video_core: Refactor state tracking (#1059)

This commit is contained in:
PabloMK7 2025-05-26 14:37:03 +02:00 committed by GitHub
parent bf587885cb
commit 5e2161d90b
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
29 changed files with 675 additions and 1701 deletions

View File

@ -821,7 +821,6 @@ void System::serialize(Archive& ar, const unsigned int file_version) {
timing->UnlockEventQueue();
memory->SetDSP(*dsp_core);
cheat_engine.Connect(cheats_pid);
gpu->Sync();
// Re-register gpu callback, because gsp service changed after service_manager got
// serialized

View File

@ -20,6 +20,7 @@ add_library(video_core STATIC
rasterizer_interface.h
renderer_base.cpp
renderer_base.h
pica/dirty_regs.h
pica/geometry_pipeline.cpp
pica/geometry_pipeline.h
pica/pica_core.cpp

View File

@ -1,4 +1,4 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -290,10 +290,6 @@ void GPU::WriteReg(VAddr addr, u32 data) {
}
}
void GPU::Sync() {
impl->renderer->Sync();
}
VideoCore::RendererBase& GPU::Renderer() {
return *impl->renderer;
}

View File

@ -1,4 +1,4 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -75,9 +75,6 @@ public:
/// Writes the provided value to the GPU virtual address.
void WriteReg(VAddr addr, u32 data);
/// Synchronizes fixed function renderer state with PICA registers.
void Sync();
/// Returns a mutable reference to the renderer.
[[nodiscard]] VideoCore::RendererBase& Renderer();

View File

@ -0,0 +1,110 @@
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include "video_core/pica/regs_internal.h"
namespace Pica {
#define M_R(base, num_bits) (((1ULL << num_bits) - 1) << (PICA_REG_INDEX(base) & 0x3f))
#define M(base) M_R(base, 1)
union DirtyRegs {
void Set(u32 reg_id) {
qwords[reg_id >> 6] |= 1ULL << (reg_id & 0x3f);
}
void Reset() {
qwords.fill(0ULL);
}
bool CheckClipping() const {
// Checks if GPUREG_FRAGOP_CLIP or GPUREG_FRAGOP_CLIP_DATAi are dirty
static constexpr u64 ClipMask = M_R(rasterizer.clip_enable, 5);
return rasterizer & ClipMask;
}
bool CheckDepth() const {
// Checks if GPUREG_DEPTHMAP_SCALE or GPUREG_DEPTHMAP_OFFSET are dirty
static constexpr u64 DepthMask =
M(rasterizer.viewport_depth_range) | M(rasterizer.viewport_depth_near_plane);
return rasterizer & DepthMask;
}
bool CheckLight(u32 index) const {
// Checks if any GPUREG_LIGHTi_* is dirty
return lights[index];
}
bool CheckFogColor() const {
// Checks if GPUREG_FOG_COLOR is dirty
static constexpr u64 FogColorMask = M(texturing.fog_color);
return texenv & FogColorMask;
}
bool CheckTexUnits() const {
// Checks if GPUREG_TEXUNITi_BORDER_COLOR or GPUREG_TEXUNITi_LOD are dirty
static constexpr u64 TexUnitMask =
M(texturing.texture0.border_color) | M(texturing.texture0.lod) |
M(texturing.texture1.border_color) | M(texturing.texture1.lod) |
M(texturing.texture2.border_color) | M(texturing.texture2.lod);
return tex_units & TexUnitMask;
}
bool CheckProctex() const {
// Checks if any GPUREG_TEXUNIT3_PROCTEXi reg is dirty
static constexpr u64 ProctexMask = M_R(texturing.proctex, 6);
return tex_units & ProctexMask;
}
bool CheckTexEnv() const {
// Checks if GPUREG_TEXENV_BUFFER_COLOR or any GPUREG_TEXENVi_COLOR reg is dirty
static constexpr u64 TexEnvMask =
M(texturing.tev_combiner_buffer_color) | M(texturing.tev_stage0.const_color) |
M(texturing.tev_stage1.const_color) | M(texturing.tev_stage2.const_color) |
M(texturing.tev_stage3.const_color) | M(texturing.tev_stage4.const_color) |
M(texturing.tev_stage5.const_color);
return texenv & TexEnvMask;
}
bool CheckLightingAmbient() const {
// Checks if GPUREG_LIGHTING_AMBIENT is dirty
static constexpr u64 LightingMask = M(lighting.global_ambient);
return light_lut & LightingMask;
}
bool CheckBlend() const {
// Checks if GPUREG_BLEND_COLOR or GPUREG_FRAGOP_ALPHA_TEST are dirty
static constexpr u64 BlendMask =
M(framebuffer.output_merger.blend_const) | M(framebuffer.output_merger.alpha_test);
return framebuffer & BlendMask;
}
bool CheckShadow() const {
// Checks if GPUREG_FRAGOP_SHADOW or GPUREG_TEXUNIT0_SHADOW are dirty
static constexpr u64 ShadowMask1 = M(framebuffer.shadow);
static constexpr u64 ShadowMask2 = M(texturing.shadow);
return (framebuffer & ShadowMask1) || (tex_units & ShadowMask2);
}
struct {
u64 misc;
u64 rasterizer;
u64 tex_units;
u64 texenv;
u64 framebuffer;
std::array<u16, 8> lights;
u64 light_lut;
u128 geo_pipeline;
u128 shader;
};
std::array<u64, 12> qwords;
};
static_assert(sizeof(DirtyRegs) == 12 * sizeof(u64));
#undef M
#undef M_R
} // namespace Pica

View File

@ -1,4 +1,4 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -149,7 +149,6 @@ void PicaCore::WriteInternalReg(u32 id, u32 value, u32 mask) {
// Track events.
if (debug_context) {
debug_context->OnEvent(DebugContext::Event::PicaCommandLoaded, &id);
SCOPE_EXIT({ debug_context->OnEvent(DebugContext::Event::PicaCommandProcessed, &id); });
}
switch (id) {
@ -363,7 +362,8 @@ void PicaCore::WriteInternalReg(u32 id, u32 value, u32 mask) {
auto& lut_config = regs.internal.lighting.lut_config;
ASSERT_MSG(lut_config.index < 256, "lut_config.index exceeded maximum value of 255!");
lighting.luts[lut_config.type][lut_config.index].raw = value;
const u32 prev = std::exchange(lighting.luts[lut_config.type][lut_config.index].raw, value);
lighting.lut_dirty |= (prev != value) << lut_config.type;
lut_config.index.Assign(lut_config.index + 1);
break;
}
@ -376,7 +376,9 @@ void PicaCore::WriteInternalReg(u32 id, u32 value, u32 mask) {
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[7]): {
fog.lut[regs.internal.texturing.fog_lut_offset % 128].raw = value;
const u32 prev =
std::exchange(fog.lut[regs.internal.texturing.fog_lut_offset % 128].raw, value);
fog.lut_dirty |= prev != value;
regs.internal.texturing.fog_lut_offset.Assign(regs.internal.texturing.fog_lut_offset + 1);
break;
}
@ -390,22 +392,28 @@ void PicaCore::WriteInternalReg(u32 id, u32 value, u32 mask) {
case PICA_REG_INDEX(texturing.proctex_lut_data[6]):
case PICA_REG_INDEX(texturing.proctex_lut_data[7]): {
auto& index = regs.internal.texturing.proctex_lut_config.index;
const auto lut_table = regs.internal.texturing.proctex_lut_config.ref_table.Value();
switch (regs.internal.texturing.proctex_lut_config.ref_table.Value()) {
const auto sync_lut = [&](auto& proctex_table) {
const u32 prev = std::exchange(proctex_table[index % proctex_table.size()].raw, value);
proctex.table_dirty |= (prev != value) << u32(lut_table);
};
switch (lut_table) {
case TexturingRegs::ProcTexLutTable::Noise:
proctex.noise_table[index % proctex.noise_table.size()].raw = value;
sync_lut(proctex.noise_table);
break;
case TexturingRegs::ProcTexLutTable::ColorMap:
proctex.color_map_table[index % proctex.color_map_table.size()].raw = value;
sync_lut(proctex.color_map_table);
break;
case TexturingRegs::ProcTexLutTable::AlphaMap:
proctex.alpha_map_table[index % proctex.alpha_map_table.size()].raw = value;
sync_lut(proctex.alpha_map_table);
break;
case TexturingRegs::ProcTexLutTable::Color:
proctex.color_table[index % proctex.color_table.size()].raw = value;
sync_lut(proctex.color_table);
break;
case TexturingRegs::ProcTexLutTable::ColorDiff:
proctex.color_diff_table[index % proctex.color_diff_table.size()].raw = value;
sync_lut(proctex.color_diff_table);
break;
}
index.Assign(index + 1);
@ -415,8 +423,11 @@ void PicaCore::WriteInternalReg(u32 id, u32 value, u32 mask) {
break;
}
// Notify the rasterizer an internal register was updated.
rasterizer->NotifyPicaRegisterChanged(id);
dirty_regs.Set(id);
if (debug_context) {
debug_context->OnEvent(DebugContext::Event::PicaCommandProcessed, &id);
}
}
void PicaCore::SubmitImmediate(u32 value) {
@ -460,8 +471,6 @@ void PicaCore::DrawImmediate() {
if (debug_context) {
debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation,
std::addressof(immediate.input_vertex));
SCOPE_EXIT(
{ debug_context->OnEvent(DebugContext::Event::FinishedPrimitiveBatch, nullptr); });
}
ShaderUnit shader_unit;
@ -486,6 +495,10 @@ void PicaCore::DrawImmediate() {
// Flush the immediate triangle.
rasterizer->DrawTriangles();
immediate.current_attribute = 0;
if (debug_context) {
debug_context->OnEvent(DebugContext::Event::FinishedPrimitiveBatch, nullptr);
}
}
void PicaCore::DrawArrays(bool is_indexed) {
@ -494,8 +507,6 @@ void PicaCore::DrawArrays(bool is_indexed) {
// Track vertex in the debug recorder.
if (debug_context) {
debug_context->OnEvent(DebugContext::Event::IncomingPrimitiveBatch, nullptr);
SCOPE_EXIT(
{ debug_context->OnEvent(DebugContext::Event::FinishedPrimitiveBatch, nullptr); });
}
const bool accelerate_draw = [this] {
@ -530,6 +541,10 @@ void PicaCore::DrawArrays(bool is_indexed) {
// Draw emitted triangles.
rasterizer->DrawTriangles();
if (debug_context) {
debug_context->OnEvent(DebugContext::Event::FinishedPrimitiveBatch, nullptr);
}
}
void PicaCore::LoadVertices(bool is_indexed) {

View File

@ -1,4 +1,4 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -6,6 +6,7 @@
#include "common/common_types.h"
#include "core/hle/service/gsp/gsp_interrupt.h"
#include "video_core/pica/dirty_regs.h"
#include "video_core/pica/geometry_pipeline.h"
#include "video_core/pica/packed_attribute.h"
#include "video_core/pica/primitive_assembly.h"
@ -118,6 +119,8 @@ public:
};
struct ProcTex {
static constexpr u8 TableAllDirty = 0xFF;
union ValueEntry {
u32 raw;
@ -168,6 +171,14 @@ public:
std::array<ValueEntry, 128> alpha_map_table;
std::array<ColorEntry, 256> color_table;
std::array<ColorDifferenceEntry, 256> color_diff_table;
union {
u8 table_dirty = TableAllDirty;
BitField<0, 1, u8> noise_lut_dirty;
BitField<2, 1, u8> color_map_dirty;
BitField<3, 1, u8> alpha_map_dirty;
BitField<4, 1, u8> lut_dirty;
BitField<5, 1, u8> diff_lut_dirty;
};
private:
friend class boost::serialization::access;
@ -178,6 +189,8 @@ public:
};
struct Lighting {
static constexpr u32 LutAllDirty = 0xFFFFFF;
union LutEntry {
// Used for raw access
u32 raw;
@ -205,6 +218,7 @@ public:
};
std::array<std::array<LutEntry, 256>, 24> luts;
u32 lut_dirty = LutAllDirty;
private:
friend class boost::serialization::access;
@ -232,6 +246,7 @@ public:
};
std::array<LutEntry, 128> lut;
bool lut_dirty = true;
private:
friend class boost::serialization::access;
@ -243,7 +258,7 @@ public:
RegsLcd regs_lcd{};
Regs regs{};
// TODO: Move these to a separate shader scheduler class
DirtyRegs dirty_regs{};
GeometryShaderUnit gs_unit;
ShaderSetup vs_setup;
ShaderSetup gs_setup;

View File

@ -1,4 +1,4 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -18,13 +18,15 @@ ShaderSetup::~ShaderSetup() = default;
void ShaderSetup::WriteUniformBoolReg(u32 value) {
const auto bits = BitSet32(value);
for (u32 i = 0; i < uniforms.b.size(); ++i) {
uniforms.b[i] = bits[i];
const bool prev = std::exchange(uniforms.b[i], bits[i]);
uniforms_dirty |= prev != bits[i];
}
}
void ShaderSetup::WriteUniformIntReg(u32 index, const Common::Vec4<u8> values) {
ASSERT(index < uniforms.i.size());
uniforms.i[index] = values;
const auto prev = std::exchange(uniforms.i[index], values);
uniforms_dirty |= prev != values;
}
std::optional<u32> ShaderSetup::WriteUniformFloatReg(ShaderRegs& config, u32 value) {
@ -41,7 +43,8 @@ std::optional<u32> ShaderSetup::WriteUniformFloatReg(ShaderRegs& config, u32 val
}
const u32 index = uniform_setup.index.Value();
uniforms.f[index] = uniform;
const auto prev = std::exchange(uniforms.f[index], uniform);
uniforms_dirty |= prev != uniform;
uniform_setup.index.Assign(index + 1);
return index;
}

View File

@ -1,4 +1,4 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -80,6 +80,7 @@ public:
SwizzleData swizzle_data{};
u32 entry_point{};
const void* cached_shader{};
bool uniforms_dirty = true;
private:
bool program_code_hash_dirty{true};

