From 2771bde17a2e732a9a223704fa4c87ba5e434491 Mon Sep 17 00:00:00 2001 From: Yuri Kunde Schlesner Date: Fri, 16 Dec 2016 21:41:38 -0800 Subject: VideoCore/Shader: Extract input vertex loading code into function --- src/video_core/command_processor.cpp | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src/video_core/command_processor.cpp') diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp index ea58e9f548..36f72393b9 100644 --- a/src/video_core/command_processor.cpp +++ b/src/video_core/command_processor.cpp @@ -149,7 +149,8 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { if (g_debug_context) g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, static_cast(&immediate_input)); - g_state.vs.Run(shader_unit, immediate_input, regs.vs.num_input_attributes + 1); + shader_unit.LoadInputVertex(immediate_input, regs.vs.num_input_attributes + 1); + g_state.vs.Run(shader_unit); Shader::OutputVertex output_vertex = shader_unit.output_registers.ToVertex(regs.vs); @@ -283,7 +284,8 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { if (g_debug_context) g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, (void*)&input); - g_state.vs.Run(shader_unit, input, loader.GetNumTotalAttributes()); + shader_unit.LoadInputVertex(input, loader.GetNumTotalAttributes()); + g_state.vs.Run(shader_unit); // Retrieve vertex from register data output_vertex = shader_unit.output_registers.ToVertex(regs.vs); -- cgit v1.2.3 From 03d2b6aa893106a1945dddf81968cb6ff52c822d Mon Sep 17 00:00:00 2001 From: Yuri Kunde Schlesner Date: Fri, 16 Dec 2016 22:30:00 -0800 Subject: VideoCore/Shader: Use only entry_point as ShaderSetup param This removes all implicit dependency of ShaderState on global PICA state. --- .../debugger/graphics/graphics_vertex_shader.cpp | 2 +- src/video_core/command_processor.cpp | 4 ++-- src/video_core/shader/shader.cpp | 16 +++++++++------- src/video_core/shader/shader.h | 4 ++-- 4 files changed, 14 insertions(+), 12 deletions(-) (limited to 'src/video_core/command_processor.cpp') diff --git a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp index 89512146ea..c556d3b151 100644 --- a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp +++ b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp @@ -518,7 +518,7 @@ void GraphicsVertexShaderWidget::Reload(bool replace_vertex_data, void* vertex_d info.labels.insert({entry_point, "main"}); // Generate debug information - debug_data = shader_setup.ProduceDebugInfo(input_vertex, num_attributes, shader_config); + debug_data = shader_setup.ProduceDebugInfo(input_vertex, num_attributes, entry_point); // Reload widget state for (int attr = 0; attr < num_attributes; ++attr) { diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp index 36f72393b9..fc224c6f20 100644 --- a/src/video_core/command_processor.cpp +++ b/src/video_core/command_processor.cpp @@ -150,7 +150,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, static_cast(&immediate_input)); shader_unit.LoadInputVertex(immediate_input, regs.vs.num_input_attributes + 1); - g_state.vs.Run(shader_unit); + g_state.vs.Run(shader_unit, regs.vs.main_offset); Shader::OutputVertex output_vertex = shader_unit.output_registers.ToVertex(regs.vs); @@ -285,7 +285,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, (void*)&input); shader_unit.LoadInputVertex(input, loader.GetNumTotalAttributes()); - g_state.vs.Run(shader_unit); + g_state.vs.Run(shader_unit, regs.vs.main_offset); // Retrieve vertex from register data output_vertex = shader_unit.output_registers.ToVertex(regs.vs); diff --git a/src/video_core/shader/shader.cpp b/src/video_core/shader/shader.cpp index 868be13604..936db05826 100644 --- a/src/video_core/shader/shader.cpp +++ b/src/video_core/shader/shader.cpp @@ -120,33 +120,35 @@ void ShaderSetup::Setup() { MICROPROFILE_DEFINE(GPU_Shader, "GPU", "Shader", MP_RGB(50, 50, 240)); -void ShaderSetup::Run(UnitState& state) { - auto& config = g_state.regs.vs; +void ShaderSetup::Run(UnitState& state, unsigned int entry_point) { + ASSERT(entry_point < 1024); MICROPROFILE_SCOPE(GPU_Shader); #ifdef ARCHITECTURE_x86_64 if (VideoCore::g_shader_jit_enabled) { - jit_shader->Run(*this, state, config.main_offset); + jit_shader->Run(*this, state, entry_point); } else { DebugData dummy_debug_data; - RunInterpreter(*this, state, dummy_debug_data, config.main_offset); + RunInterpreter(*this, state, dummy_debug_data, entry_point); } #else DebugData dummy_debug_data; - RunInterpreter(*this, state, dummy_debug_data, config.main_offset); + RunInterpreter(*this, state, dummy_debug_data, entry_point); #endif // ARCHITECTURE_x86_64 } DebugData ShaderSetup::ProduceDebugInfo(const InputVertex& input, int num_attributes, - const Regs::ShaderConfig& config) { + unsigned int entry_point) { + ASSERT(entry_point < 1024); + UnitState state; DebugData debug_data; // Setup input register table