lux/citra
1
0
Fork 0
mirror of https://git.h3cjp.net/H3cJP/citra.git synced 2024-12-27 21:56:42 +00:00
Code Issues Packages Projects Releases Wiki Activity

VideoCore: Fix channels with disk pipeline/shader cache.

Browse source
This commit is contained in:
Fernando Sahmkow 2021-11-07 17:52:45 +01:00
parent d7990c159e
commit 3f8e7a5585
11 changed files with 87 additions and 71 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

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

@ -28,12 +28,11 @@ bool ComputePipelineKey::operator==(const ComputePipelineKey& rhs) const noexcep
}
ComputePipeline::ComputePipeline(const Device& device, TextureCache& texture_cache_,
BufferCache& buffer_cache_, Tegra::MemoryManager& gpu_memory_,
Tegra::Engines::KeplerCompute& kepler_compute_,
ProgramManager& program_manager_, const Shader::Info& info_,
std::string code, std::vector<u32> code_v)
: texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, gpu_memory{gpu_memory_},
kepler_compute{kepler_compute_}, program_manager{program_manager_}, info{info_} {
BufferCache& buffer_cache_, ProgramManager& program_manager_,
const Shader::Info& info_, std::string code,
std::vector<u32> code_v)
: texture_cache{texture_cache_}, buffer_cache{buffer_cache_},
program_manager{program_manager_}, info{info_} {
switch (device.GetShaderBackend()) {
case Settings::ShaderBackend::GLSL:
source_program = CreateProgram(code, GL_COMPUTE_SHADER);
@ -86,7 +85,7 @@ void ComputePipeline::Configure() {
GLsizei texture_binding{};
GLsizei image_binding{};
const auto& qmd{kepler_compute.launch_description};
const auto& qmd{kepler_compute->launch_description};
const auto& cbufs{qmd.const_buffer_config};
const bool via_header_index{qmd.linked_tsc != 0};
const auto read_handle{[&](const auto& desc, u32 index) {
@ -101,12 +100,12 @@ void ComputePipeline::Configure() {
const u32 secondary_offset{desc.secondary_cbuf_offset + index_offset};
const GPUVAddr separate_addr{cbufs[desc.secondary_cbuf_index].Address() +
secondary_offset};
const u32 lhs_raw{gpu_memory.Read<u32>(addr)};
const u32 rhs_raw{gpu_memory.Read<u32>(separate_addr)};
const u32 lhs_raw{gpu_memory->Read<u32>(addr)};
const u32 rhs_raw{gpu_memory->Read<u32>(separate_addr)};
return TexturePair(lhs_raw | rhs_raw, via_header_index);
}
}
return TexturePair(gpu_memory.Read<u32>(addr), via_header_index);
return TexturePair(gpu_memory->Read<u32>(addr), via_header_index);
}};
const auto add_image{[&](const auto& desc, bool blacklist) {
for (u32 index = 0; index < desc.count; ++index) {

16
src/video_core/renderer_opengl/gl_compute_pipeline.h
View file

@ -49,10 +49,8 @@ static_assert(std::is_trivially_constructible_v<ComputePipelineKey>);
class ComputePipeline {
public:
explicit ComputePipeline(const Device& device, TextureCache& texture_cache_,
BufferCache& buffer_cache_, Tegra::MemoryManager& gpu_memory_,
Tegra::Engines::KeplerCompute& kepler_compute_,
ProgramManager& program_manager_, const Shader::Info& info_,
std::string code, std::vector<u32> code_v);
BufferCache& buffer_cache_, ProgramManager& program_manager_,
const Shader::Info& info_, std::string code, std::vector<u32> code_v);
void Configure();
@ -60,11 +58,17 @@ public:
return writes_global_memory;
}
void SetEngine(Tegra::Engines::KeplerCompute* kepler_compute_,
Tegra::MemoryManager* gpu_memory_) {
kepler_compute = kepler_compute_;
gpu_memory = gpu_memory_;
}
private:
TextureCache& texture_cache;
BufferCache& buffer_cache;
Tegra::MemoryManager& gpu_memory;
Tegra::Engines::KeplerCompute& kepler_compute;
Tegra::MemoryManager* gpu_memory;
Tegra::Engines::KeplerCompute* kepler_compute;
ProgramManager& program_manager;
Shader::Info info;