View File

@ -54,9 +54,7 @@ RasterizerAccelerated::HardwareVertex::HardwareVertex(const Pica::OutputVertex&
}
RasterizerAccelerated::RasterizerAccelerated(Memory::MemorySystem& memory_, Pica::PicaCore& pica_)
: memory{memory_}, pica{pica_}, regs{pica.regs.internal} {
fs_uniform_block_data.lighting_lut_dirty.fill(true);
}
: memory{memory_}, pica{pica_}, regs{pica.regs.internal} {}
/**
* This is a helper function to resolve an issue when interpolating opposite quaternions. See below
@ -130,734 +128,132 @@ RasterizerAccelerated::VertexArrayInfo RasterizerAccelerated::AnalyzeVertexArray
return {vertex_min, vertex_max, vs_input_size};
}
void RasterizerAccelerated::SyncEntireState() {
// Sync renderer-specific fixed-function state
SyncFixedState();
void RasterizerAccelerated::SyncDrawUniforms() {
auto& dirty = pica.dirty_regs;
// Sync uniforms
SyncClipPlane();
SyncDepthScale();
SyncDepthOffset();
SyncAlphaTest();
SyncCombinerColor();
auto& tev_stages = regs.texturing.GetTevStages();
for (std::size_t index = 0; index < tev_stages.size(); ++index) {
SyncTevConstColor(index, tev_stages[index]);
// The register that contains the flip bit also contains the framebuffer dimentions
// that we don't depend on. So avoid the dirty table and check manually
const bool is_flipped = regs.framebuffer.framebuffer.IsFlipped();
const bool prev_flipped = std::exchange(vs_data.flip_viewport, is_flipped);
vs_data_dirty = is_flipped != prev_flipped;
// Sync clip plane uniforms
if (dirty.CheckClipping()) {
const auto raw_clip_coef = regs.rasterizer.GetClipCoef();
vs_data.enable_clip1 = regs.rasterizer.clip_enable != 0;
vs_data.clip_coef = {raw_clip_coef.x.ToFloat32(), raw_clip_coef.y.ToFloat32(),
raw_clip_coef.z.ToFloat32(), raw_clip_coef.w.ToFloat32()};
vs_data_dirty = true;
}
SyncGlobalAmbient();
for (u32 light_index = 0; light_index < 8; light_index++) {
SyncLightSpecular0(light_index);
SyncLightSpecular1(light_index);
SyncLightDiffuse(light_index);
SyncLightAmbient(light_index);
SyncLightPosition(light_index);
SyncLightDistanceAttenuationBias(light_index);
SyncLightDistanceAttenuationScale(light_index);
// Sync depth testing uniforms
if (dirty.CheckDepth()) {
fs_data.depth_scale = f24::FromRaw(regs.rasterizer.viewport_depth_range).ToFloat32();
fs_data.depth_offset = f24::FromRaw(regs.rasterizer.viewport_depth_near_plane).ToFloat32();
fs_data_dirty = true;
}
SyncFogColor();
SyncProcTexNoise();
SyncProcTexBias();
SyncShadowBias();
SyncShadowTextureBias();
for (u32 tex_index = 0; tex_index < 3; tex_index++) {
SyncTextureLodBias(tex_index);
// Sync alpha testing and blending uniforms
if (dirty.CheckBlend()) {
fs_data.alphatest_ref = regs.framebuffer.output_merger.alpha_test.ref;
fs_data.blend_color = ColorRGBA8(regs.framebuffer.output_merger.blend_const.raw);
fs_data_dirty = true;
}
}
void RasterizerAccelerated::NotifyPicaRegisterChanged(u32 id) {
switch (id) {
// Depth modifiers
case PICA_REG_INDEX(rasterizer.viewport_depth_range):
SyncDepthScale();
break;
case PICA_REG_INDEX(rasterizer.viewport_depth_near_plane):
SyncDepthOffset();
break;
// Depth buffering
case PICA_REG_INDEX(rasterizer.depthmap_enable):
shader_dirty = true;
break;
// Shadow texture
case PICA_REG_INDEX(texturing.shadow):
SyncShadowTextureBias();
break;
// Fog state
case PICA_REG_INDEX(texturing.fog_color):
SyncFogColor();
break;
case PICA_REG_INDEX(texturing.fog_lut_data[0]):
case PICA_REG_INDEX(texturing.fog_lut_data[1]):
case PICA_REG_INDEX(texturing.fog_lut_data[2]):
case PICA_REG_INDEX(texturing.fog_lut_data[3]):
case PICA_REG_INDEX(texturing.fog_lut_data[4]):
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[7]):
fs_uniform_block_data.fog_lut_dirty = true;
break;
// ProcTex state
case PICA_REG_INDEX(texturing.proctex):
case PICA_REG_INDEX(texturing.proctex_lut):
case PICA_REG_INDEX(texturing.proctex_lut_offset):
SyncProcTexBias();
shader_dirty = true;
break;
case PICA_REG_INDEX(texturing.proctex_noise_u):
case PICA_REG_INDEX(texturing.proctex_noise_v):
case PICA_REG_INDEX(texturing.proctex_noise_frequency):
SyncProcTexNoise();
break;
case PICA_REG_INDEX(texturing.proctex_lut_data[0]):
case PICA_REG_INDEX(texturing.proctex_lut_data[1]):
case PICA_REG_INDEX(texturing.proctex_lut_data[2]):
case PICA_REG_INDEX(texturing.proctex_lut_data[3]):
case PICA_REG_INDEX(texturing.proctex_lut_data[4]):
case PICA_REG_INDEX(texturing.proctex_lut_data[5]):
case PICA_REG_INDEX(texturing.proctex_lut_data[6]):
case PICA_REG_INDEX(texturing.proctex_lut_data[7]):
using Pica::TexturingRegs;
switch (regs.texturing.proctex_lut_config.ref_table.Value()) {
case TexturingRegs::ProcTexLutTable::Noise:
fs_uniform_block_data.proctex_noise_lut_dirty = true;
break;
case TexturingRegs::ProcTexLutTable::ColorMap:
fs_uniform_block_data.proctex_color_map_dirty = true;
break;
case TexturingRegs::ProcTexLutTable::AlphaMap:
fs_uniform_block_data.proctex_alpha_map_dirty = true;
break;
case TexturingRegs::ProcTexLutTable::Color:
fs_uniform_block_data.proctex_lut_dirty = true;
break;
case TexturingRegs::ProcTexLutTable::ColorDiff:
fs_uniform_block_data.proctex_diff_lut_dirty = true;
break;
// Sync texture unit uniforms
if (dirty.CheckTexUnits()) {
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;
fs_data.tex_lod_bias[tex_index] = config.lod.bias / 256.0f;
fs_data.tex_border_color[tex_index] = ColorRGBA8(config.border_color.raw);
}
break;
// Fragment operation mode
case PICA_REG_INDEX(framebuffer.output_merger.fragment_operation_mode):
shader_dirty = true;
break;
// Alpha test
case PICA_REG_INDEX(framebuffer.output_merger.alpha_test):
SyncAlphaTest();
shader_dirty = true;
break;
case PICA_REG_INDEX(framebuffer.shadow):
SyncShadowBias();
break;
// Scissor test
case PICA_REG_INDEX(rasterizer.scissor_test.mode):
shader_dirty = true;
break;
case PICA_REG_INDEX(texturing.main_config):
shader_dirty = true;
break;
// Texture 0 type
case PICA_REG_INDEX(texturing.texture0.type):
shader_dirty = true;
break;
// TEV stages
// (This also syncs fog_mode and fog_flip which are part of tev_combiner_buffer_input)
case PICA_REG_INDEX(texturing.tev_stage0.color_source1):
case PICA_REG_INDEX(texturing.tev_stage0.color_modifier1):
case PICA_REG_INDEX(texturing.tev_stage0.color_op):
case PICA_REG_INDEX(texturing.tev_stage0.color_scale):
case PICA_REG_INDEX(texturing.tev_stage1.color_source1):
case PICA_REG_INDEX(texturing.tev_stage1.color_modifier1):
case PICA_REG_INDEX(texturing.tev_stage1.color_op):
case PICA_REG_INDEX(texturing.tev_stage1.color_scale):
case PICA_REG_INDEX(texturing.tev_stage2.color_source1):
case PICA_REG_INDEX(texturing.tev_stage2.color_modifier1):
case PICA_REG_INDEX(texturing.tev_stage2.color_op):
case PICA_REG_INDEX(texturing.tev_stage2.color_scale):
case PICA_REG_INDEX(texturing.tev_stage3.color_source1):
case PICA_REG_INDEX(texturing.tev_stage3.color_modifier1):
case PICA_REG_INDEX(texturing.tev_stage3.color_op):
case PICA_REG_INDEX(texturing.tev_stage3.color_scale):
case PICA_REG_INDEX(texturing.tev_stage4.color_source1):
case PICA_REG_INDEX(texturing.tev_stage4.color_modifier1):
case PICA_REG_INDEX(texturing.tev_stage4.color_op):
case PICA_REG_INDEX(texturing.tev_stage4.color_scale):
case PICA_REG_INDEX(texturing.tev_stage5.color_source1):
case PICA_REG_INDEX(texturing.tev_stage5.color_modifier1):
case PICA_REG_INDEX(texturing.tev_stage5.color_op):
case PICA_REG_INDEX(texturing.tev_stage5.color_scale):
case PICA_REG_INDEX(texturing.tev_combiner_buffer_input):
shader_dirty = true;
break;
case PICA_REG_INDEX(texturing.tev_stage0.const_r):
SyncTevConstColor(0, regs.texturing.tev_stage0);
break;
case PICA_REG_INDEX(texturing.tev_stage1.const_r):
SyncTevConstColor(1, regs.texturing.tev_stage1);
break;
case PICA_REG_INDEX(texturing.tev_stage2.const_r):
SyncTevConstColor(2, regs.texturing.tev_stage2);
break;
case PICA_REG_INDEX(texturing.tev_stage3.const_r):
SyncTevConstColor(3, regs.texturing.tev_stage3);
break;
case PICA_REG_INDEX(texturing.tev_stage4.const_r):
SyncTevConstColor(4, regs.texturing.tev_stage4);
break;
case PICA_REG_INDEX(texturing.tev_stage5.const_r):
SyncTevConstColor(5, regs.texturing.tev_stage5);
break;
// TEV combiner buffer color
case PICA_REG_INDEX(texturing.tev_combiner_buffer_color):
SyncCombinerColor();
break;
// Fragment lighting switches
case PICA_REG_INDEX(lighting.disable):
case PICA_REG_INDEX(lighting.max_light_index):
case PICA_REG_INDEX(lighting.config0):
case PICA_REG_INDEX(lighting.config1):
case PICA_REG_INDEX(lighting.abs_lut_input):
case PICA_REG_INDEX(lighting.lut_input):
case PICA_REG_INDEX(lighting.lut_scale):
case PICA_REG_INDEX(lighting.light_enable):
break;
// Fragment lighting specular 0 color
case PICA_REG_INDEX(lighting.light[0].specular_0):
SyncLightSpecular0(0);
break;
case PICA_REG_INDEX(lighting.light[1].specular_0):
SyncLightSpecular0(1);
break;
case PICA_REG_INDEX(lighting.light[2].specular_0):
SyncLightSpecular0(2);
break;
case PICA_REG_INDEX(lighting.light[3].specular_0):
SyncLightSpecular0(3);
break;
case PICA_REG_INDEX(lighting.light[4].specular_0):
SyncLightSpecular0(4);
break;
case PICA_REG_INDEX(lighting.light[5].specular_0):
SyncLightSpecular0(5);
break;
case PICA_REG_INDEX(lighting.light[6].specular_0):
SyncLightSpecular0(6);
break;
case PICA_REG_INDEX(lighting.light[7].specular_0):
SyncLightSpecular0(7);
break;
// Fragment lighting specular 1 color
case PICA_REG_INDEX(lighting.light[0].specular_1):
SyncLightSpecular1(0);
break;
case PICA_REG_INDEX(lighting.light[1].specular_1):
SyncLightSpecular1(1);
break;
case PICA_REG_INDEX(lighting.light[2].specular_1):
SyncLightSpecular1(2);
break;
case PICA_REG_INDEX(lighting.light[3].specular_1):
SyncLightSpecular1(3);
break;
case PICA_REG_INDEX(lighting.light[4].specular_1):
SyncLightSpecular1(4);
break;
case PICA_REG_INDEX(lighting.light[5].specular_1):
SyncLightSpecular1(5);
break;
case PICA_REG_INDEX(lighting.light[6].specular_1):
SyncLightSpecular1(6);
break;
case PICA_REG_INDEX(lighting.light[7].specular_1):
SyncLightSpecular1(7);
break;
// Fragment lighting diffuse color
case PICA_REG_INDEX(lighting.light[0].diffuse):
SyncLightDiffuse(0);
break;
case PICA_REG_INDEX(lighting.light[1].diffuse):
SyncLightDiffuse(1);
break;
case PICA_REG_INDEX(lighting.light[2].diffuse):
SyncLightDiffuse(2);
break;
case PICA_REG_INDEX(lighting.light[3].diffuse):
SyncLightDiffuse(3);
break;
case PICA_REG_INDEX(lighting.light[4].diffuse):
SyncLightDiffuse(4);
break;
case PICA_REG_INDEX(lighting.light[5].diffuse):
SyncLightDiffuse(5);
break;
case PICA_REG_INDEX(lighting.light[6].diffuse):
SyncLightDiffuse(6);
break;
case PICA_REG_INDEX(lighting.light[7].diffuse):
SyncLightDiffuse(7);
break;
// Fragment lighting ambient color
case PICA_REG_INDEX(lighting.light[0].ambient):
SyncLightAmbient(0);
break;
case PICA_REG_INDEX(lighting.light[1].ambient):
SyncLightAmbient(1);
break;
case PICA_REG_INDEX(lighting.light[2].ambient):
SyncLightAmbient(2);
break;
case PICA_REG_INDEX(lighting.light[3].ambient):
SyncLightAmbient(3);
break;
case PICA_REG_INDEX(lighting.light[4].ambient):
SyncLightAmbient(4);
break;
case PICA_REG_INDEX(lighting.light[5].ambient):
SyncLightAmbient(5);
break;
case PICA_REG_INDEX(lighting.light[6].ambient):
SyncLightAmbient(6);
break;
case PICA_REG_INDEX(lighting.light[7].ambient):
SyncLightAmbient(7);
break;
// Fragment lighting position
case PICA_REG_INDEX(lighting.light[0].x):
case PICA_REG_INDEX(lighting.light[0].z):
SyncLightPosition(0);
break;
case PICA_REG_INDEX(lighting.light[1].x):
case PICA_REG_INDEX(lighting.light[1].z):
SyncLightPosition(1);
break;
case PICA_REG_INDEX(lighting.light[2].x):
case PICA_REG_INDEX(lighting.light[2].z):
SyncLightPosition(2);
break;
case PICA_REG_INDEX(lighting.light[3].x):
case PICA_REG_INDEX(lighting.light[3].z):
SyncLightPosition(3);
break;
case PICA_REG_INDEX(lighting.light[4].x):
case PICA_REG_INDEX(lighting.light[4].z):
SyncLightPosition(4);
break;
case PICA_REG_INDEX(lighting.light[5].x):
case PICA_REG_INDEX(lighting.light[5].z):
SyncLightPosition(5);
break;
case PICA_REG_INDEX(lighting.light[6].x):
case PICA_REG_INDEX(lighting.light[6].z):
SyncLightPosition(6);
break;
case PICA_REG_INDEX(lighting.light[7].x):
case PICA_REG_INDEX(lighting.light[7].z):
SyncLightPosition(7);
break;
// Fragment spot lighting direction
case PICA_REG_INDEX(lighting.light[0].spot_x):
case PICA_REG_INDEX(lighting.light[0].spot_z):
SyncLightSpotDirection(0);
break;
case PICA_REG_INDEX(lighting.light[1].spot_x):
case PICA_REG_INDEX(lighting.light[1].spot_z):
SyncLightSpotDirection(1);
break;
case PICA_REG_INDEX(lighting.light[2].spot_x):
case PICA_REG_INDEX(lighting.light[2].spot_z):
SyncLightSpotDirection(2);
break;
case PICA_REG_INDEX(lighting.light[3].spot_x):
case PICA_REG_INDEX(lighting.light[3].spot_z):
SyncLightSpotDirection(3);
break;
case PICA_REG_INDEX(lighting.light[4].spot_x):
case PICA_REG_INDEX(lighting.light[4].spot_z):
SyncLightSpotDirection(4);
break;
case PICA_REG_INDEX(lighting.light[5].spot_x):
case PICA_REG_INDEX(lighting.light[5].spot_z):
SyncLightSpotDirection(5);
break;
case PICA_REG_INDEX(lighting.light[6].spot_x):
case PICA_REG_INDEX(lighting.light[6].spot_z):
SyncLightSpotDirection(6);
break;
case PICA_REG_INDEX(lighting.light[7].spot_x):
case PICA_REG_INDEX(lighting.light[7].spot_z):
SyncLightSpotDirection(7);
break;
// Fragment lighting light source config
case PICA_REG_INDEX(lighting.light[0].config):
case PICA_REG_INDEX(lighting.light[1].config):
case PICA_REG_INDEX(lighting.light[2].config):
case PICA_REG_INDEX(lighting.light[3].config):
case PICA_REG_INDEX(lighting.light[4].config):
case PICA_REG_INDEX(lighting.light[5].config):
case PICA_REG_INDEX(lighting.light[6].config):
case PICA_REG_INDEX(lighting.light[7].config):
shader_dirty = true;
break;
// Fragment lighting distance attenuation bias
case PICA_REG_INDEX(lighting.light[0].dist_atten_bias):
SyncLightDistanceAttenuationBias(0);
break;
case PICA_REG_INDEX(lighting.light[1].dist_atten_bias):
SyncLightDistanceAttenuationBias(1);
break;
case PICA_REG_INDEX(lighting.light[2].dist_atten_bias):
SyncLightDistanceAttenuationBias(2);
break;
case PICA_REG_INDEX(lighting.light[3].dist_atten_bias):
SyncLightDistanceAttenuationBias(3);
break;
case PICA_REG_INDEX(lighting.light[4].dist_atten_bias):
