids.Get(3), fog_factor, fog_factor, fog_factor)};
const Id fog_result{OpFMix(vec_ids.Get(3), fog_color, tex_env_rgb, fog_factor_rgb)}; const Id fog_result{OpFMix(vec_ids.Get(3), fog_color, tex_env_rgb, fog_factor_rgb)};
@ -202,7 +202,7 @@ void FragmentModule::WriteFog() {
void FragmentModule::WriteGas() { void FragmentModule::WriteGas() {
// TODO: Implement me // TODO: Implement me
telemetry.AddField(Common::Telemetry::FieldType::Session, "VideoCore_Pica_UseGasMode", true); telemetry.AddField(Common::Telemetry::FieldType::Session, "VideoCore_Pica_UseGasMode", true);
LOG_CRITICAL(Render_Vulkan, "Unimplemented gas mode"); LOG_CRITICAL(Render, "Unimplemented gas mode");
OpKill(); OpKill();
OpFunctionEnd(); OpFunctionEnd();
} }
@ -380,7 +380,7 @@ void FragmentModule::WriteLighting() {
const auto GetLightMember = [&](s32 member) -> Id { const auto GetLightMember = [&](s32 member) -> Id {
const Id member_type = member < 6 ? vec_ids.Get(3) : f32_id; const Id member_type = member < 6 ? vec_ids.Get(3) : f32_id;
const Id light_num{ConstS32(static_cast<s32>(lighting.light[light_index].num.Value()))}; const Id light_num{ConstS32(static_cast<s32>(lighting.light[light_index].num.Value()))};
return GetShaderDataMember(member_type, ConstS32(25), light_num, ConstS32(member)); return GetShaderDataMember(member_type, ConstS32(24), light_num, ConstS32(member));
}; };
// Compute light vector (directional or positional) // Compute light vector (directional or positional)
@ -583,7 +583,7 @@ void FragmentModule::WriteLighting() {
} }
// Sum final lighting result // Sum final lighting result
const Id lighting_global_ambient{GetShaderDataMember(vec_ids.Get(3), ConstS32(24))}; const Id lighting_global_ambient{GetShaderDataMember(vec_ids.Get(3), ConstS32(23))};
const Id lighting_global_ambient_rgba{ const Id lighting_global_ambient_rgba{
PadVectorF32(lighting_global_ambient, vec_ids.Get(4), 0.f)}; PadVectorF32(lighting_global_ambient, vec_ids.Get(4), 0.f)};
const Id zero_vec{ConstF32(0.f, 0.f, 0.f, 0.f)}; const Id zero_vec{ConstF32(0.f, 0.f, 0.f, 0.f)};
@ -706,7 +706,7 @@ void FragmentModule::WriteAlphaTestCondition(FramebufferRegs::CompareFunc func)
break; break;
} }
default: default:
LOG_CRITICAL(Render_Vulkan, "Unknown alpha test condition {}", func); LOG_CRITICAL(Render, "Unknown alpha test condition {}", func);
break; break;
} }
} }
@ -791,7 +791,7 @@ Id FragmentModule::AppendProcTexShiftOffset(Id v, ProcTexShift mode, ProcTexClam
case ProcTexShift::Even: case ProcTexShift::Even:
return shift(true); return shift(true);
default: default:
LOG_CRITICAL(Render_Vulkan, "Unknown shift mode {}", mode); LOG_CRITICAL(Render, "Unknown shift mode {}", mode);
return ConstF32(0.f); return ConstF32(0.f);
} }
} }
@ -819,7 +819,7 @@ Id FragmentModule::AppendProcTexClamp(Id var, ProcTexClamp mode) {
case ProcTexClamp::Pulse: case ProcTexClamp::Pulse:
return OpSelect(f32_id, OpFOrdGreaterThan(bool_id, var, ConstF32(0.5f)), one, zero); return OpSelect(f32_id, OpFOrdGreaterThan(bool_id, var, ConstF32(0.5f)), one, zero);
default: default:
LOG_CRITICAL(Render_Vulkan, "Unknown clamp mode {}", mode); LOG_CRITICAL(Render, "Unknown clamp mode {}", mode);
