diff options
Diffstat (limited to 'src/video_core')
68 files changed, 492 insertions, 184 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index ff53282c9..d23c53843 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -8,6 +8,7 @@ add_library(video_core STATIC dma_pusher.h engines/const_buffer_engine_interface.h engines/const_buffer_info.h + engines/engine_interface.h engines/engine_upload.cpp engines/engine_upload.h engines/fermi_2d.cpp diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index 510f11089..56e570994 100644 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -4,7 +4,6 @@ #pragma once -#include <array> #include <list> #include <memory> #include <mutex> @@ -89,10 +88,6 @@ public: map->MarkAsWritten(true); MarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1); } - } else { - if (map->IsWritten()) { - WriteBarrier(); - } } return {ToHandle(block), static_cast<u64>(block->GetOffset(cpu_addr))}; @@ -254,8 +249,6 @@ protected: virtual BufferType ToHandle(const OwnerBuffer& storage) = 0; - virtual void WriteBarrier() = 0; - virtual OwnerBuffer CreateBlock(VAddr cpu_addr, std::size_t size) = 0; virtual void UploadBlockData(const OwnerBuffer& buffer, std::size_t offset, std::size_t size, diff --git a/src/video_core/dma_pusher.cpp b/src/video_core/dma_pusher.cpp index 16311f05e..bdc023d54 100644 --- a/src/video_core/dma_pusher.cpp +++ b/src/video_core/dma_pusher.cpp @@ -27,6 +27,8 @@ void DmaPusher::DispatchCalls() { dma_pushbuffer_subindex = 0; + dma_state.is_last_call = true; + while (system.IsPoweredOn()) { if (!Step()) { break; @@ -82,9 +84,11 @@ bool DmaPusher::Step() { index); CallMultiMethod(&command_header.argument, max_write); dma_state.method_count -= max_write; + dma_state.is_last_call = true; index += max_write; continue; } else { + dma_state.is_last_call = dma_state.method_count <= 1; CallMethod(command_header.argument); } @@ -144,12 +148,22 @@ void DmaPusher::SetState(const CommandHeader& command_header) { } void DmaPusher::CallMethod(u32 argument) const { - gpu.CallMethod({dma_state.method, argument, dma_state.subchannel, dma_state.method_count}); + if (dma_state.method < non_puller_methods) { + gpu.CallMethod({dma_state.method, argument, dma_state.subchannel, dma_state.method_count}); + } else { + subchannels[dma_state.subchannel]->CallMethod(dma_state.method, argument, + dma_state.is_last_call); + } } void DmaPusher::CallMultiMethod(const u32* base_start, u32 num_methods) const { - gpu.CallMultiMethod(dma_state.method, dma_state.subchannel, base_start, num_methods, - dma_state.method_count); + if (dma_state.method < non_puller_methods) { + gpu.CallMultiMethod(dma_state.method, dma_state.subchannel, base_start, num_methods, + dma_state.method_count); + } else { + subchannels[dma_state.subchannel]->CallMultiMethod(dma_state.method, base_start, + num_methods, dma_state.method_count); + } } } // namespace Tegra diff --git a/src/video_core/dma_pusher.h b/src/video_core/dma_pusher.h index 6cef71306..e8b714e94 100644 --- a/src/video_core/dma_pusher.h +++ b/src/video_core/dma_pusher.h @@ -4,11 +4,13 @@ #pragma once +#include <array> #include <vector> #include <queue> #include "common/bit_field.h" #include "common/common_types.h" +#include "video_core/engines/engine_interface.h" namespace Core { class System; @@ -69,7 +71,13 @@ public: void DispatchCalls(); + void BindSubchannel(Tegra::Engines::EngineInterface* engine, u32 subchannel_id) { + subchannels[subchannel_id] = engine; + } + private: + static constexpr u32 non_puller_methods = 0x40; + static constexpr u32 max_subchannels = 8; bool Step(); void SetState(const CommandHeader& command_header); @@ -88,6 +96,7 @@ private: u32 method_count; ///< Current method count u32 length_pending; ///< Large NI command length pending bool non_incrementing; ///< Current command's NI flag + bool is_last_call; }; DmaState dma_state{}; @@ -96,6 +105,8 @@ private: GPUVAddr dma_mget{}; ///< main pushbuffer last read address bool ib_enable{true}; ///< IB mode enabled + std::array<Tegra::Engines::EngineInterface*, max_subchannels> subchannels{}; + GPU& gpu; Core::System& system; }; diff --git a/src/video_core/engines/engine_interface.h b/src/video_core/engines/engine_interface.h new file mode 100644 index 000000000..18a9db7e6 --- /dev/null +++ b/src/video_core/engines/engine_interface.h @@ -0,0 +1,22 @@ +// Copyright 2020 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include <type_traits> +#include "common/common_types.h" + +namespace Tegra::Engines { + +class EngineInterface { +public: + /// Write the value to the register identified by method. + virtual void CallMethod(u32 method, u32 method_argument, bool is_last_call) = 0; + + /// Write multiple values to the register identified by method. + virtual void CallMultiMethod(u32 method, const u32* base_start, u32 amount, + u32 methods_pending) = 0; +}; + +} // namespace Tegra::Engines diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp index 8a47614d2..ff10ff40d 100644 --- a/src/video_core/engines/fermi_2d.cpp +++ b/src/video_core/engines/fermi_2d.cpp @@ -12,13 +12,13 @@ namespace Tegra::Engines { Fermi2D::Fermi2D(VideoCore::RasterizerInterface& rasterizer) : rasterizer{rasterizer} {} -void Fermi2D::CallMethod(const GPU::MethodCall& method_call) { - ASSERT_MSG(method_call.method < Regs::NUM_REGS, +void Fermi2D::CallMethod(u32 method, u32 method_argument, bool is_last_call) { + ASSERT_MSG(method < Regs::NUM_REGS, "Invalid Fermi2D register, increase the size of the Regs structure"); - regs.reg_array[method_call.method] = method_call.argument; + regs.reg_array[method] = method_argument; - switch (method_call.method) { + switch (method) { // Trigger the surface copy on the last register write. This is blit_src_y, but this is 64-bit, // so trigger on the second 32-bit write. case FERMI2D_REG_INDEX(blit_src_y) + 1: { @@ -30,7 +30,7 @@ void Fermi2D::CallMethod(const GPU::MethodCall& method_call) { void Fermi2D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) { for (std::size_t i = 0; i < amount; i++) { - CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)}); + CallMethod(method, base_start[i], methods_pending - static_cast<u32>(i) <= 1); } } diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h index 939a5966d..8f37d053f 100644 --- a/src/video_core/engines/fermi_2d.h +++ b/src/video_core/engines/fermi_2d.h @@ -10,6 +10,7 @@ #include "common/common_funcs.h" #include "common/common_types.h" #include "common/math_util.h" +#include "video_core/engines/engine_interface.h" #include "video_core/gpu.h" namespace Tegra { @@ -31,16 +32,17 @@ namespace Tegra::Engines { #define FERMI2D_REG_INDEX(field_name) \ (offsetof(Tegra::Engines::Fermi2D::Regs, field_name) / sizeof(u32)) -class Fermi2D final { +class Fermi2D final : public EngineInterface { public: explicit Fermi2D(VideoCore::RasterizerInterface& rasterizer); ~Fermi2D() = default; /// Write the value to the register identified by method. - void CallMethod(const GPU::MethodCall& method_call); + void CallMethod(u32 method, u32 method_argument, bool is_last_call) override; /// Write multiple values to the register identified by method. - void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending); + void CallMultiMethod(u32 method, const u32* base_start, u32 amount, + u32 methods_pending) override; enum class Origin : u32 { Center = 0, diff --git a/src/video_core/engines/kepler_compute.cpp b/src/video_core/engines/kepler_compute.cpp index 00a12175f..f6237fc6a 100644 --- a/src/video_core/engines/kepler_compute.cpp +++ b/src/video_core/engines/kepler_compute.cpp @@ -24,20 +24,19 @@ KeplerCompute::KeplerCompute(Core::System& system, VideoCore::RasterizerInterfac KeplerCompute::~KeplerCompute() = default; -void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) { - ASSERT_MSG(method_call.method < Regs::NUM_REGS, +void KeplerCompute::CallMethod(u32 method, u32 method_argument, bool is_last_call) { + ASSERT_MSG(method < Regs::NUM_REGS, "Invalid KeplerCompute register, increase the size of the Regs structure"); - regs.reg_array[method_call.method] = method_call.argument; + regs.reg_array[method] = method_argument; - switch (method_call.method) { + switch (method) { case KEPLER_COMPUTE_REG_INDEX(exec_upload): { upload_state.ProcessExec(regs.exec_upload.linear != 0); break; } case KEPLER_COMPUTE_REG_INDEX(data_upload): { - const bool is_last_call = method_call.IsLastCall(); - upload_state.ProcessData(method_call.argument, is_last_call); + upload_state.ProcessData(method_argument, is_last_call); if (is_last_call) { system.GPU().Maxwell3D().OnMemoryWrite(); } @@ -54,7 +53,7 @@ void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) { void KeplerCompute::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) { for (std::size_t i = 0; i < amount; i++) { - CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)}); + CallMethod(method, base_start[i], methods_pending - static_cast<u32>(i) <= 1); } } diff --git a/src/video_core/engines/kepler_compute.h b/src/video_core/engines/kepler_compute.h index fe55fdfd0..18ceedfaf 100644 --- a/src/video_core/engines/kepler_compute.h +++ b/src/video_core/engines/kepler_compute.h @@ -11,6 +11,7 @@ #include "common/common_funcs.h" #include "common/common_types.h" #include "video_core/engines/const_buffer_engine_interface.h" +#include "video_core/engines/engine_interface.h" #include "video_core/engines/engine_upload.h" #include "video_core/engines/shader_type.h" #include "video_core/gpu.h" @@ -39,7 +40,7 @@ namespace Tegra::Engines { #define KEPLER_COMPUTE_REG_INDEX(field_name) \ (offsetof(Tegra::Engines::KeplerCompute::Regs, field_name) / sizeof(u32)) -class KeplerCompute final : public ConstBufferEngineInterface { +class KeplerCompute final : public ConstBufferEngineInterface, public EngineInterface { public: explicit KeplerCompute(Core::System& system, VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager); @@ -200,10 +201,11 @@ public: "KeplerCompute LaunchParams has wrong size"); /// Write the value to the register identified by method. - void CallMethod(const GPU::MethodCall& method_call); + void CallMethod(u32 method, u32 method_argument, bool is_last_call) override; /// Write multiple values to the register identified by method. - void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending); + void CallMultiMethod(u32 method, const u32* base_start, u32 amount, + u32 methods_pending) override; Texture::FullTextureInfo GetTexture(std::size_t offset) const; diff --git a/src/video_core/engines/kepler_memory.cpp b/src/video_core/engines/kepler_memory.cpp index 586ff15dc..dc71b2eec 100644 --- a/src/video_core/engines/kepler_memory.cpp +++ b/src/video_core/engines/kepler_memory.cpp @@ -19,20 +19,19 @@ KeplerMemory::KeplerMemory(Core::System& system, MemoryManager& memory_manager) KeplerMemory::~KeplerMemory() = default; -void KeplerMemory::CallMethod(const GPU::MethodCall& method_call) { - ASSERT_MSG(method_call.method < Regs::NUM_REGS, +void KeplerMemory::CallMethod(u32 method, u32 method_argument, bool is_last_call) { + ASSERT_MSG(method < Regs::NUM_REGS, "Invalid KeplerMemory register, increase the size of the Regs structure"); - regs.reg_array[method_call.method] = method_call.argument; + regs.reg_array[method] = method_argument; - switch (method_call.method) { + switch (method) { case KEPLERMEMORY_REG_INDEX(exec): { upload_state.ProcessExec(regs.exec.linear != 0); break; } case KEPLERMEMORY_REG_INDEX(data): { - const bool is_last_call = method_call.IsLastCall(); - upload_state.ProcessData(method_call.argument, is_last_call); + upload_state.ProcessData(method_argument, is_last_call); if (is_last_call) { system.GPU().Maxwell3D().OnMemoryWrite(); } @@ -44,7 +43,7 @@ void KeplerMemory::CallMethod(const GPU::MethodCall& method_call) { void KeplerMemory::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) { for (std::size_t i = 0; i < amount; i++) { - CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)}); + CallMethod(method, base_start[i], methods_pending - static_cast<u32>(i) <= 1); } } diff --git a/src/video_core/engines/kepler_memory.h b/src/video_core/engines/kepler_memory.h index bb26fb030..5b7f71a00 100644 --- a/src/video_core/engines/kepler_memory.h +++ b/src/video_core/engines/kepler_memory.h @@ -10,6 +10,7 @@ #include "common/bit_field.h" #include "common/common_funcs.h" #include "common/common_types.h" +#include "video_core/engines/engine_interface.h" #include "video_core/engines/engine_upload.h" #include "video_core/gpu.h" @@ -32,16 +33,17 @@ namespace Tegra::Engines { #define KEPLERMEMORY_REG_INDEX(field_name) \ (offsetof(Tegra::Engines::KeplerMemory::Regs, field_name) / sizeof(u32)) -class KeplerMemory final { +class KeplerMemory final : public EngineInterface { public: KeplerMemory(Core::System& system, MemoryManager& memory_manager); ~KeplerMemory(); /// Write the value to the register identified by method. - void CallMethod(const GPU::MethodCall& method_call); + void CallMethod(u32 method, u32 method_argument, bool is_last_call) override; /// Write multiple values to the register identified by method. - void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending); + void CallMultiMethod(u32 method, const u32* base_start, u32 amount, + u32 methods_pending) override; struct Regs { static constexpr size_t NUM_REGS = 0x7F; diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 39e3b66a2..024c9e43b 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -44,6 +44,12 @@ void Maxwell3D::InitializeRegisterDefaults() { viewport.depth_range_near = 0.0f; viewport.depth_range_far = 1.0f; } + for (auto& viewport : regs.viewport_transform) { + viewport.swizzle.x.Assign(Regs::ViewportSwizzle::PositiveX); + viewport.swizzle.y.Assign(Regs::ViewportSwizzle::PositiveY); + viewport.swizzle.z.Assign(Regs::ViewportSwizzle::PositiveZ); + viewport.swizzle.w.Assign(Regs::ViewportSwizzle::PositiveW); + } // Doom and Bomberman seems to use the uninitialized registers and just enable blend // so initialize blend registers with sane values @@ -125,12 +131,10 @@ void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u3 } } -void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) { - const u32 method = method_call.method; - +void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) { if (method == cb_data_state.current) { - regs.reg_array[method] = method_call.argument; - ProcessCBData(method_call.argument); + regs.reg_array[method] = method_argument; + ProcessCBData(method_argument); return; } else if (cb_data_state.current != null_cb_data) { FinishCBData(); @@ -153,10 +157,10 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) { executing_macro = method; } - macro_params.push_back(method_call.argument); + macro_params.push_back(method_argument); // Call the macro when there are no more parameters in the command buffer - if (method_call.IsLastCall()) { + if (is_last_call) { CallMacroMethod(executing_macro, macro_params.size(), macro_params.data()); macro_params.clear(); } @@ -166,7 +170,7 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) { ASSERT_MSG(method < Regs::NUM_REGS, "Invalid Maxwell3D register, increase the size of the Regs structure"); - u32 arg = method_call.argument; + u32 arg = method_argument; // Keep track of the register value in shadow_state when requested. if (shadow_state.shadow_ram_control == Regs::ShadowRamControl::Track || shadow_state.shadow_ram_control == Regs::ShadowRamControl::TrackWithFilter) { @@ -184,8 +188,12 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) { } switch (method) { + case MAXWELL3D_REG_INDEX(wait_for_idle): { + rasterizer.WaitForIdle(); + break; + } case MAXWELL3D_REG_INDEX(shadow_ram_control): { - shadow_state.shadow_ram_control = static_cast<Regs::ShadowRamControl>(method_call.argument); + shadow_state.shadow_ram_control = static_cast<Regs::ShadowRamControl>(method_argument); break; } case MAXWELL3D_REG_INDEX(macros.data): { @@ -268,7 +276,6 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) { break; } case MAXWELL3D_REG_INDEX(data_upload): { - const bool is_last_call = method_call.IsLastCall(); upload_state.ProcessData(arg, is_last_call); if (is_last_call) { OnMemoryWrite(); @@ -326,7 +333,7 @@ void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, } default: { for (std::size_t i = 0; i < amount; i++) { - CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)}); + CallMethod(method, base_start[i], methods_pending - static_cast<u32>(i) <= 1); } } } @@ -356,16 +363,15 @@ void Maxwell3D::StepInstance(const MMEDrawMode expected_mode, const u32 count) { StepInstance(expected_mode, count); } -void Maxwell3D::CallMethodFromMME(const GPU::MethodCall& method_call) { - const u32 method = method_call.method; +void Maxwell3D::CallMethodFromMME(u32 method, u32 method_argument) { if (mme_inline[method]) { - regs.reg_array[method] = method_call.argument; + regs.reg_array[method] = method_argument; if (method == MAXWELL3D_REG_INDEX(vertex_buffer.count) || method == MAXWELL3D_REG_INDEX(index_array.count)) { const MMEDrawMode expected_mode = method == MAXWELL3D_REG_INDEX(vertex_buffer.count) ? MMEDrawMode::Array : MMEDrawMode::Indexed; - StepInstance(expected_mode, method_call.argument); + StepInstance(expected_mode, method_argument); } else if (method == MAXWELL3D_REG_INDEX(draw.vertex_begin_gl)) { mme_draw.instance_mode = (regs.draw.instance_next != 0) || (regs.draw.instance_cont != 0); @@ -377,7 +383,7 @@ void Maxwell3D::CallMethodFromMME(const GPU::MethodCall& method_call) { if (mme_draw.current_mode != MMEDrawMode::Undefined) { FlushMMEInlineDraw(); } - CallMethod(method_call); + CallMethod(method, method_argument, true); } } diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 5e522e0d2..05dd6b39b 