vulkan: Serialize pipelines on a separate thread

This commit is contained in:
ReinUsesLisp 2021-04-01 04:09:09 -03:00 committed by ameerj
parent 8771639d1e
commit d0a529683a
2 changed files with 63 additions and 66 deletions

View File

@ -61,6 +61,33 @@ public:
~GenericEnvironment() override = default; ~GenericEnvironment() override = default;
u32 TextureBoundBuffer() const final {
return texture_bound;
}
u32 LocalMemorySize() const final {
return local_memory_size;
}
u32 SharedMemorySize() const final {
return shared_memory_size;
}
std::array<u32, 3> WorkgroupSize() const final {
return workgroup_size;
}
u64 ReadInstruction(u32 address) final {
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<u64>(program_base + address);
}
std::optional<u128> Analyze() { std::optional<u128> Analyze() {
const std::optional<u64> size{TryFindSize()}; const std::optional<u64> size{TryFindSize()};
if (!size) { if (!size) {
@ -97,26 +124,10 @@ public:
return Common::CityHash128(data.get(), size); return Common::CityHash128(data.get(), size);
} }
u64 ReadInstruction(u32 address) final {
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<u64>(program_base + address);
}
void Serialize(std::ofstream& file) const { void Serialize(std::ofstream& file) const {
const u64 code_size{static_cast<u64>(ReadSize())}; const u64 code_size{static_cast<u64>(CachedSize())};
const auto data{std::make_unique<char[]>(code_size)};
gpu_memory->ReadBlock(program_base + read_lowest, data.get(), code_size);
const u64 num_texture_types{static_cast<u64>(texture_types.size())}; const u64 num_texture_types{static_cast<u64>(texture_types.size())};
const u64 num_cbuf_values{static_cast<u64>(cbuf_values.size())}; const u64 num_cbuf_values{static_cast<u64>(cbuf_values.size())};
const u32 local_memory_size{LocalMemorySize()};
const u32 texture_bound{TextureBoundBuffer()};
file.write(reinterpret_cast<const char*>(&code_size), sizeof(code_size)) file.write(reinterpret_cast<const char*>(&code_size), sizeof(code_size))
.write(reinterpret_cast<const char*>(&num_texture_types), sizeof(num_texture_types)) .write(reinterpret_cast<const char*>(&num_texture_types), sizeof(num_texture_types))
@ -124,10 +135,10 @@ public:
.write(reinterpret_cast<const char*>(&local_memory_size), sizeof(local_memory_size)) .write(reinterpret_cast<const char*>(&local_memory_size), sizeof(local_memory_size))
.write(reinterpret_cast<const char*>(&texture_bound), sizeof(texture_bound)) .write(reinterpret_cast<const char*>(&texture_bound), sizeof(texture_bound))
.write(reinterpret_cast<const char*>(&start_address), sizeof(start_address)) .write(reinterpret_cast<const char*>(&start_address), sizeof(start_address))
.write(reinterpret_cast<const char*>(&read_lowest), sizeof(read_lowest)) .write(reinterpret_cast<const char*>(&cached_lowest), sizeof(cached_lowest))
.write(reinterpret_cast<const char*>(&read_highest), sizeof(read_highest)) .write(reinterpret_cast<const char*>(&cached_highest), sizeof(cached_highest))
.write(reinterpret_cast<const char*>(&stage), sizeof(stage)) .write(reinterpret_cast<const char*>(&stage), sizeof(stage))
.write(data.get(), code_size); .write(reinterpret_cast<const char*>(code.data()), code_size);
for (const auto [key, type] : texture_types) { for (const auto [key, type] : texture_types) {
file.write(reinterpret_cast<const char*>(&key), sizeof(key)) file.write(reinterpret_cast<const char*>(&key), sizeof(key))
.write(reinterpret_cast<const char*>(&type), sizeof(type)); .write(reinterpret_cast<const char*>(&type), sizeof(type));
@ -137,8 +148,6 @@ public:
.write(reinterpret_cast<const char*>(&type), sizeof(type)); .write(reinterpret_cast<const char*>(&type), sizeof(type));