28
src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
View file

@ -169,15 +169,15 @@ ConfigureFuncPtr ConfigureFunc(const std::array<Shader::Info, 5>& infos, u32 ena
}
} // Anonymous namespace
GraphicsPipeline::GraphicsPipeline(
const Device& device, TextureCache& texture_cache_, BufferCache& buffer_cache_,
Tegra::MemoryManager& gpu_memory_, Tegra::Engines::Maxwell3D& maxwell3d_,
ProgramManager& program_manager_, StateTracker& state_tracker_, ShaderWorker* thread_worker,
VideoCore::ShaderNotify* shader_notify, std::array<std::string, 5> sources,
std::array<std::vector<u32>, 5> sources_spirv, const std::array<const Shader::Info*, 5>& infos,
const GraphicsPipelineKey& key_)
: texture_cache{texture_cache_}, buffer_cache{buffer_cache_},
gpu_memory{gpu_memory_}, maxwell3d{maxwell3d_}, program_manager{program_manager_},
GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_cache_,
BufferCache& buffer_cache_, ProgramManager& program_manager_,
StateTracker& state_tracker_, ShaderWorker* thread_worker,
VideoCore::ShaderNotify* shader_notify,
std::array<std::string, 5> sources,
std::array<std::vector<u32>, 5> sources_spirv,
const std::array<const Shader::Info*, 5>& infos,
const GraphicsPipelineKey& key_)
: texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, program_manager{program_manager_},
state_tracker{state_tracker_}, key{key_} {
if (shader_notify) {
shader_notify->MarkShaderBuilding();
@ -285,7 +285,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
buffer_cache.runtime.SetBaseStorageBindings(base_storage_bindings);
buffer_cache.runtime.SetEnableStorageBuffers(use_storage_buffers);
const auto& regs{maxwell3d.regs};
const auto& regs{maxwell3d->regs};
const bool via_header_index{regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex};
const auto config_stage{[&](size_t stage) LAMBDA_FORCEINLINE {
const Shader::Info& info{stage_infos[stage]};
@ -299,7 +299,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
++ssbo_index;
}
}
const auto& cbufs{maxwell3d.state.shader_stages[stage].const_buffers};
const auto& cbufs{maxwell3d->state.shader_stages[stage].const_buffers};
const auto read_handle{[&](const auto& desc, u32 index) {
ASSERT(cbufs[desc.cbuf_index].enabled);
const u32 index_offset{index << desc.size_shift};
@ -312,13 +312,13 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
const u32 second_offset{desc.secondary_cbuf_offset + index_offset};
const GPUVAddr separate_addr{cbufs[desc.secondary_cbuf_index].address +
second_offset};
const u32 lhs_raw{gpu_memory.Read<u32>(addr)};
const u32 rhs_raw{gpu_memory.Read<u32>(separate_addr)};
const u32 lhs_raw{gpu_memory->Read<u32>(addr)};
const u32 rhs_raw{gpu_memory->Read<u32>(separate_addr)};
const u32 raw{lhs_raw | rhs_raw};
return TexturePair(raw, via_header_index);
}
}
return TexturePair(gpu_memory.Read<u32>(addr), via_header_index);
return TexturePair(gpu_memory->Read<u32>(addr), via_header_index);
}};
const auto add_image{[&](const auto& desc, bool blacklist) LAMBDA_FORCEINLINE {
for (u32 index = 0; index < desc.count; ++index) {

16
src/video_core/renderer_opengl/gl_graphics_pipeline.h
View file

@ -71,10 +71,9 @@ static_assert(std::is_trivially_constructible_v<GraphicsPipelineKey>);
class GraphicsPipeline {
public:
explicit GraphicsPipeline(const Device& device, TextureCache& texture_cache_,
BufferCache& buffer_cache_, Tegra::MemoryManager& gpu_memory_,
Tegra::Engines::Maxwell3D& maxwell3d_,
ProgramManager& program_manager_, StateTracker& state_tracker_,
ShaderWorker* thread_worker, VideoCore::ShaderNotify* shader_notify,
BufferCache& buffer_cache_, ProgramManager& program_manager_,
StateTracker& state_tracker_, ShaderWorker* thread_worker,
VideoCore::ShaderNotify* shader_notify,
std::array<std::string, 5> sources,
std::array<std::vector<u32>, 5> sources_spirv,
const std::array<const Shader::Info*, 5>& infos,
@ -107,6 +106,11 @@ public:
};
}
void SetEngine(Tegra::Engines::Maxwell3D* maxwell3d_, Tegra::MemoryManager* gpu_memory_) {
maxwell3d = maxwell3d_;
gpu_memory = gpu_memory_;
}
private:
template <typename Spec>
void ConfigureImpl(bool is_indexed);
@ -119,8 +123,8 @@ private:
TextureCache& texture_cache;
BufferCache& buffer_cache;
Tegra::MemoryManager& gpu_memory;
Tegra::Engines::Maxwell3D& maxwell3d;
Tegra::MemoryManager* gpu_memory;
Tegra::Engines::Maxwell3D* maxwell3d;
ProgramManager& program_manager;
StateTracker& state_tracker;
const GraphicsPipelineKey key;