SyncLightDistanceAttenuationBias(4);
break;
case PICA_REG_INDEX(lighting.light[5].dist_atten_bias):
SyncLightDistanceAttenuationBias(5);
break;
case PICA_REG_INDEX(lighting.light[6].dist_atten_bias):
SyncLightDistanceAttenuationBias(6);
break;
case PICA_REG_INDEX(lighting.light[7].dist_atten_bias):
SyncLightDistanceAttenuationBias(7);
break;
// Fragment lighting distance attenuation scale
case PICA_REG_INDEX(lighting.light[0].dist_atten_scale):
SyncLightDistanceAttenuationScale(0);
break;
case PICA_REG_INDEX(lighting.light[1].dist_atten_scale):
SyncLightDistanceAttenuationScale(1);
break;
case PICA_REG_INDEX(lighting.light[2].dist_atten_scale):
SyncLightDistanceAttenuationScale(2);
break;
case PICA_REG_INDEX(lighting.light[3].dist_atten_scale):
SyncLightDistanceAttenuationScale(3);
break;
case PICA_REG_INDEX(lighting.light[4].dist_atten_scale):
SyncLightDistanceAttenuationScale(4);
break;
case PICA_REG_INDEX(lighting.light[5].dist_atten_scale):
SyncLightDistanceAttenuationScale(5);
break;
case PICA_REG_INDEX(lighting.light[6].dist_atten_scale):
SyncLightDistanceAttenuationScale(6);
break;
case PICA_REG_INDEX(lighting.light[7].dist_atten_scale):
SyncLightDistanceAttenuationScale(7);
break;
// Fragment lighting global ambient color (emission + ambient * ambient)
case PICA_REG_INDEX(lighting.global_ambient):
SyncGlobalAmbient();
break;
// Fragment lighting lookup tables
case PICA_REG_INDEX(lighting.lut_data[0]):
case PICA_REG_INDEX(lighting.lut_data[1]):
case PICA_REG_INDEX(lighting.lut_data[2]):
case PICA_REG_INDEX(lighting.lut_data[3]):
case PICA_REG_INDEX(lighting.lut_data[4]):
case PICA_REG_INDEX(lighting.lut_data[5]):
case PICA_REG_INDEX(lighting.lut_data[6]):
case PICA_REG_INDEX(lighting.lut_data[7]): {
const auto& lut_config = regs.lighting.lut_config;
fs_uniform_block_data.lighting_lut_dirty[lut_config.type] = true;
fs_uniform_block_data.lighting_lut_dirty_any = true;
break;
fs_data_dirty = true;
}
// Texture LOD biases
case PICA_REG_INDEX(texturing.texture0.lod.bias):
SyncTextureLodBias(0);
break;
case PICA_REG_INDEX(texturing.texture1.lod.bias):
SyncTextureLodBias(1);
break;
case PICA_REG_INDEX(texturing.texture2.lod.bias):
SyncTextureLodBias(2);
break;
// Texture borders
case PICA_REG_INDEX(texturing.texture0.border_color):
SyncTextureBorderColor(0);
break;
case PICA_REG_INDEX(texturing.texture1.border_color):
SyncTextureBorderColor(1);
break;
case PICA_REG_INDEX(texturing.texture2.border_color):
SyncTextureBorderColor(2);
break;
// Clipping plane
case PICA_REG_INDEX(rasterizer.clip_enable):
case PICA_REG_INDEX(rasterizer.clip_coef[0]):
case PICA_REG_INDEX(rasterizer.clip_coef[1]):
case PICA_REG_INDEX(rasterizer.clip_coef[2]):
case PICA_REG_INDEX(rasterizer.clip_coef[3]):
SyncClipPlane();
break;
// Sync texenv uniforms
if (dirty.CheckTexEnv()) {
const auto tev_stages = regs.texturing.GetTevStages();
for (std::size_t index = 0; index < tev_stages.size(); ++index) {
fs_data.const_color[index] = ColorRGBA8(tev_stages[index].const_color);
}
fs_data.tev_combiner_buffer_color =
ColorRGBA8(regs.texturing.tev_combiner_buffer_color.raw);
fs_data_dirty = true;
}
// Forward registers that map to fixed function API features to the video backend
NotifyFixedFunctionPicaRegisterChanged(id);
}
void RasterizerAccelerated::SyncDepthScale() {
const f32 depth_scale = f24::FromRaw(regs.rasterizer.viewport_depth_range).ToFloat32();
if (depth_scale != fs_uniform_block_data.data.depth_scale) {
fs_uniform_block_data.data.depth_scale = depth_scale;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncDepthOffset() {
const f32 depth_offset = f24::FromRaw(regs.rasterizer.viewport_depth_near_plane).ToFloat32();
if (depth_offset != fs_uniform_block_data.data.depth_offset) {
fs_uniform_block_data.data.depth_offset = depth_offset;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncFogColor() {
const auto& fog_color_regs = regs.texturing.fog_color;
const Common::Vec3f fog_color = {
fog_color_regs.r.Value() / 255.0f,
fog_color_regs.g.Value() / 255.0f,
fog_color_regs.b.Value() / 255.0f,
};
if (fog_color != fs_uniform_block_data.data.fog_color) {
fs_uniform_block_data.data.fog_color = fog_color;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncProcTexNoise() {
const Common::Vec2f proctex_noise_f = {
Pica::f16::FromRaw(regs.texturing.proctex_noise_frequency.u).ToFloat32(),
Pica::f16::FromRaw(regs.texturing.proctex_noise_frequency.v).ToFloat32(),
};
const Common::Vec2f proctex_noise_a = {
regs.texturing.proctex_noise_u.amplitude / 4095.0f,
regs.texturing.proctex_noise_v.amplitude / 4095.0f,
};
const Common::Vec2f proctex_noise_p = {
Pica::f16::FromRaw(regs.texturing.proctex_noise_u.phase).ToFloat32(),
Pica::f16::FromRaw(regs.texturing.proctex_noise_v.phase).ToFloat32(),
};
if (proctex_noise_f != fs_uniform_block_data.data.proctex_noise_f ||
proctex_noise_a != fs_uniform_block_data.data.proctex_noise_a ||
proctex_noise_p != fs_uniform_block_data.data.proctex_noise_p) {
fs_uniform_block_data.data.proctex_noise_f = proctex_noise_f;
fs_uniform_block_data.data.proctex_noise_a = proctex_noise_a;
fs_uniform_block_data.data.proctex_noise_p = proctex_noise_p;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncProcTexBias() {
const auto proctex_bias = Pica::f16::FromRaw(regs.texturing.proctex.bias_low |
(regs.texturing.proctex_lut.bias_high << 8))
.ToFloat32();
if (proctex_bias != fs_uniform_block_data.data.proctex_bias) {
fs_uniform_block_data.data.proctex_bias = proctex_bias;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncAlphaTest() {
if (regs.framebuffer.output_merger.alpha_test.ref !=
static_cast<u32>(fs_uniform_block_data.data.alphatest_ref)) {
fs_uniform_block_data.data.alphatest_ref = regs.framebuffer.output_merger.alpha_test.ref;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncCombinerColor() {
const auto combiner_color = ColorRGBA8(regs.texturing.tev_combiner_buffer_color.raw);
if (combiner_color != fs_uniform_block_data.data.tev_combiner_buffer_color) {
fs_uniform_block_data.data.tev_combiner_buffer_color = combiner_color;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncTevConstColor(
const std::size_t stage_index, const Pica::TexturingRegs::TevStageConfig& tev_stage) {
const auto const_color = ColorRGBA8(tev_stage.const_color);
if (const_color == fs_uniform_block_data.data.const_color[stage_index]) {
return;
// Sync global lighting uniforms
if (dirty.CheckLightingAmbient()) {
fs_data.lighting_global_ambient = LightColor(regs.lighting.global_ambient);
fs_data_dirty = true;
}
fs_uniform_block_data.data.const_color[stage_index] = const_color;
fs_uniform_block_data.dirty = true;
}
// Sync light uniforms
for (u32 light_index = 0; light_index < 8; light_index++) {
if (!dirty.CheckLight(light_index)) {
continue;
}
void RasterizerAccelerated::SyncGlobalAmbient() {
const auto color = LightColor(regs.lighting.global_ambient);
if (color != fs_uniform_block_data.data.lighting_global_ambient) {
fs_uniform_block_data.data.lighting_global_ambient = color;
fs_uniform_block_data.dirty = true;
const auto& light = regs.lighting.light[light_index];
fs_data.light_src[light_index].specular_0 = LightColor(light.specular_0);
fs_data.light_src[light_index].specular_1 = LightColor(light.specular_1);
fs_data.light_src[light_index].diffuse = LightColor(light.diffuse);
fs_data.light_src[light_index].ambient = LightColor(light.ambient);
fs_data.light_src[light_index].position = {
Pica::f16::FromRaw(light.x).ToFloat32(),
Pica::f16::FromRaw(light.y).ToFloat32(),
Pica::f16::FromRaw(light.z).ToFloat32(),
};
fs_data.light_src[light_index].spot_direction = {
light.spot_x / 2047.0f, light.spot_y / 2047.0f, light.spot_z / 2047.0f};
fs_data.light_src[light_index].dist_atten_bias =
Pica::f20::FromRaw(light.dist_atten_bias).ToFloat32();
fs_data.light_src[light_index].dist_atten_scale =
Pica::f20::FromRaw(light.dist_atten_scale).ToFloat32();
fs_data_dirty = true;
}
}
void RasterizerAccelerated::SyncLightSpecular0(int light_index) {
const auto color = LightColor(regs.lighting.light[light_index].specular_0);
if (color != fs_uniform_block_data.data.light_src[light_index].specular_0) {
fs_uniform_block_data.data.light_src[light_index].specular_0 = color;
fs_uniform_block_data.dirty = true;
// Sync fog uniforms
if (dirty.CheckFogColor()) {
fs_data.fog_color = {
regs.texturing.fog_color.r.Value() / 255.0f,
regs.texturing.fog_color.g.Value() / 255.0f,
regs.texturing.fog_color.b.Value() / 255.0f,
};
fs_data_dirty = true;
}
}
void RasterizerAccelerated::SyncLightSpecular1(int light_index) {
const auto color = LightColor(regs.lighting.light[light_index].specular_1);
if (color != fs_uniform_block_data.data.light_src[light_index].specular_1) {
fs_uniform_block_data.data.light_src[light_index].specular_1 = color;
fs_uniform_block_data.dirty = true;
// Sync proctex uniforms
if (dirty.CheckProctex()) {
fs_data.proctex_noise_f = {
Pica::f16::FromRaw(regs.texturing.proctex_noise_frequency.u).ToFloat32(),
Pica::f16::FromRaw(regs.texturing.proctex_noise_frequency.v).ToFloat32(),
};
fs_data.proctex_noise_a = {
regs.texturing.proctex_noise_u.amplitude / 4095.0f,
regs.texturing.proctex_noise_v.amplitude / 4095.0f,
};
fs_data.proctex_noise_p = {
Pica::f16::FromRaw(regs.texturing.proctex_noise_u.phase).ToFloat32(),
Pica::f16::FromRaw(regs.texturing.proctex_noise_v.phase).ToFloat32(),
};
fs_data.proctex_bias = Pica::f16::FromRaw(regs.texturing.proctex.bias_low |
(regs.texturing.proctex_lut.bias_high << 8))
.ToFloat32();
fs_data_dirty = true;
}
}
void RasterizerAccelerated::SyncLightDiffuse(int light_index) {
const auto color = LightColor(regs.lighting.light[light_index].diffuse);
if (color != fs_uniform_block_data.data.light_src[light_index].diffuse) {
fs_uniform_block_data.data.light_src[light_index].diffuse = color;
fs_uniform_block_data.dirty = true;
// Sync shadow uniforms
if (dirty.CheckShadow()) {
const auto& shadow = regs.framebuffer.shadow;
fs_data.shadow_bias_constant = Pica::f16::FromRaw(shadow.constant).ToFloat32();
fs_data.shadow_bias_linear = Pica::f16::FromRaw(shadow.linear).ToFloat32();
fs_data.shadow_texture_bias = regs.texturing.shadow.bias << 1;
fs_data_dirty = true;
}
}
void RasterizerAccelerated::SyncLightAmbient(int light_index) {
const auto color = LightColor(regs.lighting.light[light_index].ambient);
if (color != fs_uniform_block_data.data.light_src[light_index].ambient) {
fs_uniform_block_data.data.light_src[light_index].ambient = color;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncLightPosition(int light_index) {
const Common::Vec3f position = {
Pica::f16::FromRaw(regs.lighting.light[light_index].x).ToFloat32(),
Pica::f16::FromRaw(regs.lighting.light[light_index].y).ToFloat32(),
Pica::f16::FromRaw(regs.lighting.light[light_index].z).ToFloat32(),
};
if (position != fs_uniform_block_data.data.light_src[light_index].position) {
fs_uniform_block_data.data.light_src[light_index].position = position;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncLightSpotDirection(int light_index) {
const auto& light = regs.lighting.light[light_index];
const auto spot_direction =
Common::Vec3f{light.spot_x / 2047.0f, light.spot_y / 2047.0f, light.spot_z / 2047.0f};
if (spot_direction != fs_uniform_block_data.data.light_src[light_index].spot_direction) {
fs_uniform_block_data.data.light_src[light_index].spot_direction = spot_direction;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncLightDistanceAttenuationBias(int light_index) {
const f32 dist_atten_bias =
Pica::f20::FromRaw(regs.lighting.light[light_index].dist_atten_bias).ToFloat32();
if (dist_atten_bias != fs_uniform_block_data.data.light_src[light_index].dist_atten_bias) {
fs_uniform_block_data.data.light_src[light_index].dist_atten_bias = dist_atten_bias;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncLightDistanceAttenuationScale(int light_index) {
const f32 dist_atten_scale =
Pica::f20::FromRaw(regs.lighting.light[light_index].dist_atten_scale).ToFloat32();
if (dist_atten_scale != fs_uniform_block_data.data.light_src[light_index].dist_atten_scale) {
fs_uniform_block_data.data.light_src[light_index].dist_atten_scale = dist_atten_scale;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncShadowBias() {
const auto& shadow = regs.framebuffer.shadow;
const f32 constant = Pica::f16::FromRaw(shadow.constant).ToFloat32();
const f32 linear = Pica::f16::FromRaw(shadow.linear).ToFloat32();
if (constant != fs_uniform_block_data.data.shadow_bias_constant ||
linear != fs_uniform_block_data.data.shadow_bias_linear) {
fs_uniform_block_data.data.shadow_bias_constant = constant;
fs_uniform_block_data.data.shadow_bias_linear = linear;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncShadowTextureBias() {
const s32 bias = regs.texturing.shadow.bias << 1;
if (bias != fs_uniform_block_data.data.shadow_texture_bias) {
fs_uniform_block_data.data.shadow_texture_bias = bias;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncTextureLodBias(int tex_index) {
const auto pica_textures = regs.texturing.GetTextures();
const f32 bias = pica_textures[tex_index].config.lod.bias / 256.0f;
if (bias != fs_uniform_block_data.data.tex_lod_bias[tex_index]) {
fs_uniform_block_data.data.tex_lod_bias[tex_index] = bias;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncTextureBorderColor(int tex_index) {
const auto pica_textures = regs.texturing.GetTextures();
const auto params = pica_textures[tex_index].config;
const Common::Vec4f border_color = ColorRGBA8(params.border_color.raw);
if (border_color != fs_uniform_block_data.data.tex_border_color[tex_index]) {
fs_uniform_block_data.data.tex_border_color[tex_index] = border_color;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerAccelerated::SyncClipPlane() {
const bool enable_clip1 = regs.rasterizer.clip_enable != 0;
const auto raw_clip_coef = regs.rasterizer.GetClipCoef();
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()};
if (enable_clip1 != (vs_uniform_block_data.data.enable_clip1 != 0) ||
new_clip_coef != vs_uniform_block_data.data.clip_coef) {
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;
}
// We have synched all uniforms, reset dirty state.
pica.dirty_regs.Reset();
}
} // namespace VideoCore

