From 025b20f96ae588777e3ff11083cc4184bf418af6 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Mon, 26 Apr 2021 03:53:26 -0300 Subject: shader: Move pipeline cache logic to separate files Move code to separate files to be able to reuse it from OpenGL. This greatly simplifies the pipeline cache logic on Vulkan. Transform feedback state is not yet abstracted and it's still intrusively stored inside vk_pipeline_cache. It will be moved when needed on OpenGL. --- src/video_core/shader_environment.cpp | 453 ++++++++++++++++++++++++++++++++++ 1 file changed, 453 insertions(+) create mode 100644 src/video_core/shader_environment.cpp (limited to 'src/video_core/shader_environment.cpp') diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp new file mode 100644 index 000000000..5dccc0097 --- /dev/null +++ b/src/video_core/shader_environment.cpp @@ -0,0 +1,453 @@ +// Copyright 2021 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include +#include +#include +#include +#include + +#include "common/assert.h" +#include "common/cityhash.h" +#include "common/common_types.h" +#include "common/div_ceil.h" +#include "common/fs/fs.h" +#include "common/logging/log.h" +#include "shader_recompiler/environment.h" +#include "video_core/memory_manager.h" +#include "video_core/shader_environment.h" +#include "video_core/textures/texture.h" + +namespace VideoCommon { + +constexpr std::array MAGIC_NUMBER{'y', 'u', 'z', 'u', 'c', 'a', 'c', 'h'}; +constexpr u32 CACHE_VERSION = 3; + +constexpr size_t INST_SIZE = sizeof(u64); + +using Maxwell = Tegra::Engines::Maxwell3D::Regs; + +static u64 MakeCbufKey(u32 index, u32 offset) { + return (static_cast(index) << 32) | offset; +} + +static Shader::TextureType ConvertType(const Tegra::Texture::TICEntry& entry) { + switch (entry.texture_type) { + case Tegra::Texture::TextureType::Texture1D: + return Shader::TextureType::Color1D; + case Tegra::Texture::TextureType::Texture2D: + case Tegra::Texture::TextureType::Texture2DNoMipmap: + return Shader::TextureType::Color2D; + case Tegra::Texture::TextureType::Texture3D: + return Shader::TextureType::Color3D; + case Tegra::Texture::TextureType::TextureCubemap: + return Shader::TextureType::ColorCube; + case Tegra::Texture::TextureType::Texture1DArray: + return Shader::TextureType::ColorArray1D; + case Tegra::Texture::TextureType::Texture2DArray: + return Shader::TextureType::ColorArray2D; + case Tegra::Texture::TextureType::Texture1DBuffer: + return Shader::TextureType::Buffer; + case Tegra::Texture::TextureType::TextureCubeArray: + return Shader::TextureType::ColorArrayCube; + default: + throw Shader::NotImplementedException("Unknown texture type"); + } +} + +GenericEnvironment::GenericEnvironment(Tegra::MemoryManager& gpu_memory_, GPUVAddr program_base_, + u32 start_address_) + : gpu_memory{&gpu_memory_}, program_base{program_base_} { + start_address = start_address_; +} + +GenericEnvironment::~GenericEnvironment() = default; + +u32 GenericEnvironment::TextureBoundBuffer() const { + return texture_bound; +} + +u32 GenericEnvironment::LocalMemorySize() const { + return local_memory_size; +} + +u32 GenericEnvironment::SharedMemorySize() const { + return shared_memory_size; +} + +std::array GenericEnvironment::WorkgroupSize() const { + return workgroup_size; +} + +u64 GenericEnvironment::ReadInstruction(u32 address) { + read_lowest = std::min(read_lowest, address); + read_highest = std::max(read_highest, address); + + if (address >= cached_lowest && address < cached_highest) { + return code[(address - cached_lowest) / INST_SIZE]; + } + has_unbound_instructions = true; + return gpu_memory->Read(program_base + address); +} + +std::optional GenericEnvironment::Analyze() { + const std::optional size{TryFindSize()}; + if (!size) { + return std::nullopt; + } + cached_lowest = start_address; + cached_highest = start_address + static_cast(*size); + return Common::CityHash64(reinterpret_cast(code.data()), *size); +} + +void GenericEnvironment::SetCachedSize(size_t size_bytes) { + cached_lowest = start_address; + cached_highest = start_address + static_cast(size_bytes); + code.resize(CachedSize()); + gpu_memory->ReadBlock(program_base + cached_lowest, code.data(), code.size() * sizeof(u64)); +} + +size_t GenericEnvironment::CachedSize() const noexcept { + return cached_highest - cached_lowest + INST_SIZE; +} + +size_t GenericEnvironment::ReadSize() const noexcept { + return read_highest - read_lowest + INST_SIZE; +} + +bool GenericEnvironment::CanBeSerialized() const noexcept { + return !has_unbound_instructions; +} + +u64 GenericEnvironment::CalculateHash() const { + const size_t size{ReadSize()}; + const auto data{std::make_unique(size)}; + gpu_memory->ReadBlock(program_base + read_lowest, data.get(), size); + return Common::CityHash64(data.get(), size); +} + +void GenericEnvironment::Serialize(std::ofstream& file) const { + const u64 code_size{static_cast(CachedSize())}; + const u64 num_texture_types{static_cast(texture_types.size())}; + const u64 num_cbuf_values{static_cast(cbuf_values.size())}; + + file.write(reinterpret_cast(&code_size), sizeof(code_size)) + .write(reinterpret_cast(&num_texture_types), sizeof(num_texture_types)) + .write(reinterpret_cast(&num_cbuf_values), sizeof(num_cbuf_values)) + .write(reinterpret_cast(&local_memory_size), sizeof(local_memory_size)) + .write(reinterpret_cast(&texture_bound), sizeof(texture_bound)) + .write(reinterpret_cast(&start_address), sizeof(start_address)) + .write(reinterpret_cast(&cached_lowest), sizeof(cached_lowest)) + .write(reinterpret_cast(&cached_highest), sizeof(cached_highest)) + .write(reinterpret_cast(&stage), sizeof(stage)) + .write(reinterpret_cast(code.data()), code_size); + for (const auto [key, type] : texture_types) { + file.write(reinterpret_cast(&key), sizeof(key)) + .write(reinterpret_cast(&type), sizeof(type)); + } + for (const auto [key, type] : cbuf_values) { + file.write(reinterpret_cast(&key), sizeof(key)) + .write(reinterpret_cast(&type), sizeof(type)); + } + if (stage == Shader::Stage::Compute) { + file.write(reinterpret_cast(&workgroup_size), sizeof(workgroup_size)) + .write(reinterpret_cast(&shared_memory_size), sizeof(shared_memory_size)); + } else { + file.write(reinterpret_cast(&sph), sizeof(sph)); + } +} + +std::optional GenericEnvironment::TryFindSize() { + static constexpr size_t BLOCK_SIZE = 0x1000; + static constexpr size_t MAXIMUM_SIZE = 0x100000; + + static constexpr u64 SELF_BRANCH_A = 0xE2400FFFFF87000FULL; + static constexpr u64 SELF_BRANCH_B = 0xE2400FFFFF07000FULL; + + GPUVAddr guest_addr{program_base + start_address}; + size_t offset{0}; + size_t size{BLOCK_SIZE}; + while (size <= MAXIMUM_SIZE) { + code.resize(size / INST_SIZE); + u64* const data = code.data() + offset / INST_SIZE; + gpu_memory->ReadBlock(guest_addr, data, BLOCK_SIZE); + for (size_t index = 0; index < BLOCK_SIZE; index += INST_SIZE) { + const u64 inst = data[index / INST_SIZE]; + if (inst == SELF_BRANCH_A || inst == SELF_BRANCH_B) { + return offset + index; + } + } + guest_addr += BLOCK_SIZE; + size += BLOCK_SIZE; + offset += BLOCK_SIZE; + } + return std::nullopt; +} + +Shader::TextureType GenericEnvironment::ReadTextureTypeImpl(GPUVAddr