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

@ -216,6 +216,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
return;
}
std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
pipeline->SetEngine(maxwell3d, gpu_memory);
pipeline->Configure(is_indexed);
SyncState();

11
src/video_core/renderer_opengl/gl_shader_cache.cpp
View file

@ -477,9 +477,9 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
previous_program = &program;
}
auto* const thread_worker{build_in_parallel ? workers.get() : nullptr};
return std::make_unique<GraphicsPipeline>(
device, texture_cache, buffer_cache, *gpu_memory, *maxwell3d, program_manager,
state_tracker, thread_worker, &shader_notify, sources, sources_spirv, infos, key);
return std::make_unique<GraphicsPipeline>(device, texture_cache, buffer_cache, program_manager,
state_tracker, thread_worker, &shader_notify, sources,
sources_spirv, infos, key);
} catch (Shader::Exception& exception) {
LOG_ERROR(Render_OpenGL, "{}", exception.what());
@ -533,9 +533,8 @@ std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline(
break;
}
return std::make_unique<ComputePipeline>(device, texture_cache, buffer_cache, *gpu_memory,
*kepler_compute, program_manager, program.info, code,
code_spirv);
return std::make_unique<ComputePipeline>(device, texture_cache, buffer_cache, program_manager,
program.info, code, code_spirv);
} catch (Shader::Exception& exception) {
LOG_ERROR(Render_OpenGL, "{}", exception.what());
return nullptr;

17
src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
View file

@ -215,15 +215,14 @@ ConfigureFuncPtr ConfigureFunc(const std::array<vk::ShaderModule, NUM_STAGES>& m
} // Anonymous namespace
GraphicsPipeline::GraphicsPipeline(
Tegra::Engines::Maxwell3D& maxwell3d_, Tegra::MemoryManager& gpu_memory_, Scheduler& scheduler_,
BufferCache& buffer_cache_, TextureCache& texture_cache_,
Scheduler& scheduler_, BufferCache& buffer_cache_, TextureCache& texture_cache_,
VideoCore::ShaderNotify* shader_notify, const Device& device_, DescriptorPool& descriptor_pool,
UpdateDescriptorQueue& update_descriptor_queue_, Common::ThreadWorker* worker_thread,
PipelineStatistics* pipeline_statistics, RenderPassCache& render_pass_cache,
const GraphicsPipelineCacheKey& key_, std::array<vk::ShaderModule, NUM_STAGES> stages,
const std::array<const Shader::Info*, NUM_STAGES>& infos)
: key{key_}, maxwell3d{maxwell3d_}, gpu_memory{gpu_memory_}, device{device_},
texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, scheduler{scheduler_},
: key{key_}, device{device_}, texture_cache{texture_cache_},
buffer_cache{buffer_cache_}, scheduler{scheduler_},
update_descriptor_queue{update_descriptor_queue_}, spv_modules{std::move(stages)} {
if (shader_notify) {
shader_notify->MarkShaderBuilding();
@ -288,7 +287,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
buffer_cache.SetUniformBuffersState(enabled_uniform_buffer_masks, &uniform_buffer_sizes);
const auto& regs{maxwell3d.regs};
const auto& regs{maxwell3d->regs};
const bool via_header_index{regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex};
const auto config_stage{[&](size_t stage) LAMBDA_FORCEINLINE {
const Shader::Info& info{stage_infos[stage]};
@ -302,7 +301,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
++ssbo_index;
}
}
const auto& cbufs{maxwell3d.state.shader_stages[stage].const_buffers};
const auto& cbufs{maxwell3d->state.shader_stages[stage].const_buffers};
const auto read_handle{[&](const auto& desc, u32 index) {
ASSERT(cbufs[desc.cbuf_index].enabled);
const u32 index_offset{index << desc.size_shift};
@ -315,13 +314,13 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
const u32 second_offset{desc.secondary_cbuf_offset + index_offset};
const GPUVAddr separate_addr{cbufs[desc.secondary_cbuf_index].address +
second_offset};
const u32 lhs_raw{gpu_memory.Read<u32>(addr)};
const u32 rhs_raw{gpu_memory.Read<u32>(separate_addr)};
const u32 lhs_raw{gpu_memory->Read<u32>(addr)};
const u32 rhs_raw{gpu_memory->Read<u32>(separate_addr)};
const u32 raw{lhs_raw | rhs_raw};
return TexturePair(raw, via_header_index);
}
}
return TexturePair(gpu_memory.Read<u32>(addr), via_header_index);
return TexturePair(gpu_memory->Read<u32>(addr), via_header_index);
}};
const auto add_image{[&](const auto& desc, bool blacklist) LAMBDA_FORCEINLINE {
for (u32 index = 0; index < desc.count; ++index) {