View File

@ -1,4 +1,4 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -27,105 +27,11 @@ public:
void AddTriangle(const Pica::OutputVertex& v0, const Pica::OutputVertex& v1,
const Pica::OutputVertex& v2) override;
void NotifyPicaRegisterChanged(u32 id) override;
void SyncEntireState() override;
protected:
/// Sync vertex and framgent uniforms from PICA registers
void SyncDrawUniforms();
protected:
/// Sync fixed-function pipeline state
virtual void SyncFixedState() = 0;
/// Notifies that a fixed function PICA register changed to the video backend
virtual void NotifyFixedFunctionPicaRegisterChanged(u32 id) = 0;
/// Syncs the depth scale to match the PICA register
void SyncDepthScale();
/// Syncs the depth offset to match the PICA register
void SyncDepthOffset();
/// Syncs the fog states to match the PICA register
void SyncFogColor();
/// Sync the procedural texture noise configuration to match the PICA register
void SyncProcTexNoise();
/// Sync the procedural texture bias configuration to match the PICA register
void SyncProcTexBias();
/// Syncs the alpha test states to match the PICA register
void SyncAlphaTest();
/// Syncs the TEV combiner color buffer to match the PICA register
void SyncCombinerColor();
/// Syncs the TEV constant color to match the PICA register
void SyncTevConstColor(std::size_t tev_index,
const Pica::TexturingRegs::TevStageConfig& tev_stage);
/// Syncs the lighting global ambient color to match the PICA register
void SyncGlobalAmbient();
/// Syncs the specified light's specular 0 color to match the PICA register
void SyncLightSpecular0(int light_index);
/// Syncs the specified light's specular 1 color to match the PICA register
void SyncLightSpecular1(int light_index);
/// Syncs the specified light's diffuse color to match the PICA register
void SyncLightDiffuse(int light_index);
/// Syncs the specified light's ambient color to match the PICA register
void SyncLightAmbient(int light_index);
/// Syncs the specified light's position to match the PICA register
void SyncLightPosition(int light_index);
/// Syncs the specified spot light direcition to match the PICA register
void SyncLightSpotDirection(int light_index);
/// Syncs the specified light's distance attenuation bias to match the PICA register
void SyncLightDistanceAttenuationBias(int light_index);
/// Syncs the specified light's distance attenuation scale to match the PICA register
void SyncLightDistanceAttenuationScale(int light_index);
/// Syncs the shadow rendering bias to match the PICA register
void SyncShadowBias();
/// Syncs the shadow texture bias to match the PICA register
void SyncShadowTextureBias();
/// Syncs the texture LOD bias to match the PICA register
void SyncTextureLodBias(int tex_index);
/// Syncs the texture border color to match the PICA registers
void SyncTextureBorderColor(int tex_index);
/// Syncs the clip plane state to match the PICA register
void SyncClipPlane();
protected:
/// Structure that keeps tracks of the vertex shader uniform state
struct VSUniformBlockData {
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{};
bool lighting_lut_dirty_any = true;
bool fog_lut_dirty = true;
bool proctex_noise_lut_dirty = true;
bool proctex_color_map_dirty = true;
bool proctex_alpha_map_dirty = true;
bool proctex_lut_dirty = true;
bool proctex_diff_lut_dirty = true;
bool dirty = true;
};
/// Structure that the hardware rendered vertices are composed of
struct HardwareVertex {
HardwareVertex() = default;
@ -154,21 +60,12 @@ protected:
Memory::MemorySystem& memory;
Pica::PicaCore& pica;
Pica::RegsInternal& regs;
std::vector<HardwareVertex> vertex_batch;
Pica::Shader::UserConfig user_config{};
bool shader_dirty = true;
VSUniformBlockData vs_uniform_block_data{};
FSUniformBlockData fs_uniform_block_data{};
using LightLUT = std::array<Common::Vec2f, 256>;
std::array<LightLUT, Pica::LightingRegs::NumLightingSampler> lighting_lut_data{};
std::array<Common::Vec2f, 128> fog_lut_data{};
std::array<Common::Vec2f, 128> proctex_noise_lut_data{};
std::array<Common::Vec2f, 128> proctex_color_map_data{};
std::array<Common::Vec2f, 128> proctex_alpha_map_data{};
std::array<Common::Vec4f, 256> proctex_lut_data{};
std::array<Common::Vec4f, 256> proctex_diff_lut_data{};
Pica::Shader::Generator::VSUniformData vs_data{};
Pica::Shader::Generator::FSUniformData fs_data{};
bool vs_data_dirty = true;
bool fs_data_dirty = true;
};
} // namespace VideoCore

View File

@ -39,9 +39,6 @@ public:
/// Draw the current batch of triangles
virtual void DrawTriangles() = 0;
/// Notify rasterizer that the specified PICA register has been changed
virtual void NotifyPicaRegisterChanged(u32 id) = 0;
/// Notify rasterizer that all caches should be flushed to 3DS memory
virtual void FlushAll() = 0;
@ -88,8 +85,6 @@ public:
switch_disk_resources_callback = callback;
}
virtual void SyncEntireState() {}
void SetAccurateMul(bool accurate_mul_) {
accurate_mul = accurate_mul_;
}

View File

@ -1,4 +1,4 @@
// Copyright 2014 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -60,9 +60,6 @@ public:
/// Cleans up after video dumping is ended
virtual void CleanupVideoDumping() {}
/// Synchronizes fixed function renderer state
virtual void Sync() {}
/// This is called to notify the rendering backend of a surface change
virtual void NotifySurfaceChanged() {}

