diff options
Diffstat (limited to 'src/video_core')
23 files changed, 507 insertions, 166 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 04a25da4f..db9332d00 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -156,6 +156,7 @@ if (ENABLE_VULKAN) renderer_vulkan/maxwell_to_vk.cpp renderer_vulkan/maxwell_to_vk.h renderer_vulkan/renderer_vulkan.h + renderer_vulkan/renderer_vulkan.cpp renderer_vulkan/vk_blit_screen.cpp renderer_vulkan/vk_blit_screen.h renderer_vulkan/vk_buffer_cache.cpp diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index 0510ed777..186aca61d 100644 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -101,7 +101,10 @@ public: void TickFrame() { ++epoch; while (!pending_destruction.empty()) { - if (pending_destruction.front()->GetEpoch() + 1 > epoch) { + // Delay at least 4 frames before destruction. + // This is due to triple buffering happening on some drivers. + static constexpr u64 epochs_to_destroy = 5; + if (pending_destruction.front()->GetEpoch() + epochs_to_destroy > epoch) { break; } pending_destruction.pop_front(); diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h index f443ec0fe..402869fde 100644 --- a/src/video_core/engines/shader_bytecode.h +++ b/src/video_core/engines/shader_bytecode.h @@ -624,6 +624,19 @@ enum class ShuffleOperation : u64 { Bfly = 3, // shuffleXorNV }; +enum class ShfType : u64 { + Bits32 = 0, + U64 = 2, + S64 = 3, +}; + +enum class ShfXmode : u64 { + None = 0, + HI = 1, + X = 2, + XHI = 3, +}; + union Instruction { constexpr Instruction& operator=(const Instruction& instr) { value = instr.value; @@ -776,6 +789,13 @@ union Instruction { } shr; union { + BitField<37, 2, ShfType> type; + BitField<48, 2, ShfXmode> xmode; + BitField<50, 1, u64> wrap; + BitField<20, 6, u64> immediate; + } shf; + + union { BitField<39, 5, u64> shift_amount; BitField<48, 1, u64> negate_b; BitField<49, 1, u64> negate_a; @@ -1124,6 +1144,11 @@ union Instruction { } fset; union { + BitField<47, 1, u64> ftz; + BitField<48, 4, PredCondition> cond; + } fcmp; + + union { BitField<49, 1, u64> bf; BitField<35, 3, PredCondition> cond; BitField<50, 1, u64> ftz; @@ -1703,6 +1728,7 @@ public: BFE_C, BFE_R, BFE_IMM, + BFI_RC, BFI_IMM_R, BRA, BRX, @@ -1800,6 +1826,7 @@ public: ICMP_R, ICMP_CR, ICMP_IMM, + FCMP_R, MUFU, // Multi-Function Operator RRO_C, // Range Reduction Operator RRO_R, @@ -2104,6 +2131,7 @@ private: INST("0101110100100---", Id::HSETP2_R, Type::HalfSetPredicate, "HSETP2_R"), INST("0111111-0-------", Id::HSETP2_IMM, Type::HalfSetPredicate, "HSETP2_IMM"), INST("0101110100011---", Id::HSET2_R, Type::HalfSet, "HSET2_R"), + INST("010110111010----", Id::FCMP_R, Type::Arithmetic, "FCMP_R"), INST("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"), INST("0100110010010---", Id::RRO_C, Type::Arithmetic, "RRO_C"), INST("0101110010010---", Id::RRO_R, Type::Arithmetic, "RRO_R"), @@ -2128,6 +2156,7 @@ private: INST("0100110000000---", Id::BFE_C, Type::Bfe, "BFE_C"), INST("0101110000000---", Id::BFE_R, Type::Bfe, "BFE_R"), INST("0011100-00000---", Id::BFE_IMM, Type::Bfe, "BFE_IMM"), + INST("0101001111110---", Id::BFI_RC, Type::Bfi, "BFI_RC"), INST("0011011-11110---", Id::BFI_IMM_R, Type::Bfi, "BFI_IMM_R"), INST("0100110001000---", Id::LOP_C, Type::ArithmeticInteger, "LOP_C"), INST("0101110001000---", Id::LOP_R, Type::ArithmeticInteger, "LOP_R"), diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index b9c5c41a2..062ca83b8 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -23,7 +23,7 @@ MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192)); GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer, bool is_async) : system{system}, renderer{renderer}, is_async{is_async} { auto& rasterizer{renderer.Rasterizer()}; - memory_manager = std::make_unique<Tegra::MemoryManager>(system, rasterizer); + memory_manager = std::make_unique<Tegra::MemoryManager>(system); dma_pusher = std::make_unique<Tegra::DmaPusher>(*this); maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager); fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer); diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h index 08dc96bb3..882e2d9c7 100644 --- a/src/video_core/gpu_thread.h +++ b/src/video_core/gpu_thread.h @@ -86,7 +86,7 @@ struct CommandDataContainer { struct SynchState final { std::atomic_bool is_running{true}; - using CommandQueue = Common::SPSCQueue<CommandDataContainer>; + using CommandQueue = Common::MPSCQueue<CommandDataContainer>; CommandQueue queue; u64 last_fence{}; std::atomic<u64> signaled_fence{}; diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index 11848fbce..f1d50be3e 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -9,13 +9,12 @@ #include "core/hle/kernel/process.h" #include "core/hle/kernel/vm_manager.h" #include "core/memory.h" +#include "video_core/gpu.h" #include "video_core/memory_manager.h" -#include "video_core/rasterizer_interface.h" namespace Tegra { -MemoryManager::MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer) - : rasterizer{rasterizer}, system{system} { +MemoryManager::MemoryManager(Core::System& system) : system{system} { std::fill(page_table.pointers.begin(), page_table.pointers.end(), nullptr); std::fill(page_table.attributes.begin(), page_table.attributes.end(), Common::PageType::Unmapped); @@ -84,7 +83,8 @@ GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) { const auto cpu_addr = GpuToCpuAddress(gpu_addr); ASSERT(cpu_addr); - rasterizer.FlushAndInvalidateRegion(cache_addr, aligned_size); + system.GPU().FlushAndInvalidateRegion(cache_addr, aligned_size); + UnmapRange(gpu_addr, aligned_size); ASSERT(system.CurrentProcess() ->VMManager() @@ -242,7 +242,7 @@ void MemoryManager::ReadBlock(GPUVAddr src_addr, void* dest_buffer, const std::s switch (page_table.attributes[page_index]) { case Common::PageType::Memory: { const u8* src_ptr{page_table.pointers[page_index] + page_offset}; - rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount); + system.GPU().FlushRegion(ToCacheAddr(src_ptr), copy_amount); std::memcpy(dest_buffer, src_ptr, copy_amount); break; } @@ -292,7 +292,7 @@ void MemoryManager::WriteBlock(GPUVAddr dest_addr, const void* src_buffer, const switch (page_table.attributes[page_index]) { case Common::PageType::Memory: { u8* dest_ptr{page_table.pointers[page_index] + page_offset}; - rasterizer.InvalidateRegion(ToCacheAddr(dest_ptr), copy_amount); + system.GPU().InvalidateRegion(ToCacheAddr(dest_ptr), copy_amount); std::memcpy(dest_ptr, src_buffer, copy_amount); break; } @@ -340,7 +340,7 @@ void MemoryManager::CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, const std:: switch (page_table.attributes[page_index]) { case Common::PageType::Memory: { const u8* src_ptr{page_table.pointers[page_index] + page_offset}; - rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount); + system.GPU().FlushRegion(ToCacheAddr(src_ptr), copy_amount); WriteBlock(dest_addr, src_ptr, copy_amount); break; } diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index aea010087..393447eb4 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -10,10 +10,6 @@ #include "common/common_types.h" #include "common/page_table.h" -namespace VideoCore { -class RasterizerInterface; -} - namespace Core { class System; } @@ -51,7 +47,7 @@ struct VirtualMemoryArea { class MemoryManager final { public: - explicit MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer); + explicit MemoryManager(Core::System& system); ~MemoryManager(); GPUVAddr AllocateSpace(u64 size, u64 align); @@ -176,7 +172,6 @@ private: Common::PageTable page_table{page_bits}; VMAMap vma_map; - VideoCore::RasterizerInterface& rasterizer; Core::System& system; }; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 362942e09..46a7433ea 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -248,9 +248,6 @@ void RasterizerOpenGL::SetupVertexInstances(GLuint vao) { } GLintptr RasterizerOpenGL::SetupIndexBuffer() { - if (accelerate_draw != AccelDraw::Indexed) { - return 0; - } MICROPROFILE_SCOPE(OpenGL_Index); const auto& regs = system.GPU().Maxwell3D().regs; const std::size_t size = CalculateIndexBufferSize(); @@ -546,7 +543,8 @@ void RasterizerOpenGL::Clear() { } } -void RasterizerOpenGL::DrawPrelude() { +void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) { + MICROPROFILE_SCOPE(OpenGL_Drawing); auto& gpu = system.GPU().Maxwell3D(); SyncRasterizeEnable(state); @@ -567,9 +565,6 @@ void RasterizerOpenGL::DrawPrelude() { buffer_cache.Acquire(); - // Draw the vertex batch - const bool is_indexed = accelerate_draw == AccelDraw::Indexed; - std::size_t buffer_size = CalculateVertexArraysSize(); // Add space for index buffer @@ -596,7 +591,11 @@ void RasterizerOpenGL::DrawPrelude() { // Upload vertex and index data. SetupVertexBuffer(vao); SetupVertexInstances(vao); - index_buffer_offset = SetupIndexBuffer(); + + GLintptr index_buffer_offset; + if (is_indexed) { + index_buffer_offset = SetupIndexBuffer(); + } // Prepare packed bindings. bind_ubo_pushbuffer.Setup(); @@ -630,6 +629,7 @@ void RasterizerOpenGL::DrawPrelude() { // As all cached buffers are invalidated, we need to recheck their state. gpu.dirty.ResetVertexArrays(); } + gpu.dirty.memory_general = false; shader_program_manager->ApplyTo(state); state.Apply(); @@ -637,106 +637,33 @@ void RasterizerOpenGL::DrawPrelude() { if (texture_cache.TextureBarrier()) { glTextureBarrier(); } -} - -struct DrawParams { - bool is_indexed{}; - bool is_instanced{}; - GLenum primitive_mode{}; - GLint count{}; - GLint base_vertex{}; - - // Indexed settings - GLenum index_format{}; - GLintptr index_buffer_offset{}; - - // Instanced setting - GLint num_instances{}; - GLint base_instance{}; - - void DispatchDraw() { - if (is_indexed) { - const auto index_buffer_ptr = reinterpret_cast<const void*>(index_buffer_offset); - if (is_instanced) { - glDrawElementsInstancedBaseVertexBaseInstance(primitive_mode, count, index_format, - index_buffer_ptr, num_instances, - base_vertex, base_instance); - } else { - glDrawElementsBaseVertex(primitive_mode, count, index_format, index_buffer_ptr, - base_vertex); - } - } else { - if (is_instanced) { - glDrawArraysInstancedBaseInstance(primitive_mode, base_vertex, count, num_instances, - base_instance); - } else { - glDrawArrays(primitive_mode, base_vertex, count); - } - } - } -}; - -bool RasterizerOpenGL::DrawBatch(bool is_indexed) { - accelerate_draw = is_indexed ? AccelDraw::Indexed : AccelDraw::Arrays; - MICROPROFILE_SCOPE(OpenGL_Drawing); - - DrawPrelude(); - - auto& maxwell3d = system.GPU().Maxwell3D(); - const auto& regs = maxwell3d.regs; - const auto current_instance = maxwell3d.state.current_instance; - DrawParams draw_call{}; - draw_call.is_indexed = is_indexed; - draw_call.num_instances = static_cast<GLint>(1); - draw_call.base_instance = static_cast<GLint>(current_instance); - draw_call.is_instanced = current_instance > 0; - draw_call.primitive_mode = MaxwellToGL::PrimitiveTopology(regs.draw.topology); - if (draw_call.is_indexed) { - draw_call.count = static_cast<GLint>(regs.index_array.count); - draw_call.base_vertex = static_cast<GLint>(regs.vb_element_base); - draw_call.index_format = MaxwellToGL::IndexFormat(regs.index_array.format); - draw_call.index_buffer_offset = index_buffer_offset; + const GLuint base_instance = static_cast<GLuint>(gpu.regs.vb_base_instance); + const GLsizei num_instances = + static_cast<GLsizei>(is_instanced ? gpu.mme_draw.instance_count : 1); + if (is_indexed) { + const GLenum index_format = MaxwellToGL::IndexFormat(gpu.regs.index_array.format); + const GLint base_vertex = static_cast<GLint>(gpu.regs.vb_element_base); + const GLsizei num_vertices = static_cast<GLsizei>(gpu.regs.index_array.count); + glDrawElementsInstancedBaseVertexBaseInstance( + primitive_mode, num_vertices, index_format, + reinterpret_cast<const void*>(index_buffer_offset), num_instances, base_vertex, + base_instance); } else { - draw_call.count = static_cast<GLint>(regs.vertex_buffer.count); - draw_call.base_vertex = static_cast<GLint>(regs.vertex_buffer.first); + const GLint base_vertex = static_cast<GLint>(gpu.regs.vertex_buffer.first); + const GLsizei num_vertices = static_cast<GLsizei>(gpu.regs.vertex_buffer.count); + glDrawArraysInstancedBaseInstance(primitive_mode, base_vertex, num_vertices, num_instances, + base_instance); } - draw_call.DispatchDraw(); +} - maxwell3d.dirty.memory_general = false; - accelerate_draw = AccelDraw::Disabled; +bool RasterizerOpenGL::DrawBatch(bool is_indexed) { + Draw(is_indexed, false); return true; } bool RasterizerOpenGL::DrawMultiBatch(bool is_indexed) { - accelerate_draw = is_indexed ? AccelDraw::Indexed : AccelDraw::Arrays; - - MICROPROFILE_SCOPE(OpenGL_Drawing); - - DrawPrelude(); - - auto& maxwell3d = system.GPU().Maxwell3D(); - const auto& regs = maxwell3d.regs; - const auto& draw_setup = maxwell3d.mme_draw; - DrawParams draw_call{}; - draw_call.is_indexed = is_indexed; - draw_call.num_instances = static_cast<GLint>(draw_setup.instance_count); - draw_call.base_instance = static_cast<GLint>(regs.vb_base_instance); - draw_call.is_instanced = draw_setup.instance_count > 1; - draw_call.primitive_mode = MaxwellToGL::PrimitiveTopology(regs.draw.topology); - if (draw_call.is_indexed) { - draw_call.count = static_cast<GLint>(regs.index_array.count); - draw_call.base_vertex = static_cast<GLint>(regs.vb_element_base); - draw_call.index_format = MaxwellToGL::IndexFormat(regs.index_array.format); - draw_call.index_buffer_offset = index_buffer_offset; - } else { - draw_call.count = static_cast<GLint>(regs.vertex_buffer.count); - draw_call.base_vertex = static_cast<GLint>(regs.vertex_buffer.first); - } - draw_call.DispatchDraw(); - - maxwell3d.dirty.memory_general = false; - accelerate_draw = AccelDraw::Disabled; + Draw(is_indexed, true); return true; } diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 6a27cf497..0501f3828 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -103,7 +103,7 @@ private: std::size_t size); /// Syncs all the state, shaders, render targets and textures setting before a draw call. - void DrawPrelude(); + void Draw(bool is_indexed, bool is_instanced); /// Configures the current textures to use for the draw command. void SetupDrawTextures(std::size_t stage_index, const Shader& shader); @@ -220,12 +220,7 @@ private: GLintptr SetupIndexBuffer(); - GLintptr index_buffer_offset; - void SetupShaders(GLenum primitive_mode); - - enum class AccelDraw { Disabled, Arrays, Indexed }; - AccelDraw accelerate_draw = AccelDraw::Disabled; }; } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h index ea4f35663..7ed505628 100644 --- a/src/video_core/renderer_opengl/maxwell_to_gl.h +++ b/src/video_core/renderer_opengl/maxwell_to_gl.h @@ -47,8 +47,7 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) { case Maxwell::VertexAttribute::Size::Size_10_10_10_2: return GL_UNSIGNED_INT_2_10_10_10_REV; default: - LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString()); - UNREACHABLE(); + LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString()); return {}; } case Maxwell::VertexAttribute::Type::SignedInt: @@ -72,8 +71,7 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) { case Maxwell::VertexAttribute::Size::Size_10_10_10_2: return GL_INT_2_10_10_10_REV; default: - LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString()); - UNREACHABLE(); + LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString()); return {}; } case Maxwell::VertexAttribute::Type::Float: @@ -89,13 +87,19 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) { case Maxwell::VertexAttribute::Size::Size_32_32_32_32: return GL_FLOAT; default: - LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString()); - UNREACHABLE(); + LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString()); + return {}; + } + case Maxwell::VertexAttribute::Type::UnsignedScaled: + switch (attrib.size) { + case Maxwell::VertexAttribute::Size::Size_8_8: + return GL_UNSIGNED_BYTE; + default: + LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString()); return {}; } default: - LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex type={}", attrib.TypeString()); - UNREACHABLE(); + LOG_ERROR(Render_OpenGL, "Unimplemented vertex type={}", attrib.TypeString()); return {}; } } diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp new file mode 100644 index 000000000..d5032b432 --- /dev/null +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp @@ -0,0 +1,265 @@ +// Copyright 2018 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <memory> +#include <optional> +#include <vector> + +#include <fmt/format.h> + +#include "common/assert.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" +#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/renderer_vulkan.h" +#include "video_core/renderer_vulkan/vk_blit_screen.h" +#include "video_core/renderer_vulkan/vk_device.h" +#include "video_core/renderer_vulkan/vk_memory_manager.h" +#include "video_core/renderer_vulkan/vk_rasterizer.h" +#include "video_core/renderer_vulkan/vk_resource_manager.h" +#include "video_core/renderer_vulkan/vk_scheduler.h" +#include "video_core/renderer_vulkan/vk_swapchain.h" + +namespace Vulkan { + +namespace { + +VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity_, + VkDebugUtilsMessageTypeFlagsEXT type, + const VkDebugUtilsMessengerCallbackDataEXT* data, + [[maybe_unused]] void* user_data) { + const vk::DebugUtilsMessageSeverityFlagBitsEXT severity{severity_}; + const char* message{data->pMessage}; + + if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eError) { + LOG_CRITICAL(Render_Vulkan, "{}", message); + } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eWarning) { + LOG_WARNING(Render_Vulkan, "{}", message); + } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eInfo) { + LOG_INFO(Render_Vulkan, "{}", message); + } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eVerbose) { + LOG_DEBUG(Render_Vulkan, "{}", message); + } + return VK_FALSE; +} + +std::string GetReadableVersion(u32 version) { + return fmt::format("{}.{}.{}", VK_VERSION_MAJOR(version), VK_VERSION_MINOR(version), + VK_VERSION_PATCH(version)); +} + +std::string GetDriverVersion(const VKDevice& device) { + // Extracted from + // https://github.com/SaschaWillems/vulkan.gpuinfo.org/blob/5dddea46ea1120b0df14eef8f15ff8e318e35462/functions.php#L308-L314 + const u32 version = device.GetDriverVersion(); + + if (device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) { + const u32 major = (version >> 22) & 0x3ff; + const u32 minor = (version >> 14) & 0x0ff; + const u32 secondary = (version >> 6) & 0x0ff; + const u32 tertiary = version & 0x003f; + return fmt::format("{}.{}.{}.{}", major, minor, secondary, tertiary); + } + if (device.GetDriverID() == vk::DriverIdKHR::eIntelProprietaryWindows) { + const u32 major = version >> 14; + const u32 minor = version & 0x3fff; + return fmt::format("{}.{}", major, minor); + } + + return GetReadableVersion(version); +} + +std::string BuildCommaSeparatedExtensions(std::vector<std::string> available_extensions) { + std::sort(std::begin(available_extensions), std::end(available_extensions)); + + static constexpr std::size_t AverageExtensionSize = 64; + std::string separated_extensions; + separated_extensions.reserve(available_extensions.size() * AverageExtensionSize); + + const auto end = std::end(available_extensions); + for (auto extension = std::begin(available_extensions); extension != end; ++extension) { + if (const bool is_last = extension + 1 == end; is_last) { + separated_extensions += *extension; + } else { + separated_extensions += fmt::format("{},", *extension); + } + } + return separated_extensions; +} + +} // Anonymous namespace + +RendererVulkan::RendererVulkan(Core::Frontend::EmuWindow& window, Core::System& system) + : RendererBase(window), system{system} {} + +RendererVulkan::~RendererVulkan() { + ShutDown(); +} + +void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { + const auto& layout = render_window.GetFramebufferLayout(); + if (framebuffer && layout.width > 0 && layout.height > 0 && render_window.IsShown()) { + const VAddr framebuffer_addr = framebuffer->address + framebuffer->offset; + const bool use_accelerated = + rasterizer->AccelerateDisplay(*framebuffer, framebuffer_addr, framebuffer->stride); + const bool is_srgb = use_accelerated && screen_info.is_srgb; + if (swapchain->HasFramebufferChanged(layout) || swapchain->GetSrgbState() != is_srgb) { + swapchain->Create(layout.width, layout.height, is_srgb); + blit_screen->Recreate(); + } + + scheduler->WaitWorker(); + + swapchain->AcquireNextImage(); + const