return OpFMin(f32_id, var, one); return OpFMin(f32_id, var, one);
} }
} }
@ -851,7 +851,7 @@ Id FragmentModule::AppendProcTexCombineAndMap(ProcTexCombiner combiner, Id u, Id
return OpFMin(f32_id, OpFMul(f32_id, r, ConstF32(0.5f)), ConstF32(1.f)); return OpFMin(f32_id, OpFMul(f32_id, r, ConstF32(0.5f)), ConstF32(1.f));
} }
default: default:
LOG_CRITICAL(Render_Vulkan, "Unknown combiner {}", combiner); LOG_CRITICAL(Render, "Unknown combiner {}", combiner);
return ConstF32(0.f); return ConstF32(0.f);
} }
}(); }();
@ -916,7 +916,7 @@ void FragmentModule::DefineTexSampler(u32 texture_unit) {
AddLabel(border_label); AddLabel(border_label);
const Id border_color{ const Id border_color{
GetShaderDataMember(vec_ids.Get(4), ConstS32(29), ConstU32(texture_unit))}; GetShaderDataMember(vec_ids.Get(4), ConstS32(28), ConstU32(texture_unit))};
OpReturnValue(border_color); OpReturnValue(border_color);
AddLabel(not_border_label); AddLabel(not_border_label);
@ -937,7 +937,7 @@ void FragmentModule::DefineTexSampler(u32 texture_unit) {
const Id dx_dy_max{ const Id dx_dy_max{
OpFMax(f32_id, OpCompositeExtract(f32_id, d, 0), OpCompositeExtract(f32_id, d, 1))}; OpFMax(f32_id, OpCompositeExtract(f32_id, d, 0), OpCompositeExtract(f32_id, d, 1))};
const Id lod{OpLog2(f32_id, dx_dy_max)}; const Id lod{OpLog2(f32_id, dx_dy_max)};
const Id lod_bias{GetShaderDataMember(f32_id, ConstS32(28), ConstU32(texture_unit))}; const Id lod_bias{GetShaderDataMember(f32_id, ConstS32(27), ConstU32(texture_unit))};
const Id biased_lod{OpFAdd(f32_id, lod, lod_bias)}; const Id biased_lod{OpFAdd(f32_id, lod, lod_bias)};
return OpImageSampleExplicitLod(vec_ids.Get(4), sampled_image, texcoord, return OpImageSampleExplicitLod(vec_ids.Get(4), sampled_image, texcoord,
spv::ImageOperandsMask::Lod, biased_lod); spv::ImageOperandsMask::Lod, biased_lod);
@ -976,7 +976,7 @@ void FragmentModule::DefineTexSampler(u32 texture_unit) {
// return "shadowTextureCube(texcoord0, texcoord0_w)"; // return "shadowTextureCube(texcoord0, texcoord0_w)";
break; break;
default: default:
LOG_CRITICAL(Render_Vulkan, "Unhandled texture type {:x}", state.texture0_type.Value()); LOG_CRITICAL(Render, "Unhandled texture type {:x}", state.texture0_type.Value());
UNIMPLEMENTED(); UNIMPLEMENTED();
ret_val = zero_vec; ret_val = zero_vec;
break; break;
@ -1012,7 +1012,7 @@ Id FragmentModule::ProcTexSampler() {
const Id texcoord{OpLoad(vec_ids.Get(2), texcoord_id[config.state.proctex.coord.Value()])}; const Id texcoord{OpLoad(vec_ids.Get(2), texcoord_id[config.state.proctex.coord.Value()])};
uv = OpFAbs(vec_ids.Get(2), texcoord); uv = OpFAbs(vec_ids.Get(2), texcoord);
} else { } else {
LOG_CRITICAL(Render_Vulkan, "Unexpected proctex.coord >= 3"); LOG_CRITICAL(Render, "Unexpected proctex.coord >= 3");
uv = OpFAbs(vec_ids.Get(2), OpLoad(vec_ids.Get(2), texcoord_id[0])); uv = OpFAbs(vec_ids.Get(2), OpLoad(vec_ids.Get(2), texcoord_id[0]));
} }
@ -1047,7 +1047,7 @@ Id FragmentModule::ProcTexSampler() {
// Generate noise // Generate noise
if (config.state.proctex.noise_enable) { if (config.state.proctex.noise_enable) {
