diff options
Diffstat (limited to 'src/video_core')
100 files changed, 4741 insertions, 3232 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index effe76a63..258d58eba 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -148,6 +148,7 @@ add_library(video_core STATIC textures/convert.h textures/decoders.cpp textures/decoders.h + textures/texture.cpp textures/texture.h video_core.cpp video_core.h @@ -155,7 +156,6 @@ add_library(video_core STATIC if (ENABLE_VULKAN) target_sources(video_core PRIVATE - renderer_vulkan/declarations.h renderer_vulkan/fixed_pipeline_state.cpp renderer_vulkan/fixed_pipeline_state.h renderer_vulkan/maxwell_to_vk.cpp diff --git a/src/video_core/buffer_cache/buffer_block.h b/src/video_core/buffer_cache/buffer_block.h index 4b9193182..e35ee0b67 100644 --- a/src/video_core/buffer_cache/buffer_block.h +++ b/src/video_core/buffer_cache/buffer_block.h @@ -15,37 +15,29 @@ namespace VideoCommon { class BufferBlock { public: - bool Overlaps(const CacheAddr start, const CacheAddr end) const { - return (cache_addr < end) && (cache_addr_end > start); + bool Overlaps(const VAddr start, const VAddr end) const { + return (cpu_addr < end) && (cpu_addr_end > start); } - bool IsInside(const CacheAddr other_start, const CacheAddr other_end) const { - return cache_addr <= other_start && other_end <= cache_addr_end; + bool IsInside(const VAddr other_start, const VAddr other_end) const { + return cpu_addr <= other_start && other_end <= cpu_addr_end; } - u8* GetWritableHostPtr() const { - return FromCacheAddr(cache_addr); + std::size_t GetOffset(const VAddr in_addr) { + return static_cast<std::size_t>(in_addr - cpu_addr); } - u8* GetWritableHostPtr(std::size_t offset) const { - return FromCacheAddr(cache_addr + offset); + VAddr GetCpuAddr() const { + return cpu_addr; } - std::size_t GetOffset(const CacheAddr in_addr) { - return static_cast<std::size_t>(in_addr - cache_addr); + VAddr GetCpuAddrEnd() const { + return cpu_addr_end; } - CacheAddr GetCacheAddr() const { - return cache_addr; - } - - CacheAddr GetCacheAddrEnd() const { - return cache_addr_end; - } - - void SetCacheAddr(const CacheAddr new_addr) { - cache_addr = new_addr; - cache_addr_end = new_addr + size; + void SetCpuAddr(const VAddr new_addr) { + cpu_addr = new_addr; + cpu_addr_end = new_addr + size; } std::size_t GetSize() const { @@ -61,14 +53,14 @@ public: } protected: - explicit BufferBlock(CacheAddr cache_addr, const std::size_t size) : size{size} { - SetCacheAddr(cache_addr); + explicit BufferBlock(VAddr cpu_addr, const std::size_t size) : size{size} { + SetCpuAddr(cpu_addr); } ~BufferBlock() = default; private: - CacheAddr cache_addr{}; - CacheAddr cache_addr_end{}; + VAddr cpu_addr{}; + VAddr cpu_addr_end{}; std::size_t size{}; u64 epoch{}; }; diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index 186aca61d..b57c0d4d4 100644 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -19,6 +19,7 @@ #include "common/alignment.h" #include "common/common_types.h" #include "core/core.h" +#include "core/memory.h" #include "video_core/buffer_cache/buffer_block.h" #include "video_core/buffer_cache/map_interval.h" #include "video_core/memory_manager.h" @@ -37,28 +38,45 @@ public: bool is_written = false, bool use_fast_cbuf = false) { std::lock_guard lock{mutex}; - auto& memory_manager = system.GPU().MemoryManager(); - const auto host_ptr = memory_manager.GetPointer(gpu_addr); - if (!host_ptr) { + const std::optional<VAddr> cpu_addr_opt = + system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr); + + if (!cpu_addr_opt) { return {GetEmptyBuffer(size), 0}; } - const auto cache_addr = ToCacheAddr(host_ptr); + + VAddr cpu_addr = *cpu_addr_opt; // Cache management is a big overhead, so only cache entries with a given size. // TODO: Figure out which size is the best for given games. constexpr std::size_t max_stream_size = 0x800; if (use_fast_cbuf || size < max_stream_size) { - if (!is_written && !IsRegionWritten(cache_addr, cache_addr + size - 1)) { + if (!is_written && !IsRegionWritten(cpu_addr, cpu_addr + size - 1)) { + auto& memory_manager = system.GPU().MemoryManager(); if (use_fast_cbuf) { - return ConstBufferUpload(host_ptr, size); + if (memory_manager.IsGranularRange(gpu_addr, size)) { + const auto host_ptr = memory_manager.GetPointer(gpu_addr); + return ConstBufferUpload(host_ptr, size); + } else { + staging_buffer.resize(size); + memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size); + return ConstBufferUpload(staging_buffer.data(), size); + } } else { - return StreamBufferUpload(host_ptr, size, alignment); + if (memory_manager.IsGranularRange(gpu_addr, size)) { + const auto host_ptr = memory_manager.GetPointer(gpu_addr); + return StreamBufferUpload(host_ptr, size, alignment); + } else { + staging_buffer.resize(size); + memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size); + return StreamBufferUpload(staging_buffer.data(), size, alignment); + } } } } - auto block = GetBlock(cache_addr, size); - auto map = MapAddress(block, gpu_addr, cache_addr, size); + auto block = GetBlock(cpu_addr, size); + auto map = MapAddress(block, gpu_addr, cpu_addr, size); if (is_written) { map->MarkAsModified(true, GetModifiedTicks()); if (!map->IsWritten()) { @@ -71,7 +89,7 @@ public: } } - const u64 offset = static_cast<u64>(block->GetOffset(cache_addr)); + const u64 offset = static_cast<u64>(block->GetOffset(cpu_addr)); return {ToHandle(block), offset}; } @@ -112,7 +130,7 @@ public: } /// Write any cached resources overlapping the specified region back to memory - void FlushRegion(CacheAddr addr, std::size_t size) { + void FlushRegion(VAddr addr, std::size_t size) { std::lock_guard lock{mutex}; std::vector<MapInterval> objects = GetMapsInRange(addr, size); @@ -127,7 +145,7 @@ public: } /// Mark the specified region as being invalidated - void InvalidateRegion(CacheAddr addr, u64 size) { + void InvalidateRegion(VAddr addr, u64 size) { std::lock_guard lock{mutex}; std::vector<MapInterval> objects = GetMapsInRange(addr, size); @@ -152,7 +170,7 @@ protected: virtual void WriteBarrier() = 0; - virtual TBuffer CreateBlock(CacheAddr cache_addr, std::size_t size) = 0; + virtual TBuffer CreateBlock(VAddr cpu_addr, std::size_t size) = 0; virtual void UploadBlockData(const TBuffer& buffer, std::size_t offset, std::size_t size, const u8* data) = 0; @@ -169,20 +187,17 @@ protected: /// Register an object into the cache void Register(const MapInterval& new_map, bool inherit_written = false) { - const CacheAddr cache_ptr = new_map->GetStart(); - const std::optional<VAddr> cpu_addr = - system.GPU().MemoryManager().GpuToCpuAddress(new_map->GetGpuAddress()); - if (!cache_ptr || !cpu_addr) { + const VAddr cpu_addr = new_map->GetStart(); + if (!cpu_addr) { LOG_CRITICAL(HW_GPU, "Failed to register buffer with unmapped gpu_address 0x{:016x}", new_map->GetGpuAddress()); return; } const std::size_t size = new_map->GetEnd() - new_map->GetStart(); - new_map->SetCpuAddress(*cpu_addr); new_map->MarkAsRegistered(true); const IntervalType interval{new_map->GetStart(), new_map->GetEnd()}; mapped_addresses.insert({interval, new_map}); - rasterizer.UpdatePagesCachedCount(*cpu_addr, size, 1); + rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1); if (inherit_written) { MarkRegionAsWritten(new_map->GetStart(), new_map->GetEnd() - 1); new_map->MarkAsWritten(true); @@ -192,7 +207,7 @@ protected: /// Unregisters an object from the cache void Unregister(MapInterval& map) { const std::size_t size = map->GetEnd() - map->GetStart(); - rasterizer.UpdatePagesCachedCount(map->GetCpuAddress(), size, -1); + rasterizer.UpdatePagesCachedCount(map->GetStart(), size, -1); map->MarkAsRegistered(false); if (map->IsWritten()) { UnmarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1); @@ -202,32 +217,39 @@ protected: } private: - MapInterval CreateMap(const CacheAddr start, const CacheAddr end, const GPUVAddr gpu_addr) { + MapInterval CreateMap(const VAddr start, const VAddr end, const GPUVAddr gpu_addr) { return std::make_shared<MapIntervalBase>(start, end, gpu_addr); } - MapInterval MapAddress(const TBuffer& block, const GPUVAddr gpu_addr, - const CacheAddr cache_addr, const std::size_t size) { + MapInterval MapAddress(const TBuffer& block, const GPUVAddr gpu_addr, const VAddr cpu_addr, + const std::size_t size) { - std::vector<MapInterval> overlaps = GetMapsInRange(cache_addr, size); + std::vector<MapInterval> overlaps = GetMapsInRange(cpu_addr, size); if (overlaps.empty()) { - const CacheAddr cache_addr_end = cache_addr + size; - MapInterval new_map = CreateMap(cache_addr, cache_addr_end, gpu_addr); - u8* host_ptr = FromCacheAddr(cache_addr); - UploadBlockData(block, block->GetOffset(cache_addr), size, host_ptr); + auto& memory_manager = system.GPU().MemoryManager(); + const VAddr cpu_addr_end = cpu_addr + size; + MapInterval new_map = CreateMap(cpu_addr, cpu_addr_end, gpu_addr); + if (memory_manager.IsGranularRange(gpu_addr, size)) { + u8* host_ptr = memory_manager.GetPointer(gpu_addr); + UploadBlockData(block, block->GetOffset(cpu_addr), size, host_ptr); + } else { + staging_buffer.resize(size); + memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size); + UploadBlockData(block, block->GetOffset(cpu_addr), size, staging_buffer.data()); + } Register(new_map); return new_map; } - const CacheAddr cache_addr_end = cache_addr + size; + const VAddr cpu_addr_end = cpu_addr + size; if (overlaps.size() == 1) { MapInterval& current_map = overlaps[0]; - if (current_map->IsInside(cache_addr, cache_addr_end)) { + if (current_map->IsInside(cpu_addr, cpu_addr_end)) { return current_map; } } - CacheAddr new_start = cache_addr; - CacheAddr new_end = cache_addr_end; + VAddr new_start = cpu_addr; + VAddr new_end = cpu_addr_end; bool write_inheritance = false; bool modified_inheritance = false; // Calculate new buffer parameters @@ -237,7 +259,7 @@ private: write_inheritance |= overlap->IsWritten(); modified_inheritance |= overlap->IsModified(); } - GPUVAddr new_gpu_addr = gpu_addr + new_start - cache_addr; + GPUVAddr new_gpu_addr = gpu_addr + new_start - cpu_addr; for (auto& overlap : overlaps) { Unregister(overlap); } @@ -250,7 +272,7 @@ private: return new_map; } - void UpdateBlock(const TBuffer& block, CacheAddr start, CacheAddr end, + void UpdateBlock(const TBuffer& block, VAddr start, VAddr end, std::vector<MapInterval>& overlaps) { const IntervalType