auto [fence, render_semaphore] = blit_screen->Draw(*framebuffer, use_accelerated); + + scheduler->Flush(false, render_semaphore); + + if (swapchain->Present(render_semaphore, fence)) { + blit_screen->Recreate(); + } + + render_window.SwapBuffers(); + rasterizer->TickFrame(); + } + + render_window.PollEvents(); +} + +bool RendererVulkan::Init() { + PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr{}; + render_window.RetrieveVulkanHandlers(&vkGetInstanceProcAddr, &instance, &surface); + const vk::DispatchLoaderDynamic dldi(instance, vkGetInstanceProcAddr); + + std::optional<vk::DebugUtilsMessengerEXT> callback; + if (Settings::values.renderer_debug && dldi.vkCreateDebugUtilsMessengerEXT) { + callback = CreateDebugCallback(dldi); + if (!callback) { + return false; + } + } + + if (!PickDevices(dldi)) { + if (callback) { + instance.destroy(*callback, nullptr, dldi); + } + return false; + } + debug_callback = UniqueDebugUtilsMessengerEXT( + *callback, vk::ObjectDestroy<vk::Instance, vk::DispatchLoaderDynamic>( + instance, nullptr, device->GetDispatchLoader())); + + Report(); + + memory_manager = std::make_unique<VKMemoryManager>(*device); + + resource_manager = std::make_unique<VKResourceManager>(*device); + + const auto& framebuffer = render_window.GetFramebufferLayout(); + swapchain = std::make_unique<VKSwapchain>(surface, *device); + swapchain->Create(framebuffer.width, framebuffer.height, false); + + scheduler = std::make_unique<VKScheduler>(*device, *resource_manager); + + rasterizer = std::make_unique<RasterizerVulkan>(system, render_window, screen_info, *device, + *resource_manager, *memory_manager, *scheduler); + + blit_screen = std::make_unique<VKBlitScreen>(system, render_window, *rasterizer, *device, + *resource_manager, *memory_manager, *swapchain, + *scheduler, screen_info); + + return true; +} + +void RendererVulkan::ShutDown() { + if (!device) { + return; + } + const auto dev = device->GetLogical(); + const auto& dld = device->GetDispatchLoader(); + if (dev && dld.vkDeviceWaitIdle) { + dev.waitIdle(dld); + } + + rasterizer.reset(); + blit_screen.reset(); + scheduler.reset(); + swapchain.reset(); + memory_manager.reset(); + resource_manager.reset(); + device.reset(); +} + +std::optional<vk::DebugUtilsMessengerEXT> RendererVulkan::CreateDebugCallback( + const vk::DispatchLoaderDynamic& dldi) { + const vk::DebugUtilsMessengerCreateInfoEXT callback_ci( + {}, + vk::DebugUtilsMessageSeverityFlagBitsEXT::eError | + vk::DebugUtilsMessageSeverityFlagBitsEXT::eWarning | + vk::DebugUtilsMessageSeverityFlagBitsEXT::eInfo | + vk::DebugUtilsMessageSeverityFlagBitsEXT::eVerbose, + vk::DebugUtilsMessageTypeFlagBitsEXT::eGeneral | + vk::DebugUtilsMessageTypeFlagBitsEXT::eValidation | + vk::DebugUtilsMessageTypeFlagBitsEXT::ePerformance, + &DebugCallback, nullptr); + vk::DebugUtilsMessengerEXT callback; + if (instance.createDebugUtilsMessengerEXT(&callback_ci, nullptr, &callback, dldi) != + vk::Result::eSuccess) { + LOG_ERROR(Render_Vulkan, "Failed to create debug callback"); + return {}; + } + return callback; +} + +bool RendererVulkan::PickDevices(const vk::DispatchLoaderDynamic& dldi) { + const auto devices = instance.enumeratePhysicalDevices(dldi); + + // TODO(Rodrigo): Choose device from config file + const s32 device_index = Settings::values.vulkan_device; + if (device_index < 0 || device_index >= static_cast<s32>(devices.size())) { + LOG_ERROR(Render_Vulkan, "Invalid device index {}!", device_index); + return false; + } + const vk::PhysicalDevice physical_device = devices[device_index]; + + if (!VKDevice::IsSuitable(dldi, physical_device, surface)) { + return false; + } + + device = std::make_unique<VKDevice>(dldi, physical_device, surface); + return device->Create(dldi, instance); +} + +void RendererVulkan::Report() const { + const std::string vendor_name{device->GetVendorName()}; + const std::string model_name{device->GetModelName()}; + const std::string driver_version = GetDriverVersion(*device); + const std::string driver_name = fmt::format("{} {}", vendor_name, driver_version); + + const std::string api_version = GetReadableVersion(device->GetApiVersion()); + + const std::string extensions = BuildCommaSeparatedExtensions(device->GetAvailableExtensions()); + + LOG_INFO(Render_Vulkan, "Driver: {}", driver_name); + LOG_INFO(Render_Vulkan, "Device: {}", model_name); + LOG_INFO(Render_Vulkan, "Vulkan: {}", api_version); + + auto& telemetry_session = system.TelemetrySession(); + constexpr auto field = Telemetry::FieldType::UserSystem; + telemetry_session.AddField(field, "GPU_Vendor", vendor_name); + telemetry_session.AddField(field, "GPU_Model", model_name); + telemetry_session.AddField(field, "GPU_Vulkan_Driver", driver_name); + telemetry_session.AddField(field, "GPU_Vulkan_Version", api_version); + telemetry_session.AddField(field, "GPU_Vulkan_Extensions", extensions); +} + +} // namespace Vulkan