View File

@ -164,8 +164,6 @@ RasterizerOpenGL::RasterizerOpenGL(Memory::MemorySystem& memory, Pica::PicaCore&
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, index_buffer.GetHandle());
glEnable(GL_BLEND);
SyncEntireState();
}
RasterizerOpenGL::~RasterizerOpenGL() = default;
@ -256,18 +254,113 @@ void RasterizerOpenGL::SwitchDiskResources(u64 title_id) {
}
}
void RasterizerOpenGL::SyncFixedState() {
SyncClipEnabled();
SyncCullMode();
SyncBlendEnabled();
SyncBlendFuncs();
SyncBlendColor();
SyncLogicOp();
SyncStencilTest();
SyncDepthTest();
SyncColorWriteMask();
SyncStencilWriteMask();
SyncDepthWriteMask();
void RasterizerOpenGL::SyncDrawState() {
SyncDrawUniforms();
// SyncClipEnabled();
state.clip_distance[1] = regs.rasterizer.clip_enable != 0;
// SyncCullMode();
state.cull.enabled = regs.rasterizer.cull_mode != Pica::RasterizerRegs::CullMode::KeepAll;
if (state.cull.enabled) {
state.cull.front_face =
regs.rasterizer.cull_mode == Pica::RasterizerRegs::CullMode::KeepClockWise ? GL_CW
: GL_CCW;
}
// If the framebuffer is flipped, vertex shader flips vertex y, so invert culling
const bool is_flipped = regs.framebuffer.framebuffer.IsFlipped();
state.cull.mode = is_flipped && state.cull.enabled ? GL_FRONT : GL_BACK;
// SyncBlendEnabled();
state.blend.enabled = (regs.framebuffer.output_merger.alphablend_enable == 1);
// SyncBlendFuncs();
const bool has_minmax_factor = driver.HasBlendMinMaxFactor();
state.blend.rgb_equation = PicaToGL::BlendEquation(
regs.framebuffer.output_merger.alpha_blending.blend_equation_rgb, has_minmax_factor);
state.blend.a_equation = PicaToGL::BlendEquation(
regs.framebuffer.output_merger.alpha_blending.blend_equation_a, has_minmax_factor);
state.blend.src_rgb_func =
PicaToGL::BlendFunc(regs.framebuffer.output_merger.alpha_blending.factor_source_rgb);
state.blend.dst_rgb_func =
PicaToGL::BlendFunc(regs.framebuffer.output_merger.alpha_blending.factor_dest_rgb);
state.blend.src_a_func =
PicaToGL::BlendFunc(regs.framebuffer.output_merger.alpha_blending.factor_source_a);
state.blend.dst_a_func =
PicaToGL::BlendFunc(regs.framebuffer.output_merger.alpha_blending.factor_dest_a);
if (!has_minmax_factor) {
// Blending with min/max equations is emulated in the fragment shader so
// configure blending to not modify the incoming fragment color.
emulate_minmax_blend = false;
if (state.EmulateColorBlend()) {
emulate_minmax_blend = true;
state.blend.rgb_equation = GL_FUNC_ADD;
state.blend.src_rgb_func = GL_ONE;
state.blend.dst_rgb_func = GL_ZERO;
}
if (state.EmulateAlphaBlend()) {
emulate_minmax_blend = true;
state.blend.a_equation = GL_FUNC_ADD;
state.blend.src_a_func = GL_ONE;
state.blend.dst_a_func = GL_ZERO;
}
}
// SyncBlendColor();
const auto blend_color = PicaToGL::ColorRGBA8(regs.framebuffer.output_merger.blend_const.raw);
state.blend.color.red = blend_color[0];
state.blend.color.green = blend_color[1];
state.blend.color.blue = blend_color[2];
state.blend.color.alpha = blend_color[3];
// SyncLogicOp();
// SyncColorWriteMask();
state.logic_op = PicaToGL::LogicOp(regs.framebuffer.output_merger.logic_op);
if (driver.IsOpenGLES() && !regs.framebuffer.output_merger.alphablend_enable &&
regs.framebuffer.output_merger.logic_op == Pica::FramebufferRegs::LogicOp::NoOp) {
// Color output is disabled by logic operation. We use color write mask to skip
// color but allow depth write.
state.color_mask = {};
} else {
auto is_color_write_enabled = [&](u32 value) {
return (regs.framebuffer.framebuffer.allow_color_write != 0 && value != 0) ? GL_TRUE
: GL_FALSE;
};
state.color_mask.red_enabled =
is_color_write_enabled(regs.framebuffer.output_merger.red_enable);
state.color_mask.green_enabled =
is_color_write_enabled(regs.framebuffer.output_merger.green_enable);
state.color_mask.blue_enabled =
is_color_write_enabled(regs.framebuffer.output_merger.blue_enable);
state.color_mask.alpha_enabled =
is_color_write_enabled(regs.framebuffer.output_merger.alpha_enable);
}
// SyncStencilTest();
state.stencil.test_enabled =
regs.framebuffer.output_merger.stencil_test.enable &&
regs.framebuffer.framebuffer.depth_format == Pica::FramebufferRegs::DepthFormat::D24S8;
state.stencil.test_func =
PicaToGL::CompareFunc(regs.framebuffer.output_merger.stencil_test.func);
state.stencil.test_ref = regs.framebuffer.output_merger.stencil_test.reference_value;
state.stencil.test_mask = regs.framebuffer.output_merger.stencil_test.input_mask;
state.stencil.action_stencil_fail =
PicaToGL::StencilOp(regs.framebuffer.output_merger.stencil_test.action_stencil_fail);
state.stencil.action_depth_fail =
PicaToGL::StencilOp(regs.framebuffer.output_merger.stencil_test.action_depth_fail);
state.stencil.action_depth_pass =
PicaToGL::StencilOp(regs.framebuffer.output_merger.stencil_test.action_depth_pass);
// SyncDepthTest();
state.depth.test_enabled = regs.framebuffer.output_merger.depth_test_enable == 1 ||
regs.framebuffer.output_merger.depth_write_enable == 1;
state.depth.test_func =
regs.framebuffer.output_merger.depth_test_enable == 1
? PicaToGL::CompareFunc(regs.framebuffer.output_merger.depth_test_func)
: GL_ALWAYS;
// SyncStencilWriteMask();
state.stencil.write_mask =
(regs.framebuffer.framebuffer.allow_depth_stencil_write != 0)
? static_cast<GLuint>(regs.framebuffer.output_merger.stencil_test.write_mask)
: 0;
// SyncDepthWriteMask();
state.depth.write_mask = (regs.framebuffer.framebuffer.allow_depth_stencil_write != 0 &&
regs.framebuffer.output_merger.depth_write_enable)
? GL_TRUE
: GL_FALSE;
}
void RasterizerOpenGL::SetupVertexArray(u8* array_ptr, GLintptr buffer_offset,
@ -448,6 +541,7 @@ void RasterizerOpenGL::DrawTriangles() {
bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
MICROPROFILE_SCOPE(OpenGL_Drawing);
SyncDrawState();
const bool shadow_rendering = regs.framebuffer.IsShadowRendering();
const bool has_stencil = regs.framebuffer.HasStencil();
@ -487,12 +581,6 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
state.viewport.width = static_cast<GLsizei>(viewport.width);
state.viewport.height = static_cast<GLsizei>(viewport.height);
// If the framebuffer is flipped, request vertex shader to flip vertex y
const bool is_flipped = regs.framebuffer.framebuffer.IsFlipped();
vs_uniform_block_data.dirty |= (vs_uniform_block_data.data.flip_viewport != 0) != is_flipped;
vs_uniform_block_data.data.flip_viewport = is_flipped;
state.cull.mode = is_flipped && state.cull.enabled ? GL_FRONT : GL_BACK;
// Viewport can have negative offsets or larger dimensions than our framebuffer sub-rect.
// Enable scissor test to prevent drawing outside of the framebuffer region
const auto draw_rect = fb_helper.DrawRect();
@ -504,16 +592,14 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
// Update scissor uniforms
const auto [scissor_x1, scissor_y2, scissor_x2, scissor_y1] = fb_helper.Scissor();
if (fs_uniform_block_data.data.scissor_x1 != scissor_x1 ||
fs_uniform_block_data.data.scissor_x2 != scissor_x2 ||
fs_uniform_block_data.data.scissor_y1 != scissor_y1 ||
fs_uniform_block_data.data.scissor_y2 != scissor_y2) {
if (fs_data.scissor_x1 != scissor_x1 || fs_data.scissor_x2 != scissor_x2 ||
fs_data.scissor_y1 != scissor_y1 || fs_data.scissor_y2 != scissor_y2) {
fs_uniform_block_data.data.scissor_x1 = scissor_x1;
fs_uniform_block_data.data.scissor_x2 = scissor_x2;
fs_uniform_block_data.data.scissor_y1 = scissor_y1;
fs_uniform_block_data.data.scissor_y2 = scissor_y2;
fs_uniform_block_data.dirty = true;
fs_data.scissor_x1 = scissor_x1;
fs_data.scissor_x2 = scissor_x2;
fs_data.scissor_y1 = scissor_y1;
fs_data.scissor_y2 = scissor_y2;
fs_data_dirty = true;
}
// Sync and bind the texture surfaces
@ -521,10 +607,7 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
state.Apply();
// Sync and bind the shader
if (shader_dirty) {
curr_shader_manager->UseFragmentShader(regs, user_config);
shader_dirty = false;
}
curr_shader_manager->UseFragmentShader(regs, user_config);
// Sync the LUTs within the texture buffer
SyncAndUploadLUTs();
@ -708,73 +791,6 @@ void RasterizerOpenGL::UnbindSpecial() {
state.image_shadow_buffer = 0;
}
void RasterizerOpenGL::NotifyFixedFunctionPicaRegisterChanged(u32 id) {
switch (id) {
// Clipping plane
case PICA_REG_INDEX(rasterizer.clip_enable):
SyncClipEnabled();
break;
// Culling
case PICA_REG_INDEX(rasterizer.cull_mode):
SyncCullMode();
break;
// Blending
case PICA_REG_INDEX(framebuffer.output_merger.alphablend_enable):
SyncBlendEnabled();
// Update since logic op emulation depends on alpha blend enable.
SyncLogicOp();
SyncColorWriteMask();
break;
case PICA_REG_INDEX(framebuffer.output_merger.alpha_blending):
SyncBlendFuncs();
break;
case PICA_REG_INDEX(framebuffer.output_merger.blend_const):
SyncBlendColor();
break;
// Sync GL stencil test + stencil write mask
// (Pica stencil test function register also contains a stencil write mask)
case PICA_REG_INDEX(framebuffer.output_merger.stencil_test.raw_func):
SyncStencilTest();
SyncStencilWriteMask();
break;
case PICA_REG_INDEX(framebuffer.output_merger.stencil_test.raw_op):
case PICA_REG_INDEX(framebuffer.framebuffer.depth_format):
SyncStencilTest();
break;
// Sync GL depth test + depth and color write mask
// (Pica depth test function register also contains a depth and color write mask)
case PICA_REG_INDEX(framebuffer.output_merger.depth_test_enable):
SyncDepthTest();
SyncDepthWriteMask();
SyncColorWriteMask();
break;
// Sync GL depth and stencil write mask
// (This is a dedicated combined depth / stencil write-enable register)
case PICA_REG_INDEX(framebuffer.framebuffer.allow_depth_stencil_write):
SyncDepthWriteMask();
SyncStencilWriteMask();
break;
// Sync GL color write mask
// (This is a dedicated color write-enable register)
case PICA_REG_INDEX(framebuffer.framebuffer.allow_color_write):
SyncColorWriteMask();
break;
// Logic op
case PICA_REG_INDEX(framebuffer.output_merger.logic_op):
SyncLogicOp();
// Update since color write mask is used to emulate no-op.
SyncColorWriteMask();
break;
}
}
void RasterizerOpenGL::FlushAll() {
res_cache.FlushAll();
}
@ -852,170 +868,12 @@ bool RasterizerOpenGL::AccelerateDisplay(const Pica::FramebufferConfig& config,
return true;
}
void RasterizerOpenGL::SyncClipEnabled() {
state.clip_distance[1] = regs.rasterizer.clip_enable != 0;
}
void RasterizerOpenGL::SyncCullMode() {
switch (regs.rasterizer.cull_mode) {
case Pica::RasterizerRegs::CullMode::KeepAll:
state.cull.enabled = false;
break;
case Pica::RasterizerRegs::CullMode::KeepClockWise:
state.cull.enabled = true;
state.cull.front_face = GL_CW;
break;
case Pica::RasterizerRegs::CullMode::KeepCounterClockWise:
state.cull.enabled = true;
state.cull.front_face = GL_CCW;
break;
default:
LOG_CRITICAL(Render_OpenGL, "Unknown cull mode {}",
static_cast<u32>(regs.rasterizer.cull_mode.Value()));
UNIMPLEMENTED();
break;
}
}
void RasterizerOpenGL::SyncBlendEnabled() {
state.blend.enabled = (regs.framebuffer.output_merger.alphablend_enable == 1);
}
void RasterizerOpenGL::SyncBlendFuncs() {
const bool has_minmax_factor = driver.HasBlendMinMaxFactor();
state.blend.rgb_equation = PicaToGL::BlendEquation(
regs.framebuffer.output_merger.alpha_blending.blend_equation_rgb, has_minmax_factor);
state.blend.a_equation = PicaToGL::BlendEquation(
regs.framebuffer.output_merger.alpha_blending.blend_equation_a, has_minmax_factor);
state.blend.src_rgb_func =
PicaToGL::BlendFunc(regs.framebuffer.output_merger.alpha_blending.factor_source_rgb);
state.blend.dst_rgb_func =
PicaToGL::BlendFunc(regs.framebuffer.output_merger.alpha_blending.factor_dest_rgb);
state.blend.src_a_func =
PicaToGL::BlendFunc(regs.framebuffer.output_merger.alpha_blending.factor_source_a);
state.blend.dst_a_func =
PicaToGL::BlendFunc(regs.framebuffer.output_merger.alpha_blending.factor_dest_a);
if (has_minmax_factor) {
return;
}
// Blending with min/max equations is emulated in the fragment shader so
// configure blending to not modify the incoming fragment color.
emulate_minmax_blend = false;
if (state.EmulateColorBlend()) {
emulate_minmax_blend = true;
state.blend.rgb_equation = GL_FUNC_ADD;
state.blend.src_rgb_func = GL_ONE;
state.blend.dst_rgb_func = GL_ZERO;
}
if (state.EmulateAlphaBlend()) {
emulate_minmax_blend = true;
state.blend.a_equation = GL_FUNC_ADD;
state.blend.src_a_func = GL_ONE;
state.blend.dst_a_func = GL_ZERO;
}
}
void RasterizerOpenGL::SyncBlendColor() {
const auto blend_color = PicaToGL::ColorRGBA8(regs.framebuffer.output_merger.blend_const.raw);
state.blend.color.red = blend_color[0];
state.blend.color.green = blend_color[1];
state.blend.color.blue = blend_color[2];
state.blend.color.alpha = blend_color[3];
if (blend_color != fs_uniform_block_data.data.blend_color) {
fs_uniform_block_data.data.blend_color = blend_color;
fs_uniform_block_data.dirty = true;
}
}
void RasterizerOpenGL::SyncLogicOp() {
state.logic_op = PicaToGL::LogicOp(regs.framebuffer.output_merger.logic_op);
if (driver.IsOpenGLES()) {
if (!regs.framebuffer.output_merger.alphablend_enable) {
if (regs.framebuffer.output_merger.logic_op == Pica::FramebufferRegs::LogicOp::NoOp) {
// Color output is disabled by logic operation. We use color write mask to skip
// color but allow depth write.
state.color_mask = {};
}
}
}
}
void RasterizerOpenGL::SyncColorWriteMask() {
if (driver.IsOpenGLES()) {
if (!regs.framebuffer.output_merger.alphablend_enable) {
if (regs.framebuffer.output_merger.logic_op == Pica::FramebufferRegs::LogicOp::NoOp) {
// Color output is disabled by logic operation. We use color write mask to skip
// color but allow depth write. Return early to avoid overwriting this.
return;
}
}
}
auto is_color_write_enabled = [&](u32 value) {
return (regs.framebuffer.framebuffer.allow_color_write != 0 && value != 0) ? GL_TRUE
: GL_FALSE;
};
state.color_mask.red_enabled =
is_color_write_enabled(regs.framebuffer.output_merger.red_enable);
state.color_mask.green_enabled =
is_color_write_enabled(regs.framebuffer.output_merger.green_enable);
state.color_mask.blue_enabled =
is_color_write_enabled(regs.framebuffer.output_merger.blue_enable);
state.color_mask.alpha_enabled =
is_color_write_enabled(regs.framebuffer.output_merger.alpha_enable);
}
void RasterizerOpenGL::SyncStencilWriteMask() {
state.stencil.write_mask =
(regs.framebuffer.framebuffer.allow_depth_stencil_write != 0)
? static_cast<GLuint>(regs.framebuffer.output_merger.stencil_test.write_mask)
: 0;
}
void RasterizerOpenGL::SyncDepthWriteMask() {
state.depth.write_mask = (regs.framebuffer.framebuffer.allow_depth_stencil_write != 0 &&
regs.framebuffer.output_merger.depth_write_enable)
? GL_TRUE
: GL_FALSE;
}
void RasterizerOpenGL::SyncStencilTest() {
state.stencil.test_enabled =
regs.framebuffer.output_merger.stencil_test.enable &&
regs.framebuffer.framebuffer.depth_format == Pica::FramebufferRegs::DepthFormat::D24S8;
state.stencil.test_func =
PicaToGL::CompareFunc(regs.framebuffer.output_merger.stencil_test.func);
state.stencil.test_ref = regs.framebuffer.output_merger.stencil_test.reference_value;
state.stencil.test_mask = regs.framebuffer.output_merger.stencil_test.input_mask;
state.stencil.action_stencil_fail =
PicaToGL::StencilOp(regs.framebuffer.output_merger.stencil_test.action_stencil_fail);
state.stencil.action_depth_fail =
PicaToGL::StencilOp(regs.framebuffer.output_merger.stencil_test.action_depth_fail);
state.stencil.action_depth_pass =
PicaToGL::StencilOp(regs.framebuffer.output_merger.stencil_test.action_depth_pass);
}
void RasterizerOpenGL::SyncDepthTest() {
state.depth.test_enabled = regs.framebuffer.output_merger.depth_test_enable == 1 ||
regs.framebuffer.output_merger.depth_write_enable == 1;
state.depth.test_func =
regs.framebuffer.output_merger.depth_test_enable == 1
? PicaToGL::CompareFunc(regs.framebuffer.output_merger.depth_test_func)
: GL_ALWAYS;
}
void RasterizerOpenGL::SyncAndUploadLUTsLF() {
constexpr std::size_t max_size =
sizeof(Common::Vec2f) * 256 * Pica::LightingRegs::NumLightingSampler +
sizeof(Common::Vec2f) * 128; // fog
if (!fs_uniform_block_data.lighting_lut_dirty_any && !fs_uniform_block_data.fog_lut_dirty) {
if (!pica.lighting.lut_dirty && !pica.fog.lut_dirty) {
return;
}
@ -1024,50 +882,37 @@ void RasterizerOpenGL::SyncAndUploadLUTsLF() {
const auto [buffer, offset, invalidate] =
texture_lf_buffer.Map(max_size, sizeof(Common::Vec4f));
// Sync the lighting luts
if (fs_uniform_block_data.lighting_lut_dirty_any || invalidate) {
for (unsigned index = 0; index < fs_uniform_block_data.lighting_lut_dirty.size(); index++) {
if (fs_uniform_block_data.lighting_lut_dirty[index] || invalidate) {
std::array<Common::Vec2f, 256> new_data;
const auto& source_lut = pica.lighting.luts[index];
std::transform(source_lut.begin(), source_lut.end(), new_data.begin(),
[](const auto& entry) {
return Common::Vec2f{entry.ToFloat(), entry.DiffToFloat()};
});
if (invalidate) {
pica.lighting.lut_dirty = pica.lighting.LutAllDirty;
pica.fog.lut_dirty = true;
}
if (new_data != lighting_lut_data[index] || invalidate) {
lighting_lut_data[index] = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f));
fs_uniform_block_data.data.lighting_lut_offset[index / 4][index % 4] =
static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f);
}
fs_uniform_block_data.lighting_lut_dirty[index] = false;
}
// Sync the lighting luts
while (pica.lighting.lut_dirty) {
const u32 index = std::countr_zero(pica.lighting.lut_dirty);
pica.lighting.lut_dirty &= ~(1 << index);
Common::Vec2f* new_data = reinterpret_cast<Common::Vec2f*>(buffer + bytes_used);
const auto& source_lut = pica.lighting.luts[index];
for (u32 i = 0; i < source_lut.size(); i++) {
new_data[i] = {source_lut[i].ToFloat(), source_lut[i].DiffToFloat()};
}
fs_uniform_block_data.lighting_lut_dirty_any = false;
fs_data.lighting_lut_offset[index / 4][index % 4] =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_data_dirty = true;
bytes_used += source_lut.size() * sizeof(Common::Vec2f);
}
// Sync the fog lut
if (fs_uniform_block_data.fog_lut_dirty || invalidate) {
std::array<Common::Vec2f, 128> new_data;
std::transform(
pica.fog.lut.begin(), pica.fog.lut.end(), new_data.begin(),
[](const auto& entry) { return Common::Vec2f{entry.ToFloat(), entry.DiffToFloat()}; });
if (new_data != fog_lut_data || invalidate) {
fog_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f));
fs_uniform_block_data.data.fog_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f);
if (pica.fog.lut_dirty) {
Common::Vec2f* new_data = reinterpret_cast<Common::Vec2f*>(buffer + bytes_used);
for (u32 i = 0; i < pica.fog.lut.size(); i++) {
new_data[i] = {pica.fog.lut[i].ToFloat(), pica.fog.lut[i].DiffToFloat()};
}
fs_uniform_block_data.fog_lut_dirty = false;
fs_data.fog_lut_offset = static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_data_dirty = true;
bytes_used += pica.fog.lut.size() * sizeof(Common::Vec2f);
pica.fog.lut_dirty = false;
}
texture_lf_buffer.Unmap(bytes_used);
@ -1079,10 +924,7 @@ void RasterizerOpenGL::SyncAndUploadLUTs() {
sizeof(Common::Vec4f) * 256 + // proctex
sizeof(Common::Vec4f) * 256; // proctex diff
if (!fs_uniform_block_data.proctex_noise_lut_dirty &&
!fs_uniform_block_data.proctex_color_map_dirty &&
!fs_uniform_block_data.proctex_alpha_map_dirty &&
!fs_uniform_block_data.proctex_lut_dirty && !fs_uniform_block_data.proctex_diff_lut_dirty) {
if (!pica.proctex.table_dirty) {
return;
}
@ -1090,90 +932,62 @@ void RasterizerOpenGL::SyncAndUploadLUTs() {
glBindBuffer(GL_TEXTURE_BUFFER, texture_buffer.GetHandle());
const auto [buffer, offset, invalidate] = texture_buffer.Map(max_size, sizeof(Common::Vec4f));
// helper function for SyncProcTexNoiseLUT/ColorMap/AlphaMap
const auto sync_proc_tex_value_lut =
[this, buffer = buffer, offset = offset, invalidate = invalidate, &bytes_used](
const auto& lut, std::array<Common::Vec2f, 128>& lut_data, GLint& lut_offset) {
std::array<Common::Vec2f, 128> new_data;
std::transform(lut.begin(), lut.end(), new_data.begin(), [](const auto& entry) {
return Common::Vec2f{entry.ToFloat(), entry.DiffToFloat()};
});
if (invalidate) {
pica.proctex.table_dirty = pica.proctex.TableAllDirty;
}
if (new_data != lut_data || invalidate) {
lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f));
lut_offset = static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f);
}
};
// helper function for SyncProcTexNoiseLUT/ColorMap/AlphaMap
const auto sync_proc_tex_value_lut = [&](const auto& lut, GLint& lut_offset) {
Common::Vec2f* new_data = reinterpret_cast<Common::Vec2f*>(buffer + bytes_used);
for (u32 i = 0; i < lut.size(); i++) {
new_data[i] = {lut[i].ToFloat(), lut[i].DiffToFloat()};
}
lut_offset = static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_data_dirty = true;
bytes_used += lut.size() * sizeof(Common::Vec2f);
};
// Sync the proctex noise lut
if (fs_uniform_block_data.proctex_noise_lut_dirty || invalidate) {
sync_proc_tex_value_lut(pica.proctex.noise_table, proctex_noise_lut_data,
fs_uniform_block_data.data.proctex_noise_lut_offset);
fs_uniform_block_data.proctex_noise_lut_dirty = false;
if (pica.proctex.noise_lut_dirty) {
sync_proc_tex_value_lut(pica.proctex.noise_table, fs_data.proctex_noise_lut_offset);
}
// Sync the proctex color map
if (fs_uniform_block_data.proctex_color_map_dirty || invalidate) {
sync_proc_tex_value_lut(pica.proctex.color_map_table, proctex_color_map_data,
fs_uniform_block_data.data.proctex_color_map_offset);
fs_uniform_block_data.proctex_color_map_dirty = false;
if (pica.proctex.color_map_dirty) {
sync_proc_tex_value_lut(pica.proctex.color_map_table, fs_data.proctex_color_map_offset);
}
// Sync the proctex alpha map
if (fs_uniform_block_data.proctex_alpha_map_dirty || invalidate) {
sync_proc_tex_value_lut(pica.proctex.alpha_map_table, proctex_alpha_map_data,
fs_uniform_block_data.data.proctex_alpha_map_offset);
fs_uniform_block_data.proctex_alpha_map_dirty = false;
if (pica.proctex.alpha_map_dirty) {
sync_proc_tex_value_lut(pica.proctex.alpha_map_table, fs_data.proctex_alpha_map_offset);
}
// Sync the proctex lut
if (fs_uniform_block_data.proctex_lut_dirty || invalidate) {
std::array<Common::Vec4f, 256> new_data;
std::transform(pica.proctex.color_table.begin(), pica.proctex.color_table.end(),
new_data.begin(), [](const auto& entry) {
auto rgba = entry.ToVector() / 255.0f;
return Common::Vec4f{rgba.r(), rgba.g(), rgba.b(), rgba.a()};
});
if (new_data != proctex_lut_data || invalidate) {
proctex_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec4f));
fs_uniform_block_data.data.proctex_lut_offset =
static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec4f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec4f);
if (pica.proctex.lut_dirty) {
Common::Vec4f* new_data = reinterpret_cast<Common::Vec4f*>(buffer + bytes_used);
for (u32 i = 0; i < pica.proctex.color_table.size(); i++) {
new_data[i] = pica.proctex.color_table[i].ToVector() / 255.0f;
}
fs_uniform_block_data.proctex_lut_dirty = false;
fs_data.proctex_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f));
fs_data_dirty = true;
bytes_used += pica.proctex.color_table.size() * sizeof(Common::Vec4f);
}
// Sync the proctex difference lut
if (fs_uniform_block_data.proctex_diff_lut_dirty || invalidate) {
std::array<Common::Vec4f, 256> new_data;
std::transform(pica.proctex.color_diff_table.begin(), pica.proctex.color_diff_table.end(),
new_data.begin(), [](const auto& entry) {
auto rgba = entry.ToVector() / 255.0f;
return Common::Vec4f{rgba.r(), rgba.g(), rgba.b(), rgba.a()};
});
if (new_data != proctex_diff_lut_data || invalidate) {
proctex_diff_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec4f));
fs_uniform_block_data.data.proctex_diff_lut_offset =
static_cast<GLint>((offset + bytes_used) / sizeof(Common::Vec4f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec4f);
if (pica.proctex.diff_lut_dirty) {
Common::Vec4f* new_data = reinterpret_cast<Common::Vec4f*>(buffer + bytes_used);
for (u32 i = 0; i < pica.proctex.color_diff_table.size(); i++) {
new_data[i] = pica.proctex.color_diff_table[i].ToVector() / 255.0f;
}
fs_uniform_block_data.proctex_diff_lut_dirty = false;
fs_data.proctex_diff_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f));
fs_data_dirty = true;
bytes_used += pica.proctex.color_diff_table.size() * sizeof(Common::Vec4f);
}
pica.proctex.table_dirty = 0;
texture_buffer.Unmap(bytes_used);
}
@ -1182,10 +996,8 @@ void RasterizerOpenGL::UploadUniforms(bool accelerate_draw) {
state.draw.uniform_buffer = uniform_buffer.GetHandle();
state.Apply();
const bool sync_vs_pica = accelerate_draw;
const bool sync_vs = vs_uniform_block_data.dirty;
const bool sync_fs = fs_uniform_block_data.dirty;
if (!sync_vs_pica && !sync_vs && !sync_fs) {
const bool sync_vs_pica = accelerate_draw && pica.vs_setup.uniforms_dirty;
if (!sync_vs_pica && !vs_data_dirty && !fs_data_dirty) {
return;
}
@ -1196,30 +1008,29 @@ void RasterizerOpenGL::UploadUniforms(bool accelerate_draw) {
const auto [uniforms, offset, invalidate] =
uniform_buffer.Map(uniform_size, uniform_buffer_alignment);
if (sync_vs || invalidate) {
std::memcpy(uniforms + used_bytes, &vs_uniform_block_data.data,
sizeof(vs_uniform_block_data.data));
if (vs_data_dirty || invalidate) {
std::memcpy(uniforms + used_bytes, &vs_data, sizeof(vs_data));
glBindBufferRange(GL_UNIFORM_BUFFER, UniformBindings::VSData, uniform_buffer.GetHandle(),
offset + used_bytes, sizeof(vs_uniform_block_data.data));
vs_uniform_block_data.dirty = false;
offset + used_bytes, sizeof(vs_data));
vs_data_dirty = false;
used_bytes += uniform_size_aligned_vs;
}
if (sync_fs || invalidate) {
std::memcpy(uniforms + used_bytes, &fs_uniform_block_data.data,
sizeof(fs_uniform_block_data.data));
if (fs_data_dirty || invalidate) {
std::memcpy(uniforms + used_bytes, &fs_data, sizeof(fs_data));
glBindBufferRange(GL_UNIFORM_BUFFER, UniformBindings::FSData, uniform_buffer.GetHandle(),
offset + used_bytes, sizeof(fs_uniform_block_data.data));
fs_uniform_block_data.dirty = false;
offset + used_bytes, sizeof(fs_data));
fs_data_dirty = false;
used_bytes += uniform_size_aligned_fs;
}
if (sync_vs_pica) {
if (sync_vs_pica || invalidate) {
VSPicaUniformData vs_uniforms;
vs_uniforms.uniforms.SetFromRegs(regs.vs, pica.vs_setup);
vs_uniforms.SetFromRegs(pica.vs_setup);
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));
pica.vs_setup.uniforms_dirty = false;
used_bytes += uniform_size_aligned_vs_pica;
}