const Id proctex_noise_a{GetShaderDataMember(vec_ids.Get(2), ConstS32(22))}; const Id proctex_noise_a{GetShaderDataMember(vec_ids.Get(2), ConstS32(21))};
const Id noise_coef{ProcTexNoiseCoef(uv)}; const Id noise_coef{ProcTexNoiseCoef(uv)};
uv = OpFAdd(vec_ids.Get(2), uv, uv = OpFAdd(vec_ids.Get(2), uv,
OpVectorTimesScalar(vec_ids.Get(2), proctex_noise_a, noise_coef)); OpVectorTimesScalar(vec_ids.Get(2), proctex_noise_a, noise_coef));
@ -1158,8 +1158,8 @@ Id FragmentModule::ProcTexNoiseCoef(Id x) {
return OpFma(f32_id, OpConvertSToF(f32_id, v2), ConstF32(2.f / 15.f), ConstF32(-1.f)); return OpFma(f32_id, OpConvertSToF(f32_id, v2), ConstF32(2.f / 15.f), ConstF32(-1.f));
}; };
const Id proctex_noise_f{GetShaderDataMember(vec_ids.Get(2), ConstS32(21))}; const Id proctex_noise_f{GetShaderDataMember(vec_ids.Get(2), ConstS32(20))};
const Id proctex_noise_p{GetShaderDataMember(vec_ids.Get(2), ConstS32(23))}; const Id proctex_noise_p{GetShaderDataMember(vec_ids.Get(2), ConstS32(22))};
const Id grid{OpFMul(vec_ids.Get(2), const Id grid{OpFMul(vec_ids.Get(2),
OpVectorTimesScalar(vec_ids.Get(2), proctex_noise_f, ConstF32(9.f)), OpVectorTimesScalar(vec_ids.Get(2), proctex_noise_f, ConstF32(9.f)),
OpFAbs(vec_ids.Get(2), OpFAdd(vec_ids.Get(2), x, proctex_noise_p)))}; OpFAbs(vec_ids.Get(2), OpFAdd(vec_ids.Get(2), x, proctex_noise_p)))};
@ -1245,7 +1245,7 @@ Id FragmentModule::LookupLightingLUT(Id lut_index, Id index, Id delta) {
const Id lut_index_x{OpShiftRightArithmetic(i32_id, lut_index, ConstS32(2))}; const Id lut_index_x{OpShiftRightArithmetic(i32_id, lut_index, ConstS32(2))};
const Id lut_index_y{OpBitwiseAnd(i32_id, lut_index, ConstS32(3))}; const Id lut_index_y{OpBitwiseAnd(i32_id, lut_index, ConstS32(3))};
const Id lut_offset{GetShaderDataMember(i32_id, ConstS32(19), lut_index_x, lut_index_y)}; const Id lut_offset{GetShaderDataMember(i32_id, ConstS32(18), lut_index_x, lut_index_y)};
const Id coord{OpIAdd(i32_id, lut_offset, index)}; const Id coord{OpIAdd(i32_id, lut_offset, index)};
const Id entry{ const Id entry{
OpImageFetch(vec_ids.Get(4), OpImage(image_buffer_id, texture_buffer_lut_lf), coord)}; OpImageFetch(vec_ids.Get(4), OpImage(image_buffer_id, texture_buffer_lut_lf), coord)};
@ -1274,11 +1274,11 @@ Id FragmentModule::AppendSource(TevStageConfig::Source source, s32 index) {
case Source::PreviousBuffer: case Source::PreviousBuffer:
return combiner_buffer; return combiner_buffer;
case Source::Constant: case Source::Constant:
return GetShaderDataMember(vec_ids.Get(4), ConstS32(26), ConstS32(index)); return GetShaderDataMember(vec_ids.Get(4), ConstS32(25), ConstS32(index));
case Source::Previous: case Source::Previous:
return last_tex_env_out; return last_tex_env_out;
default: default:
LOG_CRITICAL(Render_Vulkan, "Unknown source op {}", source); LOG_CRITICAL(Render, "Unknown source op {}", source);
return ConstF32(0.f, 0.f, 0.f, 0.f); return ConstF32(0.f, 0.f, 0.f, 0.f);
} }
} }
@ -1315,7 +1315,7 @@ Id FragmentModule::AppendColorModifier(TevStageConfig::ColorModifier modifier,
case ColorModifier::OneMinusSourceAlpha: case ColorModifier::OneMinusSourceAlpha:
return OpFSub(vec_ids.Get(3), one_vec, shuffle(3, 3, 3)); return OpFSub(vec_ids.Get(3), one_vec, shuffle(3, 3, 3));
default: default:
LOG_CRITICAL(Render_Vulkan, "Unknown color modifier op {}", modifier); LOG_CRITICAL(Render, "Unknown color modifier op {}", modifier);
return one_vec; return one_vec;
} }
} }
@ -1346,7 +1346,7 @@ Id FragmentModule::AppendAlphaModifier(TevStageConfig::AlphaModifier modifier,
case AlphaModifier::OneMinusSourceBlue: case AlphaModifier::OneMinusSourceBlue:
return OpFSub(f32_id, one_f32, component(2)); return OpFSub(f32_id, one_f32, component(2));
default: default:
LOG_CRITICAL(Render_Vulkan, "Unknown alpha modifier op {}", modifier); LOG_CRITICAL(Render, "Unknown alpha modifier op {}", modifier);
return one_f32; return one_f32;
} }
} }
@ -1395,7 +1395,7 @@ Id FragmentModule::AppendColorCombiner(Pica::TexturingRegs::TevStageConfig::Oper
break; break;
default: default:
color = zero_vec; color = zero_vec;
LOG_CRITICAL(Render_Vulkan, "Unknown color combiner operation: {}", operation); LOG_CRITICAL(Render, "Unknown color combiner operation: {}", operation);
break; break;
} }
@ -1435,7 +1435,7 @@ Id FragmentModule::AppendAlphaCombiner(TevStageConfig::Operation operation) {
break; break;
default: default:
color = ConstF32(0.f); color = ConstF32(0.f);
LOG_CRITICAL(Render_Vulkan, "Unknown alpha combiner operation: {}", operation); LOG_CRITICAL(Render, "Unknown alpha combiner operation: {}", operation);
break; break;
} }
@ -1485,16 +1485,15 @@ void FragmentModule::DefineUniformStructs() {
const Id shader_data_struct_id{ const Id shader_data_struct_id{
TypeStruct(i32_id, i32_id, f32_id, f32_id, f32_id, f32_id, i32_id, i32_id, i32_id, i32_id, TypeStruct(i32_id, i32_id, f32_id, f32_id, f32_id, f32_id, i32_id, i32_id, i32_id, i32_id,
i32_id, i32_id, i32_id, i32_id, i32_id, i32_id, f32_id, i32_id, u32_id, i32_id, i32_id, i32_id, i32_id, i32_id, i32_id, f32_id, i32_id,
lighting_lut_array_id, vec_ids.Get(3), vec_ids.Get(2), vec_ids.Get(2), lighting_lut_array_id, vec_ids.Get(3), vec_ids.Get(2), vec_ids.Get(2),
vec_ids.Get(2), vec_ids.Get(3), light_src_array_id, const_color_array_id, vec_ids.Get(2), vec_ids.Get(3), light_src_array_id, const_color_array_id,
vec_ids.Get(4), vec_ids.Get(3), border_color_array_id, vec_ids.Get(4))}; vec_ids.Get(4), vec_ids.Get(3), border_color_array_id, vec_ids.Get(4))};
constexpr std::array light_src_offsets{0u, 16u, 32u, 48u, 64u, 80u, 92u, 96u}; constexpr std::array light_src_offsets{0u, 16u, 32u, 48u, 64u, 80u, 92u, 96u};
constexpr std::array shader_data_offsets{0u, 4u, 8u, 12u, 16u, 20u, 24u, 28u, constexpr std::array shader_data_offsets{
32u, 36u, 40u, 44u, 48u, 52u, 56u, 60u, 0u, 4u, 8u, 12u, 16u, 20u, 24u, 28u, 32u, 36u, 40u, 44u, 48u, 52u, 56u,
64u, 68u, 72u, 80u, 176u, 192u, 200u, 208u, 60u, 64u, 68u, 80u, 176u, 192u, 200u, 208u, 224u, 240u, 1136u, 1232u, 1248u, 1264u, 1312u};
224u, 240u, 1136u, 1232u, 1248u, 1264u, 1312u};
Decorate(lighting_lut_array_id, spv::Decoration::ArrayStride, 16u); Decorate(lighting_lut_array_id, spv::Decoration::ArrayStride, 16u);
Decorate(light_src_array_id, spv::Decoration::ArrayStride, 112u); Decorate(light_src_array_id, spv::Decoration::ArrayStride, 112u);