\ No newline at end of file diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp index 939eebe83..9840f26e5 100644 --- a/src/video_core/renderer_vulkan/vk_device.cpp +++ b/src/video_core/renderer_vulkan/vk_device.cpp @@ -400,8 +400,10 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami VK_EXT_SHADER_VIEWPORT_INDEX_LAYER_EXTENSION_NAME, true); Test(extension, ext_subgroup_size_control, VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME, false); - Test(extension, nv_device_diagnostic_checkpoints, - VK_NV_DEVICE_DIAGNOSTIC_CHECKPOINTS_EXTENSION_NAME, true); + if (Settings::values.renderer_debug) { + Test(extension, nv_device_diagnostic_checkpoints, + VK_NV_DEVICE_DIAGNOSTIC_CHECKPOINTS_EXTENSION_NAME, true); + } } if (khr_shader_float16_int8) { diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index d2c6b1189..aada38702 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -571,7 +571,7 @@ RasterizerVulkan::Texceptions RasterizerVulkan::UpdateAttachments() { color_attachments[rt] = texture_cache.GetColorBufferSurface(rt, true); } if (color_attachments[rt] && WalkAttachmentOverlaps(*color_attachments[rt])) { - texceptions.set(rt); + texceptions[rt] = true; } } @@ -579,7 +579,7 @@ RasterizerVulkan::Texceptions RasterizerVulkan::UpdateAttachments() { zeta_attachment = texture_cache.GetDepthBufferSurface(true); } if (zeta_attachment && WalkAttachmentOverlaps(*zeta_attachment)) { - texceptions.set(ZETA_TEXCEPTION_INDEX); + texceptions[ZETA_TEXCEPTION_INDEX] = true; } texture_cache.GuardRenderTargets(false); @@ -1122,11 +1122,12 @@ RenderPassParams RasterizerVulkan::GetRenderPassParams(Texceptions texceptions) for (std::size_t rt = 0; rt < static_cast<std::size_t>(regs.rt_control.count); ++rt) { const auto& rendertarget = regs.rt[rt]; - if (rendertarget.Address() == 0 || rendertarget.format == Tegra::RenderTargetFormat::NONE) + if (rendertarget.Address() == 0 || rendertarget.format == Tegra::RenderTargetFormat::NONE) { continue; + } renderpass_params.color_attachments.push_back(RenderPassParams::ColorAttachment{ static_cast<u32>(rt), PixelFormatFromRenderTargetFormat(rendertarget.format), - texceptions.test(rt)}); + texceptions[rt]}); } renderpass_params.has_zeta = regs.zeta_enable; diff --git a/src/video_core/shader/ast.h b/src/video_core/shader/ast.h index a2f0044ba..cca13bcde 100644 --- a/src/video_core/shader/ast.h +++ b/src/video_core/shader/ast.h @@ -65,8 +65,8 @@ public: void DetachSegment(ASTNode start, ASTNode end); void Remove(ASTNode node); - ASTNode first{}; - ASTNode last{}; + ASTNode first; + ASTNode last; }; class ASTProgram { @@ -299,9 +299,9 @@ private: friend class ASTZipper; ASTData data; - ASTNode parent{}; - ASTNode next{}; - ASTNode previous{}; + ASTNode parent; + ASTNode next; + ASTNode previous; ASTZipper* manager{}; }; diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp index fcedd2af6..90240c765 100644 --- a/src/video_core/shader/decode/arithmetic.cpp +++ b/src/video_core/shader/decode/arithmetic.cpp @@ -21,7 +21,7 @@ u32 ShaderIR::DecodeArithmetic(NodeBlock& bb, u32 pc) { Node op_a = GetRegister(instr.gpr8); - Node op_b = [&]() -> Node { + Node op_b = [&] { if (instr.is_b_imm) { return GetImmediate19(instr); } else if (instr.is_b_gpr) { @@ -141,6 +141,15 @@ u32 ShaderIR::DecodeArithmetic(NodeBlock& bb, u32 pc) { SetRegister(bb, instr.gpr0, value); break; } + case OpCode::Id::FCMP_R: { + UNIMPLEMENTED_IF(instr.fcmp.ftz == 0); + Node op_c = GetRegister(instr.gpr39); + Node comp = GetPredicateComparisonFloat(instr.fcmp.cond, std::move(op_c), Immediate(0.0f)); + SetRegister( + bb, instr.gpr0, + Operation(OperationCode::Select, std::move(comp), std::move(op_a), std::move(op_b))); + break; + } case OpCode::Id::RRO_C: case OpCode::Id::RRO_R: case OpCode::Id::RRO_IMM: { diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp index 371fae127..e60875cc4 100644 --- a/src/video_core/shader/decode/arithmetic_integer.cpp +++ b/src/video_core/shader/decode/arithmetic_integer.cpp @@ -297,7 +297,7 @@ void ShaderIR::WriteLop3Instruction(NodeBlock& bb, Register dest, Node op_a, Nod const Node one = Immediate(1); const Node two = Immediate(2); - Node value{}; + Node value; for (u32 i = 0; i < lop_iterations; ++i) { const Node shift_amount = Immediate(i); diff --git a/src/video_core/shader/decode/bfi.cpp b/src/video_core/shader/decode/bfi.cpp index 8be1119df..f992bbe2a 100644 --- a/src/video_core/shader/decode/bfi.cpp +++ b/src/video_core/shader/decode/bfi.cpp @@ -17,10 +17,13 @@ u32 ShaderIR::DecodeBfi(NodeBlock& bb, u32 pc) { const Instruction instr = {program_code[pc]}; const auto opcode = OpCode::Decode(instr); - const auto [base, packed_shift] = [&]() -> std::tuple<Node, Node> { + const auto [packed_shift, base] = [&]() -> std::pair<Node, Node> { switch (opcode->get().GetId()) { + case OpCode::Id::BFI_RC: + return {GetRegister(instr.gpr39), + GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset)}; case OpCode::Id::BFI_IMM_R: - return {GetRegister(instr.gpr39), Immediate(instr.alu.GetSignedImm20_20())}; + return {Immediate(instr.alu.GetSignedImm20_20()), GetRegister(instr.gpr39)}; default: UNREACHABLE(); return {Immediate(0), Immediate(0)}; diff --git a/src/video_core/shader/decode/other.cpp b/src/video_core/shader/decode/other.cpp