View File

@ -58,41 +58,8 @@ public:
bool AccelerateDrawBatch(bool is_indexed) override;
private:
void SyncFixedState() 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
void SyncCullMode();
/// Syncs the blend enabled status to match the PICA register
void SyncBlendEnabled();
/// Syncs the blend functions to match the PICA register
void SyncBlendFuncs();
/// Syncs the blend color to match the PICA register
void SyncBlendColor();
/// Syncs the logic op states to match the PICA register
void SyncLogicOp();
/// Syncs the color write mask to match the PICA register state
void SyncColorWriteMask();
/// Syncs the stencil write mask to match the PICA register state
void SyncStencilWriteMask();
/// Syncs the depth write mask to match the PICA register state
void SyncDepthWriteMask();
/// Syncs the stencil test states to match the PICA register
void SyncStencilTest();
/// Syncs the depth test states to match the PICA register
void SyncDepthTest();
/// Syncs pipeline state from PICA registers
void SyncDrawState();
/// Syncs and uploads the lighting, fog and proctex LUTs
void SyncAndUploadLUTs();

View File

@ -901,8 +901,4 @@ void RendererOpenGL::CleanupVideoDumping() {
mailbox->free_cv.notify_one();
}
void RendererOpenGL::Sync() {
rasterizer.SyncEntireState();
}
} // namespace OpenGL

View File

@ -1,4 +1,4 @@
// Copyright 2022 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -53,7 +53,6 @@ public:
void TryPresent(int timeout_ms, bool is_secondary) override;
void PrepareVideoDumping() override;
void CleanupVideoDumping() override;
void Sync() override;
private:
void InitOpenGLObjects();

View File

@ -1,4 +1,4 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -35,7 +35,6 @@ public:
void SwapBuffers() override;
void TryPresent(int timeout_ms, bool is_secondary) override {}
void Sync() override {}
private:
void PrepareRenderTarget();

View File

@ -1,4 +1,4 @@
// Copyright 2015 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -27,7 +27,6 @@ public:
void AddTriangle(const Pica::OutputVertex& v0, const Pica::OutputVertex& v1,
const Pica::OutputVertex& v2) override;
void DrawTriangles() override {}
void NotifyPicaRegisterChanged(u32 id) override {}
void FlushAll() override {}
void FlushRegion(PAddr addr, u32 size) override {}
void InvalidateRegion(PAddr addr, u32 size) override {}

View File

@ -92,10 +92,6 @@ RendererVulkan::~RendererVulkan() {
}
}
void RendererVulkan::Sync() {
rasterizer.SyncEntireState();
}
void RendererVulkan::PrepareRendertarget() {
const auto& framebuffer_config = pica.regs.framebuffer_config;
const auto& regs_lcd = pica.regs_lcd;

View File

@ -1,4 +1,4 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
@ -80,7 +80,6 @@ public:
void SwapBuffers() override;
void TryPresent(int timeout_ms, bool is_secondary) override {}
void Sync() override;
private:
void ReloadPipeline();

View File

@ -100,8 +100,6 @@ bool GraphicsPipeline::TryBuild(bool wait_built) {
bool GraphicsPipeline::Build(bool fail_on_compile_required) {
MICROPROFILE_SCOPE(Vulkan_Pipeline);
const vk::Device device = instance.GetDevice();
std::array<vk::VertexInputBindingDescription, MAX_VERTEX_BINDINGS> bindings;
for (u32 i = 0; i < info.vertex_layout.binding_count; i++) {
const auto& binding = info.vertex_layout.bindings[i];
@ -273,7 +271,7 @@ bool GraphicsPipeline::Build(bool fail_on_compile_required) {
pipeline_info.flags |= vk::PipelineCreateFlagBits::eFailOnPipelineCompileRequiredEXT;
}
auto result = device.createGraphicsPipelineUnique(pipeline_cache, pipeline_info);
auto result = instance.GetDevice().createGraphicsPipelineUnique(pipeline_cache, pipeline_info);
if (result.result == vk::Result::eSuccess) {
pipeline = std::move(result.value);
} else if (result.result == vk::Result::eErrorPipelineCompileRequiredEXT) {