} }
if (stage == Shader::Stage::Compute) { if (stage == Shader::Stage::Compute) {
const std::array<u32, 3> workgroup_size{WorkgroupSize()};
const u32 shared_memory_size{SharedMemorySize()};
file.write(reinterpret_cast<const char*>(&workgroup_size), sizeof(workgroup_size)) file.write(reinterpret_cast<const char*>(&workgroup_size), sizeof(workgroup_size))
.write(reinterpret_cast<const char*>(&shared_memory_size), .write(reinterpret_cast<const char*>(&shared_memory_size),
sizeof(shared_memory_size)); sizeof(shared_memory_size));
@ -220,6 +229,11 @@ protected:
std::unordered_map<u64, Shader::TextureType> texture_types; std::unordered_map<u64, Shader::TextureType> texture_types;
std::unordered_map<u64, u32> cbuf_values; std::unordered_map<u64, u32> cbuf_values;
u32 local_memory_size{};
u32 texture_bound{};
u32 shared_memory_size{};
std::array<u32, 3> workgroup_size{};
u32 read_lowest = std::numeric_limits<u32>::max(); u32 read_lowest = std::numeric_limits<u32>::max();
u32 read_highest = 0; u32 read_highest = 0;
@ -270,6 +284,10 @@ public:
UNREACHABLE_MSG("Invalid program={}", program); UNREACHABLE_MSG("Invalid program={}", program);
break; break;
} }
const u64 local_size{sph.LocalMemorySize()};
ASSERT(local_size <= std::numeric_limits<u32>::max());
local_memory_size = static_cast<u32>(local_size);
texture_bound = maxwell3d->regs.tex_cb_index;
} }
~GraphicsEnvironment() override = default; ~GraphicsEnvironment() override = default;
@ -294,24 +312,6 @@ public:
cbuf.address, cbuf.size, cbuf_index, cbuf_offset); cbuf.address, cbuf.size, cbuf_index, cbuf_offset);
} }
u32 TextureBoundBuffer() const override {
return maxwell3d->regs.tex_cb_index;
}
u32 LocalMemorySize() const override {
const u64 size{sph.LocalMemorySize()};
ASSERT(size <= std::numeric_limits<u32>::max());
return static_cast<u32>(size);
}
u32 SharedMemorySize() const override {
throw Shader::LogicError("Requesting shared memory size in graphics stage");
}
std::array<u32, 3> WorkgroupSize() const override {
throw Shader::LogicError("Requesting workgroup size in a graphics stage");
}
private: private:
Tegra::Engines::Maxwell3D* maxwell3d{}; Tegra::Engines::Maxwell3D* maxwell3d{};
size_t stage_index{}; size_t stage_index{};
@ -325,7 +325,12 @@ public:
u32 start_address_) u32 start_address_)
: GenericEnvironment{gpu_memory_, program_base_, start_address_}, kepler_compute{ : GenericEnvironment{gpu_memory_, program_base_, start_address_}, kepler_compute{
&kepler_compute_} { &kepler_compute_} {
const auto& qmd{kepler_compute->launch_description};
stage = Shader::Stage::Compute; 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};
} }
~ComputeEnvironment() override = default; ~ComputeEnvironment() override = default;
@ -351,25 +356,6 @@ public:
cbuf.Address(), cbuf.size, cbuf_index, cbuf_offset); cbuf.Address(), cbuf.size, cbuf_index, cbuf_offset);
} }
u32 TextureBoundBuffer() const override {
return kepler_compute->regs.tex_cb_index;
}
u32 LocalMemorySize() const override {
const auto& qmd{kepler_compute->launch_description};
return qmd.local_pos_alloc;
}
u32 SharedMemorySize() const override {
const auto& qmd{kepler_compute->launch_description};
return qmd.shared_alloc;
}
std::array<u32, 3> WorkgroupSize() const override {
const auto& qmd{kepler_compute->launch_description};
return {qmd.block_dim_x, qmd.block_dim_y, qmd.block_dim_z};
}
private: private:
Tegra::Engines::KeplerCompute* kepler_compute{}; Tegra::Engines::KeplerCompute* kepler_compute{};
}; };
@ -621,7 +607,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, Tegra::GPU& gpu_,