tic_addr, u32 tic_limit, + bool via_header_index, u32 raw) { + const TextureHandle handle{raw, via_header_index}; + const GPUVAddr descriptor_addr{tic_addr + handle.image * sizeof(Tegra::Texture::TICEntry)}; + Tegra::Texture::TICEntry entry; + gpu_memory->ReadBlock(descriptor_addr, &entry, sizeof(entry)); + const Shader::TextureType result{ConvertType(entry)}; + texture_types.emplace(raw, result); + return result; +} + +GraphicsEnvironment::GraphicsEnvironment(Tegra::Engines::Maxwell3D& maxwell3d_, + Tegra::MemoryManager& gpu_memory_, + Maxwell::ShaderProgram program, GPUVAddr program_base_, + u32 start_address_) + : GenericEnvironment{gpu_memory_, program_base_, start_address_}, maxwell3d{&maxwell3d_} { + gpu_memory->ReadBlock(program_base + start_address, &sph, sizeof(sph)); + switch (program) { + case Maxwell::ShaderProgram::VertexA: + stage = Shader::Stage::VertexA; + stage_index = 0; + break; + case Maxwell::ShaderProgram::VertexB: + stage = Shader::Stage::VertexB; + stage_index = 0; + break; + case Maxwell::ShaderProgram::TesselationControl: + stage = Shader::Stage::TessellationControl; + stage_index = 1; + break; + case Maxwell::ShaderProgram::TesselationEval: + stage = Shader::Stage::TessellationEval; + stage_index = 2; + break; + case Maxwell::ShaderProgram::Geometry: + stage = Shader::Stage::Geometry; + stage_index = 3; + break; + case Maxwell::ShaderProgram::Fragment: + stage = Shader::Stage::Fragment; + stage_index = 4; + break; + default: + UNREACHABLE_MSG("Invalid program={}", program); + break; + } + const u64 local_size{sph.LocalMemorySize()}; + ASSERT(local_size <= std::numeric_limits::max()); + local_memory_size = static_cast(local_size); + texture_bound = maxwell3d->regs.tex_cb_index; +} + +u32 GraphicsEnvironment::ReadCbufValue(u32 cbuf_index, u32 cbuf_offset) { + const auto& cbuf{maxwell3d->state.shader_stages[stage_index].const_buffers[cbuf_index]}; + ASSERT(cbuf.enabled); + u32 value{}; + if (cbuf_offset < cbuf.size) { + value = gpu_memory->Read(cbuf.address + cbuf_offset); + } + cbuf_values.emplace(MakeCbufKey(cbuf_index, cbuf_offset), value); + return value; +} + +Shader::TextureType GraphicsEnvironment::ReadTextureType(u32 handle) { + const auto& regs{maxwell3d->regs}; + const bool via_header_index{regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex}; + return ReadTextureTypeImpl(regs.tic.Address(), regs.tic.limit, via_header_index, handle); +} + +ComputeEnvironment::ComputeEnvironment(Tegra::Engines::KeplerCompute& kepler_compute_, + Tegra::MemoryManager& gpu_memory_, GPUVAddr program_base_, + u32 start_address_) + : GenericEnvironment{gpu_memory_, program_base_, start_address_}, kepler_compute{ + &kepler_compute_} { + const auto& qmd{kepler_compute->launch_description}; + stage = Shader::Stage::Compute; + local_memory_size = qmd.local_pos_alloc; + texture_bound = kepler_compute->regs.tex_cb_index; + shared_memory_size = qmd.shared_alloc; + workgroup_size = {qmd.block_dim_x, qmd.block_dim_y, qmd.block_dim_z}; +} + +u32 ComputeEnvironment::ReadCbufValue(u32 cbuf_index, u32 cbuf_offset) { + const auto& qmd{kepler_compute->launch_description}; + ASSERT(((qmd.const_buffer_enable_mask.Value() >> cbuf_index) & 1) != 0); + const auto& cbuf{qmd.const_buffer_config[cbuf_index]}; + u32 value{}; + if (cbuf_offset < cbuf.size) { + value = gpu_memory->Read(cbuf.Address() + cbuf_offset); + } + cbuf_values.emplace(MakeCbufKey(cbuf_index, cbuf_offset), value); + return value; +} + +Shader::TextureType ComputeEnvironment::ReadTextureType(u32 handle) { + const auto& regs{kepler_compute->regs}; + const