View File

@ -133,8 +133,6 @@ RasterizerVulkan::RasterizerVulkan(Memory::MemorySystem& memory, Pica::PicaCore&
update_queue.AddImageSampler(utility_set, 1, 0, null_surface.ImageView(),
null_sampler.Handle());
update_queue.Flush();
SyncEntireState();
}
RasterizerVulkan::~RasterizerVulkan() = default;
@ -148,17 +146,78 @@ void RasterizerVulkan::LoadDefaultDiskResources(
pipeline_cache.LoadDiskCache();
}
void RasterizerVulkan::SyncFixedState() {
SyncCullMode();
SyncBlendEnabled();
SyncBlendFuncs();
SyncBlendColor();
SyncLogicOp();
SyncStencilTest();
SyncDepthTest();
SyncColorWriteMask();
SyncStencilWriteMask();
SyncDepthWriteMask();
void RasterizerVulkan::SyncDrawState() {
SyncDrawUniforms();
// SyncCullMode();
pipeline_info.rasterization.cull_mode.Assign(regs.rasterizer.cull_mode);
// If the framebuffer is flipped, request to also flip vulkan viewport
const bool is_flipped = regs.framebuffer.framebuffer.IsFlipped();
pipeline_info.rasterization.flip_viewport.Assign(is_flipped);
// SyncBlendEnabled();
pipeline_info.blending.blend_enable = regs.framebuffer.output_merger.alphablend_enable;
// SyncBlendFuncs();
pipeline_info.blending.color_blend_eq.Assign(
regs.framebuffer.output_merger.alpha_blending.blend_equation_rgb);
pipeline_info.blending.alpha_blend_eq.Assign(
regs.framebuffer.output_merger.alpha_blending.blend_equation_a);
pipeline_info.blending.src_color_blend_factor.Assign(
regs.framebuffer.output_merger.alpha_blending.factor_source_rgb);
pipeline_info.blending.dst_color_blend_factor.Assign(
regs.framebuffer.output_merger.alpha_blending.factor_dest_rgb);
pipeline_info.blending.src_alpha_blend_factor.Assign(
regs.framebuffer.output_merger.alpha_blending.factor_source_a);
pipeline_info.blending.dst_alpha_blend_factor.Assign(
regs.framebuffer.output_merger.alpha_blending.factor_dest_a);
// SyncBlendColor();
pipeline_info.dynamic.blend_color = regs.framebuffer.output_merger.blend_const.raw;
// SyncLogicOp();
// SyncColorWriteMask();
pipeline_info.blending.logic_op = regs.framebuffer.output_merger.logic_op;
const bool is_logic_op_emulated =
instance.NeedsLogicOpEmulation() && !regs.framebuffer.output_merger.alphablend_enable;
const bool is_logic_op_noop =
regs.framebuffer.output_merger.logic_op == Pica::FramebufferRegs::LogicOp::NoOp;
if (is_logic_op_emulated && is_logic_op_noop) {
// Color output is disabled by logic operation. We use color write mask to skip
// color but allow depth write.
pipeline_info.blending.color_write_mask = 0;
} else {
const u32 color_mask = regs.framebuffer.framebuffer.allow_color_write != 0
? (regs.framebuffer.output_merger.depth_color_mask >> 8) & 0xF
: 0;
pipeline_info.blending.color_write_mask = color_mask;
}
// SyncStencilTest();
const auto& stencil_test = regs.framebuffer.output_merger.stencil_test;
const bool test_enable = stencil_test.enable && regs.framebuffer.framebuffer.depth_format ==
Pica::FramebufferRegs::DepthFormat::D24S8;
pipeline_info.depth_stencil.stencil_test_enable.Assign(test_enable);
pipeline_info.depth_stencil.stencil_fail_op.Assign(stencil_test.action_stencil_fail);
pipeline_info.depth_stencil.stencil_pass_op.Assign(stencil_test.action_depth_pass);
pipeline_info.depth_stencil.stencil_depth_fail_op.Assign(stencil_test.action_depth_fail);
pipeline_info.depth_stencil.stencil_compare_op.Assign(stencil_test.func);
pipeline_info.dynamic.stencil_reference = stencil_test.reference_value;
pipeline_info.dynamic.stencil_compare_mask = stencil_test.input_mask;
// SyncStencilWriteMask();
pipeline_info.dynamic.stencil_write_mask =
(regs.framebuffer.framebuffer.allow_depth_stencil_write != 0)
? static_cast<u32>(regs.framebuffer.output_merger.stencil_test.write_mask)
: 0;
// SyncDepthTest();
const bool test_enabled = regs.framebuffer.output_merger.depth_test_enable == 1 ||
regs.framebuffer.output_merger.depth_write_enable == 1;
const auto compare_op = regs.framebuffer.output_merger.depth_test_enable == 1
? regs.framebuffer.output_merger.depth_test_func.Value()
: Pica::FramebufferRegs::CompareFunc::Always;
pipeline_info.depth_stencil.depth_test_enable.Assign(test_enabled);
pipeline_info.depth_stencil.depth_compare_op.Assign(compare_op);
// SyncDepthWriteMask();
const bool write_enable = (regs.framebuffer.framebuffer.allow_depth_stencil_write != 0 &&
regs.framebuffer.output_merger.depth_write_enable);
pipeline_info.depth_stencil.depth_write_enable.Assign(write_enable);
}
void RasterizerVulkan::SetupVertexArray() {
@ -463,6 +522,7 @@ void RasterizerVulkan::DrawTriangles() {
bool RasterizerVulkan::Draw(bool accelerate, bool is_indexed) {
MICROPROFILE_SCOPE(Vulkan_Drawing);
SyncDrawState();
const bool shadow_rendering = regs.framebuffer.IsShadowRendering();
const bool has_stencil = regs.framebuffer.HasStencil();
@ -487,16 +547,14 @@ bool RasterizerVulkan::Draw(bool accelerate, bool is_indexed) {
// Update scissor uniforms
const auto [scissor_x1, scissor_y2, scissor_x2, scissor_y1] = fb_helper.Scissor();
if (fs_uniform_block_data.data.scissor_x1 != scissor_x1 ||
fs_uniform_block_data.data.scissor_x2 != scissor_x2 ||
fs_uniform_block_data.data.scissor_y1 != scissor_y1 ||
fs_uniform_block_data.data.scissor_y2 != scissor_y2) {
if (fs_data.scissor_x1 != scissor_x1 || fs_data.scissor_x2 != scissor_x2 ||
fs_data.scissor_y1 != scissor_y1 || fs_data.scissor_y2 != scissor_y2) {
fs_uniform_block_data.data.scissor_x1 = scissor_x1;
fs_uniform_block_data.data.scissor_x2 = scissor_x2;
fs_uniform_block_data.data.scissor_y1 = scissor_y1;
fs_uniform_block_data.data.scissor_y2 = scissor_y2;
fs_uniform_block_data.dirty = true;
fs_data.scissor_x1 = scissor_x1;
fs_data.scissor_x2 = scissor_x2;
fs_data.scissor_y1 = scissor_y1;
fs_data.scissor_y2 = scissor_y2;
fs_data_dirty = true;
}
// Sync and bind the texture surfaces
@ -504,16 +562,7 @@ bool RasterizerVulkan::Draw(bool accelerate, bool is_indexed) {
SyncUtilityTextures(framebuffer);
// Sync and bind the shader
if (shader_dirty) {
pipeline_cache.UseFragmentShader(regs, user_config);
shader_dirty = false;
}
// If the framebuffer is flipped, request to also flip vulkan viewport
const bool is_flipped = regs.framebuffer.framebuffer.IsFlipped();
vs_uniform_block_data.dirty |= (vs_uniform_block_data.data.flip_viewport != 0) != is_flipped;
vs_uniform_block_data.data.flip_viewport = is_flipped;
pipeline_info.rasterization.flip_viewport.Assign(is_flipped);
pipeline_cache.UseFragmentShader(regs, user_config);
// Sync the LUTs within the texture buffer
SyncAndUploadLUTs();
@ -667,68 +716,6 @@ void RasterizerVulkan::BindTextureCube(const Pica::TexturingRegs::FullTextureCon
update_queue.AddImageSampler(texture_set, 0, 0, surface.ImageView(), sampler.Handle());
}
void RasterizerVulkan::NotifyFixedFunctionPicaRegisterChanged(u32 id) {
switch (id) {
// Culling
case PICA_REG_INDEX(rasterizer.cull_mode):
SyncCullMode();
break;
// Blending
case PICA_REG_INDEX(framebuffer.output_merger.alphablend_enable):
SyncBlendEnabled();
// Update since logic op emulation depends on alpha blend enable.
SyncLogicOp();
SyncColorWriteMask();
break;
case PICA_REG_INDEX(framebuffer.output_merger.alpha_blending):
SyncBlendFuncs();
break;
case PICA_REG_INDEX(framebuffer.output_merger.blend_const):
SyncBlendColor();
break;
// Sync VK stencil test + stencil write mask
// (Pica stencil test function register also contains a stencil write mask)
case PICA_REG_INDEX(framebuffer.output_merger.stencil_test.raw_func):
SyncStencilTest();
SyncStencilWriteMask();
break;
case PICA_REG_INDEX(framebuffer.output_merger.stencil_test.raw_op):
case PICA_REG_INDEX(framebuffer.framebuffer.depth_format):
SyncStencilTest();
break;
// Sync VK depth test + depth and color write mask
// (Pica depth test function register also contains a depth and color write mask)
case PICA_REG_INDEX(framebuffer.output_merger.depth_test_enable):
SyncDepthTest();
SyncDepthWriteMask();
SyncColorWriteMask();
break;
// Sync VK depth and stencil write mask
// (This is a dedicated combined depth / stencil write-enable register)
case PICA_REG_INDEX(framebuffer.framebuffer.allow_depth_stencil_write):
SyncDepthWriteMask();
SyncStencilWriteMask();
break;
// Sync VK color write mask
// (This is a dedicated color write-enable register)
case PICA_REG_INDEX(framebuffer.framebuffer.allow_color_write):
SyncColorWriteMask();
break;
// Logic op
case PICA_REG_INDEX(framebuffer.output_merger.logic_op):
SyncLogicOp();
// Update since color write mask is used to emulate no-op.
SyncColorWriteMask();
break;
}
}
void RasterizerVulkan::FlushAll() {
res_cache.FlushAll();
}
@ -825,164 +812,49 @@ void RasterizerVulkan::MakeSoftwareVertexLayout() {
}
}
void RasterizerVulkan::SyncCullMode() {
pipeline_info.rasterization.cull_mode.Assign(regs.rasterizer.cull_mode);
}
void RasterizerVulkan::SyncBlendEnabled() {
pipeline_info.blending.blend_enable = regs.framebuffer.output_merger.alphablend_enable;
}
void RasterizerVulkan::SyncBlendFuncs() {
pipeline_info.blending.color_blend_eq.Assign(
regs.framebuffer.output_merger.alpha_blending.blend_equation_rgb);
pipeline_info.blending.alpha_blend_eq.Assign(
regs.framebuffer.output_merger.alpha_blending.blend_equation_a);
pipeline_info.blending.src_color_blend_factor.Assign(
regs.framebuffer.output_merger.alpha_blending.factor_source_rgb);
pipeline_info.blending.dst_color_blend_factor.Assign(
regs.framebuffer.output_merger.alpha_blending.factor_dest_rgb);
pipeline_info.blending.src_alpha_blend_factor.Assign(
regs.framebuffer.output_merger.alpha_blending.factor_source_a);
pipeline_info.blending.dst_alpha_blend_factor.Assign(
regs.framebuffer.output_merger.alpha_blending.factor_dest_a);
}
void RasterizerVulkan::SyncBlendColor() {
pipeline_info.dynamic.blend_color = regs.framebuffer.output_merger.blend_const.raw;
}
void RasterizerVulkan::SyncLogicOp() {
if (instance.NeedsLogicOpEmulation()) {
// We need this in the fragment shader to emulate logic operations
shader_dirty = true;
}
pipeline_info.blending.logic_op = regs.framebuffer.output_merger.logic_op;
const bool is_logic_op_emulated =
instance.NeedsLogicOpEmulation() && !regs.framebuffer.output_merger.alphablend_enable;
const bool is_logic_op_noop =
regs.framebuffer.output_merger.logic_op == Pica::FramebufferRegs::LogicOp::NoOp;
if (is_logic_op_emulated && is_logic_op_noop) {
// Color output is disabled by logic operation. We use color write mask to skip
// color but allow depth write.
pipeline_info.blending.color_write_mask = 0;
}
}
void RasterizerVulkan::SyncColorWriteMask() {
const u32 color_mask = regs.framebuffer.framebuffer.allow_color_write != 0
? (regs.framebuffer.output_merger.depth_color_mask >> 8) & 0xF
: 0;
const bool is_logic_op_emulated =
instance.NeedsLogicOpEmulation() && !regs.framebuffer.output_merger.alphablend_enable;
const bool is_logic_op_noop =
regs.framebuffer.output_merger.logic_op == Pica::FramebufferRegs::LogicOp::NoOp;
if (is_logic_op_emulated && is_logic_op_noop) {
// Color output is disabled by logic operation. We use color write mask to skip
// color but allow depth write. Return early to avoid overwriting this.
return;
}
pipeline_info.blending.color_write_mask = color_mask;
}
void RasterizerVulkan::SyncStencilWriteMask() {
pipeline_info.dynamic.stencil_write_mask =
(regs.framebuffer.framebuffer.allow_depth_stencil_write != 0)
? static_cast<u32>(regs.framebuffer.output_merger.stencil_test.write_mask)
: 0;
}
void RasterizerVulkan::SyncDepthWriteMask() {
const bool write_enable = (regs.framebuffer.framebuffer.allow_depth_stencil_write != 0 &&
regs.framebuffer.output_merger.depth_write_enable);
pipeline_info.depth_stencil.depth_write_enable.Assign(write_enable);
}
void RasterizerVulkan::SyncStencilTest() {
const auto& stencil_test = regs.framebuffer.output_merger.stencil_test;
const bool test_enable = stencil_test.enable && regs.framebuffer.framebuffer.depth_format ==
Pica::FramebufferRegs::DepthFormat::D24S8;
pipeline_info.depth_stencil.stencil_test_enable.Assign(test_enable);
pipeline_info.depth_stencil.stencil_fail_op.Assign(stencil_test.action_stencil_fail);
pipeline_info.depth_stencil.stencil_pass_op.Assign(stencil_test.action_depth_pass);
pipeline_info.depth_stencil.stencil_depth_fail_op.Assign(stencil_test.action_depth_fail);
pipeline_info.depth_stencil.stencil_compare_op.Assign(stencil_test.func);
pipeline_info.dynamic.stencil_reference = stencil_test.reference_value;
pipeline_info.dynamic.stencil_compare_mask = stencil_test.input_mask;
}
void RasterizerVulkan::SyncDepthTest() {
const bool test_enabled = regs.framebuffer.output_merger.depth_test_enable == 1 ||
regs.framebuffer.output_merger.depth_write_enable == 1;
const auto compare_op = regs.framebuffer.output_merger.depth_test_enable == 1
? regs.framebuffer.output_merger.depth_test_func.Value()
: Pica::FramebufferRegs::CompareFunc::Always;
pipeline_info.depth_stencil.depth_test_enable.Assign(test_enabled);
pipeline_info.depth_stencil.depth_compare_op.Assign(compare_op);
}
void RasterizerVulkan::SyncAndUploadLUTsLF() {
constexpr std::size_t max_size =
sizeof(Common::Vec2f) * 256 * Pica::LightingRegs::NumLightingSampler +
sizeof(Common::Vec2f) * 128; // fog
if (!fs_uniform_block_data.lighting_lut_dirty_any && !fs_uniform_block_data.fog_lut_dirty) {
if (!pica.lighting.lut_dirty && !pica.fog.lut_dirty) {
return;
}
std::size_t bytes_used = 0;
auto [buffer, offset, invalidate] = texture_lf_buffer.Map(max_size, sizeof(Common::Vec4f));
// Sync the lighting luts
if (fs_uniform_block_data.lighting_lut_dirty_any || invalidate) {
for (unsigned index = 0; index < fs_uniform_block_data.lighting_lut_dirty.size(); index++) {
if (fs_uniform_block_data.lighting_lut_dirty[index] || invalidate) {
std::array<Common::Vec2f, 256> new_data;
const auto& source_lut = pica.lighting.luts[index];
std::transform(source_lut.begin(), source_lut.end(), new_data.begin(),
[](const auto& entry) {
return Common::Vec2f{entry.ToFloat(), entry.DiffToFloat()};
});
if (invalidate) {
pica.lighting.lut_dirty = pica.lighting.LutAllDirty;
pica.fog.lut_dirty = true;
}
if (new_data != lighting_lut_data[index] || invalidate) {
lighting_lut_data[index] = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f));
fs_uniform_block_data.data.lighting_lut_offset[index / 4][index % 4] =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f);
}
fs_uniform_block_data.lighting_lut_dirty[index] = false;
}
// Sync the lighting luts
while (pica.lighting.lut_dirty) {
u32 index = std::countr_zero(pica.lighting.lut_dirty);
pica.lighting.lut_dirty &= ~(1 << index);
Common::Vec2f* new_data = reinterpret_cast<Common::Vec2f*>(buffer + bytes_used);
const auto& source_lut = pica.lighting.luts[index];
for (u32 i = 0; i < source_lut.size(); i++) {
new_data[i] = {source_lut[i].ToFloat(), source_lut[i].DiffToFloat()};
}
fs_uniform_block_data.lighting_lut_dirty_any = false;
fs_data.lighting_lut_offset[index / 4][index % 4] =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_data_dirty = true;
bytes_used += source_lut.size() * sizeof(Common::Vec2f);
}
// Sync the fog lut
if (fs_uniform_block_data.fog_lut_dirty || invalidate) {
std::array<Common::Vec2f, 128> new_data;
std::transform(
pica.fog.lut.begin(), pica.fog.lut.end(), new_data.begin(),
[](const auto& entry) { return Common::Vec2f{entry.ToFloat(), entry.DiffToFloat()}; });
if (new_data != fog_lut_data || invalidate) {
fog_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f));
fs_uniform_block_data.data.fog_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f);
if (pica.fog.lut_dirty) {
Common::Vec2f* new_data = reinterpret_cast<Common::Vec2f*>(buffer + bytes_used);
for (u32 i = 0; i < pica.fog.lut.size(); i++) {
new_data[i] = {pica.fog.lut[i].ToFloat(), pica.fog.lut[i].DiffToFloat()};
}
fs_uniform_block_data.fog_lut_dirty = false;
fs_data.fog_lut_offset = static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_data_dirty = true;
bytes_used += pica.fog.lut.size() * sizeof(Common::Vec2f);
pica.fog.lut_dirty = false;
}
texture_lf_buffer.Commit(static_cast<u32>(bytes_used));
@ -995,109 +867,76 @@ void RasterizerVulkan::SyncAndUploadLUTs() {
sizeof(Common::Vec4f) * 256 + // proctex
sizeof(Common::Vec4f) * 256; // proctex diff
if (!fs_uniform_block_data.proctex_noise_lut_dirty &&
!fs_uniform_block_data.proctex_color_map_dirty &&
!fs_uniform_block_data.proctex_alpha_map_dirty &&
!fs_uniform_block_data.proctex_lut_dirty && !fs_uniform_block_data.proctex_diff_lut_dirty) {
if (!pica.proctex.lut_dirty) {
return;
}
std::size_t bytes_used = 0;
auto [buffer, offset, invalidate] = texture_buffer.Map(max_size, sizeof(Common::Vec4f));
// helper function for SyncProcTexNoiseLUT/ColorMap/AlphaMap
auto sync_proctex_value_lut =
[this, buffer = buffer, offset = offset, invalidate = invalidate,
&bytes_used](const std::array<Pica::PicaCore::ProcTex::ValueEntry, 128>& lut,
std::array<Common::Vec2f, 128>& lut_data, int& lut_offset) {
std::array<Common::Vec2f, 128> new_data;
std::transform(lut.begin(), lut.end(), new_data.begin(), [](const auto& entry) {
return Common::Vec2f{entry.ToFloat(), entry.DiffToFloat()};
});
if (invalidate) {
pica.proctex.table_dirty = pica.proctex.TableAllDirty;
}
if (new_data != lut_data || invalidate) {
lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec2f));
lut_offset = static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec2f);
// helper function for SyncProcTexNoiseLUT/ColorMap/AlphaMap
const auto sync_proctex_value_lut =
[&](const std::array<Pica::PicaCore::ProcTex::ValueEntry, 128>& lut, int& lut_offset) {
Common::Vec2f* new_data = reinterpret_cast<Common::Vec2f*>(buffer + bytes_used);
for (u32 i = 0; i < lut.size(); i++) {
new_data[i] = {lut[i].ToFloat(), lut[i].DiffToFloat()};
}
lut_offset = static_cast<int>((offset + bytes_used) / sizeof(Common::Vec2f));
fs_data_dirty = true;
bytes_used += lut.size() * sizeof(Common::Vec2f);
};
// Sync the proctex noise lut
if (fs_uniform_block_data.proctex_noise_lut_dirty || invalidate) {
sync_proctex_value_lut(proctex.noise_table, proctex_noise_lut_data,
fs_uniform_block_data.data.proctex_noise_lut_offset);
fs_uniform_block_data.proctex_noise_lut_dirty = false;
if (pica.proctex.noise_lut_dirty) {
sync_proctex_value_lut(proctex.noise_table, fs_data.proctex_noise_lut_offset);
}
// Sync the proctex color map
if (fs_uniform_block_data.proctex_color_map_dirty || invalidate) {
sync_proctex_value_lut(proctex.color_map_table, proctex_color_map_data,
fs_uniform_block_data.data.proctex_color_map_offset);
fs_uniform_block_data.proctex_color_map_dirty = false;
if (pica.proctex.color_map_dirty) {
sync_proctex_value_lut(proctex.color_map_table, fs_data.proctex_color_map_offset);
}
// Sync the proctex alpha map
if (fs_uniform_block_data.proctex_alpha_map_dirty || invalidate) {
sync_proctex_value_lut(proctex.alpha_map_table, proctex_alpha_map_data,
fs_uniform_block_data.data.proctex_alpha_map_offset);
fs_uniform_block_data.proctex_alpha_map_dirty = false;
if (pica.proctex.alpha_map_dirty) {
sync_proctex_value_lut(proctex.alpha_map_table, fs_data.proctex_alpha_map_offset);
}
// Sync the proctex lut
if (fs_uniform_block_data.proctex_lut_dirty || invalidate) {
std::array<Common::Vec4f, 256> new_data;
std::transform(proctex.color_table.begin(), proctex.color_table.end(), new_data.begin(),
[](const auto& entry) {
auto rgba = entry.ToVector() / 255.0f;
return Common::Vec4f{rgba.r(), rgba.g(), rgba.b(), rgba.a()};
});
if (new_data != proctex_lut_data || invalidate) {
proctex_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec4f));
fs_uniform_block_data.data.proctex_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec4f);
if (pica.proctex.lut_dirty) {
Common::Vec4f* new_data = reinterpret_cast<Common::Vec4f*>(buffer + bytes_used);
for (u32 i = 0; i < proctex.color_table.size(); i++) {
new_data[i] = proctex.color_table[i].ToVector() / 255.0f;
}
fs_uniform_block_data.proctex_lut_dirty = false;
fs_data.proctex_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f));
fs_data_dirty = true;
bytes_used += proctex.color_table.size() * sizeof(Common::Vec4f);
}
// Sync the proctex difference lut
if (fs_uniform_block_data.proctex_diff_lut_dirty || invalidate) {
std::array<Common::Vec4f, 256> new_data;
std::transform(proctex.color_diff_table.begin(), proctex.color_diff_table.end(),
new_data.begin(), [](const auto& entry) {
auto rgba = entry.ToVector() / 255.0f;
return Common::Vec4f{rgba.r(), rgba.g(), rgba.b(), rgba.a()};
});
if (new_data != proctex_diff_lut_data || invalidate) {
proctex_diff_lut_data = new_data;
std::memcpy(buffer + bytes_used, new_data.data(),
new_data.size() * sizeof(Common::Vec4f));
fs_uniform_block_data.data.proctex_diff_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f));
fs_uniform_block_data.dirty = true;
bytes_used += new_data.size() * sizeof(Common::Vec4f);
if (pica.proctex.diff_lut_dirty) {
Common::Vec4f* new_data = reinterpret_cast<Common::Vec4f*>(buffer + bytes_used);
for (u32 i = 0; i < proctex.color_diff_table.size(); i++) {
new_data[i] = proctex.color_diff_table[i].ToVector() / 255.0f;
}
fs_uniform_block_data.proctex_diff_lut_dirty = false;
fs_data.proctex_diff_lut_offset =
static_cast<int>((offset + bytes_used) / sizeof(Common::Vec4f));
fs_data_dirty = true;
bytes_used += proctex.color_diff_table.size() * sizeof(Common::Vec4f);
}
pica.proctex.table_dirty = 0;
texture_buffer.Commit(static_cast<u32>(bytes_used));
}
void RasterizerVulkan::UploadUniforms(bool accelerate_draw) {
const bool sync_vs_pica = accelerate_draw;
const bool sync_vs = vs_uniform_block_data.dirty;
const bool sync_fs = fs_uniform_block_data.dirty;
if (!sync_vs_pica && !sync_vs && !sync_fs) {
const bool sync_vs_pica = accelerate_draw && pica.vs_setup.uniforms_dirty;
if (!sync_vs_pica && !vs_data_dirty && !fs_data_dirty) {
return;
}
@ -1108,30 +947,26 @@ void RasterizerVulkan::UploadUniforms(bool accelerate_draw) {
u32 used_bytes = 0;
if (sync_vs || invalidate) {
std::memcpy(uniforms + used_bytes, &vs_uniform_block_data.data,
sizeof(vs_uniform_block_data.data));
if (vs_data_dirty || invalidate) {
std::memcpy(uniforms + used_bytes, &vs_data, sizeof(vs_data));
pipeline_cache.UpdateRange(1, offset + used_bytes);
vs_uniform_block_data.dirty = false;
vs_data_dirty = false;
used_bytes += uniform_size_aligned_vs;
}
if (sync_fs || invalidate) {
std::memcpy(uniforms + used_bytes, &fs_uniform_block_data.data,
sizeof(fs_uniform_block_data.data));
if (fs_data_dirty || invalidate) {
std::memcpy(uniforms + used_bytes, &fs_data, sizeof(fs_data));
pipeline_cache.UpdateRange(2, offset + used_bytes);
fs_uniform_block_data.dirty = false;
fs_data_dirty = false;
used_bytes += uniform_size_aligned_fs;
}
if (sync_vs_pica) {
if (sync_vs_pica || invalidate) {
VSPicaUniformData vs_uniforms;
vs_uniforms.uniforms.SetFromRegs(regs.vs, pica.vs_setup);
vs_uniforms.SetFromRegs(pica.vs_setup);
std::memcpy(uniforms + used_bytes, &vs_uniforms, sizeof(vs_uniforms));
pipeline_cache.UpdateRange(0, offset + used_bytes);
pica.vs_setup.uniforms_dirty = false;
used_bytes += uniform_size_aligned_vs_pica;
}

