diff options
Diffstat (limited to '')
78 files changed, 2223 insertions, 1625 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index f617665de..4742bcbe9 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -52,6 +52,8 @@ add_library(video_core STATIC engines/puller.cpp engines/puller.h framebuffer_config.h + fsr.cpp + fsr.h host1x/codecs/codec.cpp host1x/codecs/codec.h host1x/codecs/h264.cpp @@ -100,6 +102,8 @@ add_library(video_core STATIC renderer_null/null_rasterizer.h renderer_null/renderer_null.cpp renderer_null/renderer_null.h + renderer_opengl/blit_image.cpp + renderer_opengl/blit_image.h renderer_opengl/gl_buffer_cache.cpp renderer_opengl/gl_buffer_cache.h renderer_opengl/gl_compute_pipeline.cpp @@ -108,6 +112,8 @@ add_library(video_core STATIC renderer_opengl/gl_device.h renderer_opengl/gl_fence_manager.cpp renderer_opengl/gl_fence_manager.h + renderer_opengl/gl_fsr.cpp + renderer_opengl/gl_fsr.h renderer_opengl/gl_graphics_pipeline.cpp renderer_opengl/gl_graphics_pipeline.h renderer_opengl/gl_rasterizer.cpp diff --git a/src/video_core/buffer_cache/buffer_base.h b/src/video_core/buffer_cache/buffer_base.h index c47b7d866..92d77eef2 100644 --- a/src/video_core/buffer_cache/buffer_base.h +++ b/src/video_core/buffer_cache/buffer_base.h @@ -430,7 +430,7 @@ private: if (query_begin >= SizeBytes() || size < 0) { return; } - [[maybe_unused]] u64* const untracked_words = Array<Type::Untracked>(); + u64* const untracked_words = Array<Type::Untracked>(); u64* const state_words = Array<type>(); const u64 query_end = query_begin + std::min(static_cast<u64>(size), SizeBytes()); u64* const words_begin = state_words + query_begin / BYTES_PER_WORD; @@ -483,7 +483,7 @@ private: NotifyRasterizer<true>(word_index, current_bits, ~u64{0}); } // Exclude CPU modified pages when visiting GPU pages - const u64 word = current_word; + const u64 word = current_word & ~(type == Type::GPU ? untracked_words[word_index] : 0); u64 page = page_begin; page_begin = 0; @@ -531,7 +531,7 @@ private: [[nodiscard]] bool IsRegionModified(u64 offset, u64 size) const noexcept { static_assert(type != Type::Untracked); - [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>(); + const u64* const untracked_words = Array<Type::Untracked>(); const u64* const state_words = Array<type>(); const u64 num_query_words = size / BYTES_PER_WORD + 1; const u64 word_begin = offset / BYTES_PER_WORD; @@ -539,7 +539,8 @@ private: const u64 page_limit = Common::DivCeil(offset + size, BYTES_PER_PAGE); u64 page_index = (offset / BYTES_PER_PAGE) % PAGES_PER_WORD; for (u64 word_index = word_begin; word_index < word_end; ++word_index, page_index = 0) { - const u64 word = state_words[word_index]; + const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0; + const u64 word = state_words[word_index] & ~off_word; if (word == 0) { continue; } @@ -563,7 +564,7 @@ private: [[nodiscard]] std::pair<u64, u64> ModifiedRegion(u64 offset, u64 size) const noexcept { static_assert(type != Type::Untracked); - [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>(); + const u64* const untracked_words = Array<Type::Untracked>(); const u64* const state_words = Array<type>(); const u64 num_query_words = size / BYTES_PER_WORD + 1; const u64 word_begin = offset / BYTES_PER_WORD; @@ -573,7 +574,8 @@ private: u64 begin = std::numeric_limits<u64>::max(); u64 end = 0; for (u64 word_index = word_begin; word_index < word_end; ++word_index) { - const u64 word = state_words[word_index]; + const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0; + const u64 word = state_words[word_index] & ~off_word; if (word == 0) { continue; } diff --git a/src/video_core/engines/draw_manager.cpp b/src/video_core/engines/draw_manager.cpp index 2437121ce..1d22d25f1 100644 --- a/src/video_core/engines/draw_manager.cpp +++ b/src/video_core/engines/draw_manager.cpp @@ -51,6 +51,10 @@ void DrawManager::ProcessMethodCall(u32 method, u32 argument) { LOG_WARNING(HW_GPU, "(STUBBED) called"); break; } + case MAXWELL3D_REG_INDEX(draw_texture.src_y0): { + DrawTexture(); + break; + } default: break; } @@ -179,6 +183,33 @@ void DrawManager::DrawIndexSmall(u32 argument) { ProcessDraw(true, 1); } +void DrawManager::DrawTexture() { + const auto& regs{maxwell3d->regs}; + draw_texture_state.dst_x0 = static_cast<float>(regs.draw_texture.dst_x0) / 4096.f; + draw_texture_state.dst_y0 = static_cast<float>(regs.draw_texture.dst_y0) / 4096.f; + const auto dst_width = static_cast<float>(regs.draw_texture.dst_width) / 4096.f; + const auto dst_height = static_cast<float>(regs.draw_texture.dst_height) / 4096.f; + const bool lower_left{regs.window_origin.mode != + Maxwell3D::Regs::WindowOrigin::Mode::UpperLeft}; + if (lower_left) { + draw_texture_state.dst_y0 -= dst_height; + } + draw_texture_state.dst_x1 = draw_texture_state.dst_x0 + dst_width; + draw_texture_state.dst_y1 = draw_texture_state.dst_y0 + dst_height; + draw_texture_state.src_x0 = static_cast<float>(regs.draw_texture.src_x0) / 4096.f; + draw_texture_state.src_y0 = static_cast<float>(regs.draw_texture.src_y0) / 4096.f; + draw_texture_state.src_x1 = + (static_cast<float>(regs.draw_texture.dx_du) / 4294967296.f) * dst_width + + draw_texture_state.src_x0; + draw_texture_state.src_y1 = + (static_cast<float>(regs.draw_texture.dy_dv) / 4294967296.f) * dst_height + + draw_texture_state.src_y0; + draw_texture_state.src_sampler = regs.draw_texture.src_sampler; + draw_texture_state.src_texture = regs.draw_texture.src_texture; + + maxwell3d->rasterizer->DrawTexture(); +} + void DrawManager::UpdateTopology() { const auto& regs{maxwell3d->regs}; switch (regs.primitive_topology_control) { diff --git a/src/video_core/engines/draw_manager.h b/src/video_core/engines/draw_manager.h index 58d1b2d59..7c22c49f1 100644 --- a/src/video_core/engines/draw_manager.h +++ b/src/video_core/engines/draw_manager.h @@ -32,6 +32,19 @@ public: std::vector<u8> inline_index_draw_indexes; }; + struct DrawTextureState { + f32 dst_x0; + f32 dst_y0; + f32 dst_x1; + f32 dst_y1; + f32 src_x0; + f32 src_y0; + f32 src_x1; + f32 src_y1; + u32 src_sampler; + u32 src_texture; + }; + struct IndirectParams { bool is_indexed; bool include_count; @@ -64,6 +77,10 @@ public: return draw_state; } + const DrawTextureState& GetDrawTextureState() const { + return draw_texture_state; + } + IndirectParams& GetIndirectParams() { return indirect_state; } @@ -81,6 +98,8 @@ private: void DrawIndexSmall(u32 argument); + void DrawTexture(); + void UpdateTopology(); void ProcessDraw(bool draw_indexed, u32 instance_count); @@ -89,6 +108,7 @@ private: Maxwell3D* maxwell3d{}; State draw_state{}; + DrawTextureState draw_texture_state{}; IndirectParams indirect_state{}; }; } // namespace Tegra::Engines diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index 97f547789..614d61db4 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -149,6 +149,7 @@ bool Maxwell3D::IsMethodExecutable(u32 method) { case MAXWELL3D_REG_INDEX(inline_index_4x8.index0): case MAXWELL3D_REG_INDEX(vertex_array_instance_first): case MAXWELL3D_REG_INDEX(vertex_array_instance_subsequent): + case MAXWELL3D_REG_INDEX(draw_texture.src_y0): case MAXWELL3D_REG_INDEX(wait_for_idle): case MAXWELL3D_REG_INDEX(shadow_ram_control): case MAXWELL3D_REG_INDEX(load_mme.instruction_ptr): @@ -185,6 +186,7 @@ bool Maxwell3D::IsMethodExecutable(u32 method) { case MAXWELL3D_REG_INDEX(launch_dma): case MAXWELL3D_REG_INDEX(inline_data): case MAXWELL3D_REG_INDEX(fragment_barrier): + case MAXWELL3D_REG_INDEX(invalidate_texture_data_cache): case MAXWELL3D_REG_INDEX(tiled_cache_barrier): return true; default: @@ -257,7 +259,7 @@ u32 Maxwell3D::GetMaxCurrentVertices() { size_t Maxwell3D::EstimateIndexBufferSize() { GPUVAddr start_address = regs.index_buffer.StartAddress(); GPUVAddr end_address = regs.index_buffer.EndAddress(); - constexpr std::array<size_t, 4> max_sizes = { + static constexpr std::array<size_t, 4> max_sizes = { std::numeric_limits<u8>::max(), std::numeric_limits<u16>::max(), std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max()}; const size_t byte_size = regs.index_buffer.FormatSizeInBytes(); @@ -374,6 +376,9 @@ void Maxwell3D::ProcessMethodCall(u32 method, u32 argument, u32 nonshadow_argume return; case MAXWELL3D_REG_INDEX(fragment_barrier): return rasterizer->FragmentBarrier(); + case MAXWELL3D_REG_INDEX(invalidate_texture_data_cache): + rasterizer->InvalidateGPUCache(); + return rasterizer->WaitForIdle(); case MAXWELL3D_REG_INDEX(tiled_cache_barrier): return rasterizer->TiledCacheBarrier(); default: @@ -467,7 +472,7 @@ void Maxwell3D::ProcessMacroBind(u32 data) { } void Maxwell3D::ProcessFirmwareCall4() { - LOG_WARNING(HW_GPU, "(STUBBED) called"); + LOG_DEBUG(HW_GPU, "(STUBBED) called"); // Firmware call 4 is a blob that changes some registers depending on its parameters. // These registers don't affect emulation and so are stubbed by setting 0xd00 to 1. diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index 0b2fd2928..c89969bb4 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -1599,6 +1599,20 @@ public: }; static_assert(sizeof(TIRModulationCoeff) == 0x4); + struct DrawTexture { + s32 dst_x0; + s32 dst_y0; + s32 dst_width; + s32 dst_height; + s64 dx_du; + s64 dy_dv; + u32 src_sampler; + u32 src_texture; + s32 src_x0; + s32 src_y0; + }; + static_assert(sizeof(DrawTexture) == 0x30); + struct ReduceColorThreshold { union { BitField<0, 8, u32> all_hit_once; @@ -2751,7 +2765,7 @@ public: u32 reserved_sw_method2; ///< 0x102C std::array<TIRModulationCoeff, 5> tir_modulation_coeff; ///< 0x1030 std::array<u32, 15> spare_nop; ///< 0x1044 - INSERT_PADDING_BYTES_NOINIT(0x30); + DrawTexture draw_texture; ///< 0x1080 std::array<u32, 7> reserved_sw_method3_to_7; ///< 0x10B0 ReduceColorThreshold reduce_color_thresholds_unorm8; ///< 0x10CC std::array<u32, 4> reserved_sw_method10_to_13; ///< 0x10D0 diff --git a/src/video_core/fsr.cpp b/src/video_core/fsr.cpp new file mode 100644 index 000000000..5653c64fc --- /dev/null +++ b/src/video_core/fsr.cpp @@ -0,0 +1,148 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include <cmath> +#include "video_core/fsr.h" + +namespace FSR { +namespace { +// Reimplementations of the constant generating functions in ffx_fsr1.h +// GCC generated a lot of warnings when using the official header. +u32 AU1_AH1_AF1(f32 f) { + static constexpr u32 base[512]{ + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, + 0x0080, 0x0100, 0x0200, 0x0400, 0x0800, 0x0c00, 0x1000, 0x1400, 0x1800, 0x1c00, 0x2000, + 0x2400, 0x2800, 0x2c00, 0x3000, 0x3400, 0x3800, 0x3c00, 0x4000, 0x4400, 0x4800, 0x4c00, + 0x5000, 0x5400, 0x5800, 0x5c00, 0x6000, 0x6400, 0x6800, 0x6c00, 0x7000, 0x7400, 0x7800, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, + 0x7bff, 0x7bff, 0x7bff, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, + 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, + 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, + 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, + 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, + 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, + 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, + 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, + 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, + 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8001, 0x8002, 0x8004, 0x8008, + 0x8010, 0x8020, 0x8040, 0x8080, 0x8100, 0x8200, 0x8400, 0x8800, 0x8c00, 0x9000, 0x9400, + 0x9800, 0x9c00, 0xa000, 0xa400, 0xa800, 0xac00, 0xb000, 0xb400, 0xb800, 0xbc00, 0xc000, + 0xc400, 0xc800, 0xcc00, 0xd000, 0xd400, 0xd800, 0xdc00, 0xe000, 0xe400, 0xe800, 0xec00, + 0xf000, 0xf400, 0xf800, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, + }; + static constexpr s8 shift[512]{ + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x17, 0x16, + 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0f, 0x0e, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, + 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, + 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x17, + 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0f, 0x0e, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, + 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, + 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, + 0x18, 0x18, + }; + const u32 u = Common::BitCast<u32>(f); + const u32 i = u >> 23; + return base[i] + ((u & 0x7fffff) >> shift[i]); +} + +u32 AU1_AH2_AF2(f32 a[2]) { + return AU1_AH1_AF1(a[0]) + (AU1_AH1_AF1(a[1]) << 16); +} + +void FsrEasuCon(u32 con0[4], u32 con1[4], u32 con2[4], u32 con3[4], f32 inputViewportInPixelsX, + f32 inputViewportInPixelsY, f32 inputSizeInPixelsX, f32 inputSizeInPixelsY, + f32 outputSizeInPixelsX, f32 outputSizeInPixelsY) { + con0[0] = Common::BitCast<u32>(inputViewportInPixelsX / outputSizeInPixelsX); + con0[1] = Common::BitCast<u32>(inputViewportInPixelsY / outputSizeInPixelsY); + con0[2] = Common::BitCast<u32>(0.5f * inputViewportInPixelsX / outputSizeInPixelsX - 0.5f); + con0[3] = Common::BitCast<u32>(0.5f * inputViewportInPixelsY / outputSizeInPixelsY - 0.5f); + con1[0] = Common::BitCast<u32>(1.0f / inputSizeInPixelsX); + con1[1] = Common::BitCast<u32>(1.0f / inputSizeInPixelsY); + con1[2] = Common::BitCast<u32>(1.0f / inputSizeInPixelsX); + con1[3] = Common::BitCast<u32>(-1.0f / inputSizeInPixelsY); + con2[0] = Common::BitCast<u32>(-1.0f / inputSizeInPixelsX); + con2[1] = Common::BitCast<u32>(2.0f / inputSizeInPixelsY); + con2[2] = Common::BitCast<u32>(1.0f / inputSizeInPixelsX); + con2[3] = Common::BitCast<u32>(2.0f / inputSizeInPixelsY); + con3[0] = Common::BitCast<u32>(0.0f / inputSizeInPixelsX); + con3[1] = Common::BitCast<u32>(4.0f / inputSizeInPixelsY); + con3[2] = con3[3] = 0; +} +} // Anonymous namespace + +void FsrEasuConOffset(u32 con0[4], u32 con1[4], u32 con2[4], u32 con3[4], + f32 inputViewportInPixelsX, f32 inputViewportInPixelsY, + f32 inputSizeInPixelsX, f32 inputSizeInPixelsY, f32 outputSizeInPixelsX, + f32 outputSizeInPixelsY, f32 inputOffsetInPixelsX, f32 inputOffsetInPixelsY) { + FsrEasuCon(con0, con1, con2, con3, inputViewportInPixelsX, inputViewportInPixelsY, + inputSizeInPixelsX, inputSizeInPixelsY, outputSizeInPixelsX, outputSizeInPixelsY); + con0[2] = Common::BitCast<u32>(0.5f * inputViewportInPixelsX / outputSizeInPixelsX - 0.5f + + inputOffsetInPixelsX); + con0[3] = Common::BitCast<u32>(0.5f * inputViewportInPixelsY / outputSizeInPixelsY - 0.5f + + inputOffsetInPixelsY); +} + +void FsrRcasCon(u32* con, f32 sharpness) { + sharpness = std::exp2f(-sharpness); + f32 hSharp[2]{sharpness, sharpness}; + con[0] = Common::BitCast<u32>(sharpness); + con[1] = AU1_AH2_AF2(hSharp); + con[2] = 0; + con[3] = 0; +} +} // namespace FSR diff --git a/src/video_core/fsr.h b/src/video_core/fsr.h new file mode 100644 index 000000000..db0d4ec6f --- /dev/null +++ b/src/video_core/fsr.h @@ -0,0 +1,19 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include "common/bit_cast.h" +#include "common/common_types.h" + +namespace FSR { +// Reimplementations of the constant generating functions in ffx_fsr1.h +// GCC generated a lot of warnings when using the official header. +void FsrEasuConOffset(u32 con0[4], u32 con1[4], u32 con2[4], u32 con3[4], + f32 inputViewportInPixelsX, f32 inputViewportInPixelsY, + f32 inputSizeInPixelsX, f32 inputSizeInPixelsY, f32 outputSizeInPixelsX, + f32 outputSizeInPixelsY, f32 inputOffsetInPixelsX, f32 inputOffsetInPixelsY); + +void FsrRcasCon(u32* con, f32 sharpness); + +} // namespace FSR diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index c6d54be63..7024a19cf 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -99,7 +99,7 @@ struct GPU::Impl { /// Signal the ending of command list. void OnCommandListEnd() { - gpu_thread.OnCommandListEnd(); + rasterizer->ReleaseFences(); } /// Request a host GPU memory flush from the CPU. diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index 164a5252a..9c103c0d4 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp @@ -40,8 +40,6 @@ static void RunThread(std::stop_token stop_token, Core::System& system, scheduler.Push(submit_list->channel, std::move(submit_list->entries)); } else if (const auto* data = std::get_if<SwapBuffersCommand>(&next.data)) { renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr); - } else if (std::holds_alternative<OnCommandListEndCommand>(next.data)) { - rasterizer->ReleaseFences(); } else if (std::holds_alternative<GPUTickCommand>(next.data)) { system.GPU().TickWork(); } else if (const auto* flush = std::get_if<FlushRegionCommand>(&next.data)) { @@ -110,10 +108,6 @@ void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) { rasterizer->OnCPUWrite(addr, size); } -void ThreadManager::OnCommandListEnd() { - PushCommand(OnCommandListEndCommand()); -} - u64 ThreadManager::PushCommand(CommandData&& command_data, bool block) { if (!is_async) { // In synchronous GPU mode, block the caller until the command has executed diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h index c71a419c7..90bcb5958 100644 --- a/src/video_core/gpu_thread.h +++ b/src/video_core/gpu_thread.h @@ -77,16 +77,12 @@ struct FlushAndInvalidateRegionCommand final { u64 size; }; -/// Command called within the gpu, to schedule actions after a command list end -struct OnCommandListEndCommand final {}; - /// Command to make the gpu look into pending requests struct GPUTickCommand final {}; using CommandData = std::variant<std::monostate, SubmitListCommand, SwapBuffersCommand, FlushRegionCommand, - InvalidateRegionCommand, FlushAndInvalidateRegionCommand, OnCommandListEndCommand, - GPUTickCommand>; + InvalidateRegionCommand, FlushAndInvalidateRegionCommand, GPUTickCommand>; struct CommandDataContainer { CommandDataContainer() = default; @@ -134,8 +130,6 @@ public: /// Notify rasterizer that any caches of the specified region should be flushed and invalidated void FlushAndInvalidateRegion(VAddr addr, u64 size); - void OnCommandListEnd(); - void TickGPU(); private: diff --git a/src/video_core/host1x/codecs/codec.cpp b/src/video_core/host1x/codecs/codec.cpp index 42e7d6e4f..3e9022dce 100644 --- a/src/video_core/host1x/codecs/codec.cpp +++ b/src/video_core/host1x/codecs/codec.cpp @@ -152,6 +152,8 @@ bool Codec::CreateGpuAvDevice() { void Codec::InitializeAvCodecContext() { av_codec_ctx = avcodec_alloc_context3(av_codec); av_opt_set(av_codec_ctx->priv_data, "tune", "zerolatency", 0); + av_codec_ctx->thread_count = 0; + av_codec_ctx->thread_type &= ~FF_THREAD_FRAME; } void Codec::InitializeGpuDecoder() { diff --git a/src/video_core/host1x/vic.cpp b/src/video_core/host1x/vic.cpp index 36a04e4e0..10d7ef884 100644 --- a/src/video_core/host1x/vic.cpp +++ b/src/video_core/host1x/vic.cpp @@ -189,9 +189,7 @@ void Vic::WriteYUVFrame(const AVFrame* frame, const VicConfig& config) { for (std::size_t y = 0; y < frame_height; ++y) { const std::size_t src = y * stride; const std::size_t dst = y * aligned_width; - for (std::size_t x = 0; x < frame_width; ++x) { - luma_buffer[dst + x] = luma_src[src + x]; - } + std::memcpy(luma_buffer.data() + dst, luma_src + src, frame_width); } host1x.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(), luma_buffer.size()); @@ -205,15 +203,15 @@ void Vic::WriteYUVFrame(const AVFrame* frame, const VicConfig& config) { // Frame from FFmpeg software // Populate chroma buffer from both channels with interleaving. const std::size_t half_width = frame_width / 2; + u8* chroma_buffer_data = chroma_buffer.data(); const u8* chroma_b_src = frame->data[1]; const u8* chroma_r_src = frame->data[2]; for (std::size_t y = 0; y < half_height; ++y) { const std::size_t src = y * half_stride; const std::size_t dst = y * aligned_width; - for (std::size_t x = 0; x < half_width; ++x) { - chroma_buffer[dst + x * 2] = chroma_b_src[src + x]; - chroma_buffer[dst + x * 2 + 1] = chroma_r_src[src + x]; + chroma_buffer_data[dst + x * 2] = chroma_b_src[src + x]; + chroma_buffer_data[dst + x * 2 + 1] = chroma_r_src[src + x]; } } break; @@ -225,9 +223,7 @@ void Vic::WriteYUVFrame(const AVFrame* frame, const VicConfig& config) { for (std::size_t y = 0; y < half_height; ++y) { const std::size_t src = y * stride; const std::size_t dst = y * aligned_width; - for (std::size_t x = 0; x < frame_width; ++x) { - chroma_buffer[dst + x] = chroma_src[src + x]; - } + std::memcpy(chroma_buffer.data() + dst, chroma_src + src, frame_width); } break; } diff --git a/src/video_core/host_shaders/CMakeLists.txt b/src/video_core/host_shaders/CMakeLists.txt index f275b2aa9..2442c3c29 100644 --- a/src/video_core/host_shaders/CMakeLists.txt +++ b/src/video_core/host_shaders/CMakeLists.txt @@ -3,26 +3,36 @@ set(FIDELITYFX_INCLUDE_DIR ${CMAKE_SOURCE_DIR}/externals/FidelityFX-FSR/ffx-fsr) -set(GLSL_INCLUDES - fidelityfx_fsr.comp +set(FIDELITYFX_FILES ${FIDELITYFX_INCLUDE_DIR}/ffx_a.h ${FIDELITYFX_INCLUDE_DIR}/ffx_fsr1.h ) +set(GLSL_INCLUDES + fidelityfx_fsr.comp + ${FIDELITYFX_FILES} +) + set(SHADER_FILES astc_decoder.comp + blit_color_float.frag block_linear_unswizzle_2d.comp block_linear_unswizzle_3d.comp convert_abgr8_to_d24s8.frag convert_d24s8_to_abgr8.frag convert_depth_to_float.frag convert_float_to_depth.frag + convert_msaa_to_non_msaa.comp + convert_non_msaa_to_msaa.comp convert_s8d24_to_abgr8.frag full_screen_triangle.vert fxaa.frag fxaa.vert opengl_convert_s8d24.comp opengl_copy_bc4.comp + opengl_fidelityfx_fsr.frag + opengl_fidelityfx_fsr_easu.frag + opengl_fidelityfx_fsr_rcas.frag opengl_present.frag opengl_present.vert opengl_present_scaleforce.frag @@ -36,8 +46,9 @@ set(SHADER_FILES smaa_blending_weight_calculation.frag smaa_neighborhood_blending.vert smaa_neighborhood_blending.frag - vulkan_blit_color_float.frag vulkan_blit_depth_stencil.frag + vulkan_color_clear.frag + vulkan_color_clear.vert vulkan_fidelityfx_fsr_easu_fp16.comp vulkan_fidelityfx_fsr_easu_fp32.comp vulkan_fidelityfx_fsr_rcas_fp16.comp @@ -118,6 +129,25 @@ foreach(FILENAME IN ITEMS ${SHADER_FILES}) endif() endforeach() +foreach(FILEPATH IN ITEMS ${FIDELITYFX_FILES}) + get_filename_component(FILENAME ${FILEPATH} NAME) + string(REPLACE "." "_" HEADER_NAME ${FILENAME}) + set(SOURCE_FILE ${FILEPATH}) + set(SOURCE_HEADER_FILE ${SHADER_DIR}/${HEADER_NAME}.h) + add_custom_command( + OUTPUT + ${SOURCE_HEADER_FILE} + COMMAND + ${CMAKE_COMMAND} -P ${HEADER_GENERATOR} ${SOURCE_FILE} ${SOURCE_HEADER_FILE} ${INPUT_FILE} + MAIN_DEPENDENCY + ${SOURCE_FILE} + DEPENDS + ${INPUT_FILE} + # HEADER_GENERATOR should be included here but msbuild seems to assume it's always modified + ) + set(SHADER_HEADERS ${SHADER_HEADERS} ${SOURCE_HEADER_FILE}) +endforeach() + set(SHADER_SOURCES ${SHADER_FILES}) list(APPEND SHADER_SOURCES ${GLSL_INCLUDES}) diff --git a/src/video_core/host_shaders/vulkan_blit_color_float.frag b/src/video_core/host_shaders/blit_color_float.frag index c0c832296..c0c832296 100644 --- a/src/video_core/host_shaders/vulkan_blit_color_float.frag +++ b/src/video_core/host_shaders/blit_color_float.frag diff --git a/src/video_core/host_shaders/convert_msaa_to_non_msaa.comp b/src/video_core/host_shaders/convert_msaa_to_non_msaa.comp new file mode 100644 index 000000000..fc3854d18 --- /dev/null +++ b/src/video_core/host_shaders/convert_msaa_to_non_msaa.comp @@ -0,0 +1,30 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#version 450 core +layout (local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout (binding = 0, rgba8) uniform readonly restrict image2DMSArray msaa_in; +layout (binding = 1, rgba8) uniform writeonly restrict image2DArray output_img; + +void main() { + const ivec3 coords = ivec3(gl_GlobalInvocationID); + if (any(greaterThanEqual(coords, imageSize(msaa_in)))) { + return; + } + + // TODO: Specialization constants for num_samples? + const int num_samples = imageSamples(msaa_in); + for (int curr_sample = 0; curr_sample < num_samples; ++curr_sample) { + const vec4 pixel = imageLoad(msaa_in, coords, curr_sample); + + const int single_sample_x = 2 * coords.x + (curr_sample & 1); + const int single_sample_y = 2 * coords.y + ((curr_sample / 2) & 1); + const ivec3 dest_coords = ivec3(single_sample_x, single_sample_y, coords.z); + + if (any(greaterThanEqual(dest_coords, imageSize(output_img)))) { + continue; + } + imageStore(output_img, dest_coords, pixel); + } +} diff --git a/src/video_core/host_shaders/convert_non_msaa_to_msaa.comp b/src/video_core/host_shaders/convert_non_msaa_to_msaa.comp new file mode 100644 index 000000000..dedd962f1 --- /dev/null +++ b/src/video_core/host_shaders/convert_non_msaa_to_msaa.comp @@ -0,0 +1,29 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#version 450 core +layout (local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout (binding = 0, rgba8) uniform readonly restrict image2DArray img_in; +layout (binding = 1, rgba8) uniform writeonly restrict image2DMSArray output_msaa; + +void main() { + const ivec3 coords = ivec3(gl_GlobalInvocationID); + if (any(greaterThanEqual(coords, imageSize(output_msaa)))) { + return; + } + + // TODO: Specialization constants for num_samples? + const int num_samples = imageSamples(output_msaa); + for (int curr_sample = 0; curr_sample < num_samples; ++curr_sample) { + const int single_sample_x = 2 * coords.x + (curr_sample & 1); + const int single_sample_y = 2 * coords.y + ((curr_sample / 2) & 1); + const ivec3 single_coords = ivec3(single_sample_x, single_sample_y, coords.z); + + if (any(greaterThanEqual(single_coords, imageSize(img_in)))) { + continue; + } + const vec4 pixel = imageLoad(img_in, single_coords); + imageStore(output_msaa, coords, curr_sample, pixel); + } +} diff --git a/src/video_core/host_shaders/full_screen_triangle.vert b/src/video_core/host_shaders/full_screen_triangle.vert index 2c976b19f..d16d98995 100644 --- a/src/video_core/host_shaders/full_screen_triangle.vert +++ b/src/video_core/host_shaders/full_screen_triangle.vert @@ -4,13 +4,20 @@ #version 450 #ifdef VULKAN +#define VERTEX_ID gl_VertexIndex #define BEGIN_PUSH_CONSTANTS layout(push_constant) uniform PushConstants { #define END_PUSH_CONSTANTS }; #define UNIFORM(n) +#define FLIPY 1 #else // ^^^ Vulkan ^^^ // vvv OpenGL vvv +#define VERTEX_ID gl_VertexID #define BEGIN_PUSH_CONSTANTS #define END_PUSH_CONSTANTS +#define FLIPY -1 #define UNIFORM(n) layout (location = n) uniform +out gl_PerVertex { + vec4 gl_Position; +}; #endif BEGIN_PUSH_CONSTANTS @@ -21,8 +28,8 @@ END_PUSH_CONSTANTS layout(location = 0) out vec2 texcoord; void main() { - float x = float((gl_VertexIndex & 1) << 2); - float y = float((gl_VertexIndex & 2) << 1); - gl_Position = vec4(x - 1.0, y - 1.0, 0.0, 1.0); + float x = float((VERTEX_ID & 1) << 2); + float y = float((VERTEX_ID & 2) << 1); + gl_Position = vec4(x - 1.0, FLIPY * (y - 1.0), 0.0, 1.0); texcoord = fma(vec2(x, y) / 2.0, tex_scale, tex_offset); } diff --git a/src/video_core/host_shaders/opengl_fidelityfx_fsr.frag b/src/video_core/host_shaders/opengl_fidelityfx_fsr.frag new file mode 100644 index 000000000..16d22f58e --- /dev/null +++ b/src/video_core/host_shaders/opengl_fidelityfx_fsr.frag @@ -0,0 +1,108 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +//!