auto& qmd{kepler_compute->launch_description}; + return ReadTextureTypeImpl(regs.tic.Address(), regs.tic.limit, qmd.linked_tsc != 0, handle); +} + +void FileEnvironment::Deserialize(std::ifstream& file) { + u64 code_size{}; + u64 num_texture_types{}; + u64 num_cbuf_values{}; + file.read(reinterpret_cast(&code_size), sizeof(code_size)) + .read(reinterpret_cast(&num_texture_types), sizeof(num_texture_types)) + .read(reinterpret_cast(&num_cbuf_values), sizeof(num_cbuf_values)) + .read(reinterpret_cast(&local_memory_size), sizeof(local_memory_size)) + .read(reinterpret_cast(&texture_bound), sizeof(texture_bound)) + .read(reinterpret_cast(&start_address), sizeof(start_address)) + .read(reinterpret_cast(&read_lowest), sizeof(read_lowest)) + .read(reinterpret_cast(&read_highest), sizeof(read_highest)) + .read(reinterpret_cast(&stage), sizeof(stage)); + code = std::make_unique(Common::DivCeil(code_size, sizeof(u64))); + file.read(reinterpret_cast(code.get()), code_size); + for (size_t i = 0; i < num_texture_types; ++i) { + u32 key; + Shader::TextureType type; + file.read(reinterpret_cast(&key), sizeof(key)) + .read(reinterpret_cast(&type), sizeof(type)); + texture_types.emplace(key, type); + } + for (size_t i = 0; i < num_cbuf_values; ++i) { + u64 key; + u32 value; + file.read(reinterpret_cast(&key), sizeof(key)) + .read(reinterpret_cast(&value), sizeof(value)); + cbuf_values.emplace(key, value); + } + if (stage == Shader::Stage::Compute) { + file.read(reinterpret_cast(&workgroup_size), sizeof(workgroup_size)) + .read(reinterpret_cast(&shared_memory_size), sizeof(shared_memory_size)); + } else { + file.read(reinterpret_cast(&sph), sizeof(sph)); + } +} + +u64 FileEnvironment::ReadInstruction(u32 address) { + if (address < read_lowest || address > read_highest) { + throw Shader::LogicError("Out of bounds address {}", address); + } + return code[(address - read_lowest) / sizeof(u64)]; +} + +u32 FileEnvironment::ReadCbufValue(u32 cbuf_index, u32 cbuf_offset) { + const auto it{cbuf_values.find(MakeCbufKey(cbuf_index, cbuf_offset))}; + if (it == cbuf_values.end()) { + throw Shader::LogicError("Uncached read texture type"); + } + return it->second; +} + +Shader::TextureType FileEnvironment::ReadTextureType(u32 handle) { + const auto it{texture_types.find(handle)}; + if (it == texture_types.end()) { + throw Shader::LogicError("Uncached read texture type"); + } + return it->second; +} + +u32 FileEnvironment::LocalMemorySize() const { + return local_memory_size; +} + +u32 FileEnvironment::SharedMemorySize() const { + return shared_memory_size; +} + +u32 FileEnvironment::TextureBoundBuffer() const { + return texture_bound; +} + +std::array FileEnvironment::WorkgroupSize() const { + return workgroup_size; +} + +void SerializePipeline(std::span key, std::span envs, + const std::filesystem::path& filename) try { + std::ofstream file(filename, std::ios::binary | std::ios::ate | std::ios::app); + file.exceptions(std::ifstream::failbit); + if (!file.is_open()) { + LOG_ERROR(Common_Filesystem, "Failed to open pipeline cache file {}", + Common::FS::PathToUTF8String(filename)); + return; + } + if (file.tellp() == 0) { + // Write header + file.write(MAGIC_NUMBER.data(), MAGIC_NUMBER.size()) + .write(reinterpret_cast(&CACHE_VERSION), sizeof(CACHE_VERSION)); + } + if (!std::ranges::all_of(envs, &GenericEnvironment::CanBeSerialized)) { + return; + } + const u32 num_envs{static_cast(envs.size())}; + file.write(reinterpret_cast(&num_envs), sizeof(num_envs)); + for (const GenericEnvironment* const env : envs) { + env->Serialize(file); + } + file.write(key.data(), key.size_bytes()); + +} catch (const std::ios_base::failure& e) { + LOG_ERROR(Common_Filesystem, "{}", e.what()); + if (!Common::FS::RemoveFile(filename)) { + LOG_ERROR(Common_Filesystem, "Failed to delete pipeline cache file {}", + Common::FS::PathToUTF8String(filename)); + } +} + +void LoadPipelines( + std::stop_token stop_loading, const std::filesystem::path& filename, + Common::UniqueFunction load_compute, + Common::UniqueFunction> load_graphics) try { + std::ifstream file(filename, std::ios::binary | std::ios::ate); + if (!file.is_open()) { + return; + } + file.exceptions(std::ifstream::failbit); + const auto end{file.tellg()}; + file.seekg(0, std::ios::beg); + + std::array magic_number; + u32 cache_version; + file.read(magic_number.data(), magic_number.size()) + .read(reinterpret_cast(&cache_version), sizeof(cache_version)); + if (magic_number != MAGIC_NUMBER || cache_version != CACHE_VERSION) { + file.close(); + if (Common::FS::RemoveFile(filename)) { + if (magic_number != MAGIC_NUMBER) { + LOG_ERROR(Common_Filesystem, "Invalid pipeline cache file"); + } + if (cache_version != CACHE_VERSION) { + LOG_INFO(Common_Filesystem, "Deleting old pipeline cache"); + } + } else { + LOG_ERROR(Common_Filesystem, + "Invalid pipeline cache file and failed to delete it in \"{}\"", + Common::FS::PathToUTF8String(filename)); + } + return; + } + while (file.tellg() != end) { + if (stop_loading.stop_requested()) { + return; + } + u32 num_envs{}; + file.read(reinterpret_cast(&num_envs), sizeof(num_envs)); + std::vector envs(num_envs); + for (FileEnvironment& env : envs) { + env.Deserialize(file); + } + if (envs.front().ShaderStage() == Shader::Stage::Compute) { + load_compute(file, std::move(envs.front())); + } else { + load_graphics(file, std::move(envs)); + } + } + +} catch (const std::ios_base::failure& e) { + LOG_ERROR(Common_Filesystem, "{}", e.what()); + if (!Common::FS::RemoveFile(filename)) { + LOG_ERROR(Common_Filesystem, "Failed to delete pipeline cache file {}", + Common::FS::PathToUTF8String(filename)); + } +} + +} // namespace VideoCommon -- cgit v1.2.3 From d621e96d0de212cc16897eadf71e8a1b2e1eb5dc Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Sun, 23 May 2021 04:28:34 -0300 Subject: shader: Initial OpenGL implementation --- src/video_core/shader_environment.cpp | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src/video_core/shader_environment.cpp') diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp index 5dccc0097..c93174519 100644 --- a/src/video_core/shader_environment.cpp +++ b/src/video_core/shader_environment.cpp @@ -187,8 +187,8 @@ std::optional GenericEnvironment::TryFindSize() { Shader::TextureType GenericEnvironment::ReadTextureTypeImpl(GPUVAddr tic_addr, u32 tic_limit, bool via_header_index, u32 raw) { - const TextureHandle handle{raw, via_header_index}; - const GPUVAddr descriptor_addr{tic_addr + handle.image * sizeof(Tegra::Texture::TICEntry)}; + const auto handle{Tegra::Texture::TexturePair(raw, via_header_index)}; + const GPUVAddr descriptor_addr{tic_addr + handle.first * sizeof(Tegra::Texture::TICEntry)}; Tegra::Texture::TICEntry entry; gpu_memory->ReadBlock(descriptor_addr, &entry, sizeof(entry)); const Shader::TextureType result{ConvertType(entry)}; -- cgit v1.2.3 From 15bdd27cac4a0b1e6cd168272dc337cd685ef144 Mon Sep 17 00:00:00 2001 From: ameerj Date: Wed, 9 Jun 2021 23:33:48 -0400 Subject: shader_environment: Add shader_local_memory_crs_size to local memory size Fixes DOOM 2016 missing local memory --- src/video_core/shader_environment.