base_interval{start, end}; IntervalSet interval_set{}; @@ -262,13 +284,15 @@ private: for (auto& interval : interval_set) { std::size_t size = interval.upper() - interval.lower(); if (size > 0) { - u8* host_ptr = FromCacheAddr(interval.lower()); - UploadBlockData(block, block->GetOffset(interval.lower()), size, host_ptr); + staging_buffer.resize(size); + system.Memory().ReadBlockUnsafe(interval.lower(), staging_buffer.data(), size); + UploadBlockData(block, block->GetOffset(interval.lower()), size, + staging_buffer.data()); } } } - std::vector<MapInterval> GetMapsInRange(CacheAddr addr, std::size_t size) { + std::vector<MapInterval> GetMapsInRange(VAddr addr, std::size_t size) { if (size == 0) { return {}; } @@ -290,8 +314,9 @@ private: void FlushMap(MapInterval map) { std::size_t size = map->GetEnd() - map->GetStart(); TBuffer block = blocks[map->GetStart() >> block_page_bits]; - u8* host_ptr = FromCacheAddr(map->GetStart()); - DownloadBlockData(block, block->GetOffset(map->GetStart()), size, host_ptr); + staging_buffer.resize(size); + DownloadBlockData(block, block->GetOffset(map->GetStart()), size, staging_buffer.data()); + system.Memory().WriteBlockUnsafe(map->GetStart(), staging_buffer.data(), size); map->MarkAsModified(false, 0); } @@ -316,14 +341,14 @@ private: TBuffer EnlargeBlock(TBuffer buffer) { const std::size_t old_size = buffer->GetSize(); const std::size_t new_size = old_size + block_page_size; - const CacheAddr cache_addr = buffer->GetCacheAddr(); - TBuffer new_buffer = CreateBlock(cache_addr, new_size); + const VAddr cpu_addr = buffer->GetCpuAddr(); + TBuffer new_buffer = CreateBlock(cpu_addr, new_size); CopyBlock(buffer, new_buffer, 0, 0, old_size); buffer->SetEpoch(epoch); pending_destruction.push_back(buffer); - const CacheAddr cache_addr_end = cache_addr + new_size - 1; - u64 page_start = cache_addr >> block_page_bits; - const u64 page_end = cache_addr_end >> block_page_bits; + const VAddr cpu_addr_end = cpu_addr + new_size - 1; + u64 page_start = cpu_addr >> block_page_bits; + const u64 page_end = cpu_addr_end >> block_page_bits; while (page_start <= page_end) { blocks[page_start] = new_buffer; ++page_start; @@ -334,9 +359,9 @@ private: TBuffer MergeBlocks(TBuffer first, TBuffer second) { const std::size_t size_1 = first->GetSize(); const std::size_t size_2 = second->GetSize(); - const CacheAddr first_addr = first->GetCacheAddr(); - const CacheAddr second_addr = second->GetCacheAddr(); - const CacheAddr new_addr = std::min(first_addr, second_addr); + const VAddr first_addr = first->GetCpuAddr(); + const VAddr second_addr = second->GetCpuAddr(); + const VAddr new_addr = std::min(first_addr, second_addr); const std::size_t new_size = size_1 + size_2; TBuffer new_buffer = CreateBlock(new_addr, new_size); CopyBlock(first, new_buffer, 0, new_buffer->GetOffset(first_addr), size_1); @@ -345,9 +370,9 @@ private: second->SetEpoch(epoch); pending_destruction.push_back(first); pending_destruction.push_back(second); - const CacheAddr cache_addr_end = new_addr + new_size - 1; + const VAddr cpu_addr_end = new_addr + new_size - 1; u64 page_start = new_addr >> block_page_bits; - const u64 page_end = cache_addr_end >> block_page_bits; + const u64 page_end = cpu_addr_end >> block_page_bits; while (page_start <= page_end) { blocks[page_start] = new_buffer; ++page_start; @@ -355,18 +380,18 @@ private: return new_buffer; } - TBuffer GetBlock(const CacheAddr cache_addr, const std::size_t size) { + TBuffer