boost::fill(state.registers.input, Math::Vec4::AssignToAll(float24::Zero())); state.LoadInputVertex(input, num_attributes); - RunInterpreter(*this, state, debug_data, config.main_offset); + RunInterpreter(*this, state, debug_data, entry_point); return debug_data; } diff --git a/src/video_core/shader/shader.h b/src/video_core/shader/shader.h index 61becb6e5f..d21f481abb 100644 --- a/src/video_core/shader/shader.h +++ b/src/video_core/shader/shader.h @@ -191,7 +191,7 @@ struct ShaderSetup { * Runs the currently setup shader * @param state Shader unit state, must be setup per shader and per shader unit */ - void Run(UnitState& state); + void Run(UnitState& state, unsigned int entry_point); /** * Produce debug information based on the given shader and input vertex @@ -201,7 +201,7 @@ struct ShaderSetup { * @return Debug information for this shader with regards to the given vertex */ DebugData ProduceDebugInfo(const InputVertex& input, int num_attributes, - const Regs::ShaderConfig& config); + unsigned int entry_point); }; } // namespace Shader -- cgit v1.2.3 From 73e7bc2592b0c8f27aa9c55cbf35805189a666d8 Mon Sep 17 00:00:00 2001 From: Yuri Kunde Schlesner Date: Fri, 16 Dec 2016 23:21:26 -0800 Subject: VideoCore/Shader: Split shader uniform state and shader engine Currently there's only a single dummy implementation, which will be split in a following commit. --- .../debugger/graphics/graphics_vertex_shader.cpp | 4 +- src/video_core/command_processor.cpp | 13 ++++--- src/video_core/shader/shader.cpp | 44 ++++++++++++++++------ src/video_core/shader/shader.h | 17 ++++++--- src/video_core/shader/shader_interpreter.h | 1 + 5 files changed, 57 insertions(+), 22 deletions(-) (limited to 'src/video_core/command_processor.cpp') diff --git a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp index c556d3b151..7adc3ad143 100644 --- a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp +++ b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp @@ -518,7 +518,9 @@ void GraphicsVertexShaderWidget::Reload(bool replace_vertex_data, void* vertex_d info.labels.insert({entry_point, "main"}); // Generate debug information - debug_data = shader_setup.ProduceDebugInfo(input_vertex, num_attributes, entry_point); + auto* shader_engine = Pica::Shader::GetEngine(); + shader_engine->SetupBatch(&shader_setup); + debug_data = shader_engine->ProduceDebugInfo(input_vertex, num_attributes, entry_point); // Reload widget state for (int attr = 0; attr < num_attributes; ++attr) { diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp index fc224c6f20..694c9f169d 100644 --- a/src/video_core/command_processor.cpp +++ b/src/video_core/command_processor.cpp @@ -142,15 +142,16 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { MICROPROFILE_SCOPE(GPU_Drawing); immediate_attribute_id = 0; - Shader::UnitState shader_unit; - g_state.vs.Setup(); + auto* shader_engine = Shader::GetEngine(); + shader_engine->SetupBatch(&g_state.vs); // Send to vertex shader if (g_debug_context) g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, static_cast(&immediate_input)); + Shader::UnitState shader_unit; shader_unit.LoadInputVertex(immediate_input, regs.vs.num_input_attributes + 1); - g_state.vs.Run(shader_unit, regs.vs.main_offset); + shader_engine->Run(shader_unit, regs.vs.main_offset); Shader::OutputVertex output_vertex = shader_unit.output_registers.ToVertex(regs.vs); @@ -244,8 +245,10 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { unsigned int vertex_cache_pos = 0; vertex_cache_ids.fill(-1); + auto* shader_engine = Shader::GetEngine(); Shader::UnitState shader_unit; - g_state.vs.Setup(); + + shader_engine->SetupBatch(&g_state.vs); for (unsigned int index = 0; index < regs.num_vertices; ++index) { // Indexed rendering doesn't use the start offset @@ -285,7 +288,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, (void*)&input); shader_unit.LoadInputVertex(input, loader.GetNumTotalAttributes()); - g_state.vs.Run(shader_unit, regs.vs.main_offset); + shader_engine->Run(shader_unit, regs.vs.main_offset); // Retrieve vertex from register data output_vertex = shader_unit.output_registers.ToVertex(regs.vs); diff --git a/src/video_core/shader/shader.cpp b/src/video_core/shader/shader.cpp index ae696533fb..d276a1221b 100644 --- a/src/video_core/shader/shader.cpp +++ b/src/video_core/shader/shader.cpp @@ -87,6 +87,17 @@ void UnitState::LoadInputVertex(const InputVertex& input, int num_attributes) { conditional_code[1] = false; } +class MergedShaderEngine : public ShaderEngine { +public: + void SetupBatch(const ShaderSetup* setup) override; + void Run(UnitState& state, unsigned int entry_point) const override; + DebugData ProduceDebugInfo(const InputVertex& input, int num_attributes, + unsigned int entry_point) const override; + +private: + const ShaderSetup* setup = nullptr; +}; + #ifdef ARCHITECTURE_x86_64 static std::unordered_map> shader_map; static const JitShader* jit_shader; @@ -98,13 +109,17 @@ void ClearCache() { #endif // ARCHITECTURE_x86_64 } -void ShaderSetup::Setup() { +void MergedShaderEngine::SetupBatch(const ShaderSetup* setup_) { + setup = setup_; + if (setup == nullptr) + return; + #ifdef ARCHITECTURE_x86_64 if (VideoCore::g_shader_jit_enabled) { - u64 cache_key = - Common::ComputeHash64(&program_code, sizeof(program_code)) ^ - Common::ComputeHash64(&swizzle_data, sizeof(swizzle_data)); + u64 code_hash = Common::ComputeHash64(&setup->program_code, sizeof(setup->program_code)); + u64 swizzle_hash = Common::ComputeHash64(&setup->swizzle_data, sizeof(setup->swizzle_data)); + u64 cache_key = code_hash ^ swizzle_hash; auto iter = shader_map.find(cache_key); if (iter != shader_map.end()) { jit_shader = iter->second.get(); @@ -120,26 +135,28 @@ void ShaderSetup::Setup() { MICROPROFILE_DEFINE(GPU_Shader, "GPU", "Shader", MP_RGB(50, 50, 240)); -void ShaderSetup::Run(UnitState& state, unsigned int entry_point) const { +void MergedShaderEngine::Run(UnitState& state, unsigned int entry_point) const { + ASSERT(setup != nullptr); ASSERT(entry_point < 1024); MICROPROFILE_SCOPE(GPU_Shader); #ifdef ARCHITECTURE_x86_64 if (VideoCore::g_shader_jit_enabled) { - jit_shader->Run(*this, state, entry_point); + jit_shader->Run(*setup, state, entry_point); } else { DebugData dummy_debug_data; - RunInterpreter(*this, state, dummy_debug_data, entry_point); + RunInterpreter(*setup, state, dummy_debug_data, entry_point); } #else DebugData dummy_debug_data; - RunInterpreter(*this, state, dummy_debug_data, entry_point); + RunInterpreter(*setup, state, dummy_debug_data, entry_point); #endif // ARCHITECTURE_x86_64 } -DebugData ShaderSetup::ProduceDebugInfo(const InputVertex& input, int num_attributes, - unsigned int entry_point) const { +DebugData MergedShaderEngine::ProduceDebugInfo(const InputVertex& input, int num_attributes, + unsigned int entry_point) const { + ASSERT(setup != nullptr); ASSERT(entry_point < 1024); UnitState state; @@ -148,10 +165,15 @@ DebugData ShaderSetup::ProduceDebugInfo(const InputVertex& input, int num_ // Setup input register table boost::fill(state.registers.input, Math::Vec4::AssignToAll(float24::Zero())); state.LoadInputVertex(input, num_attributes); - RunInterpreter(*this, state, debug_data, entry_point); + RunInterpreter(*setup, state, debug_data, entry_point); return debug_data; } +ShaderEngine* GetEngine() { + static MergedShaderEngine merged_engine; + return &merged_engine; +} + } // namespace Shader } // namespace Pica diff --git a/src/video_core/shader/shader.h b/src/video_core/shader/shader.h index 44b9861e99..899fb26071 100644 --- a/src/video_core/shader/shader.h +++ b/src/video_core/shader/shader.h @@ -156,7 +156,6 @@ struct UnitState { void ClearCache(); struct ShaderSetup { - struct { // The float uniforms are accessed by the shader JIT using SSE instructions, and are // therefore required to be 16-byte aligned. @@ -180,18 +179,23 @@ struct ShaderSetup { std::array program_code; std::array swizzle_data; +}; + +class ShaderEngine { +public: + virtual ~ShaderEngine() = default; /** * Performs any shader unit setup that only needs to happen once per shader (as opposed to once * per vertex, which would happen within the `Run` function). */ - void Setup(); + virtual void SetupBatch(const ShaderSetup* setup) = 0; /** * Runs the currently setup shader * @param state Shader unit state, must be setup per shader and per shader unit */ - void Run(UnitState& state, unsigned int entry_point) const; + virtual void Run(UnitState& state, unsigned int entry_point) const = 0; /** * Produce debug information based on the given shader and input vertex @@ -200,10 +204,13 @@ struct ShaderSetup { * @param config Configuration object for the shader pipeline * @return Debug information for this shader with regards to the given vertex */ - DebugData ProduceDebugInfo(const InputVertex& input, int num_attributes, - unsigned int entry_point) const; + virtual DebugData ProduceDebugInfo(const InputVertex& input, int num_attributes, + unsigned int entry_point) const = 0; }; +// TODO(yuriks): Remove and make it non-global state somewhere +ShaderEngine* GetEngine(); + } // namespace Shader } // namespace Pica diff --git a/src/video_core/shader/shader_interpreter.h b/src/video_core/shader/shader_interpreter.h index d31dcd7a68..3237b50b36 100644 --- a/src/video_core/shader/shader_interpreter.h +++ b/src/video_core/shader/shader_interpreter.h @@ -8,6 +8,7 @@ namespace Pica { namespace Shader { +struct ShaderSetup; struct UnitState; template -- cgit v1.2.3 From c863fc36ce259e11da48cf01ee7887812bbe4326 Mon Sep 17 00:00:00 2001 From: Yuri Kunde Schlesner Date: Sat, 17 Dec 2016 14:38:03 -0800 Subject: Shader: Remove OutputRegisters struct --- src/video_core/command_processor.cpp | 7 ++++--- src/video_core/shader/shader.cpp | 11 ++++++----- src/video_core/shader/shader.h | 17 +++++------------ src/video_core/shader/shader_interpreter.cpp | 4 ++-- 4 files changed, 17 insertions(+), 22 deletions(-) (limited to 'src/video_core/command_processor.cpp') diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp index 694c9f169d..66d19cba03 100644 --- a/src/video_core/command_processor.cpp +++ b/src/video_core/command_processor.cpp @@ -152,8 +152,8 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { Shader::UnitState shader_unit; shader_unit.LoadInputVertex(immediate_input, regs.vs.num_input_attributes + 1); shader_engine->Run(shader_unit, regs.vs.main_offset); - Shader::OutputVertex output_vertex = - shader_unit.output_registers.ToVertex(regs.vs); + auto output_vertex = Shader::OutputVertex::FromRegisters( + shader_unit.registers.output, regs, regs.vs.output_mask); // Send to renderer using Pica::Shader::OutputVertex; @@ -291,7 +291,8 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { shader_engine->Run(shader_unit, regs.vs.main_offset); // Retrieve vertex from register data - output_vertex = shader_unit.output_registers.ToVertex(regs.vs); + output_vertex = Shader::OutputVertex::FromRegisters(shader_unit.registers.output, + regs, regs.vs.output_mask); if (is_indexed) { vertex_cache[vertex_cache_pos] = output_vertex; diff --git a/src/video_core/shader/shader.cpp b/src/video_core/shader/shader.cpp index 1662b5d387..2da50bd620 100644 --- a/src/video_core/shader/shader.cpp +++ b/src/video_core/shader/shader.cpp @@ -19,7 +19,8 @@ namespace Pica { namespace Shader { -OutputVertex OutputRegisters::ToVertex(const Regs::ShaderConfig& config) const { +OutputVertex OutputVertex::FromRegisters(Math::Vec4 output_regs[16], const Regs& regs, + u32 output_mask) { // Setup output data OutputVertex ret; // TODO(neobrain): Under some circumstances, up to 16 attributes may be output. We need to @@ -27,13 +28,13 @@ OutputVertex OutputRegisters::ToVertex(const Regs::ShaderConfig& config) const { unsigned index = 0; for (unsigned i = 0; i < 7; ++i) { - if (index >= g_state.regs.vs_output_total) + if (index >= regs.vs_output_total) break; - if ((config.output_mask & (1 << i)) == 0) + if ((output_mask & (1 << i)) == 0) continue; - const auto& output_register_map = g_state.regs.vs_output_attributes[index]; + const auto& output_register_map = regs.vs_output_attributes[index]; u32 semantics[4] = {output_register_map.map_x, output_register_map.map_y, output_register_map.map_z, output_register_map.map_w}; @@ -41,7 +42,7 @@ OutputVertex OutputRegisters::ToVertex(const Regs::ShaderConfig& config) const { for (unsigned comp = 0; comp < 4; ++comp) { float24* out = ((float24*)&ret) + semantics[comp]; if (semantics[comp] != Regs::VSOutputAttributes::INVALID) { - *out = value[i][comp]; + *out = output_regs[i][comp]; } else { // Zero output so that attributes which aren't output won't have denormals in them, // which would slow us down later. diff --git a/src/video_core/shader/shader.h b/src/video_core/shader/shader.h index 9d24104876..7d51d0044b 100644 --- a/src/video_core/shader/shader.h +++ b/src/video_core/shader/shader.h @@ -73,19 +73,13 @@ struct OutputVertex { ret.Lerp(factor, v1); return ret; } + + static OutputVertex FromRegisters(Math::Vec4 output_regs[16], const Regs& regs, + u32 output_mask); }; static_assert(std::is_pod::value, "Structure is not POD"); static_assert(sizeof(OutputVertex) == 32 * sizeof(float), "OutputVertex has invalid size"); -struct OutputRegisters { - OutputRegisters() = default; - - alignas(16) Math::Vec4 value[16]; - - OutputVertex ToVertex(const Regs::ShaderConfig& config) const; -}; -static_assert(std::is_pod::value, "Structure is not POD"); - /** * This structure contains the state information that needs to be unique for a shader unit. The 3DS * has four shader units that process shaders in parallel. At the present, Citra only implements a @@ -98,11 +92,10 @@ struct UnitState { // required to be 16-byte aligned. alignas(16) Math::Vec4 input[16]; alignas(16) Math::Vec4 temporary[16]; + alignas(16) Math::Vec4 output[16]; } registers; static_assert(std::is_pod::value, "Structure is not POD"); - OutputRegisters output_registers; - bool conditional_code[2]; // Two Address registers and one loop counter @@ -128,7 +121,7 @@ struct UnitState { static size_t OutputOffset(const DestRegister& reg) { switch (reg.GetRegisterType()) { case RegisterType::Output: - return offsetof(UnitState, output_registers.value) + + return offsetof(UnitState, registers.output) + reg.GetIndex() * sizeof(Math::Vec4); case RegisterType::Temporary: diff --git a/src/video_core/shader/shader_interpreter.cpp b/src/video_core/shader/shader_interpreter.cpp index ecc2270891..a6197c10a4 100644 --- a/src/video_core/shader/shader_interpreter.cpp +++ b/src/video_core/shader/shader_interpreter.cpp @@ -175,7 +175,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData float24* dest = (instr.common.dest.Value() < 