@ -1511,7 +1510,7 @@ void FragmentModule::DefineUniformStructs() {
shader_data_id = AddGlobalVariable( shader_data_id = AddGlobalVariable(
TypePointer(spv::StorageClass::Uniform, shader_data_struct_id), spv::StorageClass::Uniform); TypePointer(spv::StorageClass::Uniform, shader_data_struct_id), spv::StorageClass::Uniform);
Decorate(shader_data_id, spv::Decoration::DescriptorSet, 0); Decorate(shader_data_id, spv::Decoration::DescriptorSet, 0);
Decorate(shader_data_id, spv::Decoration::Binding, 1); Decorate(shader_data_id, spv::Decoration::Binding, 2);
} }
void FragmentModule::DefineInterface() { void FragmentModule::DefineInterface() {
@ -1532,9 +1531,9 @@ void FragmentModule::DefineInterface() {
image_r32_id = TypeImage(u32_id, spv::Dim::Dim2D, 0, 0, 0, 2, spv::ImageFormat::R32ui); image_r32_id = TypeImage(u32_id, spv::Dim::Dim2D, 0, 0, 0, 2, spv::ImageFormat::R32ui);
sampler_id = TypeSampler(); sampler_id = TypeSampler();
texture_buffer_lut_lf_id = DefineUniformConst(TypeSampledImage(image_buffer_id), 0, 2); texture_buffer_lut_lf_id = DefineUniformConst(TypeSampledImage(image_buffer_id), 0, 3);
texture_buffer_lut_rg_id = DefineUniformConst(TypeSampledImage(image_buffer_id), 0, 3); texture_buffer_lut_rg_id = DefineUniformConst(TypeSampledImage(image_buffer_id), 0, 4);
texture_buffer_lut_rgba_id = DefineUniformConst(TypeSampledImage(image_buffer_id), 0, 4); texture_buffer_lut_rgba_id = DefineUniformConst(TypeSampledImage(image_buffer_id), 0, 5);
tex0_id = DefineUniformConst(TypeSampledImage(image2d_id), 1, 0); tex0_id = DefineUniformConst(TypeSampledImage(image2d_id), 1, 0);
tex1_id = DefineUniformConst(TypeSampledImage(image2d_id), 1, 1); tex1_id = DefineUniformConst(TypeSampledImage(image2d_id), 1, 1);
tex2_id = DefineUniformConst(TypeSampledImage(image2d_id), 1, 2); tex2_id = DefineUniformConst(TypeSampledImage(image2d_id), 1, 2);
@ -1550,11 +1549,11 @@ void FragmentModule::DefineInterface() {
Decorate(gl_frag_depth_id, spv::Decoration::BuiltIn, spv::BuiltIn::FragDepth); Decorate(gl_frag_depth_id, spv::Decoration::BuiltIn, spv::BuiltIn::FragDepth);
} }
std::vector<u32> GenerateFragmentShaderSPV(const PicaFSConfig& config) { std::vector<u32> GenerateFragmentShader(const PicaFSConfig& config) {
auto& telemetry = Core::System::GetInstance().TelemetrySession(); auto& telemetry = Core::System::GetInstance().TelemetrySession();
FragmentModule module{telemetry, config}; FragmentModule module{telemetry, config};
module.Generate(); module.Generate();
return module.Assemble(); return module.Assemble();
} }
} // namespace Vulkan } // namespace Pica::Shader::Generator::SPIRV

10
src/video_core/renderer_vulkan/vk_shader_gen_spv.h → src/video_core/shader/generator/spv_shader_gen.h
View file

@ -7,13 +7,13 @@
#include <array> #include <array>
#include <sirit/sirit.h> #include <sirit/sirit.h>
#include "video_core/renderer_vulkan/vk_shader_gen.h" #include "video_core/shader/generator/shader_gen.h"
namespace Core { namespace Core {
class TelemetrySession; class TelemetrySession;
} }
namespace Vulkan { namespace Pica::Shader::Generator::SPIRV {