#version 460 core +#extension GL_ARB_separate_shader_objects : enable +#extension GL_ARB_shading_language_420pack : enable + +#extension GL_AMD_gpu_shader_half_float : enable +#extension GL_NV_gpu_shader5 : enable + +// FidelityFX Super Resolution Sample +// +// Copyright (c) 2021 Advanced Micro Devices, Inc. All rights reserved. +// Permission is hereby granted, free of charge, to any person obtaining a copy +// of this software and associated documentation files(the "Software"), to deal +// in the Software without restriction, including without limitation the rights +// to use, copy, modify, merge, publish, distribute, sublicense, and / or sell +// copies of the Software, and to permit persons to whom the Software is +// furnished to do so, subject to the following conditions : +// The above copyright notice and this permission notice shall be included in +// all copies or substantial portions of the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN +// THE SOFTWARE. + +layout (location = 0) uniform uvec4 constants[4]; + +#define A_GPU 1 +#define A_GLSL 1 + +#ifdef YUZU_USE_FP16 + #define A_HALF +#endif +#include "ffx_a.h" + +#ifndef YUZU_USE_FP16 + layout (binding=0) uniform sampler2D InputTexture; + #if USE_EASU + #define FSR_EASU_F 1 + AF4 FsrEasuRF(AF2 p) { AF4 res = textureGather(InputTexture, p, 0); return res; } + AF4 FsrEasuGF(AF2 p) { AF4 res = textureGather(InputTexture, p, 1); return res; } + AF4 FsrEasuBF(AF2 p) { AF4 res = textureGather(InputTexture, p, 2); return res; } + #endif + #if USE_RCAS + #define FSR_RCAS_F + AF4 FsrRcasLoadF(ASU2 p) { return texelFetch(InputTexture, ASU2(p), 0); } + void FsrRcasInputF(inout AF1 r, inout AF1 g, inout AF1 b) {} + #endif +#else + layout (binding=0) uniform sampler2D InputTexture; + #if USE_EASU + #define FSR_EASU_H 1 + AH4 FsrEasuRH(AF2 p) { AH4 res = AH4(textureGather(InputTexture, p, 0)); return res; } + AH4 FsrEasuGH(AF2 p) { AH4 res = AH4(textureGather(InputTexture, p, 1)); return res; } + AH4 FsrEasuBH(AF2 p) { AH4 res = AH4(textureGather(InputTexture, p, 2)); return res; } + #endif + #if USE_RCAS + #define FSR_RCAS_H + AH4 FsrRcasLoadH(ASW2 p) { return AH4(texelFetch(InputTexture, ASU2(p), 0)); } + void FsrRcasInputH(inout AH1 r,inout AH1 g,inout AH1 b){} + #endif +#endif + +#include "ffx_fsr1.h" + +#if USE_RCAS + layout(location = 0) in vec2 frag_texcoord; +#endif +layout (location = 0) out vec4 frag_color; + +void CurrFilter(AU2 pos) +{ +#if USE_EASU + #ifndef YUZU_USE_FP16 + AF3 c; + FsrEasuF(c, pos, constants[0], constants[1], constants[2], constants[3]); + frag_color = AF4(c, 1.0); + #else + AH3 c; + FsrEasuH(c, pos, constants[0], constants[1], constants[2], constants[3]); + frag_color = AH4(c, 1.0); + #endif +#endif +#if USE_RCAS + #ifndef YUZU_USE_FP16 + AF3 c; + FsrRcasF(c.r, c.g, c.b, pos, constants[0]); + frag_color = AF4(c, 1.0); + #else + AH3 c; + FsrRcasH(c.r, c.g, c.b, pos, constants[0]); + frag_color = AH4(c, 1.0); + #endif +#endif +} + +void main() +{ +#if USE_RCAS + CurrFilter(AU2(frag_texcoord * vec2(textureSize(InputTexture, 0)))); +#else + CurrFilter(AU2(gl_FragCoord.xy)); +#endif +} diff --git a/src/video_core/host_shaders/opengl_fidelityfx_fsr_easu.frag b/src/video_core/host_shaders/opengl_fidelityfx_fsr_easu.frag new file mode 100644 index 000000000..d39f80ac1 --- /dev/null +++ b/src/video_core/host_shaders/opengl_fidelityfx_fsr_easu.frag @@ -0,0 +1,9 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#version 460 core +#extension GL_GOOGLE_include_directive : enable + +#define USE_EASU 1 + +#include "opengl_fidelityfx_fsr.frag" diff --git a/src/video_core/host_shaders/opengl_fidelityfx_fsr_rcas.frag b/src/video_core/host_shaders/opengl_fidelityfx_fsr_rcas.frag new file mode 100644 index 000000000..cfa78ddc7 --- /dev/null +++ b/src/video_core/host_shaders/opengl_fidelityfx_fsr_rcas.frag @@ -0,0 +1,9 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#version 460 core +#extension GL_GOOGLE_include_directive : enable + +#define USE_RCAS 1 + +#include "opengl_fidelityfx_fsr.frag" diff --git a/src/video_core/host_shaders/vulkan_color_clear.frag b/src/video_core/host_shaders/vulkan_color_clear.frag new file mode 100644 index 000000000..617bf01e1 --- /dev/null +++ b/src/video_core/host_shaders/vulkan_color_clear.frag @@ -0,0 +1,14 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#version 460 core + +layout (push_constant) uniform PushConstants { + vec4 clear_color; +}; + +layout(location = 0) out vec4 color; + +void main() { + color = clear_color; +} diff --git a/src/video_core/host_shaders/vulkan_color_clear.vert b/src/video_core/host_shaders/vulkan_color_clear.vert new file mode 100644 index 000000000..d85883141 --- /dev/null +++ b/src/video_core/host_shaders/vulkan_color_clear.vert @@ -0,0 +1,10 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#version 460 core + +void main() { + float x = float((gl_VertexIndex & 1) << 2); + float y = float((gl_VertexIndex & 2) << 1); + gl_Position = vec4(x - 1.0, y - 1.0, 0.0, 1.0); +} diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index 3bcae3503..83924475b 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -6,7 +6,6 @@ #include "common/alignment.h" #include "common/assert.h" #include "common/logging/log.h" -#include "common/settings.h" #include "core/core.h" #include "core/device_memory.h" #include "core/hle/kernel/k_page_table.h" @@ -46,11 +45,6 @@ MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64 big_page_table_cpu.resize(big_page_table_size); big_page_continous.resize(big_page_table_size / continous_bits, 0); entries.resize(page_table_size / 32, 0); - if (!Settings::IsGPULevelExtreme() && Settings::IsFastmemEnabled()) { - fastmem_arena = system.DeviceMemory().buffer.VirtualBasePointer(); - } else { - fastmem_arena = nullptr; - } } MemoryManager::~MemoryManager() = default; @@ -360,7 +354,7 @@ inline void MemoryManager::MemoryOperation(GPUVAddr gpu_src_addr, std::size_t si } } -template <bool is_safe, bool use_fastmem> +template <bool is_safe> void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size, [[maybe_unused]] VideoCommon::CacheType which) const { auto set_to_zero = [&]([[maybe_unused]] std::size_t page_index, @@ -374,12 +368,8 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std: if constexpr (is_safe) { rasterizer->FlushRegion(cpu_addr_base, copy_amount, which); } - if constexpr (use_fastmem) { - std::memcpy(dest_buffer, &fastmem_arena[cpu_addr_base], copy_amount); - } else { - u8* physical = memory.GetPointer(cpu_addr_base); - std::memcpy(dest_buffer, physical, copy_amount); - } + u8* physical = memory.GetPointer(cpu_addr_base); + std::memcpy(dest_buffer, physical, copy_amount); dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount; }; auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) { @@ -388,15 +378,11 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std: if constexpr (is_safe) { rasterizer->FlushRegion(cpu_addr_base, copy_amount, which); } - if constexpr (use_fastmem) { - std::memcpy(dest_buffer, &fastmem_arena[cpu_addr_base], copy_amount); + if (!IsBigPageContinous(page_index)) [[unlikely]] { + memory.ReadBlockUnsafe(cpu_addr_base, dest_buffer, copy_amount); } else { - if (!IsBigPageContinous(page_index)) [[unlikely]] { - memory.ReadBlockUnsafe(cpu_addr_base, dest_buffer, copy_amount); - } else { - u8* physical = memory.GetPointer(cpu_addr_base); - std::memcpy(dest_buffer, physical, copy_amount); - } + u8* physical = memory.GetPointer(cpu_addr_base); + std::memcpy(dest_buffer, physical, copy_amount); } dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount; }; @@ -410,20 +396,12 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std: void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size, VideoCommon::CacheType which) const { - if (fastmem_arena) [[likely]] { - ReadBlockImpl<true, true>(gpu_src_addr, dest_buffer, size, which); - return; - } - ReadBlockImpl<true, false>(gpu_src_addr, dest_buffer, size, which); + ReadBlockImpl<true>(gpu_src_addr, dest_buffer, size, which); } void MemoryManager::ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer, const std::size_t size) const { - if (fastmem_arena) [[likely]] { - ReadBlockImpl<false, true>(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None); - return; - } - ReadBlockImpl<false, false>(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None); + ReadBlockImpl<false>(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None); } template <bool is_safe> diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index 2936364f0..cf56392ef 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -141,7 +141,7 @@ private: inline void MemoryOperation(GPUVAddr gpu_src_addr, std::size_t size, FuncMapped&& func_mapped, FuncReserved&& func_reserved, FuncUnmapped&& func_unmapped) const; - template <bool is_safe, bool use_fastmem> + template <bool is_safe> void ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size, VideoCommon::CacheType which) const; @@ -215,9 +215,8 @@ private: std::vector<u64> big_page_continous; std::vector<std::pair<VAddr, std::size_t>> page_stash{}; - u8* fastmem_arena{}; - constexpr static size_t continous_bits = 64; + static constexpr size_t continous_bits = 64; const size_t unique_identifier; std::unique_ptr<VideoCommon::InvalidationAccumulator> accumulator; diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h index 1735b6164..33e2610bc 100644 --- a/src/video_core/rasterizer_interface.h +++ b/src/video_core/rasterizer_interface.h @@ -47,6 +47,9 @@ public: /// Dispatches an indirect draw invocation virtual void DrawIndirect() {} + /// Dispatches an draw texture invocation + virtual void DrawTexture() = 0; + /// Clear the current framebuffer virtual void Clear(u32 layer_count) = 0; diff --git a/src/video_core/renderer_null/null_rasterizer.cpp b/src/video_core/renderer_null/null_rasterizer.cpp index 2c11345d7..2b5c7defa 100644 --- a/src/video_core/renderer_null/null_rasterizer.cpp +++ b/src/video_core/renderer_null/null_rasterizer.cpp @@ -21,6 +21,7 @@ RasterizerNull::RasterizerNull(Core::Memory::Memory& cpu_memory_, Tegra::GPU& gp RasterizerNull::~RasterizerNull() = default; void RasterizerNull::Draw(bool is_indexed, u32 instance_count) {} +void RasterizerNull::DrawTexture() {} void RasterizerNull::Clear(u32 layer_count) {} void RasterizerNull::DispatchCompute() {} void RasterizerNull::ResetCounter(VideoCore::QueryType type) {} diff --git a/src/video_core/renderer_null/null_rasterizer.h b/src/video_core/renderer_null/null_rasterizer.h index 2112aa70e..51f896e43 100644 --- a/src/video_core/renderer_null/null_rasterizer.h +++ b/src/video_core/renderer_null/null_rasterizer.h @@ -31,6 +31,7 @@ public: ~RasterizerNull() override; void Draw(bool is_indexed, u32 instance_count) override; + void DrawTexture() override; void Clear(u32 layer_count) override; void DispatchCompute() override; void ResetCounter(VideoCore::QueryType type) override; diff --git a/src/video_core/renderer_opengl/blit_image.cpp b/src/video_core/renderer_opengl/blit_image.cpp new file mode 100644 index 000000000..9a560a73b --- /dev/null +++ b/src/video_core/renderer_opengl/blit_image.cpp @@ -0,0 +1,59 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include <algorithm> + +#include "video_core/host_shaders/blit_color_float_frag.h" +#include "video_core/host_shaders/full_screen_triangle_vert.h" +#include "video_core/renderer_opengl/blit_image.h" +#include "video_core/renderer_opengl/gl_shader_manager.h" +#include "video_core/renderer_opengl/gl_shader_util.h" + +namespace OpenGL { + +BlitImageHelper::BlitImageHelper(ProgramManager& program_manager_) + : program_manager(program_manager_), + full_screen_vert(CreateProgram(HostShaders::FULL_SCREEN_TRIANGLE_VERT, GL_VERTEX_SHADER)), + blit_color_to_color_frag( + CreateProgram(HostShaders::BLIT_COLOR_FLOAT_FRAG, GL_FRAGMENT_SHADER)) {} + +BlitImageHelper::~BlitImageHelper() = default; + +void BlitImageHelper::BlitColor(GLuint dst_framebuffer, GLuint src_image_view, GLuint src_sampler, + const Region2D& dst_region, const Region2D& src_region, + const Extent3D& src_size) { + glEnable(GL_CULL_FACE); + glDisable(GL_COLOR_LOGIC_OP); + glDisable(GL_DEPTH_TEST); + glDisable(GL_STENCIL_TEST); + glDisable(GL_POLYGON_OFFSET_FILL); + glDisable(GL_RASTERIZER_DISCARD); + glDisable(GL_ALPHA_TEST); + glDisablei(GL_BLEND, 0); + glPolygonMode(GL_FRONT_AND_BACK, GL_FILL); + glCullFace(GL_BACK); + glFrontFace(GL_CW); + glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); + glDepthRangeIndexed(0, 0.0, 0.0); + + program_manager.BindPresentPrograms(full_screen_vert.handle, blit_color_to_color_frag.handle); + glProgramUniform2f(full_screen_vert.handle, 0, + static_cast<float>(src_region.end.x - src_region.start.x) / + static_cast<float>(src_size.width), + static_cast<float>(src_region.end.y - src_region.start.y) / + static_cast<float>(src_size.height)); + glProgramUniform2f(full_screen_vert.handle, 1, + static_cast<float>(src_region.start.x) / static_cast<float>(src_size.width), + static_cast<float>(src_region.start.y) / + static_cast<float>(src_size.height)); + glViewport(std::min(dst_region.start.x, dst_region.end.x), + std::min(dst_region.start.y, dst_region.end.y), + std::abs(dst_region.end.x - dst_region.start.x), + std::abs(dst_region.end.y - dst_region.start.y)); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, dst_framebuffer); + glBindSampler(0, src_sampler); + glBindTextureUnit(0, src_image_view); + glClear(GL_COLOR_BUFFER_BIT); + glDrawArrays(GL_TRIANGLES, 0, 3); +} +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/blit_image.h b/src/video_core/renderer_opengl/blit_image.h new file mode 100644 index 000000000..5a2b12d16 --- /dev/null +++ b/src/video_core/renderer_opengl/blit_image.h @@ -0,0 +1,38 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include <glad/glad.h> + +#include "video_core/engines/fermi_2d.h" +#include "video_core/renderer_opengl/gl_resource_manager.h" +#include "video_core/texture_cache/types.h" + +namespace OpenGL { + +using VideoCommon::Extent3D; +using VideoCommon::Offset2D; +using VideoCommon::Region2D; + +class ProgramManager; +class Framebuffer; +class ImageView; + +class BlitImageHelper { +public: + explicit BlitImageHelper(ProgramManager& program_manager); + ~BlitImageHelper(); + + void BlitColor(GLuint dst_framebuffer, GLuint src_image_view, GLuint src_sampler, + const Region2D& dst_region, const Region2D& src_region, + const Extent3D& src_size); + +private: + ProgramManager& program_manager; + + OGLProgram full_screen_vert; + OGLProgram blit_color_to_color_frag; +}; + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp index 26d066004..1a0cea9b7 100644 --- a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp @@ -30,7 +30,7 @@ bool ComputePipelineKey::operator==(const ComputePipelineKey& rhs) const noexcep ComputePipeline::ComputePipeline(const Device& device, TextureCache& texture_cache_, BufferCache& buffer_cache_, ProgramManager& program_manager_, const Shader::Info& info_, std::string code, - std::vector<u32> code_v) + std::vector<u32> code_v, bool force_context_flush) : texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, program_manager{program_manager_}, info{info_} { switch (device.GetShaderBackend()) { @@ -63,6 +63,15 @@ ComputePipeline::ComputePipeline(const Device& device, TextureCache& texture_cac writes_global_memory = !use_storage_buffers && std::ranges::any_of(info.storage_buffers_descriptors, [](const auto& desc) { return desc.is_written; }); + if (force_context_flush) { + std::scoped_lock lock{built_mutex}; + built_fence.Create(); + // Flush this context to ensure compilation commands and fence are in the GPU pipe. + glFlush(); + built_condvar.notify_one(); + } else { + is_built = true; + } } void ComputePipeline::Configure() { @@ -142,6 +151,9 @@ void ComputePipeline::Configure() { } texture_cache.FillComputeImageViews(std::span(views.data(), views.size())); + if (!is_built) { + WaitForBuild(); + } if (assembly_program.handle != 0) { program_manager.BindComputeAssemblyProgram(assembly_program.handle); } else { @@ -223,4 +235,13 @@ void ComputePipeline::Configure() { } } +void ComputePipeline::WaitForBuild() { + if (built_fence.handle == 0) { + std::unique_lock lock{built_mutex}; + built_condvar.wait(lock, [this] { return built_fence.handle != 0; }); + } + ASSERT(glClientWaitSync(built_fence.handle, 0, GL_TIMEOUT_IGNORED) != GL_WAIT_FAILED); + is_built = true; +} + } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_compute_pipeline.h b/src/video_core/renderer_opengl/gl_compute_pipeline.h index 6534dec32..9bcc72b59 100644 --- a/src/video_core/renderer_opengl/gl_compute_pipeline.h +++ b/src/video_core/renderer_opengl/gl_compute_pipeline.h @@ -50,7 +50,8 @@ class ComputePipeline { public: explicit ComputePipeline(const Device& device, TextureCache& texture_cache_, BufferCache& buffer_cache_, ProgramManager& program_manager_, - const Shader::Info& info_, std::string code, std::vector<u32> code_v); + const Shader::Info& info_, std::string code, std::vector<u32> code_v, + bool force_context_flush = false); void Configure(); @@ -65,6 +66,8 @@ public: } private: + void WaitForBuild(); + TextureCache& texture_cache; BufferCache& buffer_cache; Tegra::MemoryManager* gpu_memory; @@ -81,6 +84,11 @@ private: bool use_storage_buffers{}; bool writes_global_memory{}; + + std::mutex built_mutex; + std::condition_variable built_condvar; + OGLSync built_fence{}; + bool is_built{false}; }; } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp index cee5c3247..22ed16ebf 100644 --- a/src/video_core/renderer_opengl/gl_device.cpp +++ b/src/video_core/renderer_opengl/gl_device.cpp @@ -166,6 +166,7 @@ Device::Device(Core::Frontend::EmuWindow& emu_window) { has_shader_int64 = HasExtension(extensions, "GL_ARB_gpu_shader_int64"); has_amd_shader_half_float = GLAD_GL_AMD_gpu_shader_half_float; has_sparse_texture_2 = GLAD_GL_ARB_sparse_texture2; + has_draw_texture = GLAD_GL_NV_draw_texture; warp_size_potentially_larger_than_guest = !is_nvidia && !is_intel; need_fastmath_off = is_nvidia; can_report_memory = GLAD_GL_NVX_gpu_memory_info; diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h index 2a72d84be..3ff8cad83 100644 --- a/src/video_core/renderer_opengl/gl_device.h +++ b/src/video_core/renderer_opengl/gl_device.h @@ -4,6 +4,8 @@ #pragma once #include <cstddef> +#include <string> + #include "common/common_types.h" #include "core/frontend/emu_window.h" #include "shader_recompiler/stage.h" @@ -146,6 +148,10 @@ public: return has_sparse_texture_2; } + bool HasDrawTexture() const { + return has_draw_texture; + } + bool IsWarpSizePotentiallyLargerThanGuest() const { return warp_size_potentially_larger_than_guest; } @@ -216,6 +222,7 @@ private: bool has_shader_int64{}; bool has_amd_shader_half_float{}; bool has_sparse_texture_2{}; + bool has_draw_texture{}; bool warp_size_potentially_larger_than_guest{}; bool need_fastmath_off{}; bool has_cbuf_ftou_bug{}; diff --git a/src/video_core/renderer_opengl/gl_fsr.cpp b/src/video_core/renderer_opengl/gl_fsr.cpp new file mode 100644 index 000000000..77262dcf1 --- /dev/null +++ b/src/video_core/renderer_opengl/gl_fsr.cpp @@ -0,0 +1,101 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include "common/settings.h" +#include "video_core/fsr.h" +#include "video_core/renderer_opengl/gl_fsr.h" +#include "video_core/renderer_opengl/gl_shader_manager.h" +#include "video_core/renderer_opengl/gl_shader_util.h" + +namespace OpenGL { +using namespace FSR; + +using FsrConstants = std::array<u32, 4 * 4>; + +FSR::FSR(std::string_view fsr_vertex_source, std::string_view fsr_easu_source, + std::string_view fsr_rcas_source) + : fsr_vertex{CreateProgram(fsr_vertex_source, GL_VERTEX_SHADER)}, + fsr_easu_frag{CreateProgram(fsr_easu_source, GL_FRAGMENT_SHADER)}, + fsr_rcas_frag{CreateProgram(fsr_rcas_source, GL_FRAGMENT_SHADER)} { + glProgramUniform2f(fsr_vertex.handle, 0, 1.0f, 1.0f); + glProgramUniform2f(fsr_vertex.handle, 1, 0.0f, 0.0f); +} + +FSR::~FSR() = default; + +void FSR::Draw(ProgramManager& program_manager, const Common::Rectangle<u32>& screen, + u32 input_image_width, u32 input_image_height, + const Common::Rectangle<int>& crop_rect) { + + const auto output_image_width = screen.GetWidth(); + const auto output_image_height = screen.GetHeight(); + + if (fsr_intermediate_tex.handle) { + GLint fsr_tex_width, fsr_tex_height; + glGetTextureLevelParameteriv(fsr_intermediate_tex.handle, 0, GL_TEXTURE_WIDTH, + &fsr_tex_width); + glGetTextureLevelParameteriv(fsr_intermediate_tex.handle, 0, GL_TEXTURE_HEIGHT, + &fsr_tex_height); + if (static_cast<u32>(fsr_tex_width) != output_image_width || + static_cast<u32>(fsr_tex_height) != output_image_height) { + fsr_intermediate_tex.Release(); + } + } + if (!fsr_intermediate_tex.handle) { + fsr_intermediate_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(fsr_intermediate_tex.handle, 1, GL_RGB16F, output_image_width, + output_image_height); + glNamedFramebufferTexture(fsr_framebuffer.handle, GL_COLOR_ATTACHMENT0, + fsr_intermediate_tex.handle, 0); + } + + GLint old_draw_fb; + glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); + + glFrontFace(GL_CW); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fsr_framebuffer.handle); + glViewportIndexedf(0, 0.0f, 0.0f, static_cast<GLfloat>(output_image_width), + static_cast<GLfloat>(output_image_height)); + + FsrConstants constants; + FsrEasuConOffset( + constants.data() + 0, constants.data() + 4, constants.data() + 8, constants.data() + 12, + + static_cast<f32>(crop_rect.GetWidth()), static_cast<f32>(crop_rect.GetHeight()), + static_cast<f32>(input_image_width), static_cast<f32>(input_image_height), + static_cast<f32>(output_image_width), static_cast<f32>(output_image_height), + static_cast<f32>(crop_rect.left), static_cast<f32>(crop_rect.top)); + + glProgramUniform4uiv(fsr_easu_frag.handle, 0, sizeof(constants), std::data(constants)); + + program_manager.BindPresentPrograms(fsr_vertex.handle, fsr_easu_frag.handle); + glDrawArrays(GL_TRIANGLES, 0, 3); + + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); + glBindTextureUnit(0, fsr_intermediate_tex.handle); + + const float sharpening = + static_cast<float>(Settings::values.fsr_sharpening_slider.GetValue()) / 100.0f; + + FsrRcasCon(constants.data(), sharpening); + glProgramUniform4uiv(fsr_rcas_frag.handle, 0, sizeof(constants), std::data(constants)); +} + +void FSR::InitBuffers() { + fsr_framebuffer.Create(); +} + +void FSR::ReleaseBuffers() { + fsr_framebuffer.Release(); + fsr_intermediate_tex.Release(); +} + +const OGLProgram& FSR::GetPresentFragmentProgram() const noexcept { + return fsr_rcas_frag; +} + +bool FSR::AreBuffersInitialized() const noexcept { + return fsr_framebuffer.handle; +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_fsr.h b/src/video_core/renderer_opengl/gl_fsr.h new file mode 100644 index 000000000..1f6ae3115 --- /dev/null +++ b/src/video_core/renderer_opengl/gl_fsr.h @@ -0,0 +1,43 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include <string_view> + +#include "common/common_types.h" +#include "common/math_util.h" +#include "video_core/fsr.h" +#include "video_core/renderer_opengl/gl_resource_manager.h" + +namespace OpenGL { + +class ProgramManager; + +class FSR { +public: + explicit FSR(std::string_view fsr_vertex_source, std::string_view fsr_easu_source, + std::string_view fsr_rcas_source); + ~FSR(); + + void Draw(ProgramManager& program_manager, const Common::Rectangle<u32>& screen, + u32 input_image_width, u32 input_image_height, + const Common::Rectangle<int>& crop_rect); + + void InitBuffers(); + + void ReleaseBuffers(); + + [[nodiscard]] const OGLProgram& GetPresentFragmentProgram() const noexcept; + + [[nodiscard]] bool AreBuffersInitialized() const noexcept; + +private: + OGLFramebuffer fsr_framebuffer; + OGLProgram fsr_vertex; + OGLProgram fsr_easu_frag; + OGLProgram fsr_rcas_frag; + OGLTexture fsr_intermediate_tex; +}; + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp index c115dabe1..29491e762 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp @@ -176,7 +176,7 @@ GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_c std::array<std::string, 5> sources, std::array<std::vector<u32>, 5> sources_spirv, const std::array<const Shader::Info*, 5>& infos, - const GraphicsPipelineKey& key_) + const GraphicsPipelineKey& key_, bool force_context_flush) : texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, program_manager{program_manager_}, state_tracker{state_tracker_}, key{key_} { if (shader_notify) { @@ -231,7 +231,8 @@ GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_c const bool in_parallel = thread_worker != nullptr; const auto backend = device.GetShaderBackend(); auto func{[this, sources = std::move(sources), sources_spirv = std::move(sources_spirv), - shader_notify, backend, in_parallel](ShaderContext::Context*) mutable { + shader_notify, backend, in_parallel, + force_context_flush](ShaderContext::Context*) mutable { for (size_t stage = 0; stage < 5; ++stage) { switch (backend) { case Settings::ShaderBackend::GLSL: @@ -251,7 +252,7 @@ GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_c break; } } - if (in_parallel) { + if (force_context_flush || in_parallel) { std::scoped_lock lock{built_mutex}; built_fence.Create(); // Flush this context to ensure compilation commands and fence are in the GPU pipe. diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.h b/src/video_core/renderer_opengl/gl_graphics_pipeline.h index 1c06b3655..7bab3be0a 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.h +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.h @@ -78,7 +78,7 @@ public: std::array<std::string, 5> sources, std::array<std::vector<u32>, 5> sources_spirv, const std::array<const Shader::Info*, 5>& infos, - const GraphicsPipelineKey& key_); + const GraphicsPipelineKey& key_, bool force_context_flush = false); void Configure(bool is_indexed) { configure_func(this, is_indexed); diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 7d48af8e1..7bced675c 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -64,7 +64,8 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra shader_cache(*this, emu_window_, device, texture_cache, buffer_cache, program_manager, state_tracker, gpu.ShaderNotify()), query_cache(*this), accelerate_dma(buffer_cache), - fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache) {} + fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache), + blit_image(program_manager_) {} RasterizerOpenGL::~RasterizerOpenGL() = default; @@ -139,6 +140,7 @@ void RasterizerOpenGL::LoadDiskResources(u64 title_id, std::stop_token stop_load void RasterizerOpenGL::Clear(u32 layer_count) { MICROPROFILE_SCOPE(OpenGL_Clears); + gpu_memory->FlushCaching(); const auto& regs = maxwell3d->regs; bool use_color{}; bool use_depth{}; @@ -207,6 +209,7 @@ void RasterizerOpenGL::PrepareDraw(bool is_indexed, Func&& draw_func) { MICROPROFILE_SCOPE(OpenGL_Drawing); SCOPE_EXIT({ gpu.TickWork(); }); + gpu_memory->FlushCaching(); query_cache.UpdateCounters(); GraphicsPipeline* const pipeline{shader_cache.CurrentGraphicsPipeline()}; @@ -318,7 +321,49 @@ void RasterizerOpenGL::DrawIndirect() { buffer_cache.SetDrawIndirect(nullptr); } +void RasterizerOpenGL::DrawTexture() { + MICROPROFILE_SCOPE(OpenGL_Drawing); + + SCOPE_EXIT({ gpu.TickWork(); }); + query_cache.UpdateCounters(); + + texture_cache.SynchronizeGraphicsDescriptors(); + texture_cache.UpdateRenderTargets(false); + + SyncState(); + + const auto& draw_texture_state = maxwell3d->draw_manager->GetDrawTextureState(); + const auto& sampler = texture_cache.GetGraphicsSampler(draw_texture_state.src_sampler); + const auto& texture = texture_cache.GetImageView(draw_texture_state.src_texture); + + if (device.HasDrawTexture()) { + state_tracker.BindFramebuffer(texture_cache.GetFramebuffer()->Handle()); + + glDrawTextureNV(texture.DefaultHandle(), sampler->Handle(), draw_texture_state.dst_x0, + draw_texture_state.dst_y0, draw_texture_state.dst_x1, + draw_texture_state.dst_y1, 0, + draw_texture_state.src_x0 / static_cast<float>(texture.size.width), + draw_texture_state.src_y0 / static_cast<float>(texture.size.height), + draw_texture_state.src_x1 / static_cast<float>(texture.size.width), + draw_texture_state.src_y1 / static_cast<float>(texture.size.height)); + } else { + Region2D dst_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x0), + .y = static_cast<s32>(draw_texture_state.dst_y0)}, + Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x1), + .y = static_cast<s32>(draw_texture_state.dst_y1)}}; + Region2D src_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.src_x0), + .y = static_cast<s32>(draw_texture_state.src_y0)}, + Offset2D{.x = static_cast<s32>(draw_texture_state.src_x1), + .y = static_cast<s32>(draw_texture_state.src_y1)}}; + blit_image.BlitColor(texture_cache.GetFramebuffer()->Handle(), texture.DefaultHandle(), + sampler->Handle(), dst_region, src_region, texture.size); + } + + ++num_queued_commands; +} + void RasterizerOpenGL::DispatchCompute() { + gpu_memory->FlushCaching(); ComputePipeline* const pipeline{shader_cache.CurrentComputePipeline()}; if (!pipeline) { return; @@ -526,6 +571,7 @@ void RasterizerOpenGL::TickFrame() { } bool RasterizerOpenGL::AccelerateConditionalRendering() { + gpu_memory->FlushCaching(); if (Settings::IsGPULevelHigh()) { // Reimplement Host conditional rendering. return false; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index be4f76c18..0c45832ae 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -16,6 +16,7 @@ #include "video_core/engines/maxwell_dma.h" #include "video_core/rasterizer_accelerated.h" #include "video_core/rasterizer_interface.h" +#include "video_core/renderer_opengl/blit_image.h" #include "video_core/renderer_opengl/gl_buffer_cache.h" #include "video_core/renderer_opengl/gl_device.h" #include "video_core/renderer_opengl/gl_fence_manager.h" @@ -70,6 +71,7 @@ public: void Draw(bool