0x10) - ? &state.output_registers.value[instr.common.dest.Value().GetIndex()][0] + ? &state.registers.output[instr.common.dest.Value().GetIndex()][0] : (instr.common.dest.Value() < 0x20) ? &state.registers.temporary[instr.common.dest.Value().GetIndex()][0] : dummy_vec4_float24; @@ -518,7 +518,7 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData float24* dest = (instr.mad.dest.Value() < 0x10) - ? &state.output_registers.value[instr.mad.dest.Value().GetIndex()][0] + ? &state.registers.output[instr.mad.dest.Value().GetIndex()][0] : (instr.mad.dest.Value() < 0x20) ? &state.registers.temporary[instr.mad.dest.Value().GetIndex()][0] : dummy_vec4_float24; -- cgit v1.2.3 From 1994f1a6f27d6dcdbf6550e3dcea19c41216e4e2 Mon Sep 17 00:00:00 2001 From: Yuri Kunde Schlesner Date: Sat, 17 Dec 2016 16:06:04 -0800 Subject: VideoCore/Shader: Move per-batch ShaderEngine state into ShaderSetup --- .../debugger/graphics/graphics_vertex_shader.cpp | 5 +++-- src/video_core/command_processor.cpp | 8 +++---- src/video_core/shader/shader.h | 17 +++++++++++---- src/video_core/shader/shader_interpreter.cpp | 16 ++++++-------- src/video_core/shader/shader_interpreter.h | 11 ++++------ src/video_core/shader/shader_jit_x64.cpp | 25 +++++++++------------- src/video_core/shader/shader_jit_x64.h | 7 ++---- 7 files changed, 43 insertions(+), 46 deletions(-) (limited to 'src/video_core/command_processor.cpp') diff --git a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp index c6f807eb3f..616b34d568 100644 --- a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp +++ b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp @@ -521,8 +521,9 @@ void GraphicsVertexShaderWidget::Reload(bool replace_vertex_data, void* vertex_d // Generate debug information Pica::Shader::InterpreterEngine shader_engine; - shader_engine.SetupBatch(&shader_setup); - debug_data = shader_engine.ProduceDebugInfo(input_vertex, num_attributes, entry_point); + shader_engine.SetupBatch(shader_setup); + debug_data = + shader_engine.ProduceDebugInfo(shader_setup, input_vertex, num_attributes, entry_point); // Reload widget state for (int attr = 0; attr < num_attributes; ++attr) { diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp index 66d19cba03..c3872d06ca 100644 --- a/src/video_core/command_processor.cpp +++ b/src/video_core/command_processor.cpp @@ -143,7 +143,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { immediate_attribute_id = 0; auto* shader_engine = Shader::GetEngine(); - shader_engine->SetupBatch(&g_state.vs); + shader_engine->SetupBatch(g_state.vs); // Send to vertex shader if (g_debug_context) @@ -151,7 +151,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { static_cast(&immediate_input)); Shader::UnitState shader_unit; shader_unit.LoadInputVertex(immediate_input, regs.vs.num_input_attributes + 1); - shader_engine->Run(shader_unit, regs.vs.main_offset); + shader_engine->Run(g_state.vs, shader_unit, regs.vs.main_offset); auto output_vertex = Shader::OutputVertex::FromRegisters( shader_unit.registers.output, regs, regs.vs.output_mask); @@ -248,7 +248,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { auto* shader_engine = Shader::GetEngine(); Shader::UnitState shader_unit; - shader_engine->SetupBatch(&g_state.vs); + shader_engine->SetupBatch(g_state.vs); for (unsigned int index = 0; index < regs.num_vertices; ++index) { // Indexed rendering doesn't use the start offset @@ -288,7 +288,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, (void*)&input); shader_unit.LoadInputVertex(input, loader.GetNumTotalAttributes()); - shader_engine->Run(shader_unit, regs.vs.main_offset); + shader_engine->Run(g_state.vs, shader_unit, regs.vs.main_offset); // Retrieve vertex from register data output_vertex = Shader::OutputVertex::FromRegisters(shader_unit.registers.output, diff --git a/src/video_core/shader/shader.h b/src/video_core/shader/shader.h index 7d51d0044b..f26d2ba4f2 100644 --- a/src/video_core/shader/shader.h +++ b/src/video_core/shader/shader.h @@ -167,6 +167,12 @@ struct ShaderSetup { std::array program_code; std::array swizzle_data; + + /// Data private to ShaderEngines + struct EngineData { + /// Used by the JIT, points to a compiled shader object. + const void* cached_shader = nullptr; + } engine_data; }; class ShaderEngine { @@ -177,13 +183,16 @@ public: * Performs any shader unit setup that only needs to happen once per shader (as opposed to once * per vertex, which would happen within the `Run` function). */ - virtual void SetupBatch(const ShaderSetup* setup) = 0; + virtual void SetupBatch(ShaderSetup& setup) = 0; /** - * Runs the currently setup shader - * @param state Shader unit state, must be setup per shader and per shader unit + * Runs the currently setup shader. + * + * @param setup Shader engine state, must be setup with SetupBatch on each shader change. + * @param state Shader unit state, must be setup with input data before each shader