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -19,6 +19,7 @@ #include "common/math_util.h" #include "video_core/engines/const_buffer_engine_interface.h" #include "video_core/engines/const_buffer_info.h" +#include "video_core/engines/engine_interface.h" #include "video_core/engines/engine_upload.h" #include "video_core/engines/shader_type.h" #include "video_core/gpu.h" @@ -48,7 +49,7 @@ namespace Tegra::Engines { #define MAXWELL3D_REG_INDEX(field_name) \ (offsetof(Tegra::Engines::Maxwell3D::Regs, field_name) / sizeof(u32)) -class Maxwell3D final : public ConstBufferEngineInterface { +class Maxwell3D final : public ConstBufferEngineInterface, public EngineInterface { public: explicit Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager); @@ -575,6 +576,17 @@ public: Replay = 3, }; + enum class ViewportSwizzle : u32 { + PositiveX = 0, + NegativeX = 1, + PositiveY = 2, + NegativeY = 3, + PositiveZ = 4, + NegativeZ = 5, + PositiveW = 6, + NegativeW = 7, + }; + struct RenderTargetConfig { u32 address_high; u32 address_low; @@ -618,7 +630,14 @@ public: f32 translate_x; f32 translate_y; f32 translate_z; - INSERT_UNION_PADDING_WORDS(2); + union { + u32 raw; + BitField<0, 3, ViewportSwizzle> x; + BitField<4, 3, ViewportSwizzle> y; + BitField<8, 3, ViewportSwizzle> z; + BitField<12, 3, ViewportSwizzle> w; + } swizzle; + INSERT_UNION_PADDING_WORDS(1); Common::Rectangle<f32> GetRect() const { return { @@ -709,7 +728,9 @@ public: union { struct { - INSERT_UNION_PADDING_WORDS(0x45); + INSERT_UNION_PADDING_WORDS(0x44); + + u32 wait_for_idle; struct { u32 upload_address; @@ -1358,13 +1379,14 @@ public: u32 GetRegisterValue(u32 method) const; /// Write the value to the register identified by method. - void CallMethod(const GPU::MethodCall& method_call); + void CallMethod(u32 method, u32 method_argument, bool is_last_call) override; /// Write multiple values to the register identified by method. - void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending); + void CallMultiMethod(u32 method, const u32* base_start, u32 amount, + u32 methods_pending) override; /// Write the value to the register identified by method. - void CallMethodFromMME(const GPU::MethodCall& method_call); + void CallMethodFromMME(u32 method, u32 method_argument); void FlushMMEInlineDraw(); @@ -1536,6 +1558,7 @@ private: static_assert(offsetof(Maxwell3D::Regs, field_name) == position * 4, \ "Field " #field_name " has invalid position") +ASSERT_REG_POSITION(wait_for_idle, 0x44); ASSERT_REG_POSITION(macros, 0x45); ASSERT_REG_POSITION(shadow_ram_control, 0x49); ASSERT_REG_POSITION(upload, 0x60); diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp index 6630005b0..01d7df405 100644 --- a/src/video_core/engines/maxwell_dma.cpp +++ b/src/video_core/engines/maxwell_dma.cpp @@ -17,16 +17,16 @@ namespace Tegra::Engines { MaxwellDMA::MaxwellDMA(Core::System& system, MemoryManager& memory_manager) : system{system}, memory_manager{memory_manager} {} -void MaxwellDMA::CallMethod(const GPU::MethodCall& method_call) { - ASSERT_MSG(method_call.method < Regs::NUM_REGS, +void MaxwellDMA::CallMethod(u32 method, u32 method_argument, bool is_last_call) { + ASSERT_MSG(method < Regs::NUM_REGS, "Invalid MaxwellDMA register, increase the size of the Regs structure"); - regs.reg_array[method_call.method] = method_call.argument; + regs.reg_array[method] = method_argument; #define MAXWELLDMA_REG_INDEX(field_name) \ (offsetof(Tegra::Engines::MaxwellDMA::Regs, field_name) / sizeof(u32)) - switch (method_call.method) { + switch (method) { case MAXWELLDMA_REG_INDEX(exec): { HandleCopy(); break; @@ -39,7 +39,7 @@ void MaxwellDMA::CallMethod(const GPU::MethodCall& method_call) { void MaxwellDMA::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) { for (std::size_t i = 0; i < amount; i++) { - CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)}); + CallMethod(method, base_start[i], methods_pending - static_cast<u32>(i) <= 1); } } @@ -90,7 +90,47 @@ void MaxwellDMA::HandleCopy() { ASSERT(regs.exec.enable_2d == 1); if (regs.exec.is_dst_linear && !regs.exec.is_src_linear) { + ASSERT(regs.src_params.BlockDepth() == 0); + // Optimized path for micro copies. + if (regs.dst_pitch * regs.y_count < Texture::GetGOBSize() && regs.dst_pitch <= 64) { + const u32 bytes_per_pixel = regs.dst_pitch / regs.x_count; + const std::size_t src_size = Texture::GetGOBSize(); + const std::size_t dst_size = regs.dst_pitch * regs.y_count; + u32 pos_x = regs.src_params.pos_x; + u32 pos_y = regs.src_params.pos_y; + const u64 offset = + Texture::GetGOBOffset(regs.src_params.size_x, regs.src_params.size_y, pos_x, pos_y, + regs.src_params.BlockDepth(), bytes_per_pixel); + const u32 x_in_gob = 64 / bytes_per_pixel; + pos_x = pos_x % x_in_gob; + pos_y = pos_y % 8; + + if (read_buffer.size() < src_size) { + read_buffer.resize(src_size); + } + + if (write_buffer.size() < dst_size) { + write_buffer.resize(dst_size); + } + + if (Settings::IsGPULevelExtreme()) { + memory_manager.ReadBlock(source + offset, read_buffer.data(), src_size); + memory_manager.ReadBlock(dest, write_buffer.data(), dst_size); + } else { + memory_manager.ReadBlockUnsafe(source + offset, read_buffer.data(), src_size); + memory_manager.ReadBlockUnsafe(dest, write_buffer.data(), dst_size); + } + + Texture::UnswizzleSubrect(regs.x_count, regs.y_count, regs.dst_pitch, + regs.src_params.size_x, bytes_per_pixel, read_buffer.data(), + write_buffer.data(), regs.src_params.BlockHeight(), pos_x, + pos_y); + + memory_manager.WriteBlock(dest, write_buffer.data(), dst_size); + + return; + } // If the input is tiled and the output is linear, deswizzle the input and copy it over. const u32 bytes_per_pixel = regs.dst_pitch / regs.x_count; const std::size_t src_size = Texture::CalculateSize( diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h index c43ed8194..502dd8509 100644 --- a/src/video_core/engines/maxwell_dma.h +++ b/src/video_core/engines/maxwell_dma.h @@ -10,6 +10,7 @@ #include "common/bit_field.h" #include "common/common_funcs.h" #include "common/common_types.h" +#include "video_core/engines/engine_interface.h" #include "video_core/gpu.h" namespace Core { @@ -27,16 +28,17 @@ namespace Tegra::Engines { * https://github.com/envytools/envytools/blob/master/rnndb/fifo/gk104_copy.xml */ -class MaxwellDMA final { +class MaxwellDMA final : public EngineInterface { public: explicit MaxwellDMA(Core::System& system, MemoryManager& memory_manager); ~MaxwellDMA() = default; /// Write the value to the register identified by method. - void CallMethod(const GPU::MethodCall& method_call); + void CallMethod(u32 method, u32 method_argument, bool is_last_call) override; /// Write multiple values to the register identified by method. - void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending); + void CallMultiMethod(u32 method, const u32* base_start, u32 amount, + u32 methods_pending) override; struct Regs { static constexpr std::size_t NUM_REGS = 0x1D6; diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index b87fd873d..8eb017f65 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -299,19 +299,21 @@ void GPU::CallEngineMethod(const MethodCall& method_call) { switch (engine) { case EngineID::FERMI_TWOD_A: - fermi_2d->CallMethod(method_call); + fermi_2d->CallMethod(method_call.method, method_call.argument, method_call.IsLastCall()); break; case EngineID::MAXWELL_B: - maxwell_3d->CallMethod(method_call); + maxwell_3d->CallMethod(method_call.method, method_call.argument, method_call.IsLastCall()); break; case EngineID::KEPLER_COMPUTE_B: - kepler_compute->CallMethod(method_call); + kepler_compute->CallMethod(method_call.method, method_call.argument, + method_call.IsLastCall()); break; case EngineID::MAXWELL_DMA_COPY_A: - maxwell_dma->CallMethod(method_call); + maxwell_dma->CallMethod(method_call.method, method_call.argument, method_call.IsLastCall()); break; case EngineID::KEPLER_INLINE_TO_MEMORY_B: - kepler_memory->CallMethod(method_call); + kepler_memory->CallMethod(method_call.method, method_call.argument, + method_call.IsLastCall()); break; default: UNIMPLEMENTED_MSG("Unimplemented engine"); @@ -347,7 +349,27 @@ void GPU::ProcessBindMethod(const MethodCall& method_call) { // Bind the current subchannel to the desired engine id. LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {}", method_call.subchannel, method_call.argument); - bound_engines[method_call.subchannel] = static_cast<EngineID>(method_call.argument); + const auto engine_id = static_cast<EngineID>(method_call.argument); + bound_engines[method_call.subchannel] = static_cast<EngineID>(engine_id); + switch (engine_id) { + case EngineID::FERMI_TWOD_A: + dma_pusher->BindSubchannel(fermi_2d.get(), method_call.subchannel); + break; + case EngineID::MAXWELL_B: + dma_pusher->BindSubchannel(maxwell_3d.get(), method_call.subchannel); + break; + case EngineID::KEPLER_COMPUTE_B: + dma_pusher->BindSubchannel(kepler_compute.get(), method_call.subchannel); + break; + case EngineID::MAXWELL_DMA_COPY_A: + dma_pusher->BindSubchannel(maxwell_dma.get(), method_call.subchannel); + break; + case EngineID::KEPLER_INLINE_TO_MEMORY_B: + dma_pusher->BindSubchannel(kepler_memory.get(), method_call.subchannel); + break; + default: + UNIMPLEMENTED_MSG("Unimplemented engine {:04X}", static_cast<u32>(engine_id)); + } } void GPU::ProcessSemaphoreTriggerMethod() { diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h index dd51c95b7..a1b4c305c 100644 --- a/src/video_core/gpu.h +++ b/src/video_core/gpu.h @@ -68,6 +68,7 @@ enum class RenderTargetFormat : u32 { BGR5A1_UNORM = 0xE9, RG8_UNORM = 0xEA, RG8_SNORM = 0xEB, + RG8_UINT = 0xED, R16_UNORM = 0xEE, R16_SNORM = 0xEF, R16_SINT = 0xF0, diff --git a/src/video_core/macro_interpreter.cpp b/src/video_core/macro_interpreter.cpp index 42031d80a..947364928 100644 --- a/src/video_core/macro_interpreter.cpp +++ b/src/video_core/macro_interpreter.cpp @@ -328,7 +328,7 @@ void MacroInterpreter::SetMethodAddress(u32 address) { } void MacroInterpreter::Send(u32 value) { - maxwell3d.CallMethodFromMME({method_address.address, value}); + maxwell3d.CallMethodFromMME(method_address.address, value); // Increment the method address by the method increment. method_address.address.Assign(method_address.address.Value() + method_address.increment.Value()); diff --git a/src/video_core/morton.cpp b/src/video_core/morton.cpp index 6d522c318..836b25c1d 100644 --- a/src/video_core/morton.cpp +++ b/src/video_core/morton.cpp @@ -83,6 +83,7 @@ static constexpr ConversionArray morton_to_linear_fns = { MortonCopy<true, PixelFormat::RGBA8_SRGB>, MortonCopy<true, PixelFormat::RG8U>, MortonCopy<true, PixelFormat::RG8S>, + MortonCopy<true, PixelFormat::RG8UI>, MortonCopy<true, PixelFormat::RG32UI>, MortonCopy<true, PixelFormat::RGBX16F>, MortonCopy<true, PixelFormat::R32UI>, @@ -166,6 +167,7 @@ static constexpr ConversionArray linear_to_morton_fns = { MortonCopy<false, PixelFormat::RGBA8_SRGB>, MortonCopy<false, PixelFormat::RG8U>, MortonCopy<false, PixelFormat::RG8S>, + MortonCopy<false, PixelFormat::RG8UI>, MortonCopy<false, PixelFormat::RG32UI>, MortonCopy<false, PixelFormat::RGBX16F>, MortonCopy<false, PixelFormat::R32UI>, diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h index 603f61952..3cbdac8e7 100644 --- a/src/video_core/rasterizer_interface.h +++ b/src/video_core/rasterizer_interface.h @@ -80,6 +80,9 @@ public: /// and invalidated virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0; + /// Notify the host renderer to wait for previous primitive and compute operations. + virtual void WaitForIdle() = 0; + /// Notify the rasterizer to send all written commands to the host GPU. virtual void FlushCommands() = 0; diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp index 4efce0de7..d2cab50bd 100644 --- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp @@ -51,10 +51,6 @@ Buffer OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) { return std::make_shared<CachedBufferBlock>(cpu_addr, size); } -void OGLBufferCache::WriteBarrier() { - glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT); -} - GLuint OGLBufferCache::ToHandle(const Buffer& buffer) { return buffer->GetHandle(); } diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h index a74817857..a9e86cfc7 100644 --- a/src/video_core/renderer_opengl/gl_buffer_cache.h +++ b/src/video_core/renderer_opengl/gl_buffer_cache.h @@ -59,8 +59,6 @@ protected: GLuint ToHandle(const Buffer& buffer) override; - void WriteBarrier() override; - void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size, const u8* data) override; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 725b4c32d..69dcf952f 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -746,6 +746,17 @@ void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) { InvalidateRegion(addr, size); } +void RasterizerOpenGL::WaitForIdle() { + // Place a barrier on everything that is not framebuffer related. + // This is related to another flag that is not currently implemented. + glMemoryBarrier(GL_VERTEX_ATTRIB_ARRAY_BARRIER_BIT | GL_ELEMENT_ARRAY_BARRIER_BIT | + GL_UNIFORM_BARRIER_BIT | GL_TEXTURE_FETCH_BARRIER_BIT | + GL_SHADER_IMAGE_ACCESS_BARRIER_BIT | GL_COMMAND_BARRIER_BIT | + GL_PIXEL_BUFFER_BARRIER_BIT | GL_TEXTURE_UPDATE_BARRIER_BIT | + GL_BUFFER_UPDATE_BARRIER_BIT | GL_TRANSFORM_FEEDBACK_BARRIER_BIT | + GL_SHADER_STORAGE_BARRIER_BIT | GL_QUERY_BUFFER_BARRIER_BIT); +} + void RasterizerOpenGL::FlushCommands() { // Only flush when we have commands queued to OpenGL. if (num_queued_commands == 0) { @@ -1008,6 +1019,14 @@ void RasterizerOpenGL::SyncViewport() { const GLdouble near_depth = src.translate_z - src.scale_z * reduce_z; const GLdouble far_depth = src.translate_z + src.scale_z; glDepthRangeIndexed(static_cast<GLuint>(i), near_depth, far_depth); + + if (!GLAD_GL_NV_viewport_swizzle) { + continue; + } + glViewportSwizzleNV(static_cast<GLuint>(i), MaxwellToGL::ViewportSwizzle(src.swizzle.x), + MaxwellToGL::ViewportSwizzle(src.swizzle.y), + MaxwellToGL::ViewportSwizzle(src.swizzle.z), + MaxwellToGL::ViewportSwizzle(src.swizzle.w)); } } } diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 87249fb6f..b94c65907 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -75,6 +75,7 @@ public: void SignalSyncPoint(u32 value) override; void ReleaseFences() override; void FlushAndInvalidateRegion(VAddr addr, u64 size) override; + void WaitForIdle() override; void FlushCommands() override; void TickFrame() override; bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src, diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index 2729d1265..94fbd2a22 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -83,6 +83,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // RGBA8_SRGB {GL_RG8, GL_RG, GL_UNSIGNED_BYTE}, // RG8U {GL_RG8_SNORM, GL_RG, GL_BYTE}, // RG8S + {GL_RG8UI, GL_RG_INTEGER, GL_UNSIGNED_INT}, // RG8UI {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT}, // RG32UI {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT}, // RGBX16F {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT}, // R32UI diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h index 2c0c77c28..994ae98eb 100644 --- a/src/video_core/renderer_opengl/maxwell_to_gl.h +++ b/src/video_core/renderer_opengl/maxwell_to_gl.h @@ -503,5 +503,10 @@ inline GLenum PolygonMode(Maxwell::PolygonMode polygon_mode) { return GL_FILL; } +inline GLenum ViewportSwizzle(Maxwell::ViewportSwizzle swizzle) { + // Enumeration order matches register order. We can convert it arithmetically. + return GL_VIEWPORT_SWIZZLE_POSITIVE_X_NV + static_cast<GLenum>(swizzle); +} + } // namespace MaxwellToGL } // namespace OpenGL diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp index 7b01adf7f..568744e3c 100644 --- a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp +++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp @@ -2,6 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <algorithm> #include <cstring> #include <tuple> @@ -102,6 +103,12 @@ void FixedPipelineState::ColorBlending::Fill(const Maxwell& regs) noexcept { } } +void FixedPipelineState::ViewportSwizzles::Fill(const Maxwell& regs) noexcept { + const auto& transform = regs.viewport_transform; + std::transform(transform.begin(), transform.end(), swizzles.begin(), + [](const auto& viewport) { return static_cast<u16>(viewport.swizzle.raw); }); +} + void FixedPipelineState::BlendingAttachment::Fill(const Maxwell& regs, std::size_t index) { const auto& mask = regs.color_mask[regs.color_mask_common ? 