is_indexed, u32 instance_count) override; void DrawIndirect() override; + void DrawTexture() override; void Clear(u32 layer_count) override; void DispatchCompute() override; void ResetCounter(VideoCore::QueryType type) override; @@ -224,6 +226,8 @@ private: AccelerateDMA accelerate_dma; FenceManagerOpenGL fence_manager; + BlitImageHelper blit_image; + boost::container::static_vector<u32, MAX_IMAGE_VIEWS> image_view_indices; std::array<ImageViewId, MAX_IMAGE_VIEWS> image_view_ids; boost::container::static_vector<GLuint, MAX_TEXTURES> sampler_handles; diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index 9442193de..479bb8ba3 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -285,7 +285,7 @@ void ShaderCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading, file.read(reinterpret_cast<char*>(&key), sizeof(key)); queue_work([this, key, env = std::move(env), &state, &callback](Context* ctx) mutable { ctx->pools.ReleaseContents(); - auto pipeline{CreateComputePipeline(ctx->pools, key, env)}; + auto pipeline{CreateComputePipeline(ctx->pools, key, env, true)}; std::scoped_lock lock{state.mutex}; if (pipeline) { compute_cache.emplace(key, std::move(pipeline)); @@ -306,7 +306,7 @@ void ShaderCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading, env_ptrs.push_back(&env); } ctx->pools.ReleaseContents(); - auto pipeline{CreateGraphicsPipeline(ctx->pools, key, MakeSpan(env_ptrs), false)}; + auto pipeline{CreateGraphicsPipeline(ctx->pools, key, MakeSpan(env_ptrs), false, true)}; std::scoped_lock lock{state.mutex}; if (pipeline) { graphics_cache.emplace(key, std::move(pipeline)); @@ -438,7 +438,8 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline() { std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline( ShaderContext::ShaderPools& pools, const GraphicsPipelineKey& key, - std::span<Shader::Environment* const> envs, bool build_in_parallel) try { + std::span<Shader::Environment* const> envs, bool use_shader_workers, + bool force_context_flush) try { LOG_INFO(Render_OpenGL, "0x{:016x}", key.Hash()); size_t env_index{}; u32 total_storage_buffers{}; @@ -530,10 +531,10 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline( } previous_program = &program; } - auto* const thread_worker{build_in_parallel ? workers.get() : nullptr}; + auto* const thread_worker{use_shader_workers ? workers.get() : nullptr}; return std::make_unique<GraphicsPipeline>(device, texture_cache, buffer_cache, program_manager, state_tracker, thread_worker, &shader_notify, sources, - sources_spirv, infos, key); + sources_spirv, infos, key, force_context_flush); } catch (Shader::Exception& exception) { LOG_ERROR(Render_OpenGL, "{}", exception.what()); @@ -558,8 +559,8 @@ std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline( } std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline( - ShaderContext::ShaderPools& pools, const ComputePipelineKey& key, - Shader::Environment& env) try { + ShaderContext::ShaderPools& pools, const ComputePipelineKey& key, Shader::Environment& env, + bool force_context_flush) try { LOG_INFO(Render_OpenGL, "0x{:016x}", key.Hash()); Shader::Maxwell::Flow::CFG cfg{env, pools.flow_block, env.StartAddress()}; @@ -588,7 +589,7 @@ std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline( } return std::make_unique<ComputePipeline>(device, texture_cache, buffer_cache, program_manager, - program.info, code, code_spirv); + program.info, code, code_spirv, force_context_flush); } catch (Shader::Exception& exception) { LOG_ERROR(Render_OpenGL, "{}", exception.what()); return nullptr; diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h index f82420592..6b9732fca 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.h +++ b/src/video_core/renderer_opengl/gl_shader_cache.h @@ -50,14 +50,16 @@ private: std::unique_ptr<GraphicsPipeline> CreateGraphicsPipeline( ShaderContext::ShaderPools& pools, const GraphicsPipelineKey& key, - std::span<Shader::Environment* const> envs, bool build_in_parallel); + std::span<Shader::Environment* const> envs, bool use_shader_workers, + bool force_context_flush = false); std::unique_ptr<ComputePipeline> CreateComputePipeline(const ComputePipelineKey& key, const VideoCommon::ShaderInfo* shader); std::unique_ptr<ComputePipeline> CreateComputePipeline(ShaderContext::ShaderPools& pools, const ComputePipelineKey& key, - Shader::Environment& env); + Shader::Environment& env, + bool force_context_flush = false); std::unique_ptr<ShaderWorker> CreateWorkers() const; diff --git a/src/video_core/renderer_opengl/gl_shader_manager.cpp b/src/video_core/renderer_opengl/gl_shader_manager.cpp index d9c29d8b7..98841ae65 100644 --- a/src/video_core/renderer_opengl/gl_shader_manager.cpp +++ b/src/video_core/renderer_opengl/gl_shader_manager.cpp @@ -1,2 +1,123 @@ // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later + +#include <glad/glad.h> + +#include "video_core/renderer_opengl/gl_shader_manager.h" + +namespace OpenGL { + +static constexpr std::array ASSEMBLY_PROGRAM_ENUMS{ + GL_VERTEX_PROGRAM_NV, GL_TESS_CONTROL_PROGRAM_NV, GL_TESS_EVALUATION_PROGRAM_NV, + GL_GEOMETRY_PROGRAM_NV, GL_FRAGMENT_PROGRAM_NV, +}; + +ProgramManager::ProgramManager(const Device& device) { + glCreateProgramPipelines(1, &pipeline.handle); + if (device.UseAssemblyShaders()) { + glEnable(GL_COMPUTE_PROGRAM_NV); + } +} + +void ProgramManager::BindComputeProgram(GLuint program) { + glUseProgram(program); + is_compute_bound = true; +} + +void ProgramManager::BindComputeAssemblyProgram(GLuint program) { + if (current_assembly_compute_program != program) { + current_assembly_compute_program = program; + glBindProgramARB(GL_COMPUTE_PROGRAM_NV, program); + } + UnbindPipeline(); +} + +void ProgramManager::BindSourcePrograms(std::span<const OGLProgram, NUM_STAGES> programs) { + static constexpr std::array<GLenum, 5> stage_enums{ + GL_VERTEX_SHADER_BIT, GL_TESS_CONTROL_SHADER_BIT, GL_TESS_EVALUATION_SHADER_BIT, + GL_GEOMETRY_SHADER_BIT, GL_FRAGMENT_SHADER_BIT, + }; + for (size_t stage = 0; stage < NUM_STAGES; ++stage) { + if (current_programs[stage] != programs[stage].handle) { + current_programs[stage] = programs[stage].handle; + glUseProgramStages(pipeline.handle, stage_enums[stage], programs[stage].handle); + } + } + BindPipeline(); +} + +void ProgramManager::BindPresentPrograms(GLuint vertex, GLuint fragment) { + if (current_programs[0] != vertex) { + current_programs[0] = vertex; + glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, vertex); + } + if (current_programs[4] != fragment) { + current_programs[4] = fragment; + glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, fragment); + } + glUseProgramStages( + pipeline.handle, + GL_TESS_CONTROL_SHADER_BIT | GL_TESS_EVALUATION_SHADER_BIT | GL_GEOMETRY_SHADER_BIT, 0); + current_programs[1] = 0; + current_programs[2] = 0; + current_programs[3] = 0; + + if (current_stage_mask != 0) { + current_stage_mask = 0; + for (const GLenum program_type : ASSEMBLY_PROGRAM_ENUMS) { + glDisable(program_type); + } + } + BindPipeline(); +} + +void ProgramManager::BindAssemblyPrograms(std::span<const OGLAssemblyProgram, NUM_STAGES> programs, + u32 stage_mask) { + const u32 changed_mask = current_stage_mask ^ stage_mask; + current_stage_mask = stage_mask; + + if (changed_mask != 0) { + for (size_t stage = 0; stage < NUM_STAGES; ++stage) { + if (((changed_mask >> stage) & 1) != 0) { + if (((stage_mask >> stage) & 1) != 0) { + glEnable(ASSEMBLY_PROGRAM_ENUMS[stage]); + } else { + glDisable(ASSEMBLY_PROGRAM_ENUMS[stage]); + } + } + } + } + for (size_t stage = 0; stage < NUM_STAGES; ++stage) { + if (current_programs[stage] != programs[stage].handle) { + current_programs[stage] = programs[stage].handle; + glBindProgramARB(ASSEMBLY_PROGRAM_ENUMS[stage], programs[stage].handle); + } + } + UnbindPipeline(); +} + +void ProgramManager::RestoreGuestCompute() {} + +void ProgramManager::BindPipeline() { + if (!is_pipeline_bound) { + is_pipeline_bound = true; + glBindProgramPipeline(pipeline.handle); + } + UnbindCompute(); +} + +void ProgramManager::UnbindPipeline() { + if (is_pipeline_bound) { + is_pipeline_bound = false; + glBindProgramPipeline(0); + } + UnbindCompute(); +} + +void ProgramManager::UnbindCompute() { + if (is_compute_bound) { + is_compute_bound = false; + glUseProgram(0); + } +} +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_shader_manager.h b/src/video_core/renderer_opengl/gl_shader_manager.h index a84f5aeb3..07ffab77f 100644 --- a/src/video_core/renderer_opengl/gl_shader_manager.h +++ b/src/video_core/renderer_opengl/gl_shader_manager.h @@ -6,8 +6,6 @@ #include <array> #include <span> -#include <glad/glad.h> - #include "video_core/renderer_opengl/gl_device.h" #include "video_core/renderer_opengl/gl_resource_manager.h" @@ -16,121 +14,28 @@ namespace OpenGL { class ProgramManager { static constexpr size_t NUM_STAGES = 5; - static constexpr std::array ASSEMBLY_PROGRAM_ENUMS{ - GL_VERTEX_PROGRAM_NV, GL_TESS_CONTROL_PROGRAM_NV, GL_TESS_EVALUATION_PROGRAM_NV, - GL_GEOMETRY_PROGRAM_NV, GL_FRAGMENT_PROGRAM_NV, - }; - public: - explicit ProgramManager(const Device& device) { - glCreateProgramPipelines(1, &pipeline.handle); - if (device.UseAssemblyShaders()) { - glEnable(GL_COMPUTE_PROGRAM_NV); - } - } - - void BindComputeProgram(GLuint program) { - glUseProgram(program); - is_compute_bound = true; - } - - void BindComputeAssemblyProgram(GLuint program) { - if (current_assembly_compute_program != program) { - current_assembly_compute_program = program; - glBindProgramARB(GL_COMPUTE_PROGRAM_NV, program); - } - UnbindPipeline(); - } - - void BindSourcePrograms(std::span<const OGLProgram, NUM_STAGES> programs) { - static constexpr std::array<GLenum, 5> stage_enums{ - GL_VERTEX_SHADER_BIT, GL_TESS_CONTROL_SHADER_BIT, GL_TESS_EVALUATION_SHADER_BIT, - GL_GEOMETRY_SHADER_BIT, GL_FRAGMENT_SHADER_BIT, - }; - for (size_t stage = 0; stage < NUM_STAGES; ++stage) { - if (current_programs[stage] != programs[stage].handle) { - current_programs[stage] = programs[stage].handle; - glUseProgramStages(pipeline.handle, stage_enums[stage], programs[stage].handle); - } - } - BindPipeline(); - } - - void BindPresentPrograms(GLuint vertex, GLuint fragment) { - if (current_programs[0] != vertex) { - current_programs[0] = vertex; - glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, vertex); - } - if (current_programs[4] != fragment) { - current_programs[4] = fragment; - glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, fragment); - } - glUseProgramStages( - pipeline.handle, - GL_TESS_CONTROL_SHADER_BIT | GL_TESS_EVALUATION_SHADER_BIT | GL_GEOMETRY_SHADER_BIT, 0); - current_programs[1] = 0; - current_programs[2] = 0; - current_programs[3] = 0; - - if (current_stage_mask != 0) { - current_stage_mask = 0; - for (const GLenum program_type : ASSEMBLY_PROGRAM_ENUMS) { - glDisable(program_type); - } - } - BindPipeline(); - } + explicit ProgramManager(const Device& device); + + void BindComputeProgram(GLuint program); + + void BindComputeAssemblyProgram(GLuint program); + + void BindSourcePrograms(std::span<const OGLProgram, NUM_STAGES> programs); + + void BindPresentPrograms(GLuint vertex, GLuint fragment); void BindAssemblyPrograms(std::span<const OGLAssemblyProgram, NUM_STAGES> programs, - u32 stage_mask) { - const u32 changed_mask = current_stage_mask ^ stage_mask; - current_stage_mask = stage_mask; - - if (changed_mask != 0) { - for (size_t stage = 0; stage < NUM_STAGES; ++stage) { - if (((changed_mask >> stage) & 1) != 0) { - if (((stage_mask >> stage) & 1) != 0) { - glEnable(ASSEMBLY_PROGRAM_ENUMS[stage]); - } else { - glDisable(ASSEMBLY_PROGRAM_ENUMS[stage]); - } - } - } - } - for (size_t stage = 0; stage < NUM_STAGES; ++stage) { - if (current_programs[stage] != programs[stage].handle) { - current_programs[stage] = programs[stage].handle; - glBindProgramARB(ASSEMBLY_PROGRAM_ENUMS[stage], programs[stage].handle); - } - } - UnbindPipeline(); - } - - void RestoreGuestCompute() {} + u32 stage_mask); + + void RestoreGuestCompute(); private: - void BindPipeline() { - if (!is_pipeline_bound) { - is_pipeline_bound = true; - glBindProgramPipeline(pipeline.handle); - } - UnbindCompute(); - } - - void UnbindPipeline() { - if (is_pipeline_bound) { - is_pipeline_bound = false; - glBindProgramPipeline(0); - } - UnbindCompute(); - } - - void UnbindCompute() { - if (is_compute_bound) { - is_compute_bound = false; - glUseProgram(0); - } - } + void BindPipeline(); + + void UnbindPipeline(); + + void UnbindCompute(); OGLPipeline pipeline; bool is_pipeline_bound{}; diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index 9f7ce7414..eb6e43a08 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -557,6 +557,14 @@ void TextureCacheRuntime::CopyImage(Image& dst_image, Image& src_image, } } +void TextureCacheRuntime::CopyImageMSAA(Image& dst_image, Image& src_image, + std::span<const VideoCommon::ImageCopy> copies) { + LOG_DEBUG(Render_OpenGL, "Copying from {} samples to {} samples", src_image.info.num_samples, + dst_image.info.num_samples); + // TODO: Leverage the format conversion pass if possible/accurate. + util_shaders.CopyMSAA(dst_image, src_image, copies); +} + void TextureCacheRuntime::ReinterpretImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies) { LOG_DEBUG(Render_OpenGL, "Converting {} to {}", src.info.format, dst.info.format); diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h index 5d9d370f2..e30875496 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.h +++ b/src/video_core/renderer_opengl/gl_texture_cache.h @@ -93,12 +93,19 @@ public: return device.CanReportMemoryUsage(); } - bool ShouldReinterpret([[maybe_unused]] Image& dst, [[maybe_unused]] Image& src) { + bool ShouldReinterpret([[maybe_unused]] Image& dst, + [[maybe_unused]] Image& src) const noexcept { + return true; + } + + bool CanUploadMSAA() const noexcept { return true; } void CopyImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies); + void CopyImageMSAA(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies); + void ReinterpretImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies); void ConvertImage(Framebuffer* dst, ImageView& dst_view, ImageView& src_view) { diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index de95f2634..2a74c1d05 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -17,8 +17,14 @@ #include "core/frontend/emu_window.h" #include "core/memory.h" #include "core/telemetry_session.h" +#include "video_core/host_shaders/ffx_a_h.h" +#include "video_core/host_shaders/ffx_fsr1_h.h" +#include "video_core/host_shaders/full_screen_triangle_vert.h" #include "video_core/host_shaders/fxaa_frag.h" #include "video_core/host_shaders/fxaa_vert.h" +#include "video_core/host_shaders/opengl_fidelityfx_fsr_easu_frag.h" +#include "video_core/host_shaders/opengl_fidelityfx_fsr_frag.h" +#include "video_core/host_shaders/opengl_fidelityfx_fsr_rcas_frag.h" #include "video_core/host_shaders/opengl_present_frag.h" #include "video_core/host_shaders/opengl_present_scaleforce_frag.h" #include "video_core/host_shaders/opengl_present_vert.h" @@ -31,6 +37,7 @@ #include "video_core/host_shaders/smaa_edge_detection_vert.h" #include "video_core/host_shaders/smaa_neighborhood_blending_frag.h" #include "video_core/host_shaders/smaa_neighborhood_blending_vert.h" +#include "video_core/renderer_opengl/gl_fsr.h" #include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/gl_shader_util.h" @@ -268,12 +275,17 @@ void RendererOpenGL::InitOpenGLObjects() { fxaa_vertex = CreateProgram(HostShaders::FXAA_VERT, GL_VERTEX_SHADER); fxaa_fragment = CreateProgram(HostShaders::FXAA_FRAG, GL_FRAGMENT_SHADER); - const auto SmaaShader = [](std::string_view specialized_source, GLenum stage) { - std::string shader_source{specialized_source}; - constexpr std::string_view include_string = "#include \"opengl_smaa.glsl\""; + const auto replace_include = [](std::string& shader_source, std::string_view include_name, + std::string_view include_content) { + const std::string include_string = fmt::format("#include \"{}\"", include_name); const std::size_t pos = shader_source.find(include_string); ASSERT(pos != std::string::npos); - shader_source.replace(pos, include_string.size(), HostShaders::OPENGL_SMAA_GLSL); + shader_source.replace(pos, include_string.size(), include_content); + }; + + const auto SmaaShader = [&](std::string_view specialized_source, GLenum stage) { + std::string shader_source{specialized_source}; + replace_include(shader_source, "opengl_smaa.glsl", HostShaders::OPENGL_SMAA_GLSL); return CreateProgram(shader_source, stage); }; @@ -298,14 +310,32 @@ void RendererOpenGL::InitOpenGLObjects() { CreateProgram(fmt::format("#version 460\n{}", HostShaders::OPENGL_PRESENT_SCALEFORCE_FRAG), GL_FRAGMENT_SHADER); + std::string fsr_source{HostShaders::OPENGL_FIDELITYFX_FSR_FRAG}; + replace_include(fsr_source, "ffx_a.h", HostShaders::FFX_A_H); + replace_include(fsr_source, "ffx_fsr1.h", HostShaders::FFX_FSR1_H); + + std::string fsr_easu_frag_source{HostShaders::OPENGL_FIDELITYFX_FSR_EASU_FRAG}; + std::string fsr_rcas_frag_source{HostShaders::OPENGL_FIDELITYFX_FSR_RCAS_FRAG}; + replace_include(fsr_easu_frag_source, "opengl_fidelityfx_fsr.frag", fsr_source); + replace_include(fsr_rcas_frag_source, "opengl_fidelityfx_fsr.frag", fsr_source); + + fsr = std::make_unique<FSR>(HostShaders::FULL_SCREEN_TRIANGLE_VERT, fsr_easu_frag_source, + fsr_rcas_frag_source); + // Generate presentation sampler present_sampler.Create(); glSamplerParameteri(present_sampler.handle, GL_TEXTURE_MIN_FILTER, GL_LINEAR); glSamplerParameteri(present_sampler.handle, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); present_sampler_nn.Create(); glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_MIN_FILTER, GL_NEAREST); glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); // Generate VBO handle for drawing vertex_buffer.Create(); @@ -525,6 +555,31 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { glBindTextureUnit(0, aa_texture.handle); } + glDisablei(GL_SCISSOR_TEST, 0); + + if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { + if (!fsr->AreBuffersInitialized()) { + fsr->InitBuffers(); + } + + auto crop_rect = framebuffer_crop_rect; + if (crop_rect.GetWidth() == 0) { + crop_rect.right = framebuffer_width; + } + if (crop_rect.GetHeight() == 0) { + crop_rect.bottom = framebuffer_height; + } + crop_rect = crop_rect.Scale(Settings::values.resolution_info.up_factor); + const auto fsr_input_width = Settings::values.resolution_info.ScaleUp(framebuffer_width); + const auto fsr_input_height = Settings::values.resolution_info.ScaleUp(framebuffer_height); + glBindSampler(0, present_sampler.handle); + fsr->Draw(program_manager, layout.screen, fsr_input_width, fsr_input_height, crop_rect); + } else { + if (fsr->AreBuffersInitialized()) { + fsr->ReleaseBuffers(); + } + } + const std::array ortho_matrix = MakeOrthographicMatrix(static_cast<float>(layout.width), static_cast<float>(layout.height)); @@ -540,10 +595,7 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { case Settings::ScalingFilter::ScaleForce: return present_scaleforce_fragment.handle; case Settings::ScalingFilter::Fsr: - LOG_WARNING( - Render_OpenGL, - "FidelityFX Super Resolution is not supported in OpenGL, changing to ScaleForce"); - return present_scaleforce_fragment.handle; + return fsr->GetPresentFragmentProgram().handle; default: return present_bilinear_fragment.handle; } @@ -578,15 +630,18 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { f32 scale_u = static_cast<f32>(framebuffer_width) / static_cast<f32>(screen_info.texture.width); f32 scale_v = static_cast<f32>(framebuffer_height) / static_cast<f32>(screen_info.texture.height); - // Scale the output by the crop width/height. This is commonly used with 1280x720 rendering - // (e.g. handheld mode) on a 1920x1080 framebuffer. - if (framebuffer_crop_rect.GetWidth() > 0) { - scale_u = static_cast<f32>(framebuffer_crop_rect.GetWidth()) / - static_cast<f32>(screen_info.texture.width); - } - if (framebuffer_crop_rect.GetHeight() > 0) { - scale_v = static_cast<f32>(framebuffer_crop_rect.GetHeight()) / - static_cast<f32>(screen_info.texture.height); + + if (Settings::values.scaling_filter.GetValue() != Settings::ScalingFilter::Fsr) { + // Scale the output by the crop width/height. This is commonly used with 1280x720 rendering + // (e.g. handheld mode) on a 1920x1080 framebuffer. + if (framebuffer_crop_rect.GetWidth() > 0) { + scale_u = static_cast<f32>(framebuffer_crop_rect.GetWidth()) / + static_cast<f32>(screen_info.texture.width); + } + if (framebuffer_crop_rect.GetHeight() > 0) { + scale_v = static_cast<f32>(framebuffer_crop_rect.GetHeight()) / + static_cast<f32>(screen_info.texture.height); + } } if (Settings::values.anti_aliasing.GetValue() == Settings::AntiAliasing::Fxaa && !screen_info.was_accelerated) { @@ -612,7 +667,6 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { } else { glDisable(GL_FRAMEBUFFER_SRGB); } - glDisablei(GL_SCISSOR_TEST, 0); glViewportIndexedf(0, 0.0f, 0.0f, static_cast<GLfloat>(layout.width), static_cast<GLfloat>(layout.height)); diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index cc97d7b26..f1d5fd954 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -10,6 +10,7 @@ #include "video_core/renderer_base.h" #include "video_core/renderer_opengl/gl_device.h" +#include "video_core/renderer_opengl/gl_fsr.h" #include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_resource_manager.h" #include "video_core/renderer_opengl/gl_shader_manager.h" @@ -141,6 +142,8 @@ private: OGLTexture smaa_edges_tex; OGLTexture smaa_blend_tex; + std::unique_ptr<FSR> fsr; + /// OpenGL framebuffer data std::vector<u8> gl_framebuffer_data; diff --git a/src/video_core/renderer_opengl/util_shaders.cpp b/src/video_core/renderer_opengl/util_shaders.cpp index 404def62e..2c7ac210b 100644 --- a/src/video_core/renderer_opengl/util_shaders.cpp +++ b/src/video_core/renderer_opengl/util_shaders.cpp @@ -12,6 +12,8 @@ #include "video_core/host_shaders/astc_decoder_comp.h" #include "video_core/host_shaders/block_linear_unswizzle_2d_comp.h" #include "video_core/host_shaders/block_linear_unswizzle_3d_comp.h" +#include "video_core/host_shaders/convert_msaa_to_non_msaa_comp.h" +#include "video_core/host_shaders/convert_non_msaa_to_msaa_comp.h" #include "video_core/host_shaders/opengl_convert_s8d24_comp.h" #include "video_core/host_shaders/opengl_copy_bc4_comp.h" #include "video_core/host_shaders/pitch_unswizzle_comp.h" @@ -51,7 +53,9 @@ UtilShaders::UtilShaders(ProgramManager& program_manager_) block_linear_unswizzle_3d_program(MakeProgram(BLOCK_LINEAR_UNSWIZZLE_3D_COMP)), pitch_unswizzle_program(MakeProgram(PITCH_UNSWIZZLE_COMP)), copy_bc4_program(MakeProgram(OPENGL_COPY_BC4_COMP)), - convert_s8d24_program(MakeProgram(OPENGL_CONVERT_S8D24_COMP)) { + convert_s8d24_program(MakeProgram(OPENGL_CONVERT_S8D24_COMP)), + convert_ms_to_nonms_program(MakeProgram(CONVERT_MSAA_TO_NON_MSAA_COMP)), + convert_nonms_to_ms_program(MakeProgram(CONVERT_NON_MSAA_TO_MSAA_COMP)) { const auto swizzle_table = Tegra::Texture::MakeSwizzleTable(); swizzle_table_buffer.Create(); glNamedBufferStorage(swizzle_table_buffer.handle, sizeof(swizzle_table), &swizzle_table, 0); @@ -269,6 +273,33 @@ void UtilShaders::ConvertS8D24(Image& dst_image, std::span<const ImageCopy> copi program_manager.RestoreGuestCompute(); } +void UtilShaders::CopyMSAA(Image& dst_image, Image& src_image, + std::span<const VideoCommon::ImageCopy> copies) { + const bool is_ms_to_non_ms = src_image.info.num_samples > 1 && dst_image.info.num_samples == 1; + const auto program_handle = + is_ms_to_non_ms ? convert_ms_to_nonms_program.handle : convert_nonms_to_ms_program.handle; + program_manager.BindComputeProgram(program_handle); + + for (const ImageCopy& copy : copies) { + ASSERT(copy.src_subresource.base_layer == 0); + ASSERT(copy.src_subresource.num_layers == 1); + ASSERT(copy.dst_subresource.base_layer == 0); + ASSERT(copy.dst_subresource.num_layers == 1); + + glBindImageTexture(0, src_image.StorageHandle(), copy.src_subresource.base_level, GL_TRUE, + 0, GL_READ_ONLY, GL_RGBA8); + glBindImageTexture(1, dst_image.StorageHandle(), copy.dst_subresource.base_level, GL_TRUE, + 0, GL_WRITE_ONLY, GL_RGBA8); + + const u32 num_dispatches_x = Common::DivCeil(copy.extent.width, 8U); + const u32 num_dispatches_y = Common::DivCeil(copy.extent.height, 8U); + const u32 num_dispatches_z = copy.extent.depth; + + glDispatchCompute(num_dispatches_x, num_dispatches_y, num_dispatches_z); + } + program_manager.RestoreGuestCompute(); +} + GLenum StoreFormat(u32 bytes_per_block) { switch (bytes_per_block) { case 1: diff --git a/src/video_core/renderer_opengl/util_shaders.h b/src/video_core/renderer_opengl/util_shaders.h index 44efb6ecf..9013808e7 100644 --- a/src/video_core/renderer_opengl/util_shaders.h +++ b/src/video_core/renderer_opengl/util_shaders.h @@ -40,6 +40,9 @@ public: void ConvertS8D24(Image& dst_image, std::span<const VideoCommon::ImageCopy> copies); + void CopyMSAA(Image& dst_image, Image& src_image, + std::span<const VideoCommon::ImageCopy> copies); + private: ProgramManager& program_manager; @@ -51,6 +54,8 @@ private: OGLProgram pitch_unswizzle_program; OGLProgram copy_bc4_program; OGLProgram convert_s8d24_program; + OGLProgram convert_ms_to_nonms_program; + OGLProgram convert_nonms_to_ms_program; }; GLenum StoreFormat(u32 bytes_per_block); diff --git a/src/video_core/renderer_vulkan/blit_image.cpp b/src/video_core/renderer_vulkan/blit_image.cpp index 3f2b139e0..cf2964a3f 100644 --- a/src/video_core/renderer_vulkan/blit_image.cpp +++ b/src/video_core/renderer_vulkan/blit_image.cpp @@ -4,14 +4,16 @@ #include <algorithm> #include "common/settings.h" +#include "video_core/host_shaders/blit_color_float_frag_spv.h" #include "video_core/host_shaders/convert_abgr8_to_d24s8_frag_spv.h" #include "video_core/host_shaders/convert_d24s8_to_abgr8_frag_spv.h" #include "video_core/host_shaders/convert_depth_to_float_frag_spv.h" #include "video_core/host_shaders/convert_float_to_depth_frag_spv.h" #include "video_core/host_shaders/convert_s8d24_to_abgr8_frag_spv.h" #include "video_core/host_shaders/full_screen_triangle_vert_spv.h" -#include "video_core/host_shaders/vulkan_blit_color_float_frag_spv.h" #include "video_core/host_shaders/vulkan_blit_depth_stencil_frag_spv.h" +#include "video_core/host_shaders/vulkan_color_clear_frag_spv.h" +#include "video_core/host_shaders/vulkan_color_clear_vert_spv.h" #include "video_core/renderer_vulkan/blit_image.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/vk_scheduler.h" @@ -69,10 +71,11 @@ constexpr VkDescriptorSetLayoutCreateInfo TWO_TEXTURES_DESCRIPTOR_SET_LAYOUT_CRE .bindingCount = static_cast<u32>(TWO_TEXTURES_DESCRIPTOR_SET_LAYOUT_BINDINGS.size()), .pBindings = TWO_TEXTURES_DESCRIPTOR_SET_LAYOUT_BINDINGS.data(), }; -constexpr VkPushConstantRange PUSH_CONSTANT_RANGE{ - .stageFlags = VK_SHADER_STAGE_VERTEX_BIT, +template <VkShaderStageFlags stageFlags, size_t size> +inline constexpr VkPushConstantRange PUSH_CONSTANT_RANGE{ + .stageFlags = stageFlags, .offset = 0, - .size = sizeof(PushConstants), + .size = static_cast<u32>(size), }; constexpr VkPipelineVertexInputStateCreateInfo PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO{ .sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO, @@ -125,10 +128,8 @@ constexpr VkPipelineMultisampleStateCreateInfo PIPELINE_MULTISAMPLE_STATE_CREATE .alphaToCoverageEnable = VK_FALSE, .alphaToOneEnable = VK_FALSE, }; -constexpr std::array DYNAMIC_STATES{ - VK_DYNAMIC_STATE_VIEWPORT, - VK_DYNAMIC_STATE_SCISSOR, -}; +constexpr std::array DYNAMIC_STATES{VK_DYNAMIC_STATE_VIEWPORT, VK_DYNAMIC_STATE_SCISSOR, + VK_DYNAMIC_STATE_BLEND_CONSTANTS}; constexpr VkPipelineDynamicStateCreateInfo PIPELINE_DYNAMIC_STATE_CREATE_INFO{ .sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO, .pNext = nullptr, @@ -205,15 +206,15 @@ inline constexpr VkSamplerCreateInfo SAMPLER_CREATE_INFO{ }; constexpr VkPipelineLayoutCreateInfo PipelineLayoutCreateInfo( - const VkDescriptorSetLayout* set_layout) { + const VkDescriptorSetLayout* set_layout, vk::Span<VkPushConstantRange> push_constants) { return VkPipelineLayoutCreateInfo{ .sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO, .pNext = nullptr, .flags = 0, - .setLayoutCount = 1, + .setLayoutCount = (set_layout != nullptr ? 