scheduler{scheduler_}, descriptor_pool{descriptor_pool_}, scheduler{scheduler_}, descriptor_pool{descriptor_pool_},
update_descriptor_queue{update_descriptor_queue_}, render_pass_cache{render_pass_cache_}, update_descriptor_queue{update_descriptor_queue_}, render_pass_cache{render_pass_cache_},
buffer_cache{buffer_cache_}, texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, texture_cache{texture_cache_},
workers(11, "yuzu:PipelineBuilder") { workers(11, "yuzu:PipelineBuilder"), serialization_thread(1, "yuzu:PipelineSerialization") {
const auto& float_control{device.FloatControlProperties()}; const auto& float_control{device.FloatControlProperties()};
const VkDriverIdKHR driver_id{device.GetDriverID()}; const VkDriverIdKHR driver_id{device.GetDriverID()};
base_profile = Shader::Profile{ base_profile = Shader::Profile{
@ -796,7 +782,6 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline() {
main_pools.ReleaseContents(); main_pools.ReleaseContents();
std::array<GraphicsEnvironment, Maxwell::MaxShaderProgram> graphics_envs; std::array<GraphicsEnvironment, Maxwell::MaxShaderProgram> graphics_envs;
boost::container::static_vector<GenericEnvironment*, Maxwell::MaxShaderProgram> generic_envs;
boost::container::static_vector<Shader::Environment*, Maxwell::MaxShaderProgram> envs; boost::container::static_vector<Shader::Environment*, Maxwell::MaxShaderProgram> envs;
const GPUVAddr base_addr{maxwell3d.regs.code_address.CodeAddress()}; const GPUVAddr base_addr{maxwell3d.regs.code_address.CodeAddress()};
@ -810,13 +795,22 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline() {
env = GraphicsEnvironment{maxwell3d, gpu_memory, program, base_addr, start_address}; env = GraphicsEnvironment{maxwell3d, gpu_memory, program, base_addr, start_address};
env.SetCachedSize(shader_infos[index]->size_bytes); env.SetCachedSize(shader_infos[index]->size_bytes);
generic_envs.push_back(&env);
envs.push_back(&env); envs.push_back(&env);
} }
auto pipeline{CreateGraphicsPipeline(main_pools, graphics_key, MakeSpan(envs), true)}; auto pipeline{CreateGraphicsPipeline(main_pools, graphics_key, MakeSpan(envs), true)};
if (!pipeline_cache_filename.empty()) { if (pipeline_cache_filename.empty()) {
SerializePipeline(graphics_key, generic_envs, pipeline_cache_filename); return pipeline;
} }
serialization_thread.QueueWork([this, key = graphics_key, envs = std::move(graphics_envs)] {
boost::container::static_vector<const GenericEnvironment*, Maxwell::MaxShaderProgram>
env_ptrs;
for (size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
if (key.unique_hashes[index] != u128{}) {
env_ptrs.push_back(&envs[index]);
}
}
SerializePipeline(key, env_ptrs, pipeline_cache_filename);
});
return pipeline; return pipeline;
} }
@ -830,8 +824,10 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
main_pools.ReleaseContents(); main_pools.ReleaseContents();
auto pipeline{CreateComputePipeline(main_pools, key, env, true)}; auto pipeline{CreateComputePipeline(main_pools, key, env, true)};
if (!pipeline_cache_filename.empty()) { if (!pipeline_cache_filename.empty()) {
SerializePipeline(key, std::array<const GenericEnvironment*, 1>{&env}, serialization_thread.QueueWork([this, key, env = std::move(env)] {
pipeline_cache_filename); SerializePipeline(key, std::array<const GenericEnvironment*, 1>{&env},
pipeline_cache_filename);
});
} }
return pipeline; return pipeline;
} }

View File

@ -187,6 +187,7 @@ private:
std::string pipeline_cache_filename; std::string pipeline_cache_filename;
Common::ThreadWorker workers; Common::ThreadWorker workers;
Common::ThreadWorker serialization_thread;
}; };
} // namespace Vulkan } // namespace Vulkan