invocation. */ - virtual void Run(UnitState& state, unsigned int entry_point) const = 0; + virtual void Run(const ShaderSetup& setup, UnitState& state, + unsigned int entry_point) const = 0; }; // TODO(yuriks): Remove and make it non-global state somewhere diff --git a/src/video_core/shader/shader_interpreter.cpp b/src/video_core/shader/shader_interpreter.cpp index a6197c10a4..e44abbf1d3 100644 --- a/src/video_core/shader/shader_interpreter.cpp +++ b/src/video_core/shader/shader_interpreter.cpp @@ -652,25 +652,23 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData } } -void InterpreterEngine::SetupBatch(const ShaderSetup* setup_) { - setup = setup_; -} +void InterpreterEngine::SetupBatch(ShaderSetup& setup) {} MICROPROFILE_DECLARE(GPU_Shader); -void InterpreterEngine::Run(UnitState& state, unsigned int entry_point) const { - ASSERT(setup != nullptr); +void InterpreterEngine::Run(const ShaderSetup& setup, UnitState& state, + unsigned int entry_point) const { ASSERT(entry_point < 1024); MICROPROFILE_SCOPE(GPU_Shader); DebugData dummy_debug_data; - RunInterpreter(*setup, state, dummy_debug_data, entry_point); + RunInterpreter(setup, state, dummy_debug_data, entry_point); } -DebugData InterpreterEngine::ProduceDebugInfo(const InputVertex& input, int num_attributes, +DebugData InterpreterEngine::ProduceDebugInfo(const ShaderSetup& setup, + const InputVertex& input, int num_attributes, unsigned int entry_point) const { - ASSERT(setup != nullptr); ASSERT(entry_point < 1024); UnitState state; @@ -679,7 +677,7 @@ DebugData InterpreterEngine::ProduceDebugInfo(const InputVertex& input, in // Setup input register table boost::fill(state.registers.input, Math::Vec4::AssignToAll(float24::Zero())); state.LoadInputVertex(input, num_attributes); - RunInterpreter(*setup, state, debug_data, entry_point); + RunInterpreter(setup, state, debug_data, entry_point); return debug_data; } diff --git a/src/video_core/shader/shader_interpreter.h b/src/video_core/shader/shader_interpreter.h index c3691da700..7f94d405f7 100644 --- a/src/video_core/shader/shader_interpreter.h +++ b/src/video_core/shader/shader_interpreter.h @@ -13,8 +13,8 @@ namespace Shader { class InterpreterEngine final : public ShaderEngine { public: - void SetupBatch(const ShaderSetup* setup) override; - void Run(UnitState& state, unsigned int entry_point) const override; + void SetupBatch(ShaderSetup& setup) override; + void Run(const ShaderSetup& setup, UnitState& state, unsigned int entry_point) const override; /** * Produce debug information based on the given shader and input vertex @@ -23,11 +23,8 @@ public: * @param config Configuration object for the shader pipeline * @return Debug information for this shader with regards to the given vertex */ - DebugData ProduceDebugInfo(const InputVertex& input, int num_attributes, - unsigned int entry_point) const; - -private: - const ShaderSetup* setup = nullptr; + DebugData ProduceDebugInfo(const ShaderSetup& setup, const InputVertex& input, + int num_attributes, unsigned int entry_point) const; }; } // namespace diff --git a/src/video_core/shader/shader_jit_x64.cpp b/src/video_core/shader/shader_jit_x64.cpp index 755ae119fa..15c1d60b51 100644 --- a/src/video_core/shader/shader_jit_x64.cpp +++ b/src/video_core/shader/shader_jit_x64.cpp @@ -14,37 +14,32 @@ namespace Shader { JitX64Engine::JitX64Engine() = default; JitX64Engine::~JitX64Engine() = default; -void JitX64Engine::SetupBatch(const ShaderSetup* setup_) { - cached_shader = nullptr; - setup = setup_; - if (setup == nullptr) - return; - - u64 code_hash = Common::ComputeHash64(&setup->program_code, sizeof(setup->program_code)); - u64 swizzle_hash = Common::ComputeHash64(&setup->swizzle_data, sizeof(setup->swizzle_data)); +void JitX64Engine::SetupBatch(ShaderSetup& setup) { + u64 code_hash = Common::ComputeHash64(&setup.program_code, sizeof(setup.program_code)); + u64 swizzle_hash = Common::ComputeHash64(&setup.swizzle_data, sizeof(setup.swizzle_data)); u64 cache_key = code_hash ^ swizzle_hash; auto iter = cache.find(cache_key); if (iter != cache.end()) { - cached_shader = iter->second.get(); + setup.engine_data.cached_shader = iter->second.get(); } else { auto shader = std::make_unique(); - shader->Compile(&setup->program_code, &setup->swizzle_data); - cached_shader = shader.get(); + shader->Compile(&setup.program_code, &setup.swizzle_data); + setup.engine_data.cached_shader = shader.get(); cache.emplace_hint(iter, cache_key, std::move(shader)); } } MICROPROFILE_DECLARE(GPU_Shader); -void JitX64Engine::Run(UnitState& state, unsigned int entry_point) const { - ASSERT(setup != nullptr); - ASSERT(cached_shader != nullptr); +void JitX64Engine::Run(const ShaderSetup& setup, UnitState& state, unsigned int entry_point) const { + ASSERT(setup.engine_data.cached_shader != nullptr); ASSERT(entry_point < 1024); MICROPROFILE_SCOPE(GPU_Shader); - cached_shader->Run(*setup, state, entry_point); + const