GetBlock(const VAddr cpu_addr, const std::size_t size) { TBuffer found{}; - const CacheAddr cache_addr_end = cache_addr + size - 1; - u64 page_start = cache_addr >> block_page_bits; - const u64 page_end = cache_addr_end >> block_page_bits; + const VAddr cpu_addr_end = cpu_addr + size - 1; + u64 page_start = cpu_addr >> block_page_bits; + const u64 page_end = cpu_addr_end >> block_page_bits; while (page_start <= page_end) { auto it = blocks.find(page_start); if (it == blocks.end()) { if (found) { found = EnlargeBlock(found); } else { - const CacheAddr start_addr = (page_start << block_page_bits); + const VAddr start_addr = (page_start << block_page_bits); found = CreateBlock(start_addr, block_page_size); blocks[page_start] = found; } @@ -386,7 +411,7 @@ private: return found; } - void MarkRegionAsWritten(const CacheAddr start, const CacheAddr end) { + void MarkRegionAsWritten(const VAddr start, const VAddr end) { u64 page_start = start >> write_page_bit; const u64 page_end = end >> write_page_bit; while (page_start <= page_end) { @@ -400,7 +425,7 @@ private: } } - void UnmarkRegionAsWritten(const CacheAddr start, const CacheAddr end) { + void UnmarkRegionAsWritten(const VAddr start, const VAddr end) { u64 page_start = start >> write_page_bit; const u64 page_end = end >> write_page_bit; while (page_start <= page_end) { @@ -416,7 +441,7 @@ private: } } - bool IsRegionWritten(const CacheAddr start, const CacheAddr end) const { + bool IsRegionWritten(const VAddr start, const VAddr end) const { u64 page_start = start >> write_page_bit; const u64 page_end = end >> write_page_bit; while (page_start <= page_end) { @@ -440,8 +465,8 @@ private: u64 buffer_offset = 0; u64 buffer_offset_base = 0; - using IntervalSet = boost::icl::interval_set<CacheAddr>; - using IntervalCache = boost::icl::interval_map<CacheAddr, MapInterval>; + using IntervalSet = boost::icl::interval_set<VAddr>; + using IntervalCache = boost::icl::interval_map<VAddr, MapInterval>; using IntervalType = typename IntervalCache::interval_type; IntervalCache mapped_addresses; @@ -456,6 +481,8 @@ private: u64 epoch = 0; u64 modified_ticks = 0; + std::vector<u8> staging_buffer; + std::recursive_mutex mutex; }; diff --git a/src/video_core/buffer_cache/map_interval.h b/src/video_core/buffer_cache/map_interval.h index 3a104d5cd..b0956029d 100644 --- a/src/video_core/buffer_cache/map_interval.h +++ b/src/video_core/buffer_cache/map_interval.h @@ -11,7 +11,7 @@ namespace VideoCommon { class MapIntervalBase { public: - MapIntervalBase(const CacheAddr start, const CacheAddr end, const GPUVAddr gpu_addr) + MapIntervalBase(const VAddr start, const VAddr end, const GPUVAddr gpu_addr) : start{start}, end{end}, gpu_addr{gpu_addr} {} void SetCpuAddress(VAddr new_cpu_addr) { @@ -26,7 +26,7 @@ public: return gpu_addr; } - bool IsInside(const CacheAddr other_start, const CacheAddr other_end) const { + bool IsInside(const VAddr other_start, const VAddr other_end) const { return (start <= other_start && other_end <= end); } @@ -46,11 +46,11 @@ public: return is_registered; } - CacheAddr GetStart() const { + VAddr GetStart() const { return start; } - CacheAddr GetEnd() const { + VAddr GetEnd() const { return end; } @@ -76,8 +76,8 @@ public: } private: - CacheAddr start; - CacheAddr end; + VAddr start; + VAddr end; GPUVAddr gpu_addr; VAddr cpu_addr{}; bool is_written{}; diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index d24c9f657..5cf6a4cc3 