0 : index]; @@ -145,6 +152,7 @@ void FixedPipelineState::Fill(const Maxwell& regs) { rasterizer.Fill(regs); depth_stencil.Fill(regs); color_blending.Fill(regs); + viewport_swizzles.Fill(regs); } std::size_t FixedPipelineState::Hash() const noexcept { diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.h b/src/video_core/renderer_vulkan/fixed_pipeline_state.h index cbf55dda3..31a6398f2 100644 --- a/src/video_core/renderer_vulkan/fixed_pipeline_state.h +++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.h @@ -234,10 +234,17 @@ struct FixedPipelineState { void Fill(const Maxwell& regs) noexcept; }; + struct ViewportSwizzles { + std::array<u16, Maxwell::NumViewports> swizzles; + + void Fill(const Maxwell& regs) noexcept; + }; + VertexInput vertex_input; Rasterizer rasterizer; DepthStencil depth_stencil; ColorBlending color_blending; + ViewportSwizzles viewport_swizzles; void Fill(const Maxwell& regs); diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp index 8681b821f..12be691a5 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp @@ -160,6 +160,7 @@ struct FormatTuple { {VK_FORMAT_R8G8B8A8_SRGB, Attachable}, // RGBA8_SRGB {VK_FORMAT_R8G8_UNORM, Attachable | Storage}, // RG8U {VK_FORMAT_R8G8_SNORM, Attachable | Storage}, // RG8S + {VK_FORMAT_R8G8_UINT, Attachable | Storage}, // RG8UI {VK_FORMAT_R32G32_UINT, Attachable | Storage}, // RG32UI {VK_FORMAT_UNDEFINED}, // RGBX16F {VK_FORMAT_R32_UINT, Attachable | Storage}, // R32UI @@ -345,8 +346,6 @@ VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttrib break; case Maxwell::VertexAttribute::Type::SignedInt: switch (size) { - case Maxwell::VertexAttribute::Size::Size_16_16_16_16: - return VK_FORMAT_R16G16B16A16_SINT; case Maxwell::VertexAttribute::Size::Size_8: return VK_FORMAT_R8_SINT; case Maxwell::VertexAttribute::Size::Size_8_8: @@ -355,8 +354,22 @@ VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttrib return VK_FORMAT_R8G8B8_SINT; case Maxwell::VertexAttribute::Size::Size_8_8_8_8: return VK_FORMAT_R8G8B8A8_SINT; + case Maxwell::VertexAttribute::Size::Size_16: + return VK_FORMAT_R16_SINT; + case Maxwell::VertexAttribute::Size::Size_16_16: + return VK_FORMAT_R16G16_SINT; + case Maxwell::VertexAttribute::Size::Size_16_16_16: + return VK_FORMAT_R16G16B16_SINT; + case Maxwell::VertexAttribute::Size::Size_16_16_16_16: + return VK_FORMAT_R16G16B16A16_SINT; case Maxwell::VertexAttribute::Size::Size_32: return VK_FORMAT_R32_SINT; + case Maxwell::VertexAttribute::Size::Size_32_32: + return VK_FORMAT_R32G32_SINT; + case Maxwell::VertexAttribute::Size::Size_32_32_32: + return VK_FORMAT_R32G32B32_SINT; + case Maxwell::VertexAttribute::Size::Size_32_32_32_32: + return VK_FORMAT_R32G32B32A32_SINT; default: break; } @@ -672,4 +685,27 @@ VkComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle) { return {}; } +VkViewportCoordinateSwizzleNV ViewportSwizzle(Maxwell::ViewportSwizzle swizzle) { + switch (swizzle) { + case Maxwell::ViewportSwizzle::PositiveX: + return VK_VIEWPORT_COORDINATE_SWIZZLE_POSITIVE_X_NV; + case Maxwell::ViewportSwizzle::NegativeX: + return VK_VIEWPORT_COORDINATE_SWIZZLE_NEGATIVE_X_NV; + case Maxwell::ViewportSwizzle::PositiveY: + return VK_VIEWPORT_COORDINATE_SWIZZLE_POSITIVE_Y_NV; + case Maxwell::ViewportSwizzle::NegativeY: + return VK_VIEWPORT_COORDINATE_SWIZZLE_NEGATIVE_Y_NV; + case Maxwell::ViewportSwizzle::PositiveZ: + return VK_VIEWPORT_COORDINATE_SWIZZLE_POSITIVE_Z_NV; + case Maxwell::ViewportSwizzle::NegativeZ: + return VK_VIEWPORT_COORDINATE_SWIZZLE_NEGATIVE_Z_NV; + case Maxwell::ViewportSwizzle::PositiveW: + return VK_VIEWPORT_COORDINATE_SWIZZLE_POSITIVE_W_NV; + case Maxwell::ViewportSwizzle::NegativeW: + return VK_VIEWPORT_COORDINATE_SWIZZLE_NEGATIVE_W_NV; + } + UNREACHABLE_MSG("Invalid swizzle={}", static_cast<int>(swizzle)); + return {}; +} + } // namespace Vulkan::MaxwellToVK diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.h b/src/video_core/renderer_vulkan/maxwell_to_vk.h index 81bce4c6c..7e213452f 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.h +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.h @@ -59,4 +59,6 @@ VkCullModeFlags CullFace(Maxwell::CullFace cull_face); VkComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle); +VkViewportCoordinateSwizzleNV ViewportSwizzle(Maxwell::ViewportSwizzle swizzle); + } // namespace Vulkan::MaxwellToVK diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp index 04532f8f8..59b441943 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp @@ -12,15 +12,12 @@ #include <fmt/format.h> -#include "common/assert.h" #include "common/dynamic_library.h" #include "common/logging/log.h" #include "common/telemetry.h" #include "core/core.h" #include "core/core_timing.h" #include "core/frontend/emu_window.h" -#include "core/memory.h" -#include "core/perf_stats.h" #include "core/settings.h" #include "core/telemetry_session.h" #include "video_core/gpu.h" diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h index 18270909b..522b5bff8 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.h +++ b/src/video_core/renderer_vulkan/renderer_vulkan.h @@ -5,7 +5,6 @@ #pragma once #include <memory> -#include <optional> #include <string> #include <vector> diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.h b/src/video_core/renderer_vulkan/vk_blit_screen.h index 5eb544aea..243640fab 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.h +++ b/src/video_core/renderer_vulkan/vk_blit_screen.h @@ -4,7 +4,6 @@ #pragma once -#include <array> #include <memory> #include <tuple> diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp index 81e1de2be..5b494da8c 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp @@ -5,11 +5,7 @@ #include <algorithm> #include <cstring> #include <memory> -#include <optional> -#include <tuple> -#include "common/assert.h" -#include "common/bit_util.h" #include "core/core.h" #include "video_core/renderer_vulkan/vk_buffer_cache.h" #include "video_core/renderer_vulkan/vk_device.h" diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h index 3cd2e2774..a54583e7d 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.h +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h @@ -5,14 +5,11 @@ #pragma once #include <memory> -#include <unordered_map> -#include <vector> #include "common/common_types.h" #include "video_core/buffer_cache/buffer_cache.h" #include "video_core/rasterizer_cache.h" #include "video_core/renderer_vulkan/vk_memory_manager.h" -#include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" #include "video_core/renderer_vulkan/vk_stream_buffer.h" #include "video_core/renderer_vulkan/wrapper.h" @@ -55,8 +52,6 @@ public: protected: VkBuffer ToHandle(const Buffer& buffer) override; - void WriteBarrier() override {} - Buffer CreateBlock(VAddr cpu_addr, std::size_t size) override; void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size, diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.cpp b/src/video_core/renderer_vulkan/vk_compute_pass.cpp index 7b0268033..da71e710c 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pass.cpp @@ -6,7 +6,7 @@ #include <memory> #include <optional> #include <utility> -#include <vector> + #include "common/alignment.h" #include "common/assert.h" #include "common/common_types.h" diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.h b/src/video_core/renderer_vulkan/vk_compute_pass.h index 26bf834de..230b526bc 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.h +++ b/src/video_core/renderer_vulkan/vk_compute_pass.h @@ -6,7 +6,7 @@ #include <optional> #include <utility> -#include <vector> + #include "common/common_types.h" #include "video_core/engines/maxwell_3d.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp index 52566bb79..8e1b46277 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp @@ -2,14 +2,12 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include <memory> #include <vector> #include "video_core/renderer_vulkan/vk_compute_pipeline.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_pipeline_cache.h" -#include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_shader_decompiler.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h" diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.h b/src/video_core/renderer_vulkan/vk_compute_pipeline.h index 33b9af29e..6e2f22a4a 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pipeline.h +++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.h @@ -4,8 +4,6 @@ #pragma once -#include <memory> - #include "common/common_types.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_shader_decompiler.h" diff --git a/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp b/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp index e9d528aa6..890fd52cf 100644 --- a/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp +++ b/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp @@ -2,7 +2,6 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include <memory> #include <vector> #include "common/common_types.h" diff --git a/src/video_core/renderer_vulkan/vk_descriptor_pool.h b/src/video_core/renderer_vulkan/vk_descriptor_pool.h index ab40c70f0..9efa66bef 100644 --- a/src/video_core/renderer_vulkan/vk_descriptor_pool.h +++ b/src/video_core/renderer_vulkan/vk_descriptor_pool.h @@ -4,10 +4,8 @@ #pragma once -#include <memory> #include <vector> -#include "common/common_types.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/wrapper.h" diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp index e90c76492..f0c491d00 100644 --- a/src/video_core/renderer_vulkan/vk_device.cpp +++ b/src/video_core/renderer_vulkan/vk_device.cpp @@ -4,7 +4,6 @@ #include <bitset> #include <chrono> -#include <cstdlib> #include <optional> #include <string_view> #include <thread> @@ -95,6 +94,7 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties( VK_FORMAT_R8G8B8A8_SRGB, VK_FORMAT_R8G8_UNORM, VK_FORMAT_R8G8_SNORM, + VK_FORMAT_R8G8_UINT, VK_FORMAT_R8_UNORM, VK_FORMAT_R8_UINT, VK_FORMAT_B10G11R11_UFLOAT_PACK32, @@ -261,6 +261,10 @@ bool VKDevice::Create() { LOG_INFO(Render_Vulkan, "Device doesn't support float16 natively"); } + if (!nv_viewport_swizzle) { + LOG_INFO(Render_Vulkan, "Device doesn't support viewport swizzles"); + } + VkPhysicalDeviceUniformBufferStandardLayoutFeaturesKHR std430_layout; if (khr_uniform_buffer_standard_layout) { std430_layout.sType = @@ -294,6 +298,17 @@ bool VKDevice::Create() { LOG_INFO(Render_Vulkan, "Device doesn't support transform feedbacks"); } + VkPhysicalDeviceCustomBorderColorFeaturesEXT custom_border; + if (ext_custom_border_color) { + custom_border.