JitShader* shader = static_cast(setup.engine_data.cached_shader); + shader->Run(setup, state, entry_point); } } // namespace Shader diff --git a/src/video_core/shader/shader_jit_x64.h b/src/video_core/shader/shader_jit_x64.h index b260444774..bd30f51e24 100644 --- a/src/video_core/shader/shader_jit_x64.h +++ b/src/video_core/shader/shader_jit_x64.h @@ -19,14 +19,11 @@ public: JitX64Engine(); ~JitX64Engine() override; - void SetupBatch(const ShaderSetup* setup) override; - void Run(UnitState& state, unsigned int entry_point) const override; + void SetupBatch(ShaderSetup& setup) override; + void Run(const ShaderSetup& setup, UnitState& state, unsigned int entry_point) const override; private: - const ShaderSetup* setup = nullptr; - std::unordered_map> cache; - const JitShader* cached_shader = nullptr; }; } // namespace Shader -- cgit v1.2.3 From ab0b74b0f35be9cb2d44857cafd40cc11876697d Mon Sep 17 00:00:00 2001 From: Yuri Kunde Schlesner Date: Sat, 17 Dec 2016 16:16:02 -0800 Subject: VideoCore/Shader: Move entry_point to SetupBatch --- .../debugger/graphics/graphics_vertex_shader.cpp | 5 ++--- src/video_core/command_processor.cpp | 8 ++++---- src/video_core/shader/shader.h | 6 +++--- src/video_core/shader/shader_interpreter.cpp | 19 +++++++++---------- src/video_core/shader/shader_interpreter.h | 6 +++--- src/video_core/shader/shader_jit_x64.cpp | 10 ++++++---- src/video_core/shader/shader_jit_x64.h | 4 ++-- 7 files changed, 29 insertions(+), 29 deletions(-) (limited to 'src/video_core/command_processor.cpp') diff --git a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp index 616b34d568..f375241908 100644 --- a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp +++ b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp @@ -521,9 +521,8 @@ void GraphicsVertexShaderWidget::Reload(bool replace_vertex_data, void* vertex_d // Generate debug information Pica::Shader::InterpreterEngine shader_engine; - shader_engine.SetupBatch(shader_setup); - debug_data = - shader_engine.ProduceDebugInfo(shader_setup, input_vertex, num_attributes, entry_point); + shader_engine.SetupBatch(shader_setup, entry_point); + debug_data = shader_engine.ProduceDebugInfo(shader_setup, input_vertex, num_attributes); // Reload widget state for (int attr = 0; attr < num_attributes; ++attr) { diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp index c3872d06ca..eb79974a8a 100644 --- a/src/video_core/command_processor.cpp +++ b/src/video_core/command_processor.cpp @@ -143,7 +143,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { immediate_attribute_id = 0; auto* shader_engine = Shader::GetEngine(); - shader_engine->SetupBatch(g_state.vs); + shader_engine->SetupBatch(g_state.vs, regs.vs.main_offset); // Send to vertex shader if (g_debug_context) @@ -151,7 +151,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { static_cast(&immediate_input)); Shader::UnitState shader_unit; shader_unit.LoadInputVertex(immediate_input, regs.vs.num_input_attributes + 1); - shader_engine->Run(g_state.vs, shader_unit, regs.vs.main_offset); + shader_engine->Run(g_state.vs, shader_unit); auto output_vertex = Shader::OutputVertex::FromRegisters( shader_unit.registers.output, regs, regs.vs.output_mask); @@ -248,7 +248,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { auto* shader_engine = Shader::GetEngine(); Shader::UnitState shader_unit; - shader_engine->SetupBatch(g_state.vs); + shader_engine->SetupBatch(g_state.vs, regs.vs.main_offset); for (unsigned int index = 0; index < regs.num_vertices; ++index) { // Indexed rendering doesn't use the start offset @@ -288,7 +288,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, (void*)&input); shader_unit.LoadInputVertex(input, loader.GetNumTotalAttributes()); - shader_engine->Run(g_state.vs, shader_unit, regs.vs.main_offset); + shader_engine->Run(g_state.vs, shader_unit); // Retrieve vertex from register data output_vertex = Shader::OutputVertex::FromRegisters(shader_unit.registers.output, diff --git a/src/video_core/shader/shader.h b/src/video_core/shader/shader.h index f26d2ba4f2..44d9f76c3a 100644 --- a/src/video_core/shader/shader.h +++ b/src/video_core/shader/shader.h @@ -170,6 +170,7 @@ struct ShaderSetup { /// Data private to ShaderEngines struct EngineData { + unsigned int entry_point; /// Used by the JIT, points to a compiled shader object. const void* cached_shader = nullptr; } engine_data; @@ -183,7 +184,7 @@ public: * Performs any shader unit setup that only needs to happen once per shader (as opposed to once * per vertex, which would happen within the `Run` function). */ - virtual void SetupBatch(ShaderSetup& setup) = 0; + virtual void SetupBatch(ShaderSetup& setup, unsigned int entry_point) = 0; /** * Runs the currently setup shader. @@ -191,8 +192,7 @@ public: * @param setup Shader engine state, must be setup with SetupBatch on each shader