1u : 0u), .pSetLayouts = set_layout, - .pushConstantRangeCount = 1, - .pPushConstantRanges = &PUSH_CONSTANT_RANGE, + .pushConstantRangeCount = push_constants.size(), + .pPushConstantRanges = push_constants.data(), }; } @@ -302,8 +303,7 @@ void UpdateTwoTexturesDescriptorSet(const Device& device, VkDescriptorSet descri device.GetLogical().UpdateDescriptorSets(write_descriptor_sets, nullptr); } -void BindBlitState(vk::CommandBuffer cmdbuf, VkPipelineLayout layout, const Region2D& dst_region, - const Region2D& src_region) { +void BindBlitState(vk::CommandBuffer cmdbuf, const Region2D& dst_region) { const VkOffset2D offset{ .x = std::min(dst_region.start.x, dst_region.end.x), .y = std::min(dst_region.start.y, dst_region.end.y), @@ -325,15 +325,23 @@ void BindBlitState(vk::CommandBuffer cmdbuf, VkPipelineLayout layout, const Regi .offset = offset, .extent = extent, }; - const float scale_x = static_cast<float>(src_region.end.x - src_region.start.x); - const float scale_y = static_cast<float>(src_region.end.y - src_region.start.y); + cmdbuf.SetViewport(0, viewport); + cmdbuf.SetScissor(0, scissor); +} + +void BindBlitState(vk::CommandBuffer cmdbuf, VkPipelineLayout layout, const Region2D& dst_region, + const Region2D& src_region, const Extent3D& src_size = {1, 1, 1}) { + BindBlitState(cmdbuf, dst_region); + const float scale_x = static_cast<float>(src_region.end.x - src_region.start.x) / + static_cast<float>(src_size.width); + const float scale_y = static_cast<float>(src_region.end.y - src_region.start.y) / + static_cast<float>(src_size.height); const PushConstants push_constants{ .tex_scale = {scale_x, scale_y}, - .tex_offset = {static_cast<float>(src_region.start.x), - static_cast<float>(src_region.start.y)}, + .tex_offset = {static_cast<float>(src_region.start.x) / static_cast<float>(src_size.width), + static_cast<float>(src_region.start.y) / + static_cast<float>(src_size.height)}, }; - cmdbuf.SetViewport(0, viewport); - cmdbuf.SetScissor(0, scissor); cmdbuf.PushConstants(layout, VK_SHADER_STAGE_VERTEX_BIT, push_constants); } @@ -347,6 +355,51 @@ VkExtent2D GetConversionExtent(const ImageView& src_image_view) { .height = is_rescaled ? resolution.ScaleUp(height) : height, }; } + +void TransitionImageLayout(vk::CommandBuffer& cmdbuf, VkImage image, VkImageLayout target_layout, + VkImageLayout source_layout = VK_IMAGE_LAYOUT_GENERAL) { + constexpr VkFlags flags{VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | + VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_SHADER_READ_BIT}; + const VkImageMemoryBarrier barrier{ + .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, + .pNext = nullptr, + .srcAccessMask = flags, + .dstAccessMask = flags, + .oldLayout = source_layout, + .newLayout = target_layout, + .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, + .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, + .image = image, + .subresourceRange{ + .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, + .baseMipLevel = 0, + .levelCount = 1, + .baseArrayLayer = 0, + .layerCount = 1, + }, + }; + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, + 0, barrier); +} + +void BeginRenderPass(vk::CommandBuffer& cmdbuf, const Framebuffer* framebuffer) { + const VkRenderPass render_pass = framebuffer->RenderPass(); + const VkFramebuffer framebuffer_handle = framebuffer->Handle(); + const VkExtent2D render_area = framebuffer->RenderArea(); + const VkRenderPassBeginInfo renderpass_bi{ + .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO, + .pNext = nullptr, + .renderPass = render_pass, + .framebuffer = framebuffer_handle, + .renderArea{ + .offset{}, + .extent = render_area, + }, + .clearValueCount = 0, + .pClearValues = nullptr, + }; + cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE); +} } // Anonymous namespace BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_, @@ -360,13 +413,20 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_, descriptor_pool.Allocator(*one_texture_set_layout, TEXTURE_DESCRIPTOR_BANK_INFO<1>)}, two_textures_descriptor_allocator{ descriptor_pool.Allocator(*two_textures_set_layout, TEXTURE_DESCRIPTOR_BANK_INFO<2>)}, - one_texture_pipeline_layout(device.GetLogical().CreatePipelineLayout( - PipelineLayoutCreateInfo(one_texture_set_layout.address()))), - two_textures_pipeline_layout(device.GetLogical().CreatePipelineLayout( - PipelineLayoutCreateInfo(two_textures_set_layout.address()))), + one_texture_pipeline_layout(device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo( + one_texture_set_layout.address(), + PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_VERTEX_BIT, sizeof(PushConstants)>))), + two_textures_pipeline_layout( + device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo( + two_textures_set_layout.address(), + PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_VERTEX_BIT, sizeof(PushConstants)>))), + clear_color_pipeline_layout(device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo( + nullptr, PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(float) * 4>))), full_screen_vert(BuildShader(device, FULL_SCREEN_TRIANGLE_VERT_SPV)), - blit_color_to_color_frag(BuildShader(device, VULKAN_BLIT_COLOR_FLOAT_FRAG_SPV)), + blit_color_to_color_frag(BuildShader(device, BLIT_COLOR_FLOAT_FRAG_SPV)), blit_depth_stencil_frag(BuildShader(device, VULKAN_BLIT_DEPTH_STENCIL_FRAG_SPV)), + clear_color_vert(BuildShader(device, VULKAN_COLOR_CLEAR_VERT_SPV)), + clear_color_frag(BuildShader(device, VULKAN_COLOR_CLEAR_FRAG_SPV)), convert_depth_to_float_frag(BuildShader(device, CONVERT_DEPTH_TO_FLOAT_FRAG_SPV)), convert_float_to_depth_frag(BuildShader(device, CONVERT_FLOAT_TO_DEPTH_FRAG_SPV)), convert_abgr8_to_d24s8_frag(BuildShader(device, CONVERT_ABGR8_TO_D24S8_FRAG_SPV)), @@ -404,6 +464,32 @@ void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView scheduler.InvalidateState(); } +void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView src_image_view, + VkImage src_image, VkSampler src_sampler, + const Region2D& dst_region, const Region2D& src_region, + const Extent3D& src_size) { + const BlitImagePipelineKey key{ + .renderpass = dst_framebuffer->RenderPass(), + .operation = Tegra::Engines::Fermi2D::Operation::SrcCopy, + }; + const VkPipelineLayout layout = *one_texture_pipeline_layout; + const VkPipeline pipeline = FindOrEmplaceColorPipeline(key); + scheduler.RequestOutsideRenderPassOperationContext(); + scheduler.Record([this, dst_framebuffer, src_image_view, src_image, src_sampler, dst_region, + src_region, src_size, pipeline, layout](vk::CommandBuffer cmdbuf) { + TransitionImageLayout(cmdbuf, src_image, VK_IMAGE_LAYOUT_READ_ONLY_OPTIMAL); + BeginRenderPass(cmdbuf, dst_framebuffer); + const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit(); + UpdateOneTextureDescriptorSet(device, descriptor_set, src_sampler, src_image_view); + cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline); + cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set, + nullptr); + BindBlitState(cmdbuf, layout, dst_region, src_region, src_size); + cmdbuf.Draw(3, 1, 0, 0); + cmdbuf.EndRenderPass(); + }); +} + void BlitImageHelper::BlitDepthStencil(const Framebuffer* dst_framebuffer, VkImageView src_depth_view, VkImageView src_stencil_view, const Region2D& dst_region, const Region2D& src_region, @@ -479,6 +565,30 @@ void BlitImageHelper::ConvertS8D24ToABGR8(const Framebuffer* dst_framebuffer, ConvertDepthStencil(*convert_s8d24_to_abgr8_pipeline, dst_framebuffer, src_image_view); } +void BlitImageHelper::ClearColor(const Framebuffer* dst_framebuffer, u8 color_mask, + const std::array<f32, 4>& clear_color, + const Region2D& dst_region) { + const BlitImagePipelineKey key{ + .renderpass = dst_framebuffer->RenderPass(), + .operation = Tegra::Engines::Fermi2D::Operation::BlendPremult, + }; + const VkPipeline pipeline = FindOrEmplaceClearColorPipeline(key); + const VkPipelineLayout layout = *clear_color_pipeline_layout; + scheduler.RequestRenderpass(dst_framebuffer); + scheduler.Record( + [pipeline, layout, color_mask, clear_color, dst_region](vk::CommandBuffer cmdbuf) { + cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline); + const std::array blend_color = { + (color_mask & 0x1) ? 1.0f : 0.0f, (color_mask & 0x2) ? 1.0f : 0.0f, + (color_mask & 0x4) ? 1.0f : 0.0f, (color_mask & 0x8) ? 1.0f : 0.0f}; + cmdbuf.SetBlendConstants(blend_color.data()); + BindBlitState(cmdbuf, dst_region); + cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, clear_color); + cmdbuf.Draw(3, 1, 0, 0); + }); + scheduler.InvalidateState(); +} + void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer, const ImageView& src_image_view) { const VkPipelineLayout layout = *one_texture_pipeline_layout; @@ -654,6 +764,58 @@ VkPipeline BlitImageHelper::FindOrEmplaceDepthStencilPipeline(const BlitImagePip return *blit_depth_stencil_pipelines.back(); } +VkPipeline BlitImageHelper::FindOrEmplaceClearColorPipeline(const BlitImagePipelineKey& key) { + const auto it = std::ranges::find(clear_color_keys, key); + if (it != clear_color_keys.end()) { + return *clear_color_pipelines[std::distance(clear_color_keys.begin(), it)]; + } + clear_color_keys.push_back(key); + const std::array stages = MakeStages(*clear_color_vert, *clear_color_frag); + const VkPipelineColorBlendAttachmentState color_blend_attachment_state{ + .blendEnable = VK_TRUE, + .srcColorBlendFactor = VK_BLEND_FACTOR_CONSTANT_COLOR, + .dstColorBlendFactor = VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_COLOR, + .colorBlendOp = VK_BLEND_OP_ADD, + .srcAlphaBlendFactor = VK_BLEND_FACTOR_CONSTANT_ALPHA, + .dstAlphaBlendFactor = VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_ALPHA, + .alphaBlendOp = VK_BLEND_OP_ADD, + .colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT | + VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT, + }; + const VkPipelineColorBlendStateCreateInfo color_blend_state_generic_create_info{ + .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .logicOpEnable = VK_FALSE, + .logicOp = VK_LOGIC_OP_CLEAR, + .attachmentCount = 1, + .pAttachments = &color_blend_attachment_state, + .blendConstants = {0.0f, 0.0f, 0.0f, 0.0f}, + }; + clear_color_pipelines.push_back(device.GetLogical().CreateGraphicsPipeline({ + .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .stageCount = static_cast<u32>(stages.size()), + .pStages = stages.data(), + .pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO, + .pInputAssemblyState = &PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO, + .pTessellationState = nullptr, + .pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO, + .pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO, + .pMultisampleState = &PIPELINE_MULTISAMPLE_STATE_CREATE_INFO, + .pDepthStencilState = &PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO, + .pColorBlendState = &color_blend_state_generic_create_info, + .pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO, + .layout = *clear_color_pipeline_layout, + .renderPass = key.renderpass, + .subpass = 0, + .basePipelineHandle = VK_NULL_HANDLE, + .basePipelineIndex = 0, + })); + return *clear_color_pipelines.back(); +} + void BlitImageHelper::ConvertPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass, bool is_target_depth) { if (pipeline) { diff --git a/src/video_core/renderer_vulkan/blit_image.h b/src/video_core/renderer_vulkan/blit_image.h index 5df679fb4..2976a7d91 100644 --- a/src/video_core/renderer_vulkan/blit_image.h +++ b/src/video_core/renderer_vulkan/blit_image.h @@ -10,6 +10,8 @@ namespace Vulkan { +using VideoCommon::Extent3D; +using VideoCommon::Offset2D; using VideoCommon::Region2D; class Device; @@ -36,6 +38,10 @@ public: Tegra::Engines::Fermi2D::Filter filter, Tegra::Engines::Fermi2D::Operation operation); + void BlitColor(const Framebuffer* dst_framebuffer, VkImageView src_image_view, + VkImage src_image, VkSampler src_sampler, const Region2D& dst_region, + const Region2D& src_region, const Extent3D& src_size); + void BlitDepthStencil(const Framebuffer* dst_framebuffer, VkImageView src_depth_view, VkImageView src_stencil_view, const Region2D& dst_region, const Region2D& src_region, Tegra::Engines::Fermi2D::Filter filter, @@ -55,6 +61,9 @@ public: void ConvertS8D24ToABGR8(const Framebuffer* dst_framebuffer, ImageView& src_image_view); + void ClearColor(const Framebuffer* dst_framebuffer, u8 color_mask, + const std::array<f32, 4>& clear_color, const Region2D& dst_region); + private: void Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer, const ImageView& src_image_view); @@ -66,6 +75,8 @@ private: [[nodiscard]] VkPipeline FindOrEmplaceDepthStencilPipeline(const BlitImagePipelineKey& key); + [[nodiscard]] VkPipeline FindOrEmplaceClearColorPipeline(const BlitImagePipelineKey& key); + void ConvertPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass, bool is_target_depth); void ConvertDepthToColorPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass); @@ -91,9 +102,12 @@ private: DescriptorAllocator two_textures_descriptor_allocator; vk::PipelineLayout one_texture_pipeline_layout; vk::PipelineLayout two_textures_pipeline_layout; + vk::PipelineLayout clear_color_pipeline_layout; vk::ShaderModule full_screen_vert; vk::ShaderModule blit_color_to_color_frag; vk::ShaderModule blit_depth_stencil_frag; + vk::ShaderModule clear_color_vert; + vk::ShaderModule clear_color_frag; vk::ShaderModule convert_depth_to_float_frag; vk::ShaderModule convert_float_to_depth_frag; vk::ShaderModule convert_abgr8_to_d24s8_frag; @@ -106,6 +120,8 @@ private: std::vector<vk::Pipeline> blit_color_pipelines; std::vector<BlitImagePipelineKey> blit_depth_stencil_keys; std::vector<vk::Pipeline> blit_depth_stencil_pipelines; + std::vector<BlitImagePipelineKey> clear_color_keys; + std::vector<vk::Pipeline> clear_color_pipelines; vk::Pipeline convert_d32_to_r32_pipeline; vk::Pipeline convert_r32_to_d32_pipeline; vk::Pipeline convert_d16_to_r16_pipeline; diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp index 52855120c..2a8d9e377 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp @@ -60,22 +60,9 @@ std::string GetDriverVersion(const Device& device) { 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; +std::string BuildCommaSeparatedExtensions( + const std::set<std::string, std::less<>>& available_extensions) { + return fmt::format("{}", fmt::join(available_extensions, ",")); } } // Anonymous namespace @@ -112,6 +99,7 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_, state_tracker, scheduler) { if (Settings::values.renderer_force_max_clock.GetValue() && device.ShouldBoostClocks()) { turbo_mode.emplace(instance, dld); + scheduler.RegisterOnSubmit([this] { turbo_mode->QueueSubmitted(); }); } Report(); } catch (const vk::Exception& exception) { @@ -120,6 +108,7 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_, } RendererVulkan::~RendererVulkan() { + scheduler.RegisterOnSubmit([] {}); void(device.GetLogical().WaitIdle()); } diff --git a/src/video_core/renderer_vulkan/vk_fsr.cpp b/src/video_core/renderer_vulkan/vk_fsr.cpp index 33daa8c1c..df972cd54 100644 --- a/src/video_core/renderer_vulkan/vk_fsr.cpp +++ b/src/video_core/renderer_vulkan/vk_fsr.cpp @@ -1,12 +1,11 @@ // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later -#include <cmath> -#include "common/bit_cast.h" #include "common/common_types.h" #include "common/div_ceil.h" #include "common/settings.h" +#include "video_core/fsr.h" #include "video_core/host_shaders/vulkan_fidelityfx_fsr_easu_fp16_comp_spv.h" #include "video_core/host_shaders/vulkan_fidelityfx_fsr_easu_fp32_comp_spv.h" #include "video_core/host_shaders/vulkan_fidelityfx_fsr_rcas_fp16_comp_spv.h" @@ -17,146 +16,7 @@ #include "video_core/vulkan_common/vulkan_device.h" namespace Vulkan { -namespace { -// Reimplementations of the constant generating functions in ffx_fsr1.h -// GCC generated a lot of warnings when using the official header. -u32 AU1_AH1_AF1(f32 f) { - static constexpr u32 base[512]{ - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, - 0x0080, 0x0100, 0x0200, 0x0400, 0x0800, 0x0c00, 0x1000, 0x1400, 0x1800, 0x1c00, 0x2000, - 0x2400, 0x2800, 0x2c00, 0x3000, 0x3400, 0x3800, 0x3c00, 0x4000, 0x4400, 0x4800, 0x4c00, - 0x5000, 0x5400, 0x5800, 0x5c00, 0x6000, 0x6400, 0x6800, 0x6c00, 0x7000, 0x7400, 0x7800, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8001, 0x8002, 0x8004, 0x8008, - 0x8010, 0x8020, 0x8040, 0x8080, 0x8100, 0x8200, 0x8400, 0x8800, 0x8c00, 0x9000, 0x9400, - 0x9800, 0x9c00, 0xa000, 0xa400, 0xa800, 0xac00, 0xb000, 0xb400, 0xb800, 0xbc00, 0xc000, - 0xc400, 0xc800, 0xcc00, 0xd000, 0xd400, 0xd800, 0xdc00, 0xe000, 0xe400, 0xe800, 0xec00, - 0xf000, 0xf400, 0xf800, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - }; - static constexpr s8 shift[512]{ - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x17, 0x16, - 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0f, 0x0e, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, - 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, - 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x17, - 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0f, 0x0e, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, - 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, - 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, - }; - const u32 u = Common::BitCast<u32>(f); - const u32 i = u >> 23; - return base[i] + ((u & 0x7fffff) >> shift[i]); -} - -u32 AU1_AH2_AF2(f32 a[2]) { - return AU1_AH1_AF1(a[0]) + (AU1_AH1_AF1(a[1]) << 16); -} - -void FsrEasuCon(u32 con0[4], u32 con1[4], u32 con2[4], u32 con3[4], f32 inputViewportInPixelsX, - f32 inputViewportInPixelsY, f32 inputSizeInPixelsX, f32 inputSizeInPixelsY, - f32 outputSizeInPixelsX, f32 outputSizeInPixelsY) { - con0[0] = Common::BitCast<u32>(inputViewportInPixelsX / outputSizeInPixelsX); - con0[1] = Common::BitCast<u32>(inputViewportInPixelsY / outputSizeInPixelsY); - con0[2] = Common::BitCast<u32>(0.5f * inputViewportInPixelsX / outputSizeInPixelsX - 0.5f); - con0[3] = Common::BitCast<u32>(0.5f * inputViewportInPixelsY / outputSizeInPixelsY - 0.5f); - con1[0] = Common::BitCast<u32>(1.0f / inputSizeInPixelsX); - con1[1] = Common::BitCast<u32>(1.0f / inputSizeInPixelsY); - con1[2] = Common::BitCast<u32>(1.0f / inputSizeInPixelsX); - con1[3] = Common::BitCast<u32>(-1.0f / inputSizeInPixelsY); - con2[0] = Common::BitCast<u32>(-1.0f / inputSizeInPixelsX); - con2[1] = Common::BitCast<u32>(2.0f / inputSizeInPixelsY); - con2[2] = Common::BitCast<u32>(1.0f / inputSizeInPixelsX); - con2[3] = Common::BitCast<u32>(2.0f / inputSizeInPixelsY); - con3[0] = Common::BitCast<u32>(0.0f / inputSizeInPixelsX); - con3[1] = Common::BitCast<u32>(4.0f / inputSizeInPixelsY); - con3[2] = con3[3] = 0; -} - -void FsrEasuConOffset(u32 con0[4], u32 con1[4], u32 con2[4], u32 con3[4], - f32 inputViewportInPixelsX, f32 inputViewportInPixelsY, - f32 inputSizeInPixelsX, f32 inputSizeInPixelsY, f32 outputSizeInPixelsX, - f32 outputSizeInPixelsY, f32 inputOffsetInPixelsX, f32 inputOffsetInPixelsY) { - FsrEasuCon(con0, con1, con2, con3, inputViewportInPixelsX, inputViewportInPixelsY, - inputSizeInPixelsX, inputSizeInPixelsY, outputSizeInPixelsX, outputSizeInPixelsY); - con0[2] = Common::BitCast<u32>(0.5f * inputViewportInPixelsX / outputSizeInPixelsX - 0.5f + - inputOffsetInPixelsX); - con0[3] = Common::BitCast<u32>(0.5f * inputViewportInPixelsY / outputSizeInPixelsY - 0.5f + - inputOffsetInPixelsY); -} - -void FsrRcasCon(u32* con, f32 sharpness) { - sharpness = std::exp2f(-sharpness); - f32 hSharp[2]{sharpness, sharpness}; - con[0] = Common::BitCast<u32>(sharpness); - con[1] = AU1_AH2_AF2(hSharp); - con[2] = 0; - con[3] = 0; -} -} // Anonymous namespace +using namespace FSR; FSR::FSR(const Device& device_, MemoryAllocator& memory_allocator_, size_t image_count_, VkExtent2D output_size_) diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp index f91bb5a1d..baedc4424 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp @@ -548,31 +548,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) { static_vector<VkVertexInputBindingDescription, 32> vertex_bindings; static_vector<VkVertexInputBindingDivisorDescriptionEXT, 32> vertex_binding_divisors; static_vector<VkVertexInputAttributeDescription, 32> vertex_attributes; - if (key.state.dynamic_vertex_input) { - const size_t num_vertex_arrays = std::min( - key.state.attributes.size(), static_cast<size_t>(device.GetMaxVertexInputBindings())); - for (size_t index = 0; index < num_vertex_arrays; ++index) { - const u32 type = key.state.DynamicAttributeType(index); - if (!stage_infos[0].loads.Generic(index) || type == 0) { - continue; - } - vertex_attributes.push_back({ - .location = static_cast<u32>(index), - .binding = 0, - .format = type == 1 ? VK_FORMAT_R32_SFLOAT - : type == 2 ? VK_FORMAT_R32_SINT - : VK_FORMAT_R32_UINT, - .offset = 0, - }); - } - if (!vertex_attributes.empty()) { - vertex_bindings.push_back({ - .binding = 0, - .stride = 4, - .inputRate = VK_VERTEX_INPUT_RATE_VERTEX, - }); - } - } else { + if (!key.state.dynamic_vertex_input) { const size_t num_vertex_arrays = std::min( Maxwell::NumVertexArrays, static_cast<size_t>(device.GetMaxVertexInputBindings())); for (size_t index = 0; index < num_vertex_arrays; ++index) { diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index ed4a72166..719edbcfb 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -266,10 +266,40 @@ void RasterizerVulkan::DrawIndirect() { buffer_cache.SetDrawIndirect(nullptr); } +void RasterizerVulkan::DrawTexture() { + MICROPROFILE_SCOPE(Vulkan_Drawing); + + SCOPE_EXIT({ gpu.TickWork(); }); + FlushWork(); + + query_cache.UpdateCounters(); + + texture_cache.SynchronizeGraphicsDescriptors(); + texture_cache.UpdateRenderTargets(false); + + UpdateDynamicStates(); + + const auto& draw_texture_state = maxwell3d->draw_manager->GetDrawTextureState(); + const auto& sampler = texture_cache.GetGraphicsSampler(draw_texture_state.src_sampler); + const auto& texture = texture_cache.GetImageView(draw_texture_state.src_texture); + Region2D dst_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x0), + .y = static_cast<s32>(draw_texture_state.dst_y0)}, + Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x1), + .y = static_cast<s32>(draw_texture_state.dst_y1)}}; + Region2D src_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.src_x0), + .y = static_cast<s32>(draw_texture_state.src_y0)}, + Offset2D{.x = static_cast<s32>(draw_texture_state.src_x1), + .y = static_cast<s32>(draw_texture_state.src_y1)}}; + blit_image.BlitColor(texture_cache.GetFramebuffer(), texture.RenderTarget(), + texture.ImageHandle(), sampler->Handle(), dst_region, src_region, + texture.size); +} + void RasterizerVulkan::Clear(u32 layer_count) { MICROPROFILE_SCOPE(Vulkan_Clearing); FlushWork(); + gpu_memory->FlushCaching(); query_cache.UpdateCounters(); @@ -364,7 +394,15 @@ void RasterizerVulkan::Clear(u32 layer_count) { cmdbuf.ClearAttachments(attachment, clear_rect); }); } else { - UNIMPLEMENTED_MSG("Unimplemented Clear only the specified channel"); + u8 color_mask = static_cast<u8>(regs.clear_surface.R | regs.clear_surface.G << 1 | + regs.clear_surface.B << 2 | regs.clear_surface.A << 3); + Region2D dst_region = { + Offset2D{.x = clear_rect.rect.offset.x, .y = clear_rect.rect.offset.y}, + Offset2D{.x = clear_rect.rect.offset.x + + static_cast<s32>(clear_rect.rect.extent.width), + .y = clear_rect.rect.offset.y + + static_cast<s32>(clear_rect.rect.extent.height)}}; + blit_image.ClearColor(framebuffer, color_mask, regs.clear_color, dst_region); } } @@ -628,6 +666,7 @@ void RasterizerVulkan::TickFrame() { } bool RasterizerVulkan::AccelerateConditionalRendering() { + gpu_memory->FlushCaching(); if (Settings::IsGPULevelHigh()) { // TODO(Blinkhawk): Reimplement Host conditional rendering. return false; diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h index 472cc64d9..a0508b57c 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.h +++ b/src/video_core/renderer_vulkan/vk_rasterizer.h @@ -66,6 +66,7 @@ public: void Draw(bool is_indexed, u32 instance_count) override; void DrawIndirect() override; + void DrawTexture() override; void Clear(u32 layer_count) override; void DispatchCompute() override; void ResetCounter(VideoCore::QueryType type) override; diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp index c2e53a5d5..e03685af1 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.cpp +++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp @@ -213,6 +213,11 @@ void Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_s .signalSemaphoreCount = num_signal_semaphores, .pSignalSemaphores = signal_semaphores.data(), }; + + if (on_submit) { + on_submit(); + } + switch (const VkResult result = device.GetGraphicsQueue().Submit(submit_info)) { case VK_SUCCESS: break; diff --git a/src/video_core/renderer_vulkan/vk_scheduler.h b/src/video_core/renderer_vulkan/vk_scheduler.h index 3858c506c..bd4cb0f7e 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.h +++ b/src/video_core/renderer_vulkan/vk_scheduler.h @@ -5,6 +5,7 @@ #include <condition_variable> #include <cstddef> +#include <functional> #include <memory> #include <thread> #include <utility> @@ -66,6 +67,11 @@ public: query_cache = &query_cache_; } + // Registers a callback to perform on queue submission. + void RegisterOnSubmit(std::function<void()>&& func) { + on_submit = std::move(func); + } + /// Send work to a separate thread. template <typename T> void Record(T&& command) { @@ -216,6 +222,7 @@ private: vk::CommandBuffer current_cmdbuf; std::unique_ptr<CommandChunk> chunk; + std::function<void()> on_submit; State state; diff --git a/src/video_core/renderer_vulkan/vk_smaa.cpp b/src/video_core/renderer_vulkan/vk_smaa.cpp index 8eb735489..f8735189d 100644 --- a/src/video_core/renderer_vulkan/vk_smaa.cpp +++ b/src/video_core/renderer_vulkan/vk_smaa.cpp @@ -468,7 +468,7 @@ VkWriteDescriptorSet CreateWriteDescriptorSet(std::vector<VkDescriptorImageInfo> } void ClearColorImage(vk::CommandBuffer& cmdbuf, VkImage image) { - constexpr std::array<VkImageSubresourceRange, 1> subresources{{{ + static constexpr std::array<VkImageSubresourceRange, 1> subresources{{{ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, .baseMipLevel = 0, .levelCount = 1, @@ -528,8 +528,8 @@ SMAA::SMAA(const Device& device, MemoryAllocator& allocator, size_t image_count, } void SMAA::CreateImages() { - constexpr VkExtent2D area_extent{AREATEX_WIDTH, AREATEX_HEIGHT}; - constexpr VkExtent2D search_extent{SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT}; + static constexpr VkExtent2D area_extent{AREATEX_WIDTH, AREATEX_HEIGHT}; + static constexpr VkExtent2D search_extent{SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT}; std::tie(m_static_images[Area], m_static_buffer_commits[Area]) = CreateWrappedImage(m_device, m_allocator, area_extent, VK_FORMAT_R8G8_UNORM); @@ -586,12 +586,12 @@ void SMAA::CreateSampler() { void SMAA::CreateShaders() { // These match the order of the SMAAStage enum - constexpr std::array vert_shader_sources{ + static constexpr std::array vert_shader_sources{ ARRAY_TO_SPAN(SMAA_EDGE_DETECTION_VERT_SPV), ARRAY_TO_SPAN(SMAA_BLENDING_WEIGHT_CALCULATION_VERT_SPV), ARRAY_TO_SPAN(SMAA_NEIGHBORHOOD_BLENDING_VERT_SPV), }; - constexpr std::array frag_shader_sources{ + static constexpr std::array frag_shader_sources{ ARRAY_TO_SPAN(SMAA_EDGE_DETECTION_FRAG_SPV), ARRAY_TO_SPAN(SMAA_BLENDING_WEIGHT_CALCULATION_FRAG_SPV), ARRAY_TO_SPAN(SMAA_NEIGHBORHOOD_BLENDING_FRAG_SPV), @@ -675,8 +675,8 @@ void SMAA::UploadImages(Scheduler& scheduler) { return; } - constexpr VkExtent2D area_extent{AREATEX_WIDTH, AREATEX_HEIGHT}; - constexpr VkExtent2D search_extent{SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT}; + static constexpr VkExtent2D area_extent{AREATEX_WIDTH, AREATEX_HEIGHT}; + static constexpr VkExtent2D search_extent{SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT}; UploadImage(m_device, m_allocator, scheduler, m_static_images[Area], area_extent, VK_FORMAT_R8G8_UNORM, ARRAY_TO_SPAN(areaTexBytes)); diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index d39372ec4..9b85dfb5e 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -1230,6 +1230,11 @@ void TextureCacheRuntime::CopyImage(Image& dst, Image& src, }); } +void TextureCacheRuntime::CopyImageMSAA(Image& dst, Image& src, + std::span<const VideoCommon::ImageCopy> copies) { + UNIMPLEMENTED_MSG("Copying images with different samples is not implemented in Vulkan."); +} + u64 TextureCacheRuntime::GetDeviceLocalMemory() const { return device.GetDeviceLocalMemory(); } diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h index 1f27a3589..0ce39616f 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.h +++ b/src/video_core/renderer_vulkan/vk_texture_cache.h @@ -70,6 +70,8 @@ public: void CopyImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies); + void CopyImageMSAA(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies); + bool ShouldReinterpret(Image& dst, Image& src); void ReinterpretImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies); @@ -80,6 +82,11 @@ public: return false; } + bool CanUploadMSAA() const noexcept { + // TODO: Implement buffer to MSAA uploads + return false; + } + void AccelerateImageUpload(Image&, const StagingBufferRef&, std::span<const VideoCommon::SwizzleParameters>); @@ -106,7 +113,7 @@ public: std::optional<ASTCDecoderPass> astc_decoder_pass; const Settings::ResolutionScalingInfo& resolution; - constexpr static size_t indexing_slots = 8 * sizeof(size_t); + static constexpr size_t indexing_slots = 8 * sizeof(size_t); std::array<vk::Buffer, indexing_slots> buffers{}; std::array<std::unique_ptr<MemoryCommit>, indexing_slots> buffer_commits{}; }; diff --git a/src/video_core/renderer_vulkan/vk_turbo_mode.cpp b/src/video_core/renderer_vulkan/vk_turbo_mode.cpp index 852b86f84..db04943eb 100644 --- a/src/video_core/renderer_vulkan/vk_turbo_mode.cpp +++ b/src/video_core/renderer_vulkan/vk_turbo_mode.cpp @@ -14,11 +14,21 @@ using namespace Common::Literals; TurboMode::TurboMode(const vk::Instance& instance, const vk::InstanceDispatch& dld) : m_device{CreateDevice(instance, dld, VK_NULL_HANDLE)}, m_allocator{m_device, false} { + { + std::scoped_lock lk{m_submission_lock}; + m_submission_time = std::chrono::steady_clock::now(); + } m_thread = std::jthread([&](auto stop_token) { Run(stop_token); }); } TurboMode::~TurboMode() = default; +void TurboMode::QueueSubmitted() { + std::scoped_lock lk{m_submission_lock}; + m_submission_time = std::chrono::steady_clock::now(); + m_submission_cv.notify_one(); +} + void TurboMode::Run(std::stop_token stop_token) { auto& dld = m_device.GetLogical(); @@ -38,7 +48,7 @@ void TurboMode::Run(std::stop_token stop_token) { auto commit = m_allocator.Commit(buffer, MemoryUsage::DeviceLocal); // Create the descriptor pool to contain our descriptor. - constexpr VkDescriptorPoolSize pool_size{ + static constexpr VkDescriptorPoolSize pool_size{ .type = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, .descriptorCount = 1, }; @@ -53,7 +63,7 @@ void TurboMode::Run(std::stop_token stop_token) { }); // Create the descriptor set layout from the pool. - constexpr VkDescriptorSetLayoutBinding layout_binding{ + static constexpr VkDescriptorSetLayoutBinding layout_binding{ .binding = 0, .