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CUSTOM_BORDER_COLOR_FEATURES_EXT; + custom_border.pNext = nullptr; + custom_border.customBorderColors = VK_TRUE; + custom_border.customBorderColorWithoutFormat = VK_TRUE; + SetNext(next, custom_border); + } else { + LOG_INFO(Render_Vulkan, "Device doesn't support custom border colors"); + } + if (!ext_depth_range_unrestricted) { LOG_INFO(Render_Vulkan, "Device doesn't support depth range unrestricted"); } @@ -521,7 +536,9 @@ std::vector<const char*> VKDevice::LoadExtensions() { bool has_khr_shader_float16_int8{}; bool has_ext_subgroup_size_control{}; bool has_ext_transform_feedback{}; + bool has_ext_custom_border_color{}; for (const auto& extension : physical.EnumerateDeviceExtensionProperties()) { + Test(extension, nv_viewport_swizzle, VK_NV_VIEWPORT_SWIZZLE_EXTENSION_NAME, true); Test(extension, khr_uniform_buffer_standard_layout, VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true); Test(extension, has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME, @@ -535,6 +552,8 @@ std::vector<const char*> VKDevice::LoadExtensions() { false); Test(extension, has_ext_transform_feedback, VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME, false); + Test(extension, has_ext_custom_border_color, VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME, + false); if (Settings::values.renderer_debug) { Test(extension, nv_device_diagnostics_config, VK_NV_DEVICE_DIAGNOSTICS_CONFIG_EXTENSION_NAME, true); @@ -607,6 +626,19 @@ std::vector<const char*> VKDevice::LoadExtensions() { } } + if (has_ext_custom_border_color) { + VkPhysicalDeviceCustomBorderColorFeaturesEXT border_features; + border_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CUSTOM_BORDER_COLOR_FEATURES_EXT; + border_features.pNext = nullptr; + features.pNext = &border_features; + physical.GetFeatures2KHR(features); + + if (border_features.customBorderColors && border_features.customBorderColorWithoutFormat) { + extensions.push_back(VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME); + ext_custom_border_color = true; + } + } + return extensions; } diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h index c8640762d..6b9227b09 100644 --- a/src/video_core/renderer_vulkan/vk_device.h +++ b/src/video_core/renderer_vulkan/vk_device.h @@ -147,6 +147,11 @@ public: return is_formatless_image_load_supported; } + /// Returns true if the device supports VK_NV_viewport_swizzle. + bool IsNvViewportSwizzleSupported() const { + return nv_viewport_swizzle; + } + /// Returns true if the device supports VK_EXT_scalar_block_layout. bool IsKhrUniformBufferStandardLayoutSupported() const { return khr_uniform_buffer_standard_layout; @@ -172,6 +177,11 @@ public: return ext_transform_feedback; } + /// Returns true if the device supports VK_EXT_custom_border_color. + bool IsExtCustomBorderColorSupported() const { + return ext_custom_border_color; + } + /// Returns the vendor name reported from Vulkan. std::string_view GetVendorName() const { return vendor_name; @@ -222,11 +232,13 @@ private: bool is_float16_supported{}; ///< Support for float16 arithmetics. bool is_warp_potentially_bigger{}; ///< Host warp size can be bigger than guest. bool is_formatless_image_load_supported{}; ///< Support for shader image read without format. + bool nv_viewport_swizzle{}; ///< Support for VK_NV_viewport_swizzle. bool khr_uniform_buffer_standard_layout{}; ///< Support for std430 on UBOs. bool ext_index_type_uint8{}; ///< Support for VK_EXT_index_type_uint8. bool ext_depth_range_unrestricted{}; ///< Support for VK_EXT_depth_range_unrestricted. bool ext_shader_viewport_index_layer{}; ///< Support for VK_EXT_shader_viewport_index_layer. bool ext_transform_feedback{}; ///< Support for VK_EXT_transform_feedback. + bool ext_custom_border_color{}; ///< Support for VK_EXT_custom_border_color. bool nv_device_diagnostics_config{}; ///< Support for VK_NV_device_diagnostics_config. // Telemetry parameters diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp index b39b81b48..69b6bba00 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp @@ -2,11 +2,11 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <algorithm> #include <array> #include <cstring> #include <vector> -#include "common/assert.h" #include "common/common_types.h" #include "common/microprofile.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h" @@ -51,6 +51,23 @@ bool SupportsPrimitiveRestart(VkPrimitiveTopology topology) { topology) == std::end(unsupported_topologies); } +VkViewportSwizzleNV UnpackViewportSwizzle(u16 swizzle) { + union { + u32 raw; + BitField<0, 3, Maxwell::ViewportSwizzle> x; + BitField<4, 3, Maxwell::ViewportSwizzle> y; + BitField<8, 3, Maxwell::ViewportSwizzle> z; + BitField<12, 3, Maxwell::ViewportSwizzle> w; + } const unpacked{swizzle}; + + VkViewportSwizzleNV result; + result.x = MaxwellToVK::ViewportSwizzle(unpacked.x); + result.y = MaxwellToVK::ViewportSwizzle(unpacked.y); + result.z = MaxwellToVK::ViewportSwizzle(unpacked.z); + result.w = MaxwellToVK::ViewportSwizzle(unpacked.w); + return result; +} + } // Anonymous namespace VKGraphicsPipeline::VKGraphicsPipeline(const VKDevice& device, VKScheduler& scheduler, @@ -163,6 +180,7 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa const auto& ds = fixed_state.depth_stencil; const auto& cd = fixed_state.color_blending; const auto& rs = fixed_state.rasterizer; + const auto& viewport_swizzles = fixed_state.viewport_swizzles.swizzles; std::vector<VkVertexInputBindingDescription> vertex_bindings; std::vector<VkVertexInputBindingDivisorDescriptionEXT> vertex_binding_divisors; @@ -245,6 +263,19 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa viewport_ci.scissorCount = Maxwell::NumViewports; viewport_ci.pScissors = nullptr; + std::array<VkViewportSwizzleNV, Maxwell::NumViewports> swizzles; + std::transform(viewport_swizzles.begin(), viewport_swizzles.end(), swizzles.begin(), + UnpackViewportSwizzle); + VkPipelineViewportSwizzleStateCreateInfoNV swizzle_ci; + swizzle_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_SWIZZLE_STATE_CREATE_INFO_NV; + swizzle_ci.pNext = nullptr; + swizzle_ci.flags = 0; + swizzle_ci.viewportCount = Maxwell::NumViewports; + swizzle_ci.pViewportSwizzles = swizzles.data(); + if (device.IsNvViewportSwizzleSupported()) { + viewport_ci.pNext = &swizzle_ci; + } + VkPipelineRasterizationStateCreateInfo rasterization_ci; rasterization_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO; rasterization_ci.pNext = nullptr; diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h index 7aba70960..a1d699a6c 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h @@ -5,16 +5,13 @@ #pragma once #include <array> -#include <memory> #include <optional> -#include <unordered_map> #include <vector> #include "video_core/engines/maxwell_3d.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_renderpass_cache.h" -#include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_shader_decompiler.h" #include "video_core/renderer_vulkan/wrapper.h" diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp index 8fbd63dbc..fe45ed269 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp @@ -22,7 +22,6 @@ #include "video_core/renderer_vulkan/vk_pipeline_cache.h" #include "video_core/renderer_vulkan/vk_rasterizer.h" #include "video_core/renderer_vulkan/vk_renderpass_cache.h" -#include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h" #include "video_core/renderer_vulkan/wrapper.h" diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h index ebddafb73..0b5796fef 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h @@ -21,13 +21,11 @@ #include "video_core/renderer_vulkan/fixed_pipeline_state.h" #include "video_core/renderer_vulkan/vk_graphics_pipeline.h" #include "video_core/renderer_vulkan/vk_renderpass_cache.h" -#include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_shader_decompiler.h" #include "video_core/renderer_vulkan/wrapper.h" #include "video_core/shader/memory_util.h" #include "video_core/shader/registry.h" #include "video_core/shader/shader_ir.h" -#include "video_core/surface.h" namespace Core { class System; diff --git a/src/video_core/renderer_vulkan/vk_query_cache.cpp b/src/video_core/renderer_vulkan/vk_query_cache.cpp index 813f7c162..bc91c48cc 100644 --- a/src/video_core/renderer_vulkan/vk_query_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_query_cache.cpp @@ -4,7 +4,6 @@ #include <algorithm> #include <cstddef> -#include <cstdint> #include <utility> #include <vector> diff --git a/src/video_core/renderer_vulkan/vk_query_cache.h b/src/video_core/renderer_vulkan/vk_query_cache.h index b63784f4b..40119e6d3 100644 --- a/src/video_core/renderer_vulkan/vk_query_cache.h +++ b/src/video_core/renderer_vulkan/vk_query_cache.h @@ -5,7 +5,6 @@ #pragma once #include <cstddef> -#include <cstdint> #include <memory> #include <utility> #include <vector> diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index ccfd0e670..8b009fc22 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -9,14 +9,12 @@ #include <vector> #include <boost/container/static_vector.hpp> -#include <boost/functional/hash.hpp> #include "common/alignment.h" #include "common/assert.h" #include "common/logging/log.h" #include "common/microprofile.h" #include "core/core.h" -#include "core/memory.h" #include "core/settings.h" #include "video_core/engines/kepler_compute.h" #include "video_core/engines/maxwell_3d.h" @@ -301,7 +299,7 @@ RasterizerVulkan::RasterizerVulkan(Core::System& system, Core::Frontend::EmuWind buffer_cache(*this, system, device, memory_manager, scheduler, staging_pool), sampler_cache(device), fence_manager(system, *this, device, scheduler, texture_cache, buffer_cache, query_cache), - query_cache(system, *this, device, scheduler) { + query_cache(system, *this, device, scheduler), wfi_event{device.GetLogical().CreateEvent()} { scheduler.SetQueryCache(query_cache); } @@ -575,6 +573,26 @@ void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size) { InvalidateRegion(addr, size); } +void RasterizerVulkan::WaitForIdle() { + // Everything but wait pixel operations. This intentionally includes FRAGMENT_SHADER_BIT because + // fragment shaders can still write storage buffers. + VkPipelineStageFlags flags = + VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT | VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | + VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT | + VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT | + VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | + VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | VK_PIPELINE_STAGE_TRANSFER_BIT; + if (device.IsExtTransformFeedbackSupported()) { + flags |= VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT; + } + + scheduler.RequestOutsideRenderPassOperationContext(); + scheduler.Record([event = *wfi_event, flags](vk::CommandBuffer cmdbuf) { + cmdbuf.SetEvent(event, flags); + cmdbuf.WaitEvents(event, flags, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, {}, {}, {}); + }); +} + void RasterizerVulkan::FlushCommands() { if (draw_counter > 0) { draw_counter = 0; @@ -895,6 +913,9 @@ void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex void RasterizerVulkan::SetupIndexBuffer(BufferBindings& buffer_bindings, DrawParameters& params, bool is_indexed) { + if (params.num_vertices == 0) { + return; + } const auto& regs = system.GPU().Maxwell3D().regs; switch (regs.draw.topology) { case Maxwell::PrimitiveTopology::Quads: { diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h index d41a7929e..0ed0e48c6 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.h +++ b/src/video_core/renderer_vulkan/vk_rasterizer.h @@ -14,7 +14,6 @@ #include <boost/functional/hash.hpp> #include "common/common_types.h" -#include "video_core/memory_manager.h" #include "video_core/rasterizer_accelerated.h" #include "video_core/rasterizer_interface.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h" @@ -127,6 +126,7 @@ public: void SignalSyncPoint(u32 value) override; void ReleaseFences() override; void FlushAndInvalidateRegion(VAddr addr, u64 size) override; + void WaitForIdle() override; void FlushCommands() override; void TickFrame() override; bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src, @@ -276,6 +276,7 @@ private: vk::Buffer default_buffer; VKMemoryCommit default_buffer_commit; + vk::Event wfi_event; std::array<View, Maxwell::NumRenderTargets> color_attachments; View zeta_attachment; diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp index 07bbcf520..e6f2fa553 100644 --- a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp @@ -2,11 +2,8 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include <cstring> -#include <optional> #include <unordered_map> -#include "common/assert.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/vk_sampler_cache.h" #include "video_core/renderer_vulkan/wrapper.h" @@ -42,9 +39,18 @@ VKSamplerCache::VKSamplerCache(const VKDevice& device) : device{device} {} VKSamplerCache::~VKSamplerCache() = default; vk::Sampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) const { + const bool arbitrary_borders = device.IsExtCustomBorderColorSupported(); + const std::array color = tsc.GetBorderColor(); + + VkSamplerCustomBorderColorCreateInfoEXT border; + border.sType = VK_STRUCTURE_TYPE_SAMPLER_CUSTOM_BORDER_COLOR_CREATE_INFO_EXT; + border.pNext = nullptr; + border.format = VK_FORMAT_UNDEFINED; + std::memcpy(&border.customBorderColor, color.data(), sizeof(color)); + VkSamplerCreateInfo ci; ci.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO; - ci.pNext = nullptr; + ci.pNext = arbitrary_borders ? &border : nullptr; ci.flags = 0; ci.magFilter = MaxwellToVK::Sampler::Filter(tsc.mag_filter); ci.minFilter = MaxwellToVK::Sampler::Filter(tsc.min_filter); @@ -59,7 +65,7 @@ vk::Sampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) c ci.compareOp = MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func); ci.minLod = tsc.GetMinLod(); ci.maxLod = tsc.GetMaxLod(); - ci.borderColor = ConvertBorderColor(tsc.GetBorderColor()); + ci.borderColor = arbitrary_borders ? VK_BORDER_COLOR_INT_CUSTOM_EXT : ConvertBorderColor(color); ci.unnormalizedCoordinates = VK_FALSE; return device.GetLogical().CreateSampler(ci); } diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp index ae7ba3eb5..82ec9180e 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.cpp +++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp @@ -8,7 +8,6 @@ #include <thread> #include <utility> -#include "common/assert.h" #include "common/microprofile.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_query_cache.h" diff --git a/src/video_core/renderer_vulkan/vk_scheduler.h b/src/video_core/renderer_vulkan/vk_scheduler.h index 82a8adc69..970a65566 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.h +++ b/src/video_core/renderer_vulkan/vk_scheduler.h @@ -7,7 +7,6 @@ #include <atomic> #include <condition_variable> #include <memory> -#include <optional> #include <stack> #include <thread> #include <utility> diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.h b/src/video_core/renderer_vulkan/vk_shader_decompiler.h index ffea4709e..f4c05ac3c 100644 --- a/src/video_core/renderer_vulkan/vk_shader_decompiler.h +++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.h @@ -5,11 +5,7 @@ #pragma once #include <array> -#include <bitset> -#include <memory> #include <set> -#include <type_traits> -#include <utility> #include <vector> #include "common/common_types.h" diff --git a/src/video_core/renderer_vulkan/vk_shader_util.cpp b/src/video_core/renderer_vulkan/vk_shader_util.cpp index 784839327..112df9c71 100644 --- a/src/video_core/renderer_vulkan/vk_shader_util.cpp +++ b/src/video_core/renderer_vulkan/vk_shader_util.cpp @@ -4,8 +4,7 @@ #include <cstring> #include <memory> -#include <vector> -#include "common/alignment.h" + #include "common/assert.h" #include "common/common_types.h" #include "video_core/renderer_vulkan/vk_device.h" diff --git a/src/video_core/renderer_vulkan/vk_shader_util.h b/src/video_core/renderer_vulkan/vk_shader_util.h index be38d6697..d1d3f3cae 100644 --- a/src/video_core/renderer_vulkan/vk_shader_util.h +++ b/src/video_core/renderer_vulkan/vk_shader_util.h @@ -4,7 +4,6 @@ #pragma once -#include <vector> #include "common/common_types.h" #include "video_core/renderer_vulkan/wrapper.h" diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h index faf6418fd..3c4901437 100644 --- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h +++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h @@ -5,8 +5,6 @@ #pragma once #include <climits> -#include <unordered_map> -#include <utility> #include <vector> #include "common/common_types.h" diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index de4c23120..55f43e61b 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -10,11 +10,9 @@ #include <variant> #include <vector> -#include "common/alignment.h" #include "common/assert.h" #include "common/common_types.h" #include "core/core.h" -#include "core/memory.h" #include "video_core/engines/maxwell_3d.h" #include "video_core/morton.