change. * @param state Shader unit state, must be setup with input data before each shader invocation. */ - virtual void Run(const ShaderSetup& setup, UnitState& state, - unsigned int entry_point) const = 0; + virtual void Run(const ShaderSetup& setup, UnitState& state) const = 0; }; // TODO(yuriks): Remove and make it non-global state somewhere diff --git a/src/video_core/shader/shader_interpreter.cpp b/src/video_core/shader/shader_interpreter.cpp index e44abbf1d3..c0c89b8578 100644 --- a/src/video_core/shader/shader_interpreter.cpp +++ b/src/video_core/shader/shader_interpreter.cpp @@ -652,32 +652,31 @@ static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData } } -void InterpreterEngine::SetupBatch(ShaderSetup& setup) {} +void InterpreterEngine::SetupBatch(ShaderSetup& setup, unsigned int entry_point) { + ASSERT(entry_point < 1024); + setup.engine_data.entry_point = entry_point; +} MICROPROFILE_DECLARE(GPU_Shader); -void InterpreterEngine::Run(const ShaderSetup& setup, UnitState& state, - unsigned int entry_point) const { - ASSERT(entry_point < 1024); +void InterpreterEngine::Run(const ShaderSetup& setup, UnitState& state) const { MICROPROFILE_SCOPE(GPU_Shader); DebugData dummy_debug_data; - RunInterpreter(setup, state, dummy_debug_data, entry_point); + RunInterpreter(setup, state, dummy_debug_data, setup.engine_data.entry_point); } DebugData InterpreterEngine::ProduceDebugInfo(const ShaderSetup& setup, - const InputVertex& input, int num_attributes, - unsigned int entry_point) const { - ASSERT(entry_point < 1024); - + const InputVertex& input, + int num_attributes) const { UnitState state; DebugData debug_data; // Setup input register table boost::fill(state.registers.input, Math::Vec4::AssignToAll(float24::Zero())); state.LoadInputVertex(input, num_attributes); - RunInterpreter(setup, state, debug_data, entry_point); + RunInterpreter(setup, state, debug_data, setup.engine_data.entry_point); return debug_data; } diff --git a/src/video_core/shader/shader_interpreter.h b/src/video_core/shader/shader_interpreter.h index 7f94d405f7..d6c0e2d8c1 100644 --- a/src/video_core/shader/shader_interpreter.h +++ b/src/video_core/shader/shader_interpreter.h @@ -13,8 +13,8 @@ namespace Shader { class InterpreterEngine final : public ShaderEngine { public: - void SetupBatch(ShaderSetup& setup) override; - void Run(const ShaderSetup& setup, UnitState& state, unsigned int entry_point) const override; + void SetupBatch(ShaderSetup& setup, unsigned int entry_point) override; + void Run(const ShaderSetup& setup, UnitState& state) const override; /** * Produce debug information based on the given shader and input vertex @@ -24,7 +24,7 @@ public: * @return Debug information for this shader with regards to the given vertex */ DebugData ProduceDebugInfo(const ShaderSetup& setup, const InputVertex& input, - int num_attributes, unsigned int entry_point) const; + int num_attributes) const; }; } // namespace diff --git a/src/video_core/shader/shader_jit_x64.cpp b/src/video_core/shader/shader_jit_x64.cpp index 15c1d60b51..0ee0dd9ef7 100644 --- a/src/video_core/shader/shader_jit_x64.cpp +++ b/src/video_core/shader/shader_jit_x64.cpp @@ -14,7 +14,10 @@ namespace Shader { JitX64Engine::JitX64Engine() = default; JitX64Engine::~JitX64Engine() = default; -void JitX64Engine::SetupBatch(ShaderSetup& setup) { +void JitX64Engine::SetupBatch(ShaderSetup& setup, unsigned int entry_point) { + ASSERT(entry_point < 1024); + setup.engine_data.entry_point = entry_point; + u64 code_hash = Common::ComputeHash64(&setup.program_code, sizeof(setup.program_code)); u64 swizzle_hash = Common::ComputeHash64(&setup.swizzle_data, sizeof(setup.swizzle_data)); @@ -32,14 +35,13 @@ void JitX64Engine::SetupBatch(ShaderSetup& setup) { MICROPROFILE_DECLARE(GPU_Shader); -void JitX64Engine::Run(const ShaderSetup& setup, UnitState& state, unsigned int entry_point) const { +void JitX64Engine::Run(const ShaderSetup& setup, UnitState& state) const { ASSERT(setup.engine_data.cached_shader != nullptr); - ASSERT(entry_point < 1024); MICROPROFILE_SCOPE(GPU_Shader); const JitShader* shader = static_cast(setup.engine_data.cached_shader); - shader->Run(setup, state, entry_point); + shader->Run(setup, state, setup.engine_data.entry_point); } } // namespace Shader diff --git a/src/video_core/shader/shader_jit_x64.h b/src/video_core/shader/shader_jit_x64.h index bd30f51e24..078b2cba5d 100644 --- a/src/video_core/shader/shader_jit_x64.h +++ b/src/video_core/shader/shader_jit_x64.h @@ -19,8 +19,8 @@ public: JitX64Engine(); ~JitX64Engine() override; - void SetupBatch(ShaderSetup& setup) override; - void Run(const ShaderSetup& setup, UnitState& state, unsigned int entry_point) const override; + void SetupBatch(ShaderSetup& setup, unsigned int entry_point) override; + void Run(const ShaderSetup& setup, UnitState& state) const override; private: std::unordered_map> cache; -- cgit v1.2.3