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, .descriptorCount = 1, @@ -199,6 +209,13 @@ void TurboMode::Run(std::stop_token stop_token) { // Wait for completion. fence.Wait(); + + // Wait for the next graphics queue submission if necessary. + std::unique_lock lk{m_submission_lock}; + Common::CondvarWait(m_submission_cv, lk, stop_token, [this] { + return (std::chrono::steady_clock::now() - m_submission_time) <= + std::chrono::milliseconds{100}; + }); } } diff --git a/src/video_core/renderer_vulkan/vk_turbo_mode.h b/src/video_core/renderer_vulkan/vk_turbo_mode.h index 2060e2395..99b5ac50b 100644 --- a/src/video_core/renderer_vulkan/vk_turbo_mode.h +++ b/src/video_core/renderer_vulkan/vk_turbo_mode.h @@ -3,6 +3,9 @@ #pragma once +#include <chrono> +#include <mutex> + #include "common/polyfill_thread.h" #include "video_core/vulkan_common/vulkan_device.h" #include "video_core/vulkan_common/vulkan_memory_allocator.h" @@ -15,11 +18,17 @@ public: explicit TurboMode(const vk::Instance& instance, const vk::InstanceDispatch& dld); ~TurboMode(); + void QueueSubmitted(); + private: void Run(std::stop_token stop_token); Device m_device; MemoryAllocator m_allocator; + std::mutex m_submission_lock; + std::condition_variable_any m_submission_cv; + std::chrono::time_point<std::chrono::steady_clock> m_submission_time{}; + std::jthread m_thread; }; diff --git a/src/video_core/texture_cache/descriptor_table.h b/src/video_core/texture_cache/descriptor_table.h index ee4240288..1bad83fb4 100644 --- a/src/video_core/texture_cache/descriptor_table.h +++ b/src/video_core/texture_cache/descriptor_table.h @@ -19,9 +19,7 @@ public: explicit DescriptorTable(Tegra::MemoryManager& gpu_memory_) : gpu_memory{gpu_memory_} {} [[nodiscard]] bool Synchronize(GPUVAddr gpu_addr, u32 limit) { - [[likely]] if (current_gpu_addr == gpu_addr && current_limit == limit) { - return false; - } + [[likely]] if (current_gpu_addr == gpu_addr && current_limit == limit) { return false; } Refresh(gpu_addr, limit); return true; } diff --git a/src/video_core/texture_cache/formatter.cpp b/src/video_core/texture_cache/formatter.cpp index 418890126..30f72361d 100644 --- a/src/video_core/texture_cache/formatter.cpp +++ b/src/video_core/texture_cache/formatter.cpp @@ -22,6 +22,9 @@ std::string Name(const ImageBase& image) { const u32 num_layers = image.info.resources.layers; const u32 num_levels = image.info.resources.levels; std::string resource; + if (image.info.num_samples > 1) { + resource += fmt::format(":{}xMSAA", image.info.num_samples); + } if (num_layers > 1) { resource += fmt::format(":L{}", num_layers); } diff --git a/src/video_core/texture_cache/image_info.cpp b/src/video_core/texture_cache/image_info.cpp index 852ec2519..e9100091e 100644 --- a/src/video_core/texture_cache/image_info.cpp +++ b/src/video_core/texture_cache/image_info.cpp @@ -100,6 +100,10 @@ ImageInfo::ImageInfo(const TICEntry& config) noexcept { ASSERT_MSG(false, "Invalid texture_type={}", static_cast<int>(config.texture_type.Value())); break; } + if (num_samples > 1) { + size.width *= NumSamplesX(config.msaa_mode); + size.height *= NumSamplesY(config.msaa_mode); + } if (type != ImageType::Linear) { // FIXME: Call this without passing *this layer_stride = CalculateLayerStride(*this); diff --git a/src/video_core/texture_cache/samples_helper.h b/src/video_core/texture_cache/samples_helper.h index d552bccf0..203ac1b11 100644 --- a/src/video_core/texture_cache/samples_helper.h +++ b/src/video_core/texture_cache/samples_helper.h @@ -51,4 +51,48 @@ namespace VideoCommon { return 1; } +[[nodiscard]] inline int NumSamplesX(Tegra::Texture::MsaaMode msaa_mode) { + using Tegra::Texture::MsaaMode; + switch (msaa_mode) { + case MsaaMode::Msaa1x1: + return 1; + case MsaaMode::Msaa2x1: + case MsaaMode::Msaa2x1_D3D: + case MsaaMode::Msaa2x2: + case MsaaMode::Msaa2x2_VC4: + case MsaaMode::Msaa2x2_VC12: + return 2; + case MsaaMode::Msaa4x2: + case MsaaMode::Msaa4x2_D3D: + case MsaaMode::Msaa4x2_VC8: + case MsaaMode::Msaa4x2_VC24: + case MsaaMode::Msaa4x4: + return 4; + } + ASSERT_MSG(false, "Invalid MSAA mode={}", static_cast<int>(msaa_mode)); + return 1; +} + +[[nodiscard]] inline int NumSamplesY(Tegra::Texture::MsaaMode msaa_mode) { + using Tegra::Texture::MsaaMode; + switch (msaa_mode) { + case MsaaMode::Msaa1x1: + case MsaaMode::Msaa2x1: + case MsaaMode::Msaa2x1_D3D: + return 1; + case MsaaMode::Msaa2x2: + case MsaaMode::Msaa2x2_VC4: + case MsaaMode::Msaa2x2_VC12: + case MsaaMode::Msaa4x2: + case MsaaMode::Msaa4x2_D3D: + case MsaaMode::Msaa4x2_VC8: + case MsaaMode::Msaa4x2_VC24: + return 2; + case MsaaMode::Msaa4x4: + return 4; + } + ASSERT_MSG(false, "Invalid MSAA mode={}", static_cast<int>(msaa_mode)); + return 1; +} + } // namespace VideoCommon diff --git a/src/video_core/texture_cache/slot_vector.h b/src/video_core/texture_cache/slot_vector.h index 1e2aad76a..9df6a2903 100644 --- a/src/video_core/texture_cache/slot_vector.h +++ b/src/video_core/texture_cache/slot_vector.h @@ -29,7 +29,7 @@ struct SlotId { }; template <class T> -requires std::is_nothrow_move_assignable_v<T> && std::is_nothrow_move_constructible_v<T> + requires std::is_nothrow_move_assignable_v<T> && std::is_nothrow_move_constructible_v<T> class SlotVector { public: class Iterator { diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 87152c8e9..3e2cbb0b0 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -149,6 +149,13 @@ typename P::ImageView& TextureCache<P>::GetImageView(ImageViewId id) noexcept { } template <class P> +typename P::ImageView& TextureCache<P>::GetImageView(u32 index) noexcept { + const auto image_view_id = VisitImageView(channel_state->graphics_image_table, + channel_state->graphics_image_view_ids, index); + return slot_image_views[image_view_id]; +} + +template <class P> void TextureCache<P>::MarkModification(ImageId id) noexcept { MarkModification(slot_images[id]); } @@ -766,7 +773,7 @@ void TextureCache<P>::RefreshContents(Image& image, ImageId image_id) { image.flags &= ~ImageFlagBits::CpuModified; TrackImage(image, image_id); - if (image.info.num_samples > 1) { + if (image.info.num_samples > 1 && !runtime.CanUploadMSAA()) { LOG_WARNING(HW_GPU, "MSAA image uploads are not implemented"); return; } @@ -1160,14 +1167,14 @@ ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VA if (True(overlap.flags & ImageFlagBits::GpuModified)) { new_image.flags |= ImageFlagBits::GpuModified; } + const auto& resolution = Settings::values.resolution_info; + const SubresourceBase base = new_image.TryFindBase(overlap.gpu_addr).value(); + const u32 up_scale = can_rescale ? resolution.up_scale : 1; + const u32 down_shift = can_rescale ? resolution.down_shift : 0; + auto copies = MakeShrinkImageCopies(new_info, overlap.info, base, up_scale, down_shift); if (overlap.info.num_samples != new_image.info.num_samples) { - LOG_WARNING(HW_GPU, "Copying between images with different samples is not implemented"); + runtime.CopyImageMSAA(new_image, overlap, std::move(copies)); } else { - const auto& resolution = Settings::values.resolution_info; - const SubresourceBase base = new_image.TryFindBase(overlap.gpu_addr).value(); - const u32 up_scale = can_rescale ? resolution.up_scale : 1; - const u32 down_shift = can_rescale ? resolution.down_shift : 0; - auto copies = MakeShrinkImageCopies(new_info, overlap.info, base, up_scale, down_shift); runtime.CopyImage(new_image, overlap, std::move(copies)); } if (True(overlap.flags & ImageFlagBits::Tracked)) { diff --git a/src/video_core/texture_cache/texture_cache_base.h b/src/video_core/texture_cache/texture_cache_base.h index 4eea1f609..485eaabaa 100644 --- a/src/video_core/texture_cache/texture_cache_base.h +++ b/src/video_core/texture_cache/texture_cache_base.h @@ -129,6 +129,9 @@ public: /// Return a reference to the given image view id [[nodiscard]] ImageView& GetImageView(ImageViewId id) noexcept; + /// Get the imageview from the graphics descriptor table in the specified index + [[nodiscard]] ImageView& GetImageView(u32 index) noexcept; + /// Mark an image as modified from the GPU void MarkModification(ImageId id) noexcept; diff --git a/src/video_core/texture_cache/util.cpp b/src/video_core/texture_cache/util.cpp index 03acc68d9..697f86641 100644 --- a/src/video_core/texture_cache/util.cpp +++ b/src/video_core/texture_cache/util.cpp @@ -573,10 +573,6 @@ u32 CalculateUnswizzledSizeBytes(const ImageInfo& info) noexcept { if (info.type == ImageType::Buffer) { return info.size.width * BytesPerBlock(info.format); } - if (info.num_samples > 1) { - // Multisample images can't be uploaded or downloaded to the host - return 0; - } if (info.type == ImageType::Linear) { return info.pitch * Common::DivCeil(info.size.height, DefaultBlockHeight(info.format)); } @@ -703,7 +699,6 @@ ImageViewType RenderTargetImageViewType(const ImageInfo& info) noexcept { std::vector<ImageCopy> MakeShrinkImageCopies(const ImageInfo& dst, const ImageInfo& src, SubresourceBase base, u32 up_scale, u32 down_shift) { ASSERT(dst.resources.levels >= src.resources.levels); - ASSERT(dst.num_samples == src.num_samples); const bool is_dst_3d = dst.type == ImageType::e3D; if (is_dst_3d) { diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 59120cd09..95bcdd37b 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -29,7 +29,7 @@ constexpr u32 pdep(u32 value) { template <u32 mask, u32 incr_amount> void incrpdep(u32& value) { - constexpr u32 swizzled_incr = pdep<mask>(incr_amount); + static constexpr u32 swizzled_incr = pdep<mask>(incr_amount); value = ((value | ~mask) + swizzled_incr) & mask; } diff --git a/src/video_core/vulkan_common/nsight_aftermath_tracker.cpp b/src/video_core/vulkan_common/nsight_aftermath_tracker.cpp index 85f1d13e0..5fa0d9620 100644 --- a/src/video_core/vulkan_common/nsight_aftermath_tracker.cpp +++ b/src/video_core/vulkan_common/nsight_aftermath_tracker.cpp @@ -57,7 +57,7 @@ NsightAftermathTracker::NsightAftermathTracker() { if (!GFSDK_Aftermath_SUCCEED(GFSDK_Aftermath_EnableGpuCrashDumps( GFSDK_Aftermath_Version_API, GFSDK_Aftermath_GpuCrashDumpWatchedApiFlags_Vulkan, GFSDK_Aftermath_GpuCrashDumpFeatureFlags_Default, GpuCrashDumpCallback, - ShaderDebugInfoCallback, CrashDumpDescriptionCallback, this))) { + ShaderDebugInfoCallback, CrashDumpDescriptionCallback, nullptr, this))) { LOG_ERROR(Render_Vulkan, "GFSDK_Aftermath_EnableGpuCrashDumps failed"); return; } @@ -83,7 +83,7 @@ void NsightAftermathTracker::SaveShader(std::span<const u32> spirv) const { std::scoped_lock lock{mutex}; - GFSDK_Aftermath_ShaderHash hash; + GFSDK_Aftermath_ShaderBinaryHash hash; if (!GFSDK_Aftermath_SUCCEED( GFSDK_Aftermath_GetShaderHashSpirv(GFSDK_Aftermath_Version_API, &shader, &hash))) { LOG_ERROR(Render_Vulkan, "Failed to hash SPIR-V module"); @@ -121,8 +121,8 @@ void NsightAftermathTracker::OnGpuCrashDumpCallback(const void* gpu_crash_dump, u32 json_size = 0; if (!GFSDK_Aftermath_SUCCEED(GFSDK_Aftermath_GpuCrashDump_GenerateJSON( decoder, GFSDK_Aftermath_GpuCrashDumpDecoderFlags_ALL_INFO, - GFSDK_Aftermath_GpuCrashDumpFormatterFlags_NONE, nullptr, nullptr, nullptr, nullptr, - this, &json_size))) { + GFSDK_Aftermath_GpuCrashDumpFormatterFlags_NONE, nullptr, nullptr, nullptr, this, + &json_size))) { LOG_ERROR(Render_Vulkan, "Failed to generate JSON"); return; } diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp index fd1c5a683..23d922e5d 100644 --- a/src/video_core/vulkan_common/vulkan_device.cpp +++ b/src/video_core/vulkan_common/vulkan_device.cpp @@ -74,30 +74,6 @@ enum class NvidiaArchitecture { VoltaOrOlder, }; -constexpr std::array REQUIRED_EXTENSIONS{ - VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME, - VK_EXT_ROBUSTNESS_2_EXTENSION_NAME, -#ifdef _WIN32 - VK_KHR_EXTERNAL_MEMORY_WIN32_EXTENSION_NAME, -#endif -#ifdef __unix__ - VK_KHR_EXTERNAL_MEMORY_FD_EXTENSION_NAME, -#endif -}; - -constexpr std::array REQUIRED_EXTENSIONS_BEFORE_1_2{ - VK_KHR_TIMELINE_SEMAPHORE_EXTENSION_NAME, - VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME, - VK_KHR_8BIT_STORAGE_EXTENSION_NAME, - VK_KHR_SHADER_FLOAT_CONTROLS_EXTENSION_NAME, - VK_KHR_SAMPLER_MIRROR_CLAMP_TO_EDGE_EXTENSION_NAME, - VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME, -}; - -constexpr std::array REQUIRED_EXTENSIONS_BEFORE_1_3{ - VK_EXT_SHADER_DEMOTE_TO_HELPER_INVOCATION_EXTENSION_NAME, -}; - template <typename T> void SetNext(void**& next, T& data) { *next = &data; @@ -286,24 +262,9 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(vk::Physica return format_properties; } -std::vector<std::string> GetSupportedExtensions(vk::PhysicalDevice physical) { - const std::vector extensions = physical.EnumerateDeviceExtensionProperties(); - std::vector<std::string> supported_extensions; - supported_extensions.reserve(extensions.size()); - for (const auto& extension : extensions) { - supported_extensions.emplace_back(extension.extensionName); - } - return supported_extensions; -} - -bool IsExtensionSupported(std::span<const std::string> supported_extensions, - std::string_view extension) { - return std::ranges::find(supported_extensions, extension) != supported_extensions.end(); -} - NvidiaArchitecture GetNvidiaArchitecture(vk::PhysicalDevice physical, - std::span<const std::string> exts) { - if (IsExtensionSupported(exts, VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME)) { + const std::set<std::string, std::less<>>& exts) { + if (exts.contains(VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME)) { VkPhysicalDeviceFragmentShadingRatePropertiesKHR shading_rate_props{}; shading_rate_props.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FRAGMENT_SHADING_RATE_PROPERTIES_KHR; @@ -316,423 +277,55 @@ NvidiaArchitecture GetNvidiaArchitecture(vk::PhysicalDevice physical, return NvidiaArchitecture::AmpereOrNewer; } } - if (IsExtensionSupported(exts, VK_NV_SHADING_RATE_IMAGE_EXTENSION_NAME)) { + if (exts.contains(VK_NV_SHADING_RATE_IMAGE_EXTENSION_NAME)) { return NvidiaArchitecture::Turing; } return NvidiaArchitecture::VoltaOrOlder; } + +std::vector<const char*> ExtensionListForVulkan( + const std::set<std::string, std::less<>>& extensions) { + std::vector<const char*> output; + for (const auto& extension : extensions) { + output.push_back(extension.c_str()); + } + return output; +} + } // Anonymous namespace Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR surface, const vk::InstanceDispatch& dld_) - : instance{instance_}, dld{dld_}, physical{physical_}, properties{physical.GetProperties()}, - instance_version{properties.apiVersion}, supported_extensions{GetSupportedExtensions( - physical)}, + : instance{instance_}, dld{dld_}, physical{physical_}, format_properties(GetFormatProperties(physical)) { - CheckSuitability(surface != nullptr); - SetupFamilies(surface); - SetupFeatures(); - SetupProperties(); - - const auto queue_cis = GetDeviceQueueCreateInfos(); - const std::vector extensions = LoadExtensions(surface != nullptr); - - VkPhysicalDeviceFeatures2 features2{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2, - .pNext = nullptr, - .features{ - .robustBufferAccess = true, - .fullDrawIndexUint32 = false, - .imageCubeArray = true, - .independentBlend = true, - .geometryShader = true, - .tessellationShader = true, - .sampleRateShading = true, - .dualSrcBlend = true, - .logicOp = true, - .multiDrawIndirect = true, - .drawIndirectFirstInstance = true, - .depthClamp = true, - .depthBiasClamp = true, - .fillModeNonSolid = true, - .depthBounds = is_depth_bounds_supported, - .wideLines = true, - .largePoints = true, - .alphaToOne = false, - .multiViewport = true, - .samplerAnisotropy = true, - .textureCompressionETC2 = false, - .textureCompressionASTC_LDR = is_optimal_astc_supported, - .textureCompressionBC = false, - .occlusionQueryPrecise = true, - .pipelineStatisticsQuery = false, - .vertexPipelineStoresAndAtomics = true, - .fragmentStoresAndAtomics = true, - .shaderTessellationAndGeometryPointSize = false, - .shaderImageGatherExtended = true, - .shaderStorageImageExtendedFormats = false, - .shaderStorageImageMultisample = is_shader_storage_image_multisample, - .shaderStorageImageReadWithoutFormat = is_formatless_image_load_supported, - .shaderStorageImageWriteWithoutFormat = true, - .shaderUniformBufferArrayDynamicIndexing = false, - .shaderSampledImageArrayDynamicIndexing = false, - .shaderStorageBufferArrayDynamicIndexing = false, - .shaderStorageImageArrayDynamicIndexing = false, - .shaderClipDistance = true, - .shaderCullDistance = true, - .shaderFloat64 = is_shader_float64_supported, - .shaderInt64 = is_shader_int64_supported, - .shaderInt16 = is_shader_int16_supported, - .shaderResourceResidency = false, - .shaderResourceMinLod = false, - .sparseBinding = false, - .sparseResidencyBuffer = false, - .sparseResidencyImage2D = false, - .sparseResidencyImage3D = false, - .sparseResidency2Samples = false, - .sparseResidency4Samples = false, - .sparseResidency8Samples = false, - .sparseResidency16Samples = false, - .sparseResidencyAliased = false, - .variableMultisampleRate = false, - .inheritedQueries = false, - }, - }; - const void* first_next = &features2; - void** next = &features2.pNext; - - VkPhysicalDeviceTimelineSemaphoreFeatures timeline_semaphore{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TIMELINE_SEMAPHORE_FEATURES, - .pNext = nullptr, - .timelineSemaphore = true, - }; - SetNext(next, timeline_semaphore); - - VkPhysicalDevice16BitStorageFeatures bit16_storage{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES, - .pNext = nullptr, - .storageBuffer16BitAccess = true, - .uniformAndStorageBuffer16BitAccess = true, - .storagePushConstant16 = false, - .storageInputOutput16 = false, - }; - SetNext(next, bit16_storage); - - VkPhysicalDevice8BitStorageFeatures bit8_storage{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_8BIT_STORAGE_FEATURES, - .pNext = nullptr, - .storageBuffer8BitAccess = true, - .uniformAndStorageBuffer8BitAccess = true, - .storagePushConstant8 = false, - }; - SetNext(next, bit8_storage); - - VkPhysicalDeviceRobustness2FeaturesEXT robustness2{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ROBUSTNESS_2_FEATURES_EXT, - .pNext = nullptr, - .robustBufferAccess2 = true, - .robustImageAccess2 = true, - .nullDescriptor = true, - }; - SetNext(next, robustness2); - - VkPhysicalDeviceHostQueryResetFeatures host_query_reset{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_HOST_QUERY_RESET_FEATURES, - .pNext = nullptr, - .hostQueryReset = true, - }; - SetNext(next, host_query_reset); - - VkPhysicalDeviceVariablePointerFeatures variable_pointers{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VARIABLE_POINTERS_FEATURES, - .pNext = nullptr, - .variablePointersStorageBuffer = VK_TRUE, - .variablePointers = VK_TRUE, - }; - SetNext(next, variable_pointers); - - VkPhysicalDeviceShaderDemoteToHelperInvocationFeatures demote{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DEMOTE_TO_HELPER_INVOCATION_FEATURES, - .pNext = nullptr, - .shaderDemoteToHelperInvocation = true, - }; - SetNext(next, demote); - - VkPhysicalDeviceShaderDrawParametersFeatures draw_parameters{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DRAW_PARAMETER_FEATURES, - .pNext = nullptr, - .shaderDrawParameters = true, - }; - SetNext(next, draw_parameters); - - VkPhysicalDeviceShaderFloat16Int8Features float16_int8; - if (is_int8_supported || is_float16_supported) { - float16_int8 = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_FLOAT16_INT8_FEATURES, - .pNext = nullptr, - .shaderFloat16 = is_float16_supported, - .shaderInt8 = is_int8_supported, - }; - SetNext(next, float16_int8); - } - if (!is_float16_supported) { - LOG_INFO(Render_Vulkan, "Device doesn't support float16 natively"); - } - if (!is_int8_supported) { - LOG_INFO(Render_Vulkan, "Device doesn't support int8 natively"); - } - - if (!nv_viewport_swizzle) { - LOG_INFO(Render_Vulkan, "Device doesn't support viewport swizzles"); - } - - if (!nv_viewport_array2) { - LOG_INFO(Render_Vulkan, "Device doesn't support viewport masks"); - } - - if (!nv_geometry_shader_passthrough) { - LOG_INFO(Render_Vulkan, "Device doesn't support passthrough geometry shaders"); - } + // Get suitability and device properties. + const bool is_suitable = GetSuitability(surface != nullptr); - VkPhysicalDeviceUniformBufferStandardLayoutFeatures std430_layout; - if (khr_uniform_buffer_standard_layout) { - std430_layout = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_UNIFORM_BUFFER_STANDARD_LAYOUT_FEATURES, - .pNext = nullptr, - .uniformBufferStandardLayout = true, - }; - SetNext(next, std430_layout); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support packed UBOs"); - } - - VkPhysicalDeviceIndexTypeUint8FeaturesEXT index_type_uint8; - if (ext_index_type_uint8) { - index_type_uint8 = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_INDEX_TYPE_UINT8_FEATURES_EXT, - .pNext = nullptr, - .indexTypeUint8 = true, - }; - SetNext(next, index_type_uint8); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support uint8 indexes"); - } - - VkPhysicalDevicePrimitiveTopologyListRestartFeaturesEXT primitive_topology_list_restart; - if (is_topology_list_restart_supported || is_patch_list_restart_supported) { - primitive_topology_list_restart = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PRIMITIVE_TOPOLOGY_LIST_RESTART_FEATURES_EXT, - .pNext = nullptr, - .primitiveTopologyListRestart = is_topology_list_restart_supported, - .primitiveTopologyPatchListRestart = is_patch_list_restart_supported, - }; - SetNext(next, primitive_topology_list_restart); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support list topology primitive restart"); - } - - VkPhysicalDeviceTransformFeedbackFeaturesEXT transform_feedback; - if (ext_transform_feedback) { - transform_feedback = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT, - .pNext = nullptr, - .transformFeedback = true, - .geometryStreams = true, - }; - SetNext(next, transform_feedback); - } else { - 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, - .pNext = nullptr, - .customBorderColors = VK_TRUE, - .customBorderColorWithoutFormat = VK_TRUE, - }; - SetNext(next, custom_border); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support custom border colors"); - } - - VkPhysicalDeviceExtendedDynamicStateFeaturesEXT dynamic_state; - if (ext_extended_dynamic_state) { - dynamic_state = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_FEATURES_EXT, - .pNext = nullptr, - .extendedDynamicState = VK_TRUE, - }; - SetNext(next, dynamic_state); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state"); - } - - VkPhysicalDeviceExtendedDynamicState2FeaturesEXT dynamic_state_2; - if (ext_extended_dynamic_state_2) { - dynamic_state_2 = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_2_FEATURES_EXT, - .pNext = nullptr, - .extendedDynamicState2 = VK_TRUE, - .extendedDynamicState2LogicOp = ext_extended_dynamic_state_2_extra ? VK_TRUE : VK_FALSE, - .extendedDynamicState2PatchControlPoints = VK_FALSE, - }; - SetNext(next, dynamic_state_2); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state 2"); - } - - VkPhysicalDeviceExtendedDynamicState3FeaturesEXT dynamic_state_3; - if (ext_extended_dynamic_state_3) { - dynamic_state_3 = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_3_FEATURES_EXT, - .pNext = nullptr, - .extendedDynamicState3TessellationDomainOrigin = VK_FALSE, - .extendedDynamicState3DepthClampEnable = - ext_extended_dynamic_state_3_enables ? VK_TRUE : VK_FALSE, - .extendedDynamicState3PolygonMode = VK_FALSE, - .extendedDynamicState3RasterizationSamples = VK_FALSE, - .extendedDynamicState3SampleMask = VK_FALSE, - .extendedDynamicState3AlphaToCoverageEnable = VK_FALSE, - .extendedDynamicState3AlphaToOneEnable = VK_FALSE, - .extendedDynamicState3LogicOpEnable = - ext_extended_dynamic_state_3_enables ? VK_TRUE : VK_FALSE, - .extendedDynamicState3ColorBlendEnable = - ext_extended_dynamic_state_3_blend ? VK_TRUE : VK_FALSE, - .extendedDynamicState3ColorBlendEquation = - ext_extended_dynamic_state_3_blend ? VK_TRUE : VK_FALSE, - .extendedDynamicState3ColorWriteMask = - ext_extended_dynamic_state_3_blend ? VK_TRUE : VK_FALSE, - .extendedDynamicState3RasterizationStream = VK_FALSE, - .extendedDynamicState3ConservativeRasterizationMode = VK_FALSE, - .extendedDynamicState3ExtraPrimitiveOverestimationSize = VK_FALSE, - .extendedDynamicState3DepthClipEnable = VK_FALSE, - .extendedDynamicState3SampleLocationsEnable = VK_FALSE, - .extendedDynamicState3ColorBlendAdvanced = VK_FALSE, - .extendedDynamicState3ProvokingVertexMode = VK_FALSE, - .extendedDynamicState3LineRasterizationMode = VK_FALSE, - .extendedDynamicState3LineStippleEnable = VK_FALSE, - .extendedDynamicState3DepthClipNegativeOneToOne = VK_FALSE, - .extendedDynamicState3ViewportWScalingEnable = VK_FALSE, - .extendedDynamicState3ViewportSwizzle = VK_FALSE, - .extendedDynamicState3CoverageToColorEnable = VK_FALSE, - .extendedDynamicState3CoverageToColorLocation = VK_FALSE, - .extendedDynamicState3CoverageModulationMode = VK_FALSE, - .extendedDynamicState3CoverageModulationTableEnable = VK_FALSE, - .extendedDynamicState3CoverageModulationTable = VK_FALSE, - .extendedDynamicState3CoverageReductionMode = VK_FALSE, - .extendedDynamicState3RepresentativeFragmentTestEnable = VK_FALSE, - .extendedDynamicState3ShadingRateImageEnable = VK_FALSE, - }; - SetNext(next, dynamic_state_3); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state 3"); - } - - VkPhysicalDeviceLineRasterizationFeaturesEXT line_raster; - if (ext_line_rasterization) { - line_raster = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_LINE_RASTERIZATION_FEATURES_EXT, - .pNext = nullptr, - .rectangularLines = VK_TRUE, - .bresenhamLines = VK_FALSE, - .smoothLines = VK_TRUE, - .stippledRectangularLines = VK_FALSE, - .stippledBresenhamLines = VK_FALSE, - .stippledSmoothLines = VK_FALSE, - }; - SetNext(next, line_raster); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support