using Sirit::Id; using Sirit::Id;
@ -41,7 +41,7 @@ public:
void Generate(); void Generate();
private: private:
/// Undos the vulkan perspective transformation and applies the PICA one /// Undos the host perspective transformation and applies the PICA one
void WriteDepth(); void WriteDepth();
/// Emits code to emulate the scissor rectangle /// Emits code to emulate the scissor rectangle
@ -289,6 +289,6 @@ private:
* @param separable_shader generates shader that can be used for separate shader object * @param separable_shader generates shader that can be used for separate shader object
* @returns String of the shader source code * @returns String of the shader source code
*/ */
std::vector<u32> GenerateFragmentShaderSPV(const PicaFSConfig& config); std::vector<u32> GenerateFragmentShader(const PicaFSConfig& config);
} // namespace Vulkan } // namespace Pica::Shader::Generator::SPIRV

80
src/video_core/shader/shader_uniforms.cpp
View file

@ -1,80 +0,0 @@
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <algorithm>
#include "video_core/shader/shader.h"
#include "video_core/shader/shader_uniforms.h"
namespace Pica::Shader {
void PicaUniformsData::SetFromRegs(const Pica::ShaderRegs& regs,
const Pica::Shader::ShaderSetup& setup) {
std::transform(std::begin(setup.uniforms.b), std::end(setup.uniforms.b), std::begin(bools),
[](bool value) -> BoolAligned { return {value ? 1 : 0}; });
std::transform(std::begin(regs.int_uniforms), std::end(regs.int_uniforms), std::begin(i),
[](const auto& value) -> Common::Vec4u {
return {value.x.Value(), value.y.Value(), value.z.Value(), value.w.Value()};
});
std::transform(std::begin(setup.uniforms.f), std::end(setup.uniforms.f), std::begin(f),
[](const auto& value) -> Common::Vec4f {
return {value.x.ToFloat32(), value.y.ToFloat32(), value.z.ToFloat32(),
value.w.ToFloat32()};
});
}
constexpr std::string_view UniformBlockDefFormat = R"(
#define NUM_TEV_STAGES 6
#define NUM_LIGHTS 8
#define NUM_LIGHTING_SAMPLERS 24
struct LightSrc {{
vec3 specular_0;
vec3 specular_1;
vec3 diffuse;
vec3 ambient;
vec3 position;
vec3 spot_direction;
float dist_atten_bias;
float dist_atten_scale;
}};
layout ({}std140) uniform shader_data {{
int framebuffer_scale;
int alphatest_ref;
float depth_scale;
float depth_offset;
float shadow_bias_constant;
float shadow_bias_linear;
int scissor_x1;
int scissor_y1;
int scissor_x2;
int scissor_y2;
int fog_lut_offset;
int proctex_noise_lut_offset;
int proctex_color_map_offset;
int proctex_alpha_map_offset;
int proctex_lut_offset;
int proctex_diff_lut_offset;
float proctex_bias;
int shadow_texture_bias;
bool enable_clip1;
ivec4 lighting_lut_offset[NUM_LIGHTING_SAMPLERS / 4];
vec3 fog_color;
vec2 proctex_noise_f;
vec2 proctex_noise_a;
vec2 proctex_noise_p;
vec3 lighting_global_ambient;
LightSrc light_src[NUM_LIGHTS];
vec4 const_color[NUM_TEV_STAGES];
vec4 tev_combiner_buffer_color;
vec3 tex_lod_bias;
vec4 tex_border_color[3];
vec4 clip_coef;
vec4 blend_color;
}};
)";
std::string BuildShaderUniformDefinitions(const std::string& extra_layout_parameters) {
return fmt::format(UniformBlockDefFormat, extra_layout_parameters);
}
} // namespace Pica::Shader