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/video_core/shader_environment.cpp') diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp index c93174519..a7a57a36f 100644 --- a/src/video_core/shader_environment.cpp +++ b/src/video_core/shader_environment.cpp @@ -69,7 +69,7 @@ u32 GenericEnvironment::TextureBoundBuffer() const { } u32 GenericEnvironment::LocalMemorySize() const { - return local_memory_size; + return local_memory_size + sph.common3.shader_local_memory_crs_size; } u32 GenericEnvironment::SharedMemorySize() const { -- cgit v1.2.3 From 5befc0bf872058315c4f81bf58dcd173db2589fd Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Thu, 10 Jun 2021 02:27:00 -0300 Subject: shader_environment: Fix local memory size calculations --- src/video_core/shader_environment.cpp | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src/video_core/shader_environment.cpp') diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp index a7a57a36f..6243cd176 100644 --- a/src/video_core/shader_environment.cpp +++ b/src/video_core/shader_environment.cpp @@ -69,7 +69,7 @@ u32 GenericEnvironment::TextureBoundBuffer() const { } u32 GenericEnvironment::LocalMemorySize() const { - return local_memory_size + sph.common3.shader_local_memory_crs_size; + return local_memory_size; } u32 GenericEnvironment::SharedMemorySize() const { @@ -233,7 +233,7 @@ GraphicsEnvironment::GraphicsEnvironment(Tegra::Engines::Maxwell3D& maxwell3d_, } const u64 local_size{sph.LocalMemorySize()}; ASSERT(local_size <= std::numeric_limits::max()); - local_memory_size = static_cast(local_size); + local_memory_size = static_cast(local_size) + sph.common3.shader_local_memory_crs_size; texture_bound = maxwell3d->regs.tex_cb_index; } @@ -261,7 +261,7 @@ ComputeEnvironment::ComputeEnvironment(Tegra::Engines::KeplerCompute& kepler_com &kepler_compute_} { const auto& qmd{kepler_compute->launch_description}; stage = Shader::Stage::Compute; - local_memory_size = qmd.local_pos_alloc; + local_memory_size = qmd.local_pos_alloc + qmd.local_crs_alloc; texture_bound = kepler_compute->regs.tex_cb_index; shared_memory_size = qmd.shared_alloc; workgroup_size = {qmd.block_dim_x, qmd.block_dim_y, qmd.block_dim_z}; -- cgit v1.2.3 From 395bed3a0af90a53be44e81eadd06f4931c8e933 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Wed, 23 Jun 2021 02:41:00 -0300 Subject: shader: Unify shader stage types --- src/video_core/shader_environment.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/video_core/shader_environment.cpp') diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp index 6243cd176..d463e2b56 100644 --- a/src/video_core/shader_environment.cpp +++ b/src/video_core/shader_environment.cpp @@ -22,7 +22,7 @@ namespace VideoCommon { constexpr std::array MAGIC_NUMBER{'y', 'u', 'z', 'u', 'c', 'a', 'c', 'h'}; -constexpr u32 CACHE_VERSION = 3; +constexpr u32 CACHE_VERSION = 4; constexpr size_t INST_SIZE = sizeof(u64); -- cgit v1.2.3 From 7dafa96ab59892b7f1fbffdb61e4326e6443955f Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Thu, 24 Jun 2021 02:41:09 -0300 Subject: shader: Rework varyings and implement passthrough geometry shaders Put all varyings into a single std::bitset with helpers to access it. Implement passthrough geometry shaders using host's. --- src/video_core/shader_environment.cpp | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'src/video_core/shader_environment.cpp') diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp index d463e2b56..429cab30d 100644 --- a/src/video_core/shader_environment.cpp +++ b/src/video_core/shader_environment.cpp @@ -22,7 +22,7 @@ namespace VideoCommon { constexpr std::array MAGIC_NUMBER{'y', 'u', 'z', 'u', 'c', 'a', 'c', 'h'}; -constexpr u32 CACHE_VERSION = 4; +constexpr u32 CACHE_VERSION = 5; constexpr size_t INST_SIZE = sizeof(u64); @@ -155,6 +155,10 @@ void GenericEnvironment::Serialize(std::ofstream& file) const { .write(reinterpret_cast(&shared_memory_size), sizeof(shared_memory_size)); } else { file.write(reinterpret_cast(&sph), sizeof(sph)); + if (stage == Shader::Stage::Geometry) { + file.write(reinterpret_cast(&gp_passthrough_mask), + sizeof(gp_passthrough_mask)); + } } } @@ -202,6 +206,7 @@ GraphicsEnvironment::GraphicsEnvironment(Tegra::Engines::Maxwell3D& maxwell3d_, u32 start_address_) : GenericEnvironment{gpu_memory_, program_base_, start_address_}, maxwell3d{&maxwell3d_} { gpu_memory->ReadBlock(program_base + start_address, &sph, sizeof(sph)); + gp_passthrough_mask = maxwell3d->regs.gp_passthrough_mask; switch (program) { case Maxwell::ShaderProgram::VertexA: stage = Shader::Stage::VertexA; @@ -319,6 +324,9 @@ void FileEnvironment::Deserialize(std::ifstream& file) { .read(reinterpret_cast(&shared_memory_size), sizeof(shared_memory_size)); } else { file.read(reinterpret_cast(&sph), sizeof(sph)); + if (stage == Shader::Stage::Geometry) { + file.read(reinterpret_cast(&gp_passthrough_mask), sizeof(gp_passthrough_mask)); + } } } -- cgit v1.2.3 From 258f35515d61d01049d2e433146cab808837bb7d Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Sun, 18 Jul 2021 21:07:12 -0300 Subject: shader_environment: Receive cache version from outside This allows us invalidating OpenGL and Vulkan separately in the future. --- src/video_core/shader_environment.cpp | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'src/video_core/shader_environment.cpp') diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp index 429cab30d..8a4581c19 100644 --- a/src/video_core/shader_environment.cpp +++ b/src/video_core/shader_environment.cpp @@ -22,7 +22,6 @@ namespace VideoCommon { constexpr std::array MAGIC_NUMBER{'y', 'u', 'z', 'u', 'c', 'a', 'c', 'h'}; -constexpr u32 CACHE_VERSION = 5; constexpr size_t INST_SIZE = sizeof(u64); @@ -370,7 +369,7 @@ std::array FileEnvironment::WorkgroupSize() const { } void SerializePipeline(std::span key, std::span envs, - const std::filesystem::path& filename) try { + const std::filesystem::path& filename, u32 cache_version) try { std::ofstream file(filename, std::ios::binary | std::ios::ate | std::ios::app); file.exceptions(std::ifstream::failbit); if (!file.is_open()) { @@ -381,7 +380,7 @@ void SerializePipeline(std::span key, std::span(&CACHE_VERSION), sizeof(CACHE_VERSION)); + .write(reinterpret_cast(&cache_version), sizeof(cache_version)); } if (!std::ranges::all_of(envs, &GenericEnvironment::CanBeSerialized)) { return; @@ -402,7 +401,7 @@ void SerializePipeline(std::span key, std::span load_compute, Common::UniqueFunction> load_graphics) try { std::ifstream file(filename, std::ios::binary | std::ios::ate); @@ -417,13 +416,13 @@ void LoadPipelines( u32 cache_version; file.read(magic_number.data(), magic_number.size()) .read(reinterpret_cast(&cache_version), sizeof(cache_version)); - if (magic_number != MAGIC_NUMBER || cache_version != CACHE_VERSION) { + if (magic_number != MAGIC_NUMBER || cache_version != expected_cache_version) { file.close(); if (Common::FS::RemoveFile(filename)) { if (magic_number != MAGIC_NUMBER) { LOG_ERROR(Common_Filesystem, "Invalid pipeline cache file"); } - if (cache_version != CACHE_VERSION) { + if (cache_version != expected_cache_version) { LOG_INFO(Common_Filesystem, "Deleting old pipeline cache"); } } else { -- cgit v1.2.3