smooth lines"); - } - - if (!ext_conservative_rasterization) { - LOG_INFO(Render_Vulkan, "Device doesn't support conservative rasterization"); - } - - VkPhysicalDeviceProvokingVertexFeaturesEXT provoking_vertex; - if (ext_provoking_vertex) { - provoking_vertex = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROVOKING_VERTEX_FEATURES_EXT, - .pNext = nullptr, - .provokingVertexLast = VK_TRUE, - .transformFeedbackPreservesProvokingVertex = VK_TRUE, - }; - SetNext(next, provoking_vertex); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support provoking vertex last"); - } - - VkPhysicalDeviceVertexInputDynamicStateFeaturesEXT vertex_input_dynamic; - if (ext_vertex_input_dynamic_state) { - vertex_input_dynamic = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_INPUT_DYNAMIC_STATE_FEATURES_EXT, - .pNext = nullptr, - .vertexInputDynamicState = VK_TRUE, - }; - SetNext(next, vertex_input_dynamic); - } else { - LOG_INFO(Render_Vulkan, "Device doesn't support vertex input dynamic state"); - } - - VkPhysicalDeviceShaderAtomicInt64Features atomic_int64; - if (ext_shader_atomic_int64) { - atomic_int64 = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_ATOMIC_INT64_FEATURES, - .pNext = nullptr, - .shaderBufferInt64Atomics = VK_TRUE, - .shaderSharedInt64Atomics = VK_TRUE, - }; - SetNext(next, atomic_int64); - } + const VkDriverId driver_id = properties.driver.driverID; + const bool is_radv = driver_id == VK_DRIVER_ID_MESA_RADV; + const bool is_amd_driver = + driver_id == VK_DRIVER_ID_AMD_PROPRIETARY || driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE; + const bool is_amd = is_amd_driver || is_radv; + const bool is_intel_windows = driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS; + const bool is_intel_anv = driver_id == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA; + const bool is_nvidia = driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY; + const bool is_mvk = driver_id == VK_DRIVER_ID_MOLTENVK; - VkPhysicalDeviceWorkgroupMemoryExplicitLayoutFeaturesKHR workgroup_layout; - if (khr_workgroup_memory_explicit_layout && is_shader_int16_supported) { - workgroup_layout = { - .sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_WORKGROUP_MEMORY_EXPLICIT_LAYOUT_FEATURES_KHR, - .pNext = nullptr, - .workgroupMemoryExplicitLayout = VK_TRUE, - .workgroupMemoryExplicitLayoutScalarBlockLayout = VK_TRUE, - .workgroupMemoryExplicitLayout8BitAccess = VK_TRUE, - .workgroupMemoryExplicitLayout16BitAccess = VK_TRUE, - }; - SetNext(next, workgroup_layout); - } else if (khr_workgroup_memory_explicit_layout) { - // TODO(lat9nq): Find a proper fix for this - LOG_WARNING(Render_Vulkan, "Disabling VK_KHR_workgroup_memory_explicit_layout due to a " - "yuzu bug when host driver does not support 16-bit integers"); - khr_workgroup_memory_explicit_layout = false; + if (is_mvk && !is_suitable) { + LOG_WARNING(Render_Vulkan, "Unsuitable driver is MoltenVK, continuing anyway"); + } else if (!is_suitable) { + throw vk::Exception(VK_ERROR_INCOMPATIBLE_DRIVER); } - VkPhysicalDevicePipelineExecutablePropertiesFeaturesKHR executable_properties; - if (khr_pipeline_executable_properties) { - LOG_INFO(Render_Vulkan, "Enabling shader feedback, expect slower shader build times"); - executable_properties = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PIPELINE_EXECUTABLE_PROPERTIES_FEATURES_KHR, - .pNext = nullptr, - .pipelineExecutableInfo = VK_TRUE, - }; - SetNext(next, executable_properties); - } - - if (!ext_depth_range_unrestricted) { - LOG_INFO(Render_Vulkan, "Device doesn't support depth range unrestricted"); - } + SetupFamilies(surface); + const auto queue_cis = GetDeviceQueueCreateInfos(); - VkPhysicalDeviceDepthClipControlFeaturesEXT depth_clip_control_features; - if (ext_depth_clip_control) { - depth_clip_control_features = { - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DEPTH_CLIP_CONTROL_FEATURES_EXT, - .pNext = nullptr, - .depthClipControl = VK_TRUE, - }; - SetNext(next, depth_clip_control_features); - } + // GetSuitability has already configured the linked list of features for us. + // Reuse it here. + const void* first_next = &features2; - VkDeviceDiagnosticsConfigCreateInfoNV diagnostics_nv; - if (Settings::values.enable_nsight_aftermath && nv_device_diagnostics_config) { + VkDeviceDiagnosticsConfigCreateInfoNV diagnostics_nv{}; + if (Settings::values.enable_nsight_aftermath && extensions.device_diagnostics_config) { nsight_aftermath_tracker = std::make_unique<NsightAftermathTracker>(); diagnostics_nv = { @@ -744,33 +337,39 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR }; first_next = &diagnostics_nv; } - logical = vk::Device::Create(physical, queue_cis, extensions, first_next, dld); - is_integrated = properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU; - is_virtual = properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_VIRTUAL_GPU; - is_non_gpu = properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_OTHER || - properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_CPU; + is_blit_depth_stencil_supported = TestDepthStencilBlits(); + is_optimal_astc_supported = ComputeIsOptimalAstcSupported(); + is_warp_potentially_bigger = !extensions.subgroup_size_control || + properties.subgroup_size_control.maxSubgroupSize > GuestWarpSize; + + is_integrated = properties.properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU; + is_virtual = properties.properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_VIRTUAL_GPU; + is_non_gpu = properties.properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_OTHER || + properties.properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_CPU; + + supports_d24_depth = + IsFormatSupported(VK_FORMAT_D24_UNORM_S8_UINT, + VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT, FormatType::Optimal); CollectPhysicalMemoryInfo(); - CollectTelemetryParameters(); CollectToolingInfo(); - if (driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY_KHR) { - const u32 nv_major_version = (properties.driverVersion >> 22) & 0x3ff; - + if (is_nvidia) { + const u32 nv_major_version = (properties.properties.driverVersion >> 22) & 0x3ff; const auto arch = GetNvidiaArchitecture(physical, supported_extensions); switch (arch) { case NvidiaArchitecture::AmpereOrNewer: - LOG_WARNING(Render_Vulkan, "Blacklisting Ampere devices from float16 math"); - is_float16_supported = false; + LOG_WARNING(Render_Vulkan, "Ampere and newer have broken float16 math"); + features.shader_float16_int8.shaderFloat16 = false; break; case NvidiaArchitecture::Turing: break; case NvidiaArchitecture::VoltaOrOlder: if (nv_major_version < 527) { - LOG_WARNING(Render_Vulkan, - "Blacklisting Volta and older from VK_KHR_push_descriptor"); - khr_push_descriptor = false; + LOG_WARNING(Render_Vulkan, "Volta and older have broken VK_KHR_push_descriptor"); + extensions.push_descriptor = false; + loaded_extensions.erase(VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME); } break; } @@ -779,75 +378,75 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR cant_blit_msaa = true; } } - const bool is_radv = driver_id == VK_DRIVER_ID_MESA_RADV; - if (ext_extended_dynamic_state && is_radv) { + if (extensions.extended_dynamic_state && is_radv) { // Mask driver version variant - const u32 version = (properties.driverVersion << 3) >> 3; + const u32 version = (properties.properties.driverVersion << 3) >> 3; if (version < VK_MAKE_API_VERSION(0, 21, 2, 0)) { LOG_WARNING(Render_Vulkan, "RADV versions older than 21.2 have broken VK_EXT_extended_dynamic_state"); - ext_extended_dynamic_state = false; + extensions.extended_dynamic_state = false; + loaded_extensions.erase(VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME); } } - if (ext_vertex_input_dynamic_state && is_radv) { + if (extensions.extended_dynamic_state2 && is_radv) { + const u32 version = (properties.properties.driverVersion << 3) >> 3; + if (version < VK_MAKE_API_VERSION(0, 22, 3, 1)) { + LOG_WARNING( + Render_Vulkan, + "RADV versions older than 22.3.1 have broken VK_EXT_extended_dynamic_state2"); + features.extended_dynamic_state2.extendedDynamicState2 = false; + features.extended_dynamic_state2.extendedDynamicState2LogicOp = false; + features.extended_dynamic_state2.extendedDynamicState2PatchControlPoints = false; + extensions.extended_dynamic_state2 = false; + loaded_extensions.erase(VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME); + } + } + if (extensions.vertex_input_dynamic_state && is_radv) { // TODO(ameerj): Blacklist only offending driver versions // TODO(ameerj): Confirm if RDNA1 is affected const bool is_rdna2 = - IsExtensionSupported(supported_extensions, VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME); + supported_extensions.contains(VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME); if (is_rdna2) { LOG_WARNING(Render_Vulkan, "RADV has broken VK_EXT_vertex_input_dynamic_state on RDNA2 hardware"); - ext_vertex_input_dynamic_state = false; + extensions.vertex_input_dynamic_state = false; + loaded_extensions.erase(VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); } } - if (ext_extended_dynamic_state_2 && is_radv) { - const u32 version = (properties.driverVersion << 3) >> 3; - if (version < VK_MAKE_API_VERSION(0, 22, 3, 1)) { - LOG_WARNING( - Render_Vulkan, - "RADV versions older than 22.3.1 have broken VK_EXT_extended_dynamic_state2"); - ext_extended_dynamic_state_2 = false; - ext_extended_dynamic_state_2_extra = false; - } - } - sets_per_pool = 64; - const bool is_amd = - driver_id == VK_DRIVER_ID_AMD_PROPRIETARY || driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE; - if (is_amd) { + sets_per_pool = 64; + if (is_amd_driver) { // AMD drivers need a higher amount of Sets per Pool in certain circunstances like in XC2. sets_per_pool = 96; // Disable VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT on AMD GCN4 and lower as it is broken. - if (!is_float16_supported) { - LOG_WARNING( - Render_Vulkan, - "AMD GCN4 and earlier do not properly support VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT"); + if (!features.shader_float16_int8.shaderFloat16) { + LOG_WARNING(Render_Vulkan, + "AMD GCN4 and earlier have broken VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT"); has_broken_cube_compatibility = true; } } - const bool is_amd_or_radv = is_amd || is_radv; - if (ext_sampler_filter_minmax && is_amd_or_radv) { + if (extensions.sampler_filter_minmax && is_amd) { // Disable ext_sampler_filter_minmax on AMD GCN4 and lower as it is broken. - if (!is_float16_supported) { + if (!features.shader_float16_int8.shaderFloat16) { LOG_WARNING(Render_Vulkan, - "Blacklisting AMD GCN4 and earlier for VK_EXT_sampler_filter_minmax"); - ext_sampler_filter_minmax = false; + "AMD GCN4 and earlier have broken VK_EXT_sampler_filter_minmax"); + extensions.sampler_filter_minmax = false; + loaded_extensions.erase(VK_EXT_SAMPLER_FILTER_MINMAX_EXTENSION_NAME); } } - const bool is_intel_windows = driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS; - const bool is_intel_anv = driver_id == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA; - if (ext_vertex_input_dynamic_state && is_intel_windows) { - const u32 version = (properties.driverVersion << 3) >> 3; + if (extensions.vertex_input_dynamic_state && is_intel_windows) { + const u32 version = (properties.properties.driverVersion << 3) >> 3; if (version < VK_MAKE_API_VERSION(27, 20, 100, 0)) { - LOG_WARNING(Render_Vulkan, "Blacklisting Intel for VK_EXT_vertex_input_dynamic_state"); - ext_vertex_input_dynamic_state = false; + LOG_WARNING(Render_Vulkan, "Intel has broken VK_EXT_vertex_input_dynamic_state"); + extensions.vertex_input_dynamic_state = false; + loaded_extensions.erase(VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); } } - if (is_float16_supported && is_intel_windows) { + if (features.shader_float16_int8.shaderFloat16 && is_intel_windows) { // Intel's compiler crashes when using fp16 on Astral Chain, disable it for the time being. - LOG_WARNING(Render_Vulkan, "Blacklisting Intel proprietary from float16 math"); - is_float16_supported = false; + LOG_WARNING(Render_Vulkan, "Intel has broken float16 math"); + features.shader_float16_int8.shaderFloat16 = false; } if (is_intel_windows) { LOG_WARNING(Render_Vulkan, "Intel proprietary drivers do not support MSAA image blits"); @@ -857,10 +456,17 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR LOG_WARNING(Render_Vulkan, "ANV driver does not support native BGR format"); must_emulate_bgr565 = true; } + if (is_mvk) { + LOG_WARNING(Render_Vulkan, + "MVK driver breaks when using more than 16 vertex attributes/bindings"); + properties.properties.limits.maxVertexInputAttributes = + std::min(properties.properties.limits.maxVertexInputAttributes, 16U); + properties.properties.limits.maxVertexInputBindings = + std::min(properties.properties.limits.maxVertexInputBindings, 16U); + } - supports_d24_depth = - IsFormatSupported(VK_FORMAT_D24_UNORM_S8_UINT, - VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT, FormatType::Optimal); + logical = vk::Device::Create(physical, queue_cis, ExtensionListForVulkan(loaded_extensions), + first_next, dld); graphics_queue = logical.GetQueue(graphics_family); present_queue = logical.GetQueue(present_family); @@ -915,7 +521,7 @@ void Device::SaveShader(std::span<const u32> spirv) const { } } -bool Device::IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const { +bool Device::ComputeIsOptimalAstcSupported() const { // Disable for now to avoid converting ASTC twice. static constexpr std::array astc_formats = { VK_FORMAT_ASTC_4x4_UNORM_BLOCK, VK_FORMAT_ASTC_4x4_SRGB_BLOCK, @@ -933,7 +539,7 @@ bool Device::IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) co VK_FORMAT_ASTC_12x10_UNORM_BLOCK, VK_FORMAT_ASTC_12x10_SRGB_BLOCK, VK_FORMAT_ASTC_12x12_UNORM_BLOCK, VK_FORMAT_ASTC_12x12_SRGB_BLOCK, }; - if (!features.textureCompressionASTC_LDR) { + if (!features.features.textureCompressionASTC_LDR) { return false; } const auto format_feature_usage{ @@ -971,7 +577,7 @@ bool Device::IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags want } std::string Device::GetDriverName() const { - switch (driver_id) { + switch (properties.driver.driverID) { case VK_DRIVER_ID_AMD_PROPRIETARY: return "AMD"; case VK_DRIVER_ID_AMD_OPEN_SOURCE: @@ -987,522 +593,336 @@ std::string Device::GetDriverName() const { case VK_DRIVER_ID_MESA_LLVMPIPE: return "LAVAPIPE"; default: - return vendor_name; + return properties.driver.driverName; } } bool Device::ShouldBoostClocks() const { + const auto driver_id = properties.driver.driverID; + const auto vendor_id = properties.properties.vendorID; + const auto device_id = properties.properties.deviceID; + const bool validated_driver = driver_id == VK_DRIVER_ID_AMD_PROPRIETARY || driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE || driver_id == VK_DRIVER_ID_MESA_RADV || driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY || driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS || driver_id == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA; - const bool is_steam_deck = properties.vendorID == 0x1002 && properties.deviceID == 0x163F; + const bool is_steam_deck = vendor_id == 0x1002 && device_id == 0x163F; return validated_driver && !is_steam_deck; } -static std::vector<const char*> ExtensionsRequiredForInstanceVersion(u32 available_version) { - std::vector<const char*> extensions{REQUIRED_EXTENSIONS.begin(), REQUIRED_EXTENSIONS.end()}; +bool Device::GetSuitability(bool requires_swapchain) { + // Assume we will be suitable. + bool suitable = true; - if (available_version < VK_API_VERSION_1_2) { - extensions.insert(extensions.end(), REQUIRED_EXTENSIONS_BEFORE_1_2.begin(), - REQUIRED_EXTENSIONS_BEFORE_1_2.end()); - } - - if (available_version < VK_API_VERSION_1_3) { - extensions.insert(extensions.end(), REQUIRED_EXTENSIONS_BEFORE_1_3.begin(), - REQUIRED_EXTENSIONS_BEFORE_1_3.end()); - } - - return extensions; -} + // Configure properties. + properties.properties = physical.GetProperties(); -void Device::CheckSuitability(bool requires_swapchain) const { - std::vector<const char*> required_extensions = - ExtensionsRequiredForInstanceVersion(instance_version); - std::vector<const char*> available_extensions; + // Set instance version. + instance_version = properties.properties.apiVersion; - if (requires_swapchain) { - required_extensions.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME); - } + // Minimum of API version 1.1 is required. (This is well-supported.) + ASSERT(instance_version >= VK_API_VERSION_1_1); + // Get available extensions. auto extension_properties = physical.EnumerateDeviceExtensionProperties(); + // Get the set of supported extensions. + supported_extensions.clear(); for (const VkExtensionProperties& property : extension_properties) { - available_extensions.push_back(property.extensionName); + supported_extensions.insert(property.extensionName); } - bool has_all_required_extensions = true; - for (const char* requirement_name : required_extensions) { - const bool found = - std::ranges::any_of(available_extensions, [&](const char* extension_name) { - return std::strcmp(requirement_name, extension_name) == 0; - }); + // Generate list of extensions to load. + loaded_extensions.clear(); - if (!found) { - LOG_ERROR(Render_Vulkan, "Missing required extension: {}", requirement_name); - has_all_required_extensions = false; - } +#define EXTENSION(prefix, macro_name, var_name) \ + if (supported_extensions.contains(VK_##prefix##_##macro_name##_EXTENSION_NAME)) { \ + loaded_extensions.insert(VK_##prefix##_##macro_name##_EXTENSION_NAME); \ + extensions.var_name = true; \ } - - if (!has_all_required_extensions) { - throw vk::Exception(VK_ERROR_EXTENSION_NOT_PRESENT); +#define FEATURE_EXTENSION(prefix, struct_name, macro_name, var_name) \ + if (supported_extensions.contains(VK_##prefix##_##macro_name##_EXTENSION_NAME)) { \ + loaded_extensions.insert(VK_##prefix##_##macro_name##_EXTENSION_NAME); \ + extensions.var_name = true; \ } - struct LimitTuple { - u32 minimum; - u32 value; - const char* name; - }; - const VkPhysicalDeviceLimits& limits{properties.limits}; - const std::array limits_report{ - LimitTuple{65536, limits.maxUniformBufferRange, "maxUniformBufferRange"}, - LimitTuple{16, limits.maxViewports, "maxViewports"}, - LimitTuple{8, limits.maxColorAttachments, "maxColorAttachments"}, - LimitTuple{8, limits.maxClipDistances, "maxClipDistances"}, - }; - for (const auto& tuple : limits_report) { - if (tuple.value < tuple.minimum) { - LOG_ERROR(Render_Vulkan, "{} has to be {} or greater but it is {}", tuple.name, - tuple.minimum, tuple.value); - throw vk::Exception(VK_ERROR_FEATURE_NOT_PRESENT); - } + if (instance_version < VK_API_VERSION_1_2) { + FOR_EACH_VK_FEATURE_1_2(FEATURE_EXTENSION); + } + if (instance_version < VK_API_VERSION_1_3) { + FOR_EACH_VK_FEATURE_1_3(FEATURE_EXTENSION); } - VkPhysicalDeviceShaderDemoteToHelperInvocationFeatures demote{}; - demote.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DEMOTE_TO_HELPER_INVOCATION_FEATURES; - demote.pNext = nullptr; - VkPhysicalDeviceVariablePointerFeatures variable_pointers{}; - variable_pointers.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VARIABLE_POINTERS_FEATURES; - variable_pointers.pNext = &demote; + FOR_EACH_VK_FEATURE_EXT(FEATURE_EXTENSION); + FOR_EACH_VK_EXTENSION(EXTENSION); +#ifdef _WIN32 + FOR_EACH_VK_EXTENSION_WIN32(EXTENSION); +#endif - VkPhysicalDeviceRobustness2FeaturesEXT robustness2{}; - robustness2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ROBUSTNESS_2_FEATURES_EXT; - robustness2.pNext = &variable_pointers; +#undef FEATURE_EXTENSION +#undef EXTENSION - VkPhysicalDeviceTimelineSemaphoreFeatures timeline_semaphore{}; - timeline_semaphore.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TIMELINE_SEMAPHORE_FEATURES; - timeline_semaphore.pNext = &robustness2; + // Some extensions are mandatory. Check those. +#define CHECK_EXTENSION(extension_name) \ + if (!loaded_extensions.contains(extension_name)) { \ + LOG_ERROR(Render_Vulkan, "Missing required extension {}", extension_name); \ + suitable = false; \ + } - VkPhysicalDevice16BitStorageFeatures bit16_storage{}; - bit16_storage.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES; - bit16_storage.pNext = &timeline_semaphore; +#define LOG_EXTENSION(extension_name) \ + if (!loaded_extensions.contains(extension_name)) { \ + LOG_INFO(Render_Vulkan, "Device doesn't support extension {}", extension_name); \ + } - VkPhysicalDevice8BitStorageFeatures bit8_storage{}; - bit8_storage.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_8BIT_STORAGE_FEATURES; - bit8_storage.pNext = &bit16_storage; + FOR_EACH_VK_RECOMMENDED_EXTENSION(LOG_EXTENSION); + FOR_EACH_VK_MANDATORY_EXTENSION(CHECK_EXTENSION); +#ifdef _WIN32 + FOR_EACH_VK_MANDATORY_EXTENSION_WIN32(CHECK_EXTENSION); +#else + FOR_EACH_VK_MANDATORY_EXTENSION_GENERIC(CHECK_EXTENSION); +#endif - VkPhysicalDeviceHostQueryResetFeatures host_query_reset{}; - host_query_reset.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_HOST_QUERY_RESET_FEATURES; - host_query_reset.pNext = &bit8_storage; + if (requires_swapchain) { + CHECK_EXTENSION(VK_KHR_SWAPCHAIN_EXTENSION_NAME); + } - VkPhysicalDeviceShaderDrawParametersFeatures draw_parameters{}; - draw_parameters.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DRAW_PARAMETER_FEATURES; - draw_parameters.pNext = &host_query_reset; +#undef LOG_EXTENSION +#undef CHECK_EXTENSION - VkPhysicalDeviceFeatures2 features2{}; + // Generate the linked list of features to test. features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; - features2.pNext = &draw_parameters; - physical.GetFeatures2(features2); + // Set next pointer. + void** next = &features2.pNext; - const VkPhysicalDeviceFeatures& features{features2.features}; - std::array feature_report{ - std::make_pair(features.robustBufferAccess, "robustBufferAccess"), - std::make_pair(features.vertexPipelineStoresAndAtomics, "vertexPipelineStoresAndAtomics"), - std::make_pair(features.imageCubeArray, "imageCubeArray"), - std::make_pair(features.independentBlend, "independentBlend"), - std::make_pair(features.multiDrawIndirect, "multiDrawIndirect"), - std::make_pair(features.drawIndirectFirstInstance, "drawIndirectFirstInstance"), - std::make_pair(features.depthClamp, "depthClamp"), - std::make_pair(features.samplerAnisotropy, "samplerAnisotropy"), - std::make_pair(features.largePoints, "largePoints"), - std::make_pair(features.multiViewport, "multiViewport"), - std::make_pair(features.depthBiasClamp, "depthBiasClamp"), - std::make_pair(features.fillModeNonSolid, "fillModeNonSolid"), - std::make_pair(features.wideLines, "wideLines"), - std::make_pair(features.geometryShader, "geometryShader"), - std::make_pair(features.tessellationShader, "tessellationShader"), - std::make_pair(features.sampleRateShading, "sampleRateShading"), - std::make_pair(features.dualSrcBlend, "dualSrcBlend"), - std::make_pair(features.logicOp, "logicOp"), - std::make_pair(features.occlusionQueryPrecise, "occlusionQueryPrecise"), - std::make_pair(features.fragmentStoresAndAtomics, "fragmentStoresAndAtomics"), - std::make_pair(features.shaderImageGatherExtended, "shaderImageGatherExtended"), - std::make_pair(features.shaderStorageImageWriteWithoutFormat, - "shaderStorageImageWriteWithoutFormat"), - std::make_pair(features.shaderClipDistance, "shaderClipDistance"), - std::make_pair(features.shaderCullDistance, "shaderCullDistance"), - std::make_pair(variable_pointers.variablePointers, "variablePointers"), - std::make_pair(variable_pointers.variablePointersStorageBuffer, - "variablePointersStorageBuffer"), - std::make_pair(robustness2.robustBufferAccess2, "robustBufferAccess2"), - std::make_pair(robustness2.robustImageAccess2, "robustImageAccess2"), - std::make_pair(robustness2.nullDescriptor, "nullDescriptor"), - std::make_pair(demote.shaderDemoteToHelperInvocation, "shaderDemoteToHelperInvocation"), - std::make_pair(timeline_semaphore.timelineSemaphore, "timelineSemaphore"), - std::make_pair(bit16_storage.storageBuffer16BitAccess, "storageBuffer16BitAccess"), - std::make_pair(bit16_storage.uniformAndStorageBuffer16BitAccess, - "uniformAndStorageBuffer16BitAccess"), - std::make_pair(bit8_storage.storageBuffer8BitAccess, "storageBuffer8BitAccess"), - std::make_pair(bit8_storage.uniformAndStorageBuffer8BitAccess, - "uniformAndStorageBuffer8BitAccess"), - std::make_pair(host_query_reset.hostQueryReset, "hostQueryReset"), - std::make_pair(draw_parameters.shaderDrawParameters, "shaderDrawParameters"), - }; + // Test all features we know about. If the feature is not available in core at our + // current API version, and was not enabled by an extension, skip testing the feature. + // We set the structure sType explicitly here as it is zeroed by the constructor. +#define FEATURE(prefix, struct_name, macro_name, var_name) \ + features.var_name.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_##macro_name##_FEATURES; \ + SetNext(next, features.var_name); - bool has_all_required_features = true; - for (const auto& [is_supported, name] : feature_report) { - if (!is_supported) { - LOG_ERROR(Render_Vulkan, "Missing required feature: {}", name); - has_all_required_features = false; - } +#define EXT_FEATURE(prefix, struct_name, macro_name, var_name) \ + if (extensions.var_name) { \ + features.var_name.sType = \ + VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_##macro_name##_FEATURES_##prefix; \ + SetNext(next, features.var_name); \ } - if (!has_all_required_features) { - throw vk::Exception(VK_ERROR_FEATURE_NOT_PRESENT); + FOR_EACH_VK_FEATURE_1_1(FEATURE); + FOR_EACH_VK_FEATURE_EXT(EXT_FEATURE); + if (instance_version >= VK_API_VERSION_1_2) { + FOR_EACH_VK_FEATURE_1_2(FEATURE); + } else { + FOR_EACH_VK_FEATURE_1_2(EXT_FEATURE); } -} - -std::vector<const char*> Device::LoadExtensions(bool requires_surface) { - std::vector<const char*> extensions = ExtensionsRequiredForInstanceVersion(instance_version); - if (requires_surface) { - extensions.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME); + if (instance_version >= VK_API_VERSION_1_3) { + FOR_EACH_VK_FEATURE_1_3(FEATURE); + } else { + FOR_EACH_VK_FEATURE_1_3(EXT_FEATURE); } - bool has_khr_shader_float16_int8{}; - bool has_khr_workgroup_memory_explicit_layout{}; - bool has_khr_pipeline_executable_properties{}; - bool has_khr_image_format_list{}; - bool has_khr_swapchain_mutable_format{}; - bool has_ext_subgroup_size_control{}; - bool has_ext_transform_feedback{}; - bool has_ext_custom_border_color{}; - bool has_ext_extended_dynamic_state{}; - bool has_ext_extended_dynamic_state_2{}; - bool has_ext_extended_dynamic_state_3{}; - bool has_ext_shader_atomic_int64{}; - bool has_ext_provoking_vertex{}; - bool has_ext_vertex_input_dynamic_state{}; - bool has_ext_line_rasterization{}; - bool has_ext_primitive_topology_list_restart{}; - bool has_ext_depth_clip_control{}; - for (const std::string& extension : supported_extensions) { - const auto test = [&](std::optional<std::reference_wrapper<bool>> status, const char* name, - bool push) { - if (extension != name) { - return; - } - if (push) { - extensions.push_back(name); - } - if (status) { - status->get() = true; - } - }; - test(nv_viewport_swizzle, VK_NV_VIEWPORT_SWIZZLE_EXTENSION_NAME, true); - test(nv_viewport_array2, VK_NV_VIEWPORT_ARRAY2_EXTENSION_NAME, true); - test(nv_geometry_shader_passthrough, VK_NV_GEOMETRY_SHADER_PASSTHROUGH_EXTENSION_NAME, - true); - test(khr_uniform_buffer_standard_layout, - VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true); - test(khr_spirv_1_4, VK_KHR_SPIRV_1_4_EXTENSION_NAME, true); - test(khr_push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME, true); - test(has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME, false); - test(khr_draw_indirect_count, VK_KHR_DRAW_INDIRECT_COUNT_EXTENSION_NAME, true); - test(ext_depth_range_unrestricted, VK_EXT_DEPTH_RANGE_UNRESTRICTED_EXTENSION_NAME, true); - test(ext_index_type_uint8, VK_EXT_INDEX_TYPE_UINT8_EXTENSION_NAME, true); - test(has_ext_primitive_topology_list_restart, - VK_EXT_PRIMITIVE_TOPOLOGY_LIST_RESTART_EXTENSION_NAME, true); - test(ext_sampler_filter_minmax, VK_EXT_SAMPLER_FILTER_MINMAX_EXTENSION_NAME, true); - test(ext_shader_viewport_index_layer, VK_EXT_SHADER_VIEWPORT_INDEX_LAYER_EXTENSION_NAME, - true); - test(ext_tooling_info, VK_EXT_TOOLING_INFO_EXTENSION_NAME, true); - test(ext_shader_stencil_export, VK_EXT_SHADER_STENCIL_EXPORT_EXTENSION_NAME, true); - test(ext_conservative_rasterization, VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME, - true); - test(has_ext_depth_clip_control, VK_EXT_DEPTH_CLIP_CONTROL_EXTENSION_NAME, false); - test(has_ext_transform_feedback, VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME, false); - test(has_ext_custom_border_color, VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME, false); - test(has_ext_extended_dynamic_state, VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME, false); - test(has_ext_extended_dynamic_state_2, VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME, - false); - test(has_ext_extended_dynamic_state_3, VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME, - false); - test(has_ext_subgroup_size_control, VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME, true); - test(has_ext_provoking_vertex, VK_EXT_PROVOKING_VERTEX_EXTENSION_NAME, false); - test(has_ext_vertex_input_dynamic_state, VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME, - false); - test(has_ext_shader_atomic_int64, VK_KHR_SHADER_ATOMIC_INT64_EXTENSION_NAME, false); - test(has_khr_workgroup_memory_explicit_layout, - VK_KHR_WORKGROUP_MEMORY_EXPLICIT_LAYOUT_EXTENSION_NAME, false); - test(has_khr_image_format_list, VK_KHR_IMAGE_FORMAT_LIST_EXTENSION_NAME, false); - test(has_khr_swapchain_mutable_format, VK_KHR_SWAPCHAIN_MUTABLE_FORMAT_EXTENSION_NAME, - false); - test(has_ext_line_rasterization, VK_EXT_LINE_RASTERIZATION_EXTENSION_NAME, false); - test(ext_memory_budget, VK_EXT_MEMORY_BUDGET_EXTENSION_NAME, true); - if (Settings::values.enable_nsight_aftermath) { - test(nv_device_diagnostics_config, VK_NV_DEVICE_DIAGNOSTICS_CONFIG_EXTENSION_NAME, - true); - } - if (Settings::values.renderer_shader_feedback) { - test(has_khr_pipeline_executable_properties, - VK_KHR_PIPELINE_EXECUTABLE_PROPERTIES_EXTENSION_NAME, false); - } - } - VkPhysicalDeviceFeatures2 features{}; - features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; - - VkPhysicalDeviceProperties2 physical_properties{}; - physical_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2; - - if (has_khr_shader_float16_int8) { - VkPhysicalDeviceShaderFloat16Int8Features float16_int8_features; - float16_int8_features.