index 7321698b2..4944e9d69 100644 --- a/src/video_core/shader/decode/other.cpp +++ b/src/video_core/shader/decode/other.cpp @@ -69,13 +69,16 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { case OpCode::Id::MOV_SYS: { const Node value = [this, instr] { switch (instr.sys20) { + case SystemVariable::LaneId: + LOG_WARNING(HW_GPU, "MOV_SYS instruction with LaneId is incomplete"); + return Immediate(0U); case SystemVariable::InvocationId: return Operation(OperationCode::InvocationId); case SystemVariable::Ydirection: return Operation(OperationCode::YNegate); case SystemVariable::InvocationInfo: LOG_WARNING(HW_GPU, "MOV_SYS instruction with InvocationInfo is incomplete"); - return Immediate(0u); + return Immediate(0U); case SystemVariable::Tid: { Node value = Immediate(0); value = BitfieldInsert(value, Operation(OperationCode::LocalInvocationIdX), 0, 9); @@ -188,7 +191,7 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T, "SYNC condition code used: {}", static_cast<u32>(cc)); - if (disable_flow_stack) { + if (decompiled) { break; } @@ -200,7 +203,7 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { const Tegra::Shader::ConditionCode cc = instr.flow_condition_code; UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T, "BRK condition code used: {}", static_cast<u32>(cc)); - if (disable_flow_stack) { + if (decompiled) { break; } diff --git a/src/video_core/shader/decode/shift.cpp b/src/video_core/shader/decode/shift.cpp index d419e9c45..3b391d3e6 100644 --- a/src/video_core/shader/decode/shift.cpp +++ b/src/video_core/shader/decode/shift.cpp @@ -10,8 +10,80 @@ namespace VideoCommon::Shader { +using std::move; using Tegra::Shader::Instruction; using Tegra::Shader::OpCode; +using Tegra::Shader::ShfType; +using Tegra::Shader::ShfXmode; + +namespace { + +Node IsFull(Node shift) { + return Operation(OperationCode::LogicalIEqual, move(shift), Immediate(32)); +} + +Node Shift(OperationCode opcode, Node value, Node shift) { + Node is_full = Operation(OperationCode::LogicalIEqual, shift, Immediate(32)); + Node shifted = Operation(opcode, move(value), shift); + return Operation(OperationCode::Select, IsFull(move(shift)), Immediate(0), move(shifted)); +} + +Node ClampShift(Node shift, s32 size = 32) { + shift = Operation(OperationCode::IMax, move(shift), Immediate(0)); + return Operation(OperationCode::IMin, move(shift), Immediate(size)); +} + +Node WrapShift(Node shift, s32 size = 32) { + return Operation(OperationCode::UBitwiseAnd, move(shift), Immediate(size - 1)); +} + +Node ShiftRight(Node low, Node high, Node shift, Node low_shift, ShfType type) { + // These values are used when the shift value is less than 32 + Node less_low = Shift(OperationCode::ILogicalShiftRight, low, shift); + Node less_high = Shift(OperationCode::ILogicalShiftLeft, high, low_shift); + Node less = Operation(OperationCode::IBitwiseOr, move(less_high), move(less_low)); + + if (type == ShfType::Bits32) { + // On 32 bit shifts we are either full (shifting 32) or shifting less than 32 bits + return Operation(OperationCode::Select, IsFull(move(shift)), move(high), move(less)); + } + + // And these when it's larger than or 32 + const bool is_signed = type == ShfType::S64; + const auto opcode = SignedToUnsignedCode(OperationCode::IArithmeticShiftRight, is_signed); + Node reduced = Operation(OperationCode::IAdd, shift, Immediate(-32)); + Node greater = Shift(opcode, high, move(reduced)); + + Node is_less = Operation(OperationCode::LogicalILessThan, shift, Immediate(32)); + Node is_zero = Operation(OperationCode::LogicalIEqual, move(shift), Immediate(0)); + + Node value = Operation(OperationCode::Select, move(is_less), move(less), move(greater)); + return Operation(OperationCode::Select, move(is_zero), move(high), move(value)); +} + +Node ShiftLeft(Node low, Node high, Node shift, Node low_shift, ShfType type) { + // These values are used when the shift value is less than 32 + Node less_low = Operation(OperationCode::ILogicalShiftRight, low, low_shift); + Node less_high = Operation(OperationCode::ILogicalShiftLeft, high, shift); + Node less = Operation(OperationCode::IBitwiseOr, move(less_low), move(less_high)); + + if (type == ShfType::Bits32) { + // On 32 bit shifts we are either full (shifting 32) or shifting less than 32 bits + return Operation(OperationCode::Select, IsFull(move(shift)), move(low), move(less)); + } + + // And these when it's larger than or 32 + Node reduced = Operation(OperationCode::IAdd, shift, Immediate(-32)); + Node greater = Shift(OperationCode::ILogicalShiftLeft, move(low), move(reduced)); + + Node is_less = Operation(OperationCode::LogicalILessThan, shift, Immediate(32)); + Node is_zero = Operation(OperationCode::LogicalIEqual, move(shift), Immediate(0)); + + Node value = Operation(OperationCode::Select, move(is_less), move(less), move(greater)); + return Operation(OperationCode::Select, move(is_zero), move(high), move(value)); +} + +} // Anonymous namespace u32 ShaderIR::DecodeShift(NodeBlock& bb, u32 pc) { const Instruction instr = {program_code[pc]}; @@ -28,29 +100,48 @@ u32 ShaderIR::DecodeShift(NodeBlock& bb, u32 