28
src/video_core/renderer_vulkan/vk_graphics_pipeline.h
View file

@ -69,15 +69,16 @@ class GraphicsPipeline {
static constexpr size_t NUM_STAGES = Tegra::Engines::Maxwell3D::Regs::MaxShaderStage;
public:
explicit GraphicsPipeline(
Tegra::Engines::Maxwell3D& maxwell3d, Tegra::MemoryManager& gpu_memory,
Scheduler& scheduler, BufferCache& buffer_cache, TextureCache& texture_cache,
VideoCore::ShaderNotify* shader_notify, const Device& device,
DescriptorPool& descriptor_pool, UpdateDescriptorQueue& update_descriptor_queue,
Common::ThreadWorker* worker_thread, PipelineStatistics* pipeline_statistics,
RenderPassCache& render_pass_cache, const GraphicsPipelineCacheKey& key,
std::array<vk::ShaderModule, NUM_STAGES> stages,
const std::array<const Shader::Info*, NUM_STAGES>& infos);
explicit GraphicsPipeline(Scheduler& scheduler, BufferCache& buffer_cache,
TextureCache& texture_cache, VideoCore::ShaderNotify* shader_notify,
const Device& device, DescriptorPool& descriptor_pool,
UpdateDescriptorQueue& update_descriptor_queue,
Common::ThreadWorker* worker_thread,
PipelineStatistics* pipeline_statistics,
RenderPassCache& render_pass_cache,
const GraphicsPipelineCacheKey& key,
std::array<vk::ShaderModule, NUM_STAGES> stages,
const std::array<const Shader::Info*, NUM_STAGES>& infos);
GraphicsPipeline& operator=(GraphicsPipeline&&) noexcept = delete;
GraphicsPipeline(GraphicsPipeline&&) noexcept = delete;
@ -109,6 +110,11 @@ public:
return [](GraphicsPipeline* pl, bool is_indexed) { pl->ConfigureImpl<Spec>(is_indexed); };
}
void SetEngine(Tegra::Engines::Maxwell3D* maxwell3d_, Tegra::MemoryManager* gpu_memory_) {
maxwell3d = maxwell3d_;
gpu_memory = gpu_memory_;
}
private:
template <typename Spec>
void ConfigureImpl(bool is_indexed);
@ -120,8 +126,8 @@ private:
void Validate();
const GraphicsPipelineCacheKey key;
Tegra::Engines::Maxwell3D& maxwell3d;
Tegra::MemoryManager& gpu_memory;
Tegra::Engines::Maxwell3D* maxwell3d;
Tegra::MemoryManager* gpu_memory;
const Device& device;
TextureCache& texture_cache;
BufferCache& buffer_cache;

8
src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
View file

@ -555,10 +555,10 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
previous_stage = &program;
}
Common::ThreadWorker* const thread_worker{build_in_parallel ? &workers : nullptr};
return std::make_unique<GraphicsPipeline>(
*maxwell3d, *gpu_memory, scheduler, buffer_cache, texture_cache, &shader_notify, device,
descriptor_pool, update_descriptor_queue, thread_worker, statistics, render_pass_cache, key,
std::move(modules), infos);
return std::make_unique<GraphicsPipeline>(scheduler, buffer_cache, texture_cache,
&shader_notify, device, descriptor_pool,
update_descriptor_queue, thread_worker, statistics,
render_pass_cache, key, std::move(modules), infos);
} catch (const Shader::Exception& exception) {
LOG_ERROR(Render_Vulkan, "{}", exception.what());

2
src/video_core/renderer_vulkan/vk_rasterizer.cpp
View file

@ -190,6 +190,8 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
return;
}
std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
// update engine as channel may be different.
pipeline->SetEngine(maxwell3d, gpu_memory);
pipeline->Configure(is_indexed);
BeginTransformFeedback();

12
src/video_core/texture_cache/texture_cache.h
View file

@ -885,12 +885,14 @@ void TextureCache<P>::InvalidateScale(Image& image) {
}
image.image_view_ids.clear();
image.image_view_infos.clear();
if constexpr (ENABLE_VALIDATION) {
std::ranges::fill(state->graphics_image_view_ids, CORRUPT_ID);
std::ranges::fill(state->compute_image_view_ids, CORRUPT_ID);
for (auto& this_state : channel_storage) {
if constexpr (ENABLE_VALIDATION) {
std::ranges::fill(this_state.graphics_image_view_ids, CORRUPT_ID);
std::ranges::fill(this_state.compute_image_view_ids, CORRUPT_ID);
}
this_state.graphics_image_table.Invalidate();
this_state.compute_image_table.Invalidate();
}
state->graphics_image_table.Invalidate();
state->compute_image_table.Invalidate();
has_deleted_images = true;
}