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_FLOAT16_INT8_FEATURES; - float16_int8_features.pNext = nullptr; - features.pNext = &float16_int8_features; - - physical.GetFeatures2(features); - is_float16_supported = float16_int8_features.shaderFloat16; - is_int8_supported = float16_int8_features.shaderInt8; - extensions.push_back(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME); - } - if (has_ext_subgroup_size_control) { - VkPhysicalDeviceSubgroupSizeControlFeatures subgroup_features; - subgroup_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_FEATURES; - subgroup_features.pNext = nullptr; - features.pNext = &subgroup_features; - physical.GetFeatures2(features); - - VkPhysicalDeviceSubgroupSizeControlProperties subgroup_properties; - subgroup_properties.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_PROPERTIES; - subgroup_properties.pNext = nullptr; - physical_properties.pNext = &subgroup_properties; - physical.GetProperties2(physical_properties); +#undef EXT_FEATURE +#undef FEATURE - is_warp_potentially_bigger = subgroup_properties.maxSubgroupSize > GuestWarpSize; + // Perform the feature test. + physical.GetFeatures2(features2); + features.features = features2.features; - if (subgroup_features.subgroupSizeControl && - subgroup_properties.minSubgroupSize <= GuestWarpSize && - subgroup_properties.maxSubgroupSize >= GuestWarpSize) { - extensions.push_back(VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME); - guest_warp_stages = subgroup_properties.requiredSubgroupSizeStages; - ext_subgroup_size_control = true; - } - } else { - is_warp_potentially_bigger = true; - } - if (has_ext_provoking_vertex) { - VkPhysicalDeviceProvokingVertexFeaturesEXT provoking_vertex; - provoking_vertex.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROVOKING_VERTEX_FEATURES_EXT; - provoking_vertex.pNext = nullptr; - features.pNext = &provoking_vertex; - physical.GetFeatures2(features); - - if (provoking_vertex.provokingVertexLast && - provoking_vertex.transformFeedbackPreservesProvokingVertex) { - extensions.push_back(VK_EXT_PROVOKING_VERTEX_EXTENSION_NAME); - ext_provoking_vertex = true; - } + // Some features are mandatory. Check those. +#define CHECK_FEATURE(feature, name) \ + if (!features.feature.name) { \ + LOG_ERROR(Render_Vulkan, "Missing required feature {}", #name); \ + suitable = false; \ } - if (has_ext_vertex_input_dynamic_state) { - VkPhysicalDeviceVertexInputDynamicStateFeaturesEXT vertex_input; - vertex_input.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_INPUT_DYNAMIC_STATE_FEATURES_EXT; - vertex_input.pNext = nullptr; - features.pNext = &vertex_input; - physical.GetFeatures2(features); - - if (vertex_input.vertexInputDynamicState) { - extensions.push_back(VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); - ext_vertex_input_dynamic_state = true; - } - } - if (has_ext_shader_atomic_int64) { - VkPhysicalDeviceShaderAtomicInt64Features atomic_int64; - atomic_int64.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_ATOMIC_INT64_FEATURES; - atomic_int64.pNext = nullptr; - features.pNext = &atomic_int64; - physical.GetFeatures2(features); - - if (atomic_int64.shaderBufferInt64Atomics && atomic_int64.shaderSharedInt64Atomics) { - extensions.push_back(VK_KHR_SHADER_ATOMIC_INT64_EXTENSION_NAME); - ext_shader_atomic_int64 = true; - } - } - if (has_ext_transform_feedback) { - VkPhysicalDeviceTransformFeedbackFeaturesEXT tfb_features; - tfb_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT; - tfb_features.pNext = nullptr; - features.pNext = &tfb_features; - physical.GetFeatures2(features); - - VkPhysicalDeviceTransformFeedbackPropertiesEXT tfb_properties; - tfb_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_PROPERTIES_EXT; - tfb_properties.pNext = nullptr; - physical_properties.pNext = &tfb_properties; - physical.GetProperties2(physical_properties); - if (tfb_features.transformFeedback && tfb_features.geometryStreams && - tfb_properties.maxTransformFeedbackStreams >= 4 && - tfb_properties.maxTransformFeedbackBuffers && tfb_properties.transformFeedbackQueries && - tfb_properties.transformFeedbackDraw) { - extensions.push_back(VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME); - ext_transform_feedback = true; - } - } - 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.GetFeatures2(features); - - if (border_features.customBorderColors && border_features.customBorderColorWithoutFormat) { - extensions.push_back(VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME); - ext_custom_border_color = true; - } - } - if (has_ext_extended_dynamic_state) { - VkPhysicalDeviceExtendedDynamicStateFeaturesEXT extended_dynamic_state; - extended_dynamic_state.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_FEATURES_EXT; - extended_dynamic_state.pNext = nullptr; - features.pNext = &extended_dynamic_state; - physical.GetFeatures2(features); - - if (extended_dynamic_state.extendedDynamicState) { - extensions.push_back(VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME); - ext_extended_dynamic_state = true; - } - } - if (has_ext_extended_dynamic_state_2) { - VkPhysicalDeviceExtendedDynamicState2FeaturesEXT extended_dynamic_state_2; - extended_dynamic_state_2.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_2_FEATURES_EXT; - extended_dynamic_state_2.pNext = nullptr; - features.pNext = &extended_dynamic_state_2; - physical.GetFeatures2(features); - - if (extended_dynamic_state_2.extendedDynamicState2) { - extensions.push_back(VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME); - ext_extended_dynamic_state_2 = true; - ext_extended_dynamic_state_2_extra = - extended_dynamic_state_2.extendedDynamicState2LogicOp; - } +#define LOG_FEATURE(feature, name) \ + if (!features.feature.name) { \ + LOG_INFO(Render_Vulkan, "Device doesn't support feature {}", #name); \ } - if (has_ext_extended_dynamic_state_3) { - VkPhysicalDeviceExtendedDynamicState3FeaturesEXT extended_dynamic_state_3; - extended_dynamic_state_3.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_3_FEATURES_EXT; - extended_dynamic_state_3.pNext = nullptr; - features.pNext = &extended_dynamic_state_3; - physical.GetFeatures2(features); - - ext_extended_dynamic_state_3_blend = - extended_dynamic_state_3.extendedDynamicState3ColorBlendEnable && - extended_dynamic_state_3.extendedDynamicState3ColorBlendEquation && - extended_dynamic_state_3.extendedDynamicState3ColorWriteMask; - - ext_extended_dynamic_state_3_enables = - extended_dynamic_state_3.extendedDynamicState3DepthClampEnable && - extended_dynamic_state_3.extendedDynamicState3LogicOpEnable; - - ext_extended_dynamic_state_3 = - ext_extended_dynamic_state_3_blend || ext_extended_dynamic_state_3_enables; - if (ext_extended_dynamic_state_3) { - extensions.push_back(VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME); - } + + FOR_EACH_VK_RECOMMENDED_FEATURE(LOG_FEATURE); + FOR_EACH_VK_MANDATORY_FEATURE(CHECK_FEATURE); + +#undef LOG_FEATURE +#undef CHECK_FEATURE + + // Generate linked list of properties. + properties2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2; + + // Set next pointer. + next = &properties2.pNext; + + // Get driver info. + properties.driver.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES; + SetNext(next, properties.driver); + + // Retrieve relevant extension properties. + if (extensions.shader_float_controls) { + properties.float_controls.sType = + VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT_CONTROLS_PROPERTIES; + SetNext(next, properties.float_controls); } - if (has_ext_line_rasterization) { - VkPhysicalDeviceLineRasterizationFeaturesEXT line_raster; - line_raster.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_LINE_RASTERIZATION_FEATURES_EXT; - line_raster.pNext = nullptr; - features.pNext = &line_raster; - physical.GetFeatures2(features); - if (line_raster.rectangularLines && line_raster.smoothLines) { - extensions.push_back(VK_EXT_LINE_RASTERIZATION_EXTENSION_NAME); - ext_line_rasterization = true; - } + if (extensions.push_descriptor) { + properties.push_descriptor.sType = + VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PUSH_DESCRIPTOR_PROPERTIES_KHR; + SetNext(next, properties.push_descriptor); } - if (has_ext_depth_clip_control) { - VkPhysicalDeviceDepthClipControlFeaturesEXT depth_clip_control_features; - depth_clip_control_features.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DEPTH_CLIP_CONTROL_FEATURES_EXT; - depth_clip_control_features.pNext = nullptr; - features.pNext = &depth_clip_control_features; - physical.GetFeatures2(features); - - if (depth_clip_control_features.depthClipControl) { - extensions.push_back(VK_EXT_DEPTH_CLIP_CONTROL_EXTENSION_NAME); - ext_depth_clip_control = true; - } + if (extensions.subgroup_size_control) { + properties.subgroup_size_control.sType = + VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_PROPERTIES; + SetNext(next, properties.subgroup_size_control); } - if (has_khr_workgroup_memory_explicit_layout) { - VkPhysicalDeviceWorkgroupMemoryExplicitLayoutFeaturesKHR layout; - layout.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_WORKGROUP_MEMORY_EXPLICIT_LAYOUT_FEATURES_KHR; - layout.pNext = nullptr; - features.pNext = &layout; - physical.GetFeatures2(features); - - if (layout.workgroupMemoryExplicitLayout && - layout.workgroupMemoryExplicitLayout8BitAccess && - layout.workgroupMemoryExplicitLayout16BitAccess && - layout.workgroupMemoryExplicitLayoutScalarBlockLayout) { - extensions.push_back(VK_KHR_WORKGROUP_MEMORY_EXPLICIT_LAYOUT_EXTENSION_NAME); - khr_workgroup_memory_explicit_layout = true; - } + if (extensions.transform_feedback) { + properties.transform_feedback.sType = + VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_PROPERTIES_EXT; + SetNext(next, properties.transform_feedback); } - if (has_khr_pipeline_executable_properties) { - VkPhysicalDevicePipelineExecutablePropertiesFeaturesKHR executable_properties; - executable_properties.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PIPELINE_EXECUTABLE_PROPERTIES_FEATURES_KHR; - executable_properties.pNext = nullptr; - features.pNext = &executable_properties; - physical.GetFeatures2(features); - - if (executable_properties.pipelineExecutableInfo) { - extensions.push_back(VK_KHR_PIPELINE_EXECUTABLE_PROPERTIES_EXTENSION_NAME); - khr_pipeline_executable_properties = true; + + // Perform the property fetch. + physical.GetProperties2(properties2); + properties.properties = properties2.properties; + + // Unload extensions if feature support is insufficient. + RemoveUnsuitableExtensions(); + + // Check limits. + struct Limit { + u32 minimum; + u32 value; + const char* name; + }; + + const VkPhysicalDeviceLimits& limits{properties.properties.limits}; + const std::array limits_report{ + Limit{65536, limits.maxUniformBufferRange, "maxUniformBufferRange"}, + Limit{16, limits.maxViewports, "maxViewports"}, + Limit{8, limits.maxColorAttachments, "maxColorAttachments"}, + Limit{8, limits.maxClipDistances, "maxClipDistances"}, + }; + + for (const auto& [min, value, name] : limits_report) { + if (value < min) { + LOG_ERROR(Render_Vulkan, "{} has to be {} or greater but it is {}", name, min, value); + suitable = false; } } - if (has_ext_primitive_topology_list_restart) { - VkPhysicalDevicePrimitiveTopologyListRestartFeaturesEXT primitive_topology_list_restart{}; - primitive_topology_list_restart.sType = - VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PRIMITIVE_TOPOLOGY_LIST_RESTART_FEATURES_EXT; - primitive_topology_list_restart.pNext = nullptr; - features.pNext = &primitive_topology_list_restart; - physical.GetFeatures2(features); - - is_topology_list_restart_supported = - primitive_topology_list_restart.primitiveTopologyListRestart; - is_patch_list_restart_supported = - primitive_topology_list_restart.primitiveTopologyPatchListRestart; - } - if (requires_surface && has_khr_image_format_list && has_khr_swapchain_mutable_format) { - extensions.push_back(VK_KHR_IMAGE_FORMAT_LIST_EXTENSION_NAME); - extensions.push_back(VK_KHR_SWAPCHAIN_MUTABLE_FORMAT_EXTENSION_NAME); - khr_swapchain_mutable_format = true; - } - if (khr_push_descriptor) { - VkPhysicalDevicePushDescriptorPropertiesKHR push_descriptor; - push_descriptor.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PUSH_DESCRIPTOR_PROPERTIES_KHR; - push_descriptor.pNext = nullptr; - physical_properties.pNext = &push_descriptor; - physical.GetProperties2(physical_properties); + // Return whether we were suitable. + return suitable; +} - max_push_descriptors = push_descriptor.maxPushDescriptors; +void Device::RemoveExtensionIfUnsuitable(bool is_suitable, const std::string& extension_name) { + if (loaded_extensions.contains(extension_name) && !is_suitable) { + LOG_WARNING(Render_Vulkan, "Removing unsuitable extension {}", extension_name); + loaded_extensions.erase(extension_name); } +} - has_null_descriptor = true; - - return extensions; +void Device::RemoveUnsuitableExtensions() { + // VK_EXT_custom_border_color + extensions.custom_border_color = features.custom_border_color.customBorderColors && + features.custom_border_color.customBorderColorWithoutFormat; + RemoveExtensionIfUnsuitable(extensions.custom_border_color, + VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME); + + // VK_EXT_depth_clip_control + extensions.depth_clip_control = features.depth_clip_control.depthClipControl; + RemoveExtensionIfUnsuitable(extensions.depth_clip_control, + VK_EXT_DEPTH_CLIP_CONTROL_EXTENSION_NAME); + + // VK_EXT_extended_dynamic_state + extensions.extended_dynamic_state = features.extended_dynamic_state.extendedDynamicState; + RemoveExtensionIfUnsuitable(extensions.extended_dynamic_state, + VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME); + + // VK_EXT_extended_dynamic_state2 + extensions.extended_dynamic_state2 = features.extended_dynamic_state2.extendedDynamicState2; + RemoveExtensionIfUnsuitable(extensions.extended_dynamic_state2, + VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME); + + // VK_EXT_extended_dynamic_state3 + dynamic_state3_blending = + features.extended_dynamic_state3.extendedDynamicState3ColorBlendEnable && + features.extended_dynamic_state3.extendedDynamicState3ColorBlendEquation && + features.extended_dynamic_state3.extendedDynamicState3ColorWriteMask; + dynamic_state3_enables = + features.extended_dynamic_state3.extendedDynamicState3DepthClampEnable && + features.extended_dynamic_state3.extendedDynamicState3LogicOpEnable; + + extensions.extended_dynamic_state3 = dynamic_state3_blending || dynamic_state3_enables; + dynamic_state3_blending = dynamic_state3_blending && extensions.extended_dynamic_state3; + dynamic_state3_enables = dynamic_state3_enables && extensions.extended_dynamic_state3; + RemoveExtensionIfUnsuitable(extensions.extended_dynamic_state3, + VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME); + + // VK_EXT_provoking_vertex + extensions.provoking_vertex = + features.provoking_vertex.provokingVertexLast && + features.provoking_vertex.transformFeedbackPreservesProvokingVertex; + RemoveExtensionIfUnsuitable(extensions.provoking_vertex, + VK_EXT_PROVOKING_VERTEX_EXTENSION_NAME); + + // VK_KHR_shader_atomic_int64 + extensions.shader_atomic_int64 = features.shader_atomic_int64.shaderBufferInt64Atomics && + features.shader_atomic_int64.shaderSharedInt64Atomics; + RemoveExtensionIfUnsuitable(extensions.shader_atomic_int64, + VK_KHR_SHADER_ATOMIC_INT64_EXTENSION_NAME); + + // VK_EXT_shader_demote_to_helper_invocation + extensions.shader_demote_to_helper_invocation = + features.shader_demote_to_helper_invocation.shaderDemoteToHelperInvocation; + RemoveExtensionIfUnsuitable(extensions.shader_demote_to_helper_invocation, + VK_EXT_SHADER_DEMOTE_TO_HELPER_INVOCATION_EXTENSION_NAME); + + // VK_EXT_subgroup_size_control + extensions.subgroup_size_control = + features.subgroup_size_control.subgroupSizeControl && + properties.subgroup_size_control.minSubgroupSize <= GuestWarpSize && + properties.subgroup_size_control.maxSubgroupSize >= GuestWarpSize; + RemoveExtensionIfUnsuitable(extensions.subgroup_size_control, + VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME); + + // VK_EXT_transform_feedback + extensions.transform_feedback = + features.transform_feedback.transformFeedback && + features.transform_feedback.geometryStreams && + properties.transform_feedback.maxTransformFeedbackStreams >= 4 && + properties.transform_feedback.maxTransformFeedbackBuffers > 0 && + properties.transform_feedback.transformFeedbackQueries && + properties.transform_feedback.transformFeedbackDraw; + RemoveExtensionIfUnsuitable(extensions.transform_feedback, + VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME); + + // VK_EXT_vertex_input_dynamic_state + extensions.vertex_input_dynamic_state = + features.vertex_input_dynamic_state.vertexInputDynamicState; + RemoveExtensionIfUnsuitable(extensions.vertex_input_dynamic_state, + VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); + + // VK_KHR_pipeline_executable_properties + if (Settings::values.renderer_shader_feedback.GetValue()) { + extensions.pipeline_executable_properties = + features.pipeline_executable_properties.pipelineExecutableInfo; + RemoveExtensionIfUnsuitable(extensions.pipeline_executable_properties, + VK_KHR_PIPELINE_EXECUTABLE_PROPERTIES_EXTENSION_NAME); + } else { + extensions.pipeline_executable_properties = false; + loaded_extensions.erase(VK_KHR_PIPELINE_EXECUTABLE_PROPERTIES_EXTENSION_NAME); + } + + // VK_KHR_workgroup_memory_explicit_layout + extensions.workgroup_memory_explicit_layout = + features.features.shaderInt16 && + features.workgroup_memory_explicit_layout.workgroupMemoryExplicitLayout && + features.workgroup_memory_explicit_layout.workgroupMemoryExplicitLayout8BitAccess && + features.workgroup_memory_explicit_layout.workgroupMemoryExplicitLayout16BitAccess && + features.workgroup_memory_explicit_layout.workgroupMemoryExplicitLayoutScalarBlockLayout; + RemoveExtensionIfUnsuitable(extensions.workgroup_memory_explicit_layout, + VK_KHR_WORKGROUP_MEMORY_EXPLICIT_LAYOUT_EXTENSION_NAME); } void Device::SetupFamilies(VkSurfaceKHR surface) { @@ -1540,53 +960,6 @@ void Device::SetupFamilies(VkSurfaceKHR surface) { } } -void Device::SetupFeatures() { - const VkPhysicalDeviceFeatures features{physical.GetFeatures()}; - is_depth_bounds_supported = features.depthBounds; - is_formatless_image_load_supported = features.shaderStorageImageReadWithoutFormat; - is_shader_float64_supported = features.shaderFloat64; - is_shader_int64_supported = features.shaderInt64; - is_shader_int16_supported = features.shaderInt16; - is_shader_storage_image_multisample = features.shaderStorageImageMultisample; - is_blit_depth_stencil_supported = TestDepthStencilBlits(); - is_optimal_astc_supported = IsOptimalAstcSupported(features); - - const VkPhysicalDeviceLimits& limits{properties.limits}; - max_vertex_input_attributes = limits.maxVertexInputAttributes; - max_vertex_input_bindings = limits.maxVertexInputBindings; -} - -void Device::SetupProperties() { - float_controls.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT_CONTROLS_PROPERTIES; - - VkPhysicalDeviceProperties2KHR properties2{}; - properties2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2; - properties2.pNext = &float_controls; - - physical.GetProperties2(properties2); -} - -void Device::CollectTelemetryParameters() { - VkPhysicalDeviceDriverProperties driver{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES, - .pNext = nullptr, - .driverID = {}, - .driverName = {}, - .driverInfo = {}, - .conformanceVersion = {}, - }; - - VkPhysicalDeviceProperties2 device_properties{ - .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2, - .pNext = &driver, - .properties = {}, - }; - physical.GetProperties2(device_properties); - - driver_id = driver.driverID; - vendor_name = driver.driverName; -} - u64 Device::GetDeviceMemoryUsage() const { VkPhysicalDeviceMemoryBudgetPropertiesEXT budget; budget.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_BUDGET_PROPERTIES_EXT; @@ -1602,7 +975,8 @@ u64 Device::GetDeviceMemoryUsage() const { void Device::CollectPhysicalMemoryInfo() { VkPhysicalDeviceMemoryBudgetPropertiesEXT budget{}; budget.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_BUDGET_PROPERTIES_EXT; - const auto mem_info = physical.GetMemoryProperties(ext_memory_budget ? &budget : nullptr); + const auto mem_info = + physical.GetMemoryProperties(extensions.memory_budget ? &budget : nullptr); const auto& mem_properties = mem_info.memoryProperties; const size_t num_properties = mem_properties.memoryHeapCount; device_access_memory = 0; @@ -1618,7 +992,7 @@ void Device::CollectPhysicalMemoryInfo() { if (is_heap_local) { local_memory += mem_properties.memoryHeaps[element].size; } - if (ext_memory_budget) { + if (extensions.memory_budget) { device_initial_usage += budget.heapUsage[element]; device_access_memory += budget.heapBudget[element]; continue; @@ -1634,7 +1008,7 @@ void Device::CollectPhysicalMemoryInfo() { } void Device::CollectToolingInfo() { - if (!ext_tooling_info) { + if (!extensions.tooling_info) { return; } auto tools{physical.GetPhysicalDeviceToolProperties()}; diff --git a/src/video_core/vulkan_common/vulkan_device.h b/src/video_core/vulkan_common/vulkan_device.h index 4bc267163..0662a2d9f 100644 --- a/src/video_core/vulkan_common/vulkan_device.h +++ b/src/video_core/vulkan_common/vulkan_device.h @@ -3,6 +3,7 @@ #pragma once +#include <set> #include <span> #include <string> #include <unordered_map> @@ -11,6 +12,156 @@ #include "common/common_types.h" #include "video_core/vulkan_common/vulkan_wrapper.h" +// Define all features which may be used by the implementation here. +// Vulkan version in the macro describes the minimum version required for feature availability. +// If the Vulkan version is lower than the required version, the named extension is required. +#define FOR_EACH_VK_FEATURE_1_1(FEATURE) \ + FEATURE(EXT, SubgroupSizeControl, SUBGROUP_SIZE_CONTROL, subgroup_size_control) \ + FEATURE(KHR, 16BitStorage, 16BIT_STORAGE, bit16_storage) \ + FEATURE(KHR, ShaderAtomicInt64, SHADER_ATOMIC_INT64, shader_atomic_int64) \ + FEATURE(KHR, ShaderDrawParameters, SHADER_DRAW_PARAMETERS, shader_draw_parameters) \ + FEATURE(KHR, ShaderFloat16Int8, SHADER_FLOAT16_INT8, shader_float16_int8) \ + FEATURE(KHR, UniformBufferStandardLayout, UNIFORM_BUFFER_STANDARD_LAYOUT, \ + uniform_buffer_standard_layout) \ + FEATURE(KHR, VariablePointer, VARIABLE_POINTERS, variable_pointer) + +#define FOR_EACH_VK_FEATURE_1_2(FEATURE) \ + FEATURE(EXT, HostQueryReset, HOST_QUERY_RESET, host_query_reset) \ + FEATURE(KHR, 8BitStorage, 8BIT_STORAGE, bit8_storage) \ + FEATURE(KHR, TimelineSemaphore, TIMELINE_SEMAPHORE, timeline_semaphore) + +#define FOR_EACH_VK_FEATURE_1_3(FEATURE) \ + FEATURE(EXT, ShaderDemoteToHelperInvocation, SHADER_DEMOTE_TO_HELPER_INVOCATION, \ + shader_demote_to_helper_invocation) + +// Define all features which may be used by the implementation and require an extension here. +#define FOR_EACH_VK_FEATURE_EXT(FEATURE) \ + FEATURE(EXT, CustomBorderColor, CUSTOM_BORDER_COLOR, custom_border_color) \ + FEATURE(EXT, DepthClipControl, DEPTH_CLIP_CONTROL, depth_clip_control) \ + FEATURE(EXT, ExtendedDynamicState, EXTENDED_DYNAMIC_STATE, extended_dynamic_state) \ + FEATURE(EXT, ExtendedDynamicState2, EXTENDED_DYNAMIC_STATE_2, extended_dynamic_state2) \ + FEATURE(EXT, ExtendedDynamicState3, EXTENDED_DYNAMIC_STATE_3, extended_dynamic_state3) \ + FEATURE(EXT, IndexTypeUint8, INDEX_TYPE_UINT8, index_type_uint8) \ + FEATURE(EXT, LineRasterization, LINE_RASTERIZATION, line_rasterization) \ + FEATURE(EXT, PrimitiveTopologyListRestart, PRIMITIVE_TOPOLOGY_LIST_RESTART, \ + primitive_topology_list_restart) \ + FEATURE(EXT, ProvokingVertex, PROVOKING_VERTEX, provoking_vertex) \ + FEATURE(EXT, Robustness2, ROBUSTNESS_2, robustness2) \ + FEATURE(EXT, TransformFeedback, TRANSFORM_FEEDBACK, transform_feedback) \ + FEATURE(EXT, VertexInputDynamicState, VERTEX_INPUT_DYNAMIC_STATE, vertex_input_dynamic_state) \ + FEATURE(KHR, PipelineExecutableProperties, PIPELINE_EXECUTABLE_PROPERTIES, \ + pipeline_executable_properties) \ + FEATURE(KHR, WorkgroupMemoryExplicitLayout, WORKGROUP_MEMORY_EXPLICIT_LAYOUT, \ + workgroup_memory_explicit_layout) + +// Define miscellaneous extensions which may be used by the implementation here. +#define FOR_EACH_VK_EXTENSION(EXTENSION) \ + EXTENSION(EXT, CONSERVATIVE_RASTERIZATION, conservative_rasterization) \ + EXTENSION(EXT, DEPTH_RANGE_UNRESTRICTED, depth_range_unrestricted) \ + EXTENSION(EXT, MEMORY_BUDGET, memory_budget) \ + EXTENSION(EXT, ROBUSTNESS_2, robustness_2) \ + EXTENSION(EXT, SAMPLER_FILTER_MINMAX, sampler_filter_minmax) \ + EXTENSION(EXT, SHADER_STENCIL_EXPORT, shader_stencil_export) \ + EXTENSION(EXT, SHADER_VIEWPORT_INDEX_LAYER, shader_viewport_index_layer) \ + EXTENSION(EXT, TOOLING_INFO, tooling_info) \ + EXTENSION(EXT, VERTEX_ATTRIBUTE_DIVISOR, vertex_attribute_divisor) \ + EXTENSION(KHR, DRAW_INDIRECT_COUNT, draw_indirect_count) \ + EXTENSION(KHR, DRIVER_PROPERTIES, driver_properties) \ + EXTENSION(KHR, EXTERNAL_MEMORY_FD, external_memory_fd) \ + EXTENSION(KHR, PUSH_DESCRIPTOR, push_descriptor) \ + EXTENSION(KHR, SAMPLER_MIRROR_CLAMP_TO_EDGE, sampler_mirror_clamp_to_edge) \ + EXTENSION(KHR, SHADER_FLOAT_CONTROLS, shader_float_controls) \ + EXTENSION(KHR, SPIRV_1_4, spirv_1_4) \ + EXTENSION(KHR, SWAPCHAIN, swapchain) \ + EXTENSION(KHR, SWAPCHAIN_MUTABLE_FORMAT, swapchain_mutable_format) \ + EXTENSION(NV, DEVICE_DIAGNOSTICS_CONFIG, device_diagnostics_config) \ + EXTENSION(NV, GEOMETRY_SHADER_PASSTHROUGH, geometry_shader_passthrough) \ + EXTENSION(NV, VIEWPORT_ARRAY2, viewport_array2) \ + EXTENSION(NV, VIEWPORT_SWIZZLE, viewport_swizzle) + +#define FOR_EACH_VK_EXTENSION_WIN32(EXTENSION) \ + EXTENSION(KHR, EXTERNAL_MEMORY_WIN32, external_memory_win32) + +// Define extensions which must be supported. +#define FOR_EACH_VK_MANDATORY_EXTENSION(EXTENSION_NAME) \ + EXTENSION_NAME(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME) \ + EXTENSION_NAME(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME) \ + EXTENSION_NAME(VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME) \ + EXTENSION_NAME(VK_KHR_SAMPLER_MIRROR_CLAMP_TO_EDGE_EXTENSION_NAME) \ + EXTENSION_NAME(VK_KHR_SHADER_FLOAT_CONTROLS_EXTENSION_NAME) + +#define FOR_EACH_VK_MANDATORY_EXTENSION_GENERIC(EXTENSION_NAME) \ + EXTENSION_NAME(VK_KHR_EXTERNAL_MEMORY_FD_EXTENSION_NAME) + +#define FOR_EACH_VK_MANDATORY_EXTENSION_WIN32(EXTENSION_NAME) \ + EXTENSION_NAME(VK_KHR_EXTERNAL_MEMORY_WIN32_EXTENSION_NAME) + +// Define extensions where the absence of the extension may result in a degraded experience. +#define FOR_EACH_VK_RECOMMENDED_EXTENSION(EXTENSION_NAME) \ + EXTENSION_NAME(VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME) \ + EXTENSION_NAME(VK_EXT_DEPTH_RANGE_UNRESTRICTED_EXTENSION_NAME) \ + EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME) \ + EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME) \ + EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME) \ + EXTENSION_NAME(VK_EXT_LINE_RASTERIZATION_EXTENSION_NAME) \ + EXTENSION_NAME(VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME) \ + EXTENSION_NAME(VK_NV_GEOMETRY_SHADER_PASSTHROUGH_EXTENSION_NAME) \ + EXTENSION_NAME(VK_NV_VIEWPORT_ARRAY2_EXTENSION_NAME) \ + EXTENSION_NAME(VK_NV_VIEWPORT_SWIZZLE_EXTENSION_NAME) + +// Define features which must be supported. +#define FOR_EACH_VK_MANDATORY_FEATURE(FEATURE_NAME) \ + FEATURE_NAME(bit16_storage, storageBuffer16BitAccess) \ + FEATURE_NAME(bit16_storage, uniformAndStorageBuffer16BitAccess) \ + FEATURE_NAME(bit8_storage, storageBuffer8BitAccess) \ + FEATURE_NAME(bit8_storage, uniformAndStorageBuffer8BitAccess) \ + FEATURE_NAME(features, depthBiasClamp) \ + FEATURE_NAME(features, depthClamp) \ + FEATURE_NAME(features, drawIndirectFirstInstance) \ + FEATURE_NAME(features, dualSrcBlend) \ + FEATURE_NAME(features, fillModeNonSolid) \ + FEATURE_NAME(features, fragmentStoresAndAtomics) \ + FEATURE_NAME(features, geometryShader) \ + FEATURE_NAME(features, imageCubeArray) \ + FEATURE_NAME(features, independentBlend) \ + FEATURE_NAME(features, largePoints) \ + FEATURE_NAME(features, logicOp) \ + FEATURE_NAME(features, multiDrawIndirect) \ + FEATURE_NAME(features, multiViewport) \ + FEATURE_NAME(features, occlusionQueryPrecise) \ + FEATURE_NAME(features, robustBufferAccess) \ + FEATURE_NAME(features, samplerAnisotropy) \ + FEATURE_NAME(features, sampleRateShading) \ + FEATURE_NAME(features, shaderClipDistance) \ + FEATURE_NAME(features, shaderCullDistance) \ + FEATURE_NAME(features, shaderImageGatherExtended) \ + FEATURE_NAME(features, shaderStorageImageWriteWithoutFormat) \ + FEATURE_NAME(features, tessellationShader) \ + FEATURE_NAME(features, vertexPipelineStoresAndAtomics) \ + FEATURE_NAME(features, wideLines) \ + FEATURE_NAME(host_query_reset, hostQueryReset) \ + FEATURE_NAME(robustness2, nullDescriptor) \ + FEATURE_NAME(robustness2, robustBufferAccess2) \ + FEATURE_NAME(robustness2, robustImageAccess2) \ + FEATURE_NAME(shader_demote_to_helper_invocation, shaderDemoteToHelperInvocation) \ + FEATURE_NAME(shader_draw_parameters, shaderDrawParameters) \ + FEATURE_NAME(timeline_semaphore, timelineSemaphore) \ + FEATURE_NAME(variable_pointer, variablePointers) \ + FEATURE_NAME(variable_pointer, variablePointersStorageBuffer) + +// Define features where the absence of the feature may result in a degraded experience. +#define FOR_EACH_VK_RECOMMENDED_FEATURE(FEATURE_NAME) \ + FEATURE_NAME(custom_border_color, customBorderColors) \ + FEATURE_NAME(extended_dynamic_state, extendedDynamicState) \ + FEATURE_NAME(index_type_uint8, indexTypeUint8) \ + FEATURE_NAME(primitive_topology_list_restart, primitiveTopologyListRestart) \ + FEATURE_NAME(provoking_vertex, provokingVertexLast) \ + FEATURE_NAME(shader_float16_int8, shaderFloat16) \ + FEATURE_NAME(shader_float16_int8, shaderInt8) \ + FEATURE_NAME(transform_feedback, transformFeedback) \ + FEATURE_NAME(uniform_buffer_standard_layout, uniformBufferStandardLayout) \ + FEATURE_NAME(vertex_input_dynamic_state, vertexInputDynamicState) + namespace Vulkan { class NsightAftermathTracker; @@ -88,69 +239,69 @@ public: /// Returns the current Vulkan API version provided in Vulkan-formatted version numbers. u32 ApiVersion() const { - return properties.apiVersion; + return properties.properties.apiVersion; } /// Returns the current driver version provided in Vulkan-formatted version numbers. u32 GetDriverVersion() const { - return properties.driverVersion; + return properties.properties.driverVersion; } /// Returns the device name. std::string_view GetModelName() const { - return properties.deviceName; + return properties.properties.deviceName; } /// Returns the driver ID. VkDriverIdKHR GetDriverID() const { - return driver_id; + return properties.driver.driverID; } bool ShouldBoostClocks() const; /// Returns uniform buffer alignment requeriment. VkDeviceSize GetUniformBufferAlignment() const { - return properties.limits.minUniformBufferOffsetAlignment; + return properties.properties.limits.minUniformBufferOffsetAlignment; } /// Returns storage alignment requeriment. VkDeviceSize GetStorageBufferAlignment() const { - return properties.limits.minStorageBufferOffsetAlignment; + return properties.properties.limits.minStorageBufferOffsetAlignment; } /// Returns the maximum range for storage buffers. VkDeviceSize GetMaxStorageBufferRange() const { - return properties.limits.maxStorageBufferRange; + return properties.properties.limits.maxStorageBufferRange; } /// Returns the maximum size for push constants. VkDeviceSize GetMaxPushConstantsSize() const { - return properties.limits.maxPushConstantsSize; + return properties.properties.limits.maxPushConstantsSize; } /// Returns the maximum size for shared memory. u32 GetMaxComputeSharedMemorySize() const { - return properties.limits.maxComputeSharedMemorySize; + return properties.properties.limits.maxComputeSharedMemorySize; } /// Returns float control properties of the device. const VkPhysicalDeviceFloatControlsPropertiesKHR& FloatControlProperties() const { - return float_controls; + return properties.float_controls; } /// Returns true if ASTC is natively supported. bool IsOptimalAstcSupported() const { - return is_optimal_astc_supported; + return features.features.textureCompressionASTC_LDR; } /// Returns true if the device supports float16 natively. bool IsFloat16Supported() const { - return is_float16_supported; + return features.shader_float16_int8.shaderFloat16; } /// Returns true if the device supports int8 natively. bool IsInt8Supported() const { - return is_int8_supported; + return features.shader_float16_int8.shaderInt8; } /// Returns true if the device warp size can potentially be bigger than guest's warp size. @@ -160,32 +311,32 @@ public: /// Returns true if the device can be forced to use the guest warp size. bool IsGuestWarpSizeSupported(VkShaderStageFlagBits stage) const { - return guest_warp_stages & stage; + return properties.subgroup_size_control.requiredSubgroupSizeStages & stage; } /// Returns the maximum number of push descriptors. u32 MaxPushDescriptors() const { - return max_push_descriptors; + return properties.push_descriptor.maxPushDescriptors; } /// Returns true if formatless image load is supported. bool IsFormatlessImageLoadSupported() const { - return is_formatless_image_load_supported; + return features.features.shaderStorageImageReadWithoutFormat; } /// Returns true if shader int64 is supported. bool IsShaderInt64Supported() const { - return is_shader_int64_supported; + return features.features.shaderInt64; } /// Returns true if shader int16 is supported. bool IsShaderInt16Supported() const { - return is_shader_int16_supported; + return features.features.shaderInt16; } // Returns true if depth bounds is supported. bool IsDepthBoundsSupported() const { - return is_depth_bounds_supported; + return features.features.depthBounds; } /// Returns true when blitting from and to depth stencil images is supported. @@ -195,151 +346,151 @@ public: /// Returns true if the device supports VK_NV_viewport_swizzle. bool IsNvViewportSwizzleSupported() const { - return nv_viewport_swizzle; + return extensions.viewport_swizzle; } /// Returns true if the device supports VK_NV_viewport_array2. bool IsNvViewportArray2Supported() const { - return nv_viewport_array2; + return extensions.viewport_array2; } /// Returns true if the device supports VK_NV_geometry_shader_passthrough. bool IsNvGeometryShaderPassthroughSupported() const { - return nv_geometry_shader_passthrough; + return extensions.geometry_shader_passthrough; } /// Returns true if the device supports VK_KHR_uniform_buffer_standard_layout. bool IsKhrUniformBufferStandardLayoutSupported() const { - return khr_uniform_buffer_standard_layout; + return extensions.uniform_buffer_standard_layout; } /// Returns true if the device supports VK_KHR_push_descriptor. bool IsKhrPushDescriptorSupported() const { - return khr_push_descriptor; + return extensions.push_descriptor; } /// Returns true if VK_KHR_pipeline_executable_properties is enabled. bool IsKhrPipelineExecutablePropertiesEnabled() const { - return khr_pipeline_executable_properties; + return extensions.pipeline_executable_properties; } /// Returns true if VK_KHR_swapchain_mutable_format is enabled. bool IsKhrSwapchainMutableFormatEnabled() const { - return khr_swapchain_mutable_format; + return extensions.swapchain_mutable_format; } /// Returns true if the device supports VK_KHR_workgroup_memory_explicit_layout. bool IsKhrWorkgroupMemoryExplicitLayoutSupported() const { - return khr_workgroup_memory_explicit_layout; + return extensions.workgroup_memory_explicit_layout; } /// Returns true if the device supports VK_EXT_primitive_topology_list_restart. bool IsTopologyListPrimitiveRestartSupported() const { - return is_topology_list_restart_supported; + return features.primitive_topology_list_restart.primitiveTopologyListRestart; } /// Returns true if the device supports VK_EXT_primitive_topology_list_restart. bool IsPatchListPrimitiveRestartSupported() const { - return is_patch_list_restart_supported; + return features.primitive_topology_list_restart.primitiveTopologyPatchListRestart; } /// Returns true if the device supports VK_EXT_index_type_uint8. bool IsExtIndexTypeUint8Supported() const { - return ext_index_type_uint8; + return extensions.index_type_uint8; } /// Returns true if the device supports VK_EXT_sampler_filter_minmax. bool IsExtSamplerFilterMinmaxSupported() const { - return ext_sampler_filter_minmax; + return extensions.sampler_filter_minmax; } /// Returns true if the device supports VK_EXT_depth_range_unrestricted. bool IsExtDepthRangeUnrestrictedSupported() const { - return ext_depth_range_unrestricted; + return extensions.depth_range_unrestricted; } /// Returns true if the device supports VK_EXT_depth_clip_control. bool IsExtDepthClipControlSupported() const { - return ext_depth_clip_control; + return extensions.depth_clip_control; } /// Returns true if the device supports VK_EXT_shader_viewport_index_layer. bool IsExtShaderViewportIndexLayerSupported() const { - return ext_shader_viewport_index_layer; + return extensions.shader_viewport_index_layer; } /// Returns true if the device supports VK_EXT_subgroup_size_control. bool IsExtSubgroupSizeControlSupported() const { - return ext_subgroup_size_control; + return extensions.subgroup_size_control; } /// Returns true if the device supports VK_EXT_transform_feedback. bool IsExtTransformFeedbackSupported() const { - return ext_transform_feedback; + return extensions.transform_feedback; } /// Returns true if the device supports VK_EXT_custom_border_color. bool IsExtCustomBorderColorSupported() const { - return ext_custom_border_color; + return extensions.custom_border_color; } /// Returns true if the device supports VK_EXT_extended_dynamic_state. bool IsExtExtendedDynamicStateSupported() const { - return ext_extended_dynamic_state; + return extensions.extended_dynamic_state; } /// Returns true if the device supports VK_EXT_extended_dynamic_state2. bool IsExtExtendedDynamicState2Supported() const { - return ext_extended_dynamic_state_2; + return extensions.extended_dynamic_state2; } bool IsExtExtendedDynamicState2ExtrasSupported() const { - return ext_extended_dynamic_state_2_extra; + return features.extended_dynamic_state2.extendedDynamicState2LogicOp; } /// Returns true if the device supports VK_EXT_extended_dynamic_state3. bool IsExtExtendedDynamicState3Supported() const { - return ext_extended_dynamic_state_3; + return extensions.extended_dynamic_state3; } /// Returns true if the device supports VK_EXT_extended_dynamic_state3. bool IsExtExtendedDynamicState3BlendingSupported() const { - return ext_extended_dynamic_state_3_blend; + return dynamic_state3_blending; } /// Returns true if the device supports VK_EXT_extended_dynamic_state3. bool IsExtExtendedDynamicState3EnablesSupported() const { - return ext_extended_dynamic_state_3_enables; + return dynamic_state3_enables; } /// Returns true if the device supports VK_EXT_line_rasterization. bool IsExtLineRasterizationSupported() const { - return ext_line_rasterization; + return extensions.line_rasterization; } /// Returns true if the device supports VK_EXT_vertex_input_dynamic_state. bool IsExtVertexInputDynamicStateSupported() const { - return ext_vertex_input_dynamic_state; + return extensions.vertex_input_dynamic_state; } /// Returns true if the device supports VK_EXT_shader_stencil_export. bool IsExtShaderStencilExportSupported() const { - return ext_shader_stencil_export; + return extensions.shader_stencil_export; } /// Returns true if the device supports VK_EXT_conservative_rasterization. bool IsExtConservativeRasterizationSupported() const { - return ext_conservative_rasterization; + return extensions.conservative_rasterization; } /// Returns true if the device supports VK_EXT_provoking_vertex. bool IsExtProvokingVertexSupported() const { - return ext_provoking_vertex; + return extensions.provoking_vertex; } /// Returns true if the device supports VK_KHR_shader_atomic_int64. bool IsExtShaderAtomicInt64Supported() const { - return ext_shader_atomic_int64; + return extensions.shader_atomic_int64; } /// Returns the minimum supported version of SPIR-V. @@ -347,7 +498,7 @@ public: if (instance_version >= VK_API_VERSION_1_3) { return 0x00010600U; } - if (khr_spirv_1_4) { + if (extensions.spirv_1_4) { return 0x00010400U; } return 0x00010000U; @@ -365,11 +516,11 @@ public: /// Returns the vendor name reported from Vulkan. std::string_view GetVendorName() const { - return vendor_name; + return properties.driver.driverName; } /// Returns the list of available extensions. - const std::vector<std::string>& GetAvailableExtensions() const { + const std::set<std::string, std::less<>>& GetAvailableExtensions() const { return supported_extensions; } @@ -378,7 +529,7 @@ public: } bool CanReportMemoryUsage() const { - return ext_memory_budget; + return extensions.memory_budget; } u64 GetDeviceMemoryUsage() const; @@ -400,36 +551,29 @@ public: } bool HasNullDescriptor() const { - return has_null_descriptor; + return features.robustness2.nullDescriptor; } u32 GetMaxVertexInputAttributes() const { - return max_vertex_input_attributes; + return properties.properties.limits.maxVertexInputAttributes; } u32 GetMaxVertexInputBindings() const { - return max_vertex_input_bindings; + return properties.properties.limits.maxVertexInputBindings; } private: - /// Checks if the physical device is suitable. - void CheckSuitability(bool requires_swapchain) const; + /// Checks if the physical device is suitable and configures the object state + /// with all necessary info about its properties. + bool GetSuitability(bool requires_swapchain); - /// Loads extensions into a vector and stores available ones in this object. - std::vector<const char*> LoadExtensions(bool requires_surface); + // Remove extensions which have incomplete feature support. + void RemoveUnsuitableExtensions(); + void RemoveExtensionIfUnsuitable(bool is_suitable, const std::string& extension_name); /// Sets up queue families. void SetupFamilies(VkSurfaceKHR surface); - /// Sets up device features. - void SetupFeatures(); - - /// Sets up device properties. - void SetupProperties(); - - /// Collects telemetry information from the device. - void CollectTelemetryParameters(); - /// Collects information about attached tools. void CollectToolingInfo(); @@ -440,91 +584,93 @@ private: std::vector<VkDeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const; /// Returns true if ASTC textures are natively supported. - bool IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const; + bool ComputeIsOptimalAstcSupported() const; /// Returns true if the device natively supports blitting depth stencil images. bool TestDepthStencilBlits() const; - VkInstance instance; ///< Vulkan instance. - vk::DeviceDispatch dld; ///< Device function pointers. - vk::PhysicalDevice physical; ///< Physical device. - VkPhysicalDeviceProperties properties; ///< Device properties. - VkPhysicalDeviceFloatControlsPropertiesKHR float_controls{}; ///< Float control properties. - vk::Device logical; ///< Logical device. - vk::Queue graphics_queue; ///< Main graphics queue. - vk::Queue present_queue; ///< Main present queue. - u32 instance_version{}; ///< Vulkan onstance version. - u32 graphics_family{}; ///< Main graphics queue family index. - u32 present_family{}; ///< Main present queue family index. - VkDriverIdKHR driver_id{}; ///< Driver ID. - VkShaderStageFlags guest_warp_stages{}; ///< Stages where the guest warp size can be forced. - u64 device_access_memory{}; ///< Total size of device local memory in bytes. - u32 max_push_descriptors{}; ///< Maximum number of push descriptors - u32 sets_per_pool{}; ///< Sets per Description Pool - bool is_optimal_astc_supported{}; ///< Support for native ASTC. - bool is_float16_supported{}; ///< Support for float16 arithmetic. - bool is_int8_supported{}; ///< Support for int8 arithmetic. - 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 is_depth_bounds_supported{}; ///< Support for depth bounds. - bool is_shader_float64_supported{}; ///< Support for float64. - bool is_shader_int64_supported{}; ///< Support for int64. - bool is_shader_int16_supported{}; ///< Support for int16. - bool is_shader_storage_image_multisample{}; ///< Support for image operations on MSAA images. - bool is_blit_depth_stencil_supported{}; ///< Support for blitting from and to depth stencil. - bool is_topology_list_restart_supported{}; ///< Support for primitive restart with list - ///< topologies. - bool is_patch_list_restart_supported{}; ///< Support for primitive restart with list patch. - bool is_integrated{}; ///< Is GPU an iGPU. - bool is_virtual{}; ///< Is GPU a virtual GPU. - bool is_non_gpu{}; ///< Is SoftwareRasterizer, FPGA, non-GPU device. - bool nv_viewport_swizzle{}; ///< Support for VK_NV_viewport_swizzle. - bool nv_viewport_array2{}; ///< Support for VK_NV_viewport_array2. - bool nv_geometry_shader_passthrough{}; ///< Support for VK_NV_geometry_shader_passthrough. - bool khr_draw_indirect_count{}; ///< Support for VK_KHR_draw_indirect_count. - bool khr_uniform_buffer_standard_layout{}; ///< Support for scalar uniform buffer layouts. - bool khr_spirv_1_4{}; ///< Support for VK_KHR_spirv_1_4. - bool khr_workgroup_memory_explicit_layout{}; ///< Support for explicit workgroup layouts. - bool khr_push_descriptor{}; ///< Support for VK_KHR_push_descritor. - bool khr_pipeline_executable_properties{}; ///< Support for executable properties. - bool khr_swapchain_mutable_format{}; ///< Support for VK_KHR_swapchain_mutable_format. - bool ext_index_type_uint8{}; ///< Support for VK_EXT_index_type_uint8. - bool ext_sampler_filter_minmax{}; ///< Support for VK_EXT_sampler_filter_minmax. - bool ext_depth_clip_control{}; ///< Support for VK_EXT_depth_clip_control - 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_tooling_info{}; ///< Support for VK_EXT_tooling_info. - bool ext_subgroup_size_control{}; ///< Support for VK_EXT_subgroup_size_control. - bool ext_transform_feedback{}; ///< Support for VK_EXT_transform_feedback. - bool ext_custom_border_color{}; ///< Support for VK_EXT_custom_border_color. - bool ext_extended_dynamic_state{}; ///< Support for VK_EXT_extended_dynamic_state. - bool ext_extended_dynamic_state_2{}; ///< Support for VK_EXT_extended_dynamic_state2. - bool ext_extended_dynamic_state_2_extra{}; ///< Support for VK_EXT_extended_dynamic_state2. - bool ext_extended_dynamic_state_3{}; ///< Support for VK_EXT_extended_dynamic_state3. - bool ext_extended_dynamic_state_3_blend{}; ///< Support for VK_EXT_extended_dynamic_state3. - bool ext_extended_dynamic_state_3_enables{}; ///< Support for VK_EXT_extended_dynamic_state3. - bool ext_line_rasterization{}; ///< Support for VK_EXT_line_rasterization. - bool ext_vertex_input_dynamic_state{}; ///< Support for VK_EXT_vertex_input_dynamic_state. - bool ext_shader_stencil_export{}; ///< Support for VK_EXT_shader_stencil_export. - bool ext_shader_atomic_int64{}; ///< Support for VK_KHR_shader_atomic_int64. - bool ext_conservative_rasterization{}; ///< Support for VK_EXT_conservative_rasterization. - bool ext_provoking_vertex{}; ///< Support for VK_EXT_provoking_vertex. - bool ext_memory_budget{}; ///< Support for VK_EXT_memory_budget. - bool nv_device_diagnostics_config{}; ///< Support for VK_NV_device_diagnostics_config. - bool has_broken_cube_compatibility{}; ///< Has broken cube compatiblity bit - bool has_renderdoc{}; ///< Has RenderDoc attached - bool has_nsight_graphics{}; ///< Has Nsight Graphics attached - bool supports_d24_depth{}; ///< Supports D24 depth buffers. - bool cant_blit_msaa{}; ///< Does not support MSAA<->MSAA blitting. - bool must_emulate_bgr565{}; ///< Emulates BGR565 by swizzling RGB565 format. - bool has_null_descriptor{}; ///< Has support for null descriptors. - u32 max_vertex_input_attributes{}; ///< Max vertex input attributes in pipeline - u32 max_vertex_input_bindings{}; ///< Max vertex input buffers in pipeline +private: + VkInstance instance; ///< Vulkan instance. + vk::DeviceDispatch dld; ///< Device function pointers. + vk::PhysicalDevice physical; ///< Physical device. + vk::Device logical; ///< Logical device. + vk::Queue graphics_queue; ///< Main graphics queue. + vk::Queue present_queue; ///< Main present queue. + u32 instance_version{}; ///< Vulkan instance version. + u32 graphics_family{}; ///< Main graphics queue family index. + u32 present_family{}; ///< Main present queue family index. + + struct Extensions { +#define EXTENSION(prefix, macro_name, var_name) bool var_name{}; +#define FEATURE(prefix, struct_name, macro_name, var_name) bool var_name{}; + + FOR_EACH_VK_FEATURE_1_1(FEATURE); + FOR_EACH_VK_FEATURE_1_2(FEATURE); + FOR_EACH_VK_FEATURE_1_3(FEATURE); + FOR_EACH_VK_FEATURE_EXT(FEATURE); + FOR_EACH_VK_EXTENSION(EXTENSION); + FOR_EACH_VK_EXTENSION_WIN32(EXTENSION); + +#undef EXTENSION +#undef FEATURE + }; + + struct Features { +#define FEATURE_CORE(prefix, struct_name, macro_name, var_name) \ + VkPhysicalDevice##struct_name##Features var_name{}; +#define FEATURE_EXT(prefix, struct_name, macro_name, var_name) \ + VkPhysicalDevice##struct_name##Features##prefix var_name{}; + + FOR_EACH_VK_FEATURE_1_1(FEATURE_CORE); + FOR_EACH_VK_FEATURE_1_2(FEATURE_CORE); + FOR_EACH_VK_FEATURE_1_3(FEATURE_CORE); + FOR_EACH_VK_FEATURE_EXT(FEATURE_EXT); + +#undef FEATURE_CORE +#undef FEATURE_EXT + + VkPhysicalDeviceFeatures features{}; + }; + + struct Properties { + VkPhysicalDeviceDriverProperties driver{}; + VkPhysicalDeviceFloatControlsProperties float_controls{}; + VkPhysicalDevicePushDescriptorPropertiesKHR push_descriptor{}; + VkPhysicalDeviceSubgroupSizeControlProperties subgroup_size_control{}; + VkPhysicalDeviceTransformFeedbackPropertiesEXT transform_feedback{}; + + VkPhysicalDeviceProperties properties{}; + }; + + Extensions extensions{}; + Features features{}; + Properties properties{}; + + VkPhysicalDeviceFeatures2 features2{}; + VkPhysicalDeviceProperties2 properties2{}; + + // Misc features + bool is_optimal_astc_supported{}; ///< Support for all guest ASTC formats. + bool is_blit_depth_stencil_supported{}; ///< Support for blitting from and to depth stencil. + bool is_warp_potentially_bigger{}; ///< Host warp size can be bigger than guest. + bool is_integrated{}; ///< Is GPU an iGPU. + bool is_virtual{}; ///< Is GPU a virtual GPU. + bool is_non_gpu{}; ///< Is SoftwareRasterizer, FPGA, non-GPU device. + bool has_broken_cube_compatibility{}; ///< Has broken cube compatiblity bit + bool has_renderdoc{}; ///< Has RenderDoc attached + bool has_nsight_graphics{}; ///< Has Nsight Graphics attached + bool supports_d24_depth{}; ///< Supports D24 depth buffers. + bool cant_blit_msaa{}; ///< Does not support MSAA<->MSAA blitting. + bool must_emulate_bgr565{}; ///< Emulates BGR565 by swizzling RGB565 format. + bool dynamic_state3_blending{}; ///< Has all blending features of dynamic_state3. + bool dynamic_state3_enables{}; ///< Has all enables features of dynamic_state3. + u64 device_access_memory{}; ///< Total size of device local memory in bytes. + u32 sets_per_pool{}; ///< Sets per Description Pool // Telemetry parameters - std::string vendor_name; ///< Device's driver name. - std::vector<std::string> supported_extensions; ///< Reported Vulkan extensions. - std::vector<size_t> valid_heap_memory; ///< Heaps used. + std::set<std::string, std::less<>> supported_extensions; ///< Reported Vulkan extensions. + std::set<std::string, std::less<>> loaded_extensions; ///< Loaded Vulkan extensions. + std::vector<size_t> valid_heap_memory; ///< Heaps used. /// Format properties dictionary. std::unordered_map<VkFormat, VkFormatProperties> format_properties; diff --git a/src/video_core/vulkan_common/vulkan_wrapper.cpp b/src/video_core/vulkan_common/vulkan_wrapper.cpp index 61be1fce1..486d4dfaf 100644 --- a/src/video_core/vulkan_common/vulkan_wrapper.cpp +++ b/src/video_core/vulkan_common/vulkan_wrapper.cpp @@ -96,8 +96,8 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept { X(vkCmdDrawIndexed); X(vkCmdDrawIndirect); X(vkCmdDrawIndexedIndirect); - X(vkCmdDrawIndirectCountKHR); - X(vkCmdDrawIndexedIndirectCountKHR); + X(vkCmdDrawIndirectCount); + X(vkCmdDrawIndexedIndirectCount); X(vkCmdEndQuery); X(vkCmdEndRenderPass); X(vkCmdEndTransformFeedbackEXT); @@ -221,6 +221,12 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept { if (!dld.vkResetQueryPool) { Proc(dld.vkResetQueryPool, dld, "vkResetQueryPoolEXT", device); } + + // Support for draw indirect with count is optional in Vulkan 1.2 + if (!dld.vkCmdDrawIndirectCount) { + Proc(dld.vkCmdDrawIndirectCount, dld, "vkCmdDrawIndirectCountKHR", device); + Proc(dld.vkCmdDrawIndexedIndirectCount, dld, "vkCmdDrawIndexedIndirectCountKHR", device); + } #undef X } diff --git a/src/video_core/vulkan_common/vulkan_wrapper.h b/src/video_core/vulkan_common/vulkan_wrapper.h index 412779b51..e86f661cb 100644 --- a/src/video_core/vulkan_common/vulkan_wrapper.h +++ b/src/video_core/vulkan_common/vulkan_wrapper.h @@ -215,8 +215,8 @@ struct DeviceDispatch : InstanceDispatch { PFN_vkCmdDrawIndexed vkCmdDrawIndexed{}; PFN_vkCmdDrawIndirect vkCmdDrawIndirect{}; PFN_vkCmdDrawIndexedIndirect vkCmdDrawIndexedIndirect{}; - PFN_vkCmdDrawIndirectCountKHR vkCmdDrawIndirectCountKHR{}; - PFN_vkCmdDrawIndexedIndirectCountKHR vkCmdDrawIndexedIndirectCountKHR{}; + PFN_vkCmdDrawIndirectCount vkCmdDrawIndirectCount{}; + PFN_vkCmdDrawIndexedIndirectCount vkCmdDrawIndexedIndirectCount{}; PFN_vkCmdEndDebugUtilsLabelEXT vkCmdEndDebugUtilsLabelEXT{}; PFN_vkCmdEndQuery vkCmdEndQuery{}; PFN_vkCmdEndRenderPass vkCmdEndRenderPass{}; @@ -1065,15 +1065,15 @@ public: void DrawIndirectCount(VkBuffer src_buffer, VkDeviceSize src_offset, VkBuffer count_buffer, VkDeviceSize count_offset, u32 draw_count, u32 stride) const noexcept { - dld->vkCmdDrawIndirectCountKHR(handle, src_buffer, src_offset, count_buffer, count_offset, - draw_count, stride); + dld->vkCmdDrawIndirectCount(handle, src_buffer, src_offset, count_buffer, count_offset, + draw_count, stride); } void DrawIndexedIndirectCount(VkBuffer src_buffer, VkDeviceSize src_offset, VkBuffer count_buffer, VkDeviceSize count_offset, u32 draw_count, u32 stride) const noexcept { - dld->vkCmdDrawIndexedIndirectCountKHR(handle, src_buffer, src_offset, count_buffer, - count_offset, draw_count, stride); + dld->vkCmdDrawIndexedIndirectCount(handle, src_buffer, src_offset, count_buffer, + count_offset, draw_count, stride); } void ClearAttachments(Span<VkClearAttachment> attachments, |