pc) { } }(); - switch (opcode->get().GetId()) { + switch (const auto opid = opcode->get().GetId(); opid) { case OpCode::Id::SHR_C: case OpCode::Id::SHR_R: case OpCode::Id::SHR_IMM: { - if (instr.shr.wrap) { - op_b = Operation(OperationCode::UBitwiseAnd, std::move(op_b), Immediate(0x1f)); - } else { - op_b = Operation(OperationCode::IMax, std::move(op_b), Immediate(0)); - op_b = Operation(OperationCode::IMin, std::move(op_b), Immediate(31)); - } + op_b = instr.shr.wrap ? WrapShift(move(op_b)) : ClampShift(move(op_b)); Node value = SignedOperation(OperationCode::IArithmeticShiftRight, instr.shift.is_signed, - std::move(op_a), std::move(op_b)); + move(op_a), move(op_b)); SetInternalFlagsFromInteger(bb, value, instr.generates_cc); - SetRegister(bb, instr.gpr0, std::move(value)); + SetRegister(bb, instr.gpr0, move(value)); break; } case OpCode::Id::SHL_C: case OpCode::Id::SHL_R: case OpCode::Id::SHL_IMM: { - const Node value = Operation(OperationCode::ILogicalShiftLeft, op_a, op_b); + Node value = Operation(OperationCode::ILogicalShiftLeft, op_a, op_b); SetInternalFlagsFromInteger(bb, value, instr.generates_cc); - SetRegister(bb, instr.gpr0, value); + SetRegister(bb, instr.gpr0, move(value)); + break; + } + case OpCode::Id::SHF_RIGHT_R: + case OpCode::Id::SHF_RIGHT_IMM: + case OpCode::Id::SHF_LEFT_R: + case OpCode::Id::SHF_LEFT_IMM: { + UNIMPLEMENTED_IF(instr.generates_cc); + UNIMPLEMENTED_IF_MSG(instr.shf.xmode != ShfXmode::None, "xmode={}", + static_cast<int>(instr.shf.xmode.Value())); + + if (instr.is_b_imm) { + op_b = Immediate(static_cast<u32>(instr.shf.immediate)); + } + const s32 size = instr.shf.type == ShfType::Bits32 ? 32 : 64; + Node shift = instr.shf.wrap ? WrapShift(move(op_b), size) : ClampShift(move(op_b), size); + + Node negated_shift = Operation(OperationCode::INegate, shift); + Node low_shift = Operation(OperationCode::IAdd, move(negated_shift), Immediate(32)); + + const bool is_right = opid == OpCode::Id::SHF_RIGHT_R || opid == OpCode::Id::SHF_RIGHT_IMM; + Node value = (is_right ? ShiftRight : ShiftLeft)( + move(op_a), GetRegister(instr.gpr39), move(shift), move(low_shift), instr.shf.type); + + SetRegister(bb, instr.gpr0, move(value)); break; } default: diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp index d980535b1..351c8c2f1 100644 --- a/src/video_core/shader/decode/texture.cpp +++ b/src/video_core/shader/decode/texture.cpp @@ -630,7 +630,7 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type, aoffi = GetAoffiCoordinates(GetRegister(parameter_register++), coord_count, false); } - Node dc{}; + Node dc; if (depth_compare) { // Depth is always stored in the register signaled by gpr20 or in the next register if lod // or bias are used @@ -666,7 +666,7 @@ Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type, const Node array = is_array ? GetRegister(array_register) : nullptr; - Node dc{}; + Node dc; if (depth_compare) { // Depth is always stored in the register signaled by gpr20 or in the next register if lod // or bias are used diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h index 5f83403db..a0a7b9111 100644 --- a/src/video_core/shader/node.h +++ b/src/video_core/shader/node.h @@ -443,9 +443,9 @@ struct MetaTexture { std::vector<Node> derivates; Node bias; Node lod; - Node component{}; + Node component; u32 element{}; - Node index{}; + Node index; }; struct MetaImage { diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp index ea39bca54..face8c943 100644 --- a/src/video_core/shader/track.cpp +++ b/src/video_core/shader/track.cpp @@ -41,8 +41,8 @@ std::optional<std::pair<Node, Node>> DecoupleIndirectRead(const OperationNode& o if (operation.GetCode() != OperationCode::UAdd) { return std::nullopt; } - Node gpr{}; - Node offset{}; + Node gpr; + Node offset; ASSERT(operation.GetOperandsCount() == 2); for (std::size_t i = 0; i < operation.GetOperandsCount(); i++) { Node operand = operation[i]; diff --git a/src/video_core/video_core.cpp b/src/video_core/video_core.cpp index 8e947394c..a5f81a8a0 100644 --- a/src/video_core/video_core.cpp +++ b/src/video_core/video_core.cpp @@ -3,19 +3,32 @@ // Refer to the license.txt file included. #include <memory> +#include "common/logging/log.h" #include "core/core.h" #include "core/settings.h" #include "video_core/gpu_asynch.h" #include "video_core/gpu_synch.h" #include "video_core/renderer_base.h" #include "video_core/renderer_opengl/renderer_opengl.h" +#ifdef HAS_VULKAN +#include "video_core/renderer_vulkan/renderer_vulkan.h" +#endif #include "video_core/video_core.h" namespace VideoCore { std::unique_ptr<RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_window, Core::System& system) { - return std::make_unique<OpenGL::RendererOpenGL>(emu_window, system); + switch (Settings::values.renderer_backend) { + case Settings::RendererBackend::OpenGL: + return std::make_unique<OpenGL::RendererOpenGL>(emu_window, system); +#ifdef HAS_VULKAN + case Settings::RendererBackend::Vulkan: + return std::make_unique<Vulkan::RendererVulkan>(emu_window, system); +#endif + default: + return nullptr; + } } std::unique_ptr<Tegra::GPU> CreateGPU(Core::System& system) { |