View File

@ -61,40 +61,9 @@ public:
u32 pixel_stride, ScreenInfo& screen_info);
bool AccelerateDrawBatch(bool is_indexed) override;
void SyncFixedState() override;
private:
void NotifyFixedFunctionPicaRegisterChanged(u32 id) override;
/// Syncs the cull mode to match the PICA register
void SyncCullMode();
/// Syncs the blend enabled status to match the PICA register
void SyncBlendEnabled();
/// Syncs the blend functions to match the PICA register
void SyncBlendFuncs();
/// Syncs the blend color to match the PICA register
void SyncBlendColor();
/// Syncs the logic op states to match the PICA register
void SyncLogicOp();
/// Syncs the color write mask to match the PICA register state
void SyncColorWriteMask();
/// Syncs the stencil write mask to match the PICA register state
void SyncStencilWriteMask();
/// Syncs the depth write mask to match the PICA register state
void SyncDepthWriteMask();
/// Syncs the stencil test states to match the PICA register
void SyncStencilTest();
/// Syncs the depth test states to match the PICA register
void SyncDepthTest();
/// Syncs pipeline state from PICA registers
void SyncDrawState();
/// Syncs and uploads the lighting, fog and proctex LUTs
void SyncAndUploadLUTs();

View File

@ -1,8 +1,7 @@
// Copyright 2024 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <limits>
#include "common/assert.h"
#include "video_core/rasterizer_cache/pixel_format.h"
#include "video_core/renderer_vulkan/vk_instance.h"

View File

@ -1,8 +1,7 @@
// Copyright 2017 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <exception>
#include <map>
#include <set>
#include <string>
@ -359,8 +358,8 @@ private:
}
/// Generates code representing a bool uniform
std::string GetUniformBool(u32 index) const {
return fmt::format("uniforms.b[{}]", index);
std::string GetUniformBool(u32 index, bool invert_test = false) const {
return fmt::format("(uniforms.b & {}u) {} 0u", 1 << index, invert_test ? "==" : "!=");
}
/**
@ -673,9 +672,8 @@ private:
if (instr.opcode.Value() == OpCode::Id::JMPC) {
condition = EvaluateCondition(instr.flow_control);
} else {
bool invert_test = instr.flow_control.num_instructions & 1;
condition = (invert_test ? "!" : "") +
GetUniformBool(instr.flow_control.bool_uniform_id);
const bool invert_test = instr.flow_control.num_instructions & 1;
condition = GetUniformBool(instr.flow_control.bool_uniform_id, invert_test);
}
shader.AddLine("if ({}) {{", condition);

View File

@ -15,19 +15,15 @@ using VSOutputAttributes = Pica::RasterizerRegs::VSOutputAttributes;
namespace Pica::Shader::Generator::GLSL {
constexpr std::string_view VSPicaUniformBlockDef = R"(
struct pica_uniforms {
bool b[16];
uvec4 i[4];
vec4 f[96];
};
#ifdef VULKAN
layout (set = 0, binding = 0, std140) uniform vs_pica_data {
#else
layout (binding = 0, std140) uniform vs_pica_data {
#endif
pica_uniforms uniforms;
};
uint b;
uvec4 i[4];
vec4 f[96];
} uniforms;
)";
constexpr std::string_view VSUniformBlockDef = R"(

View File

@ -1,26 +1,26 @@
// Copyright 2023 Citra Emulator Project
// Copyright Citra Emulator Project / Azahar Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <algorithm>
#include "video_core/pica/regs_shader.h"
#include "video_core/pica/shader_setup.h"
#include "video_core/shader/generator/shader_uniforms.h"
namespace Pica::Shader::Generator {
void PicaUniformsData::SetFromRegs(const Pica::ShaderRegs& regs, const Pica::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()};
});
void VSPicaUniformData::SetFromRegs(const Pica::ShaderSetup& setup) {
b = 0;
for (u32 j = 0; j < setup.uniforms.b.size(); j++) {
b |= setup.uniforms.b[j] << j;
}
for (u32 j = 0; j < setup.uniforms.i.size(); j++) {
const auto& value = setup.uniforms.i[j];
i[j] = Common::MakeVec<u32>(value.x, value.y, value.z, value.w);
}
for (u32 j = 0; j < setup.uniforms.f.size(); j++) {
const auto& value = setup.uniforms.f[j];
f[j] = Common::MakeVec<f32>(value.x.ToFloat32(), value.y.ToFloat32(), value.z.ToFloat32(),
value.w.ToFloat32());
}
}
} // namespace Pica::Shader::Generator

View File

@ -8,7 +8,6 @@
#include "video_core/pica/regs_lighting.h"
namespace Pica {
struct ShaderRegs;
struct ShaderSetup;
} // namespace Pica
@ -69,22 +68,6 @@ static_assert(sizeof(FSUniformData) == 0x530,
static_assert(sizeof(FSUniformData) < 16384,
"UniformData structure must be less than 16kb as per the OpenGL spec");
/**
* Uniform struct for the Uniform Buffer Object that contains PICA vertex/geometry shader uniforms.
* NOTE: the same rule from UniformData also applies here.
*/
struct PicaUniformsData {
void SetFromRegs(const ShaderRegs& regs, const ShaderSetup& setup);
struct BoolAligned {
alignas(16) int b;
};
std::array<BoolAligned, 16> bools;
alignas(16) std::array<Common::Vec4u, 4> i;
alignas(16) std::array<Common::Vec4f, 96> f;
};
struct VSUniformData {
u32 enable_clip1;
u32 flip_viewport;
@ -95,10 +78,18 @@ static_assert(sizeof(VSUniformData) == 32,
static_assert(sizeof(VSUniformData) < 16384,
"VSUniformData structure must be less than 16kb as per the OpenGL spec");
/**
* Uniform struct for the Uniform Buffer Object that contains PICA vertex/geometry shader uniforms.
* NOTE: the same rule from UniformData also applies here.
*/
struct VSPicaUniformData {
alignas(16) PicaUniformsData uniforms;
void SetFromRegs(const ShaderSetup& setup);
u32 b;
alignas(16) std::array<Common::Vec4u, 4> i;
alignas(16) std::array<Common::Vec4f, 96> f;
};
static_assert(sizeof(VSPicaUniformData) == 1856,
static_assert(sizeof(VSPicaUniformData) == 1616,
"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");