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" @@ -26,7 +24,6 @@ #include "video_core/renderer_vulkan/vk_texture_cache.h" #include "video_core/renderer_vulkan/wrapper.h" #include "video_core/surface.h" -#include "video_core/textures/convert.h" namespace Vulkan { diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h index 115595f28..f211ccb1e 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.h +++ b/src/video_core/renderer_vulkan/vk_texture_cache.h @@ -7,19 +7,13 @@ #include <memory> #include <unordered_map> -#include "common/assert.h" #include "common/common_types.h" -#include "common/logging/log.h" -#include "common/math_util.h" -#include "video_core/gpu.h" -#include "video_core/rasterizer_cache.h" #include "video_core/renderer_vulkan/vk_image.h" #include "video_core/renderer_vulkan/vk_memory_manager.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/wrapper.h" #include "video_core/texture_cache/surface_base.h" #include "video_core/texture_cache/texture_cache.h" -#include "video_core/textures/decoders.h" namespace Core { class System; diff --git a/src/video_core/renderer_vulkan/vk_update_descriptor.h b/src/video_core/renderer_vulkan/vk_update_descriptor.h index 6ba2c9997..cc7e3dff4 100644 --- a/src/video_core/renderer_vulkan/vk_update_descriptor.h +++ b/src/video_core/renderer_vulkan/vk_update_descriptor.h @@ -4,7 +4,6 @@ #pragma once -#include <type_traits> #include <variant> #include <boost/container/static_vector.hpp> diff --git a/src/video_core/renderer_vulkan/wrapper.cpp b/src/video_core/renderer_vulkan/wrapper.cpp index 7f5bc1404..2ce9b0626 100644 --- a/src/video_core/renderer_vulkan/wrapper.cpp +++ b/src/video_core/renderer_vulkan/wrapper.cpp @@ -87,6 +87,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept { X(vkCmdSetStencilReference); X(vkCmdSetStencilWriteMask); X(vkCmdSetViewport); + X(vkCmdWaitEvents); X(vkCreateBuffer); X(vkCreateBufferView); X(vkCreateCommandPool); diff --git a/src/video_core/renderer_vulkan/wrapper.h b/src/video_core/renderer_vulkan/wrapper.h index bda16a2cb..98937a77a 100644 --- a/src/video_core/renderer_vulkan/wrapper.h +++ b/src/video_core/renderer_vulkan/wrapper.h @@ -205,6 +205,7 @@ struct DeviceDispatch : public InstanceDispatch { PFN_vkCmdSetStencilReference vkCmdSetStencilReference; PFN_vkCmdSetStencilWriteMask vkCmdSetStencilWriteMask; PFN_vkCmdSetViewport vkCmdSetViewport; + PFN_vkCmdWaitEvents vkCmdWaitEvents; PFN_vkCreateBuffer vkCreateBuffer; PFN_vkCreateBufferView vkCreateBufferView; PFN_vkCreateCommandPool vkCreateCommandPool; @@ -958,6 +959,15 @@ public: dld->vkCmdSetEvent(handle, event, stage_flags); } + void WaitEvents(Span<VkEvent> events, VkPipelineStageFlags src_stage_mask, + VkPipelineStageFlags dst_stage_mask, Span<VkMemoryBarrier> memory_barriers, + Span<VkBufferMemoryBarrier> buffer_barriers, + Span<VkImageMemoryBarrier> image_barriers) const noexcept { + dld->vkCmdWaitEvents(handle, events.size(), events.data(), src_stage_mask, dst_stage_mask, + memory_barriers.size(), memory_barriers.data(), buffer_barriers.size(), + buffer_barriers.data(), image_barriers.size(), image_barriers.data()); + } + void BindTransformFeedbackBuffersEXT(u32 first, u32 count, const VkBuffer* buffers, const VkDeviceSize* offsets, const VkDeviceSize* sizes) const noexcept { diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp index cc7181229..bbe93903c 100644 --- a/src/video_core/surface.cpp +++ b/src/video_core/surface.cpp @@ -145,6 +145,8 @@ PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format) return PixelFormat::RG8U; case Tegra::RenderTargetFormat::RG8_SNORM: return PixelFormat::RG8S; + case Tegra::RenderTargetFormat::RG8_UINT: + return PixelFormat::RG8UI; case Tegra::RenderTargetFormat::R16_FLOAT: return PixelFormat::R16F; case Tegra::RenderTargetFormat::R16_UNORM: diff --git a/src/video_core/surface.h b/src/video_core/surface.h index e0acd44d3..6da6a1b97 100644 --- a/src/video_core/surface.h +++ b/src/video_core/surface.h @@ -57,51 +57,52 @@ enum class PixelFormat { RGBA8_SRGB = 39, RG8U = 40, RG8S = 41, - RG32UI = 42, - RGBX16F = 43, - R32UI = 44, - R32I = 45, - ASTC_2D_8X8 = 46, - ASTC_2D_8X5 = 47, - ASTC_2D_5X4 = 48, - BGRA8_SRGB = 49, - DXT1_SRGB = 50, - DXT23_SRGB = 51, - DXT45_SRGB = 52, - BC7U_SRGB = 53, - R4G4B4A4U = 54, - ASTC_2D_4X4_SRGB = 55, - ASTC_2D_8X8_SRGB = 56, - ASTC_2D_8X5_SRGB = 57, - ASTC_2D_5X4_SRGB = 58, - ASTC_2D_5X5 = 59, - ASTC_2D_5X5_SRGB = 60, - ASTC_2D_10X8 = 61, - ASTC_2D_10X8_SRGB = 62, - ASTC_2D_6X6 = 63, - ASTC_2D_6X6_SRGB = 64, - ASTC_2D_10X10 = 65, - ASTC_2D_10X10_SRGB = 66, - ASTC_2D_12X12 = 67, - ASTC_2D_12X12_SRGB = 68, - ASTC_2D_8X6 = 69, - ASTC_2D_8X6_SRGB = 70, - ASTC_2D_6X5 = 71, - ASTC_2D_6X5_SRGB = 72, - E5B9G9R9F = 73, + RG8UI = 42, + RG32UI = 43, + RGBX16F = 44, + R32UI = 45, + R32I = 46, + ASTC_2D_8X8 = 47, + ASTC_2D_8X5 = 48, + ASTC_2D_5X4 = 49, + BGRA8_SRGB = 50, + DXT1_SRGB = 51, + DXT23_SRGB = 52, + DXT45_SRGB = 53, + BC7U_SRGB = 54, + R4G4B4A4U = 55, + ASTC_2D_4X4_SRGB = 56, + ASTC_2D_8X8_SRGB = 57, + ASTC_2D_8X5_SRGB = 58, + ASTC_2D_5X4_SRGB = 59, + ASTC_2D_5X5 = 60, + ASTC_2D_5X5_SRGB = 61, + ASTC_2D_10X8 = 62, + ASTC_2D_10X8_SRGB = 63, + ASTC_2D_6X6 = 64, + ASTC_2D_6X6_SRGB = 65, + ASTC_2D_10X10 = 66, + ASTC_2D_10X10_SRGB = 67, + ASTC_2D_12X12 = 68, + ASTC_2D_12X12_SRGB = 69, + ASTC_2D_8X6 = 70, + ASTC_2D_8X6_SRGB = 71, + ASTC_2D_6X5 = 72, + ASTC_2D_6X5_SRGB = 73, + E5B9G9R9F = 74, MaxColorFormat, // Depth formats - Z32F = 74, - Z16 = 75, + Z32F = 75, + Z16 = 76, MaxDepthFormat, // DepthStencil formats - Z24S8 = 76, - S8Z24 = 77, - Z32FS8 = 78, + Z24S8 = 77, + S8Z24 = 78, + Z32FS8 = 79, MaxDepthStencilFormat, @@ -171,6 +172,7 @@ constexpr std::array<u32, MaxPixelFormat> compression_factor_shift_table = {{ 0, // RGBA8_SRGB 0, // RG8U 0, // RG8S + 0, // RG8UI 0, // RG32UI 0, // RGBX16F 0, // R32UI @@ -269,6 +271,7 @@ constexpr std::array<u32, MaxPixelFormat> block_width_table = {{ 1, // RGBA8_SRGB 1, // RG8U 1, // RG8S + 1, // RG8UI 1, // RG32UI 1, // RGBX16F 1, // R32UI @@ -359,6 +362,7 @@ constexpr std::array<u32, MaxPixelFormat> block_height_table = {{ 1, // RGBA8_SRGB 1, // RG8U 1, // RG8S + 1, // RG8UI 1, // RG32UI 1, // RGBX16F 1, // R32UI @@ -449,6 +453,7 @@ constexpr std::array<u32, MaxPixelFormat> bpp_table = {{ 32, // RGBA8_SRGB 16, // RG8U 16, // RG8S + 16, // RG8UI 64, // RG32UI 64, // RGBX16F 32, // R32UI diff --git a/src/video_core/texture_cache/format_lookup_table.cpp b/src/video_core/texture_cache/format_lookup_table.cpp index 25d2ee2e8..7032e0059 100644 --- a/src/video_core/texture_cache/format_lookup_table.cpp +++ b/src/video_core/texture_cache/format_lookup_table.cpp @@ -41,7 +41,7 @@ struct Table { ComponentType alpha_component; bool is_srgb; }; -constexpr std::array<Table, 76> DefinitionTable = {{ +constexpr std::array<Table, 77> DefinitionTable = {{ {TextureFormat::A8R8G8B8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ABGR8U}, {TextureFormat::A8R8G8B8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::ABGR8S}, {TextureFormat::A8R8G8B8, C, UINT, UINT, UINT, UINT, PixelFormat::ABGR8UI}, @@ -60,6 +60,7 @@ constexpr std::array<Table, 76> DefinitionTable = {{ {TextureFormat::G8R8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RG8U}, {TextureFormat::G8R8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RG8S}, + {TextureFormat::G8R8, C, UINT, UINT, UINT, UINT, PixelFormat::RG8UI}, {TextureFormat::R16_G16_B16_A16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RGBA16S}, {TextureFormat::R16_G16_B16_A16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RGBA16U}, diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index fae8638ec..548e4c3fe 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -382,4 +382,18 @@ std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height } } +u64 GetGOBOffset(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height, + u32 bytes_per_pixel) { + auto div_ceil = [](const u32 x, const u32 y) { return ((x + y - 1) / y); }; + const u32 gobs_in_block = 1 << block_height; + const u32 y_blocks = gob_size_y << block_height; + const u32 x_per_gob = gob_size_x / bytes_per_pixel; + const u32 x_blocks = div_ceil(width, x_per_gob); + const u32 block_size = gob_size * gobs_in_block; + const u32 stride = block_size * x_blocks; + const u32 base = (dst_y / y_blocks) * stride + (dst_x / x_per_gob) * block_size; + const u32 relative_y = dst_y % y_blocks; + return base + (relative_y / gob_size_y) * gob_size; +} + } // namespace Tegra::Texture diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h index 9f2d6d308..06f3ebf87 100644 --- a/src/video_core/textures/decoders.h +++ b/src/video_core/textures/decoders.h @@ -59,4 +59,8 @@ void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 void SwizzleKepler(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height, std::size_t copy_size, const u8* source_data, u8* swizzle_data); +/// Obtains the offset of the gob for positions 'dst_x' & 'dst_y' +u64 GetGOBOffset(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height, + u32 bytes_per_pixel); + } // namespace Tegra::Texture |