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -303,6 +303,10 @@ public: return (type == Type::SignedNorm) || (type == Type::UnsignedNorm); } + bool IsConstant() const { + return constant; + } + bool IsValid() const { return size != Size::Invalid; } @@ -312,6 +316,35 @@ public: } }; + struct MsaaSampleLocation { + union { + BitField<0, 4, u32> x0; + BitField<4, 4, u32> y0; + BitField<8, 4, u32> x1; + BitField<12, 4, u32> y1; + BitField<16, 4, u32> x2; + BitField<20, 4, u32> y2; + BitField<24, 4, u32> x3; + BitField<28, 4, u32> y3; + }; + + constexpr std::pair<u32, u32> Location(int index) const { + switch (index) { + case 0: + return {x0, y0}; + case 1: + return {x1, y1}; + case 2: + return {x2, y2}; + case 3: + return {x3, y3}; + default: + UNREACHABLE(); + return {0, 0}; + } + } + }; + enum class DepthMode : u32 { MinusOneToOne = 0, ZeroToOne = 1, @@ -793,7 +826,13 @@ public: u32 rt_separate_frag_data; - INSERT_UNION_PADDING_WORDS(0xC); + INSERT_UNION_PADDING_WORDS(0x1); + + u32 multisample_raster_enable; + u32 multisample_raster_samples; + std::array<u32, 4> multisample_sample_mask; + + INSERT_UNION_PADDING_WORDS(0x5); struct { u32 address_high; @@ -830,7 +869,16 @@ public: std::array<VertexAttribute, NumVertexAttributes> vertex_attrib_format; - INSERT_UNION_PADDING_WORDS(0xF); + std::array<MsaaSampleLocation, 4> multisample_sample_locations; + + INSERT_UNION_PADDING_WORDS(0x2); + + union { + BitField<0, 1, u32> enable; + BitField<4, 3, u32> target; + } multisample_coverage_to_color; + + INSERT_UNION_PADDING_WORDS(0x8); struct { union { @@ -922,7 +970,10 @@ public: BitField<4, 1, u32> triangle_rast_flip; } screen_y_control; - INSERT_UNION_PADDING_WORDS(0x21); + float line_width_smooth; + float line_width_aliased; + + INSERT_UNION_PADDING_WORDS(0x1F); u32 vb_element_base; u32 vb_base_instance; @@ -943,7 +994,7 @@ public: CounterReset counter_reset; - INSERT_UNION_PADDING_WORDS(0x1); + u32 multisample_enable; u32 zeta_enable; @@ -980,7 +1031,7 @@ public: float polygon_offset_factor; - INSERT_UNION_PADDING_WORDS(0x1); + u32 line_smooth_enable; struct { u32 tic_address_high; @@ -1007,7 +1058,11 @@ public: float polygon_offset_units; - INSERT_UNION_PADDING_WORDS(0x11); + INSERT_UNION_PADDING_WORDS(0x4); + + Tegra::Texture::MsaaMode multisample_mode; + + INSERT_UNION_PADDING_WORDS(0xC); union { BitField<2, 1, u32> coord_origin; @@ -1507,12 +1562,17 @@ ASSERT_REG_POSITION(stencil_back_func_ref, 0x3D5); ASSERT_REG_POSITION(stencil_back_mask, 0x3D6); ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7); ASSERT_REG_POSITION(color_mask_common, 0x3E4); -ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB); ASSERT_REG_POSITION(depth_bounds, 0x3E7); +ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB); +ASSERT_REG_POSITION(multisample_raster_enable, 0x3ED); +ASSERT_REG_POSITION(multisample_raster_samples, 0x3EE); +ASSERT_REG_POSITION(multisample_sample_mask, 0x3EF); ASSERT_REG_POSITION(zeta, 0x3F8); ASSERT_REG_POSITION(clear_flags, 0x43E); ASSERT_REG_POSITION(fill_rectangle, 0x44F); ASSERT_REG_POSITION(vertex_attrib_format, 0x458); +ASSERT_REG_POSITION(multisample_sample_locations, 0x478); +ASSERT_REG_POSITION(multisample_coverage_to_color, 0x47E); ASSERT_REG_POSITION(rt_control, 0x487); ASSERT_REG_POSITION(zeta_width, 0x48a); ASSERT_REG_POSITION(zeta_height, 0x48b); @@ -1538,6 +1598,8 @@ ASSERT_REG_POSITION(stencil_front_func_mask, 0x4E6); ASSERT_REG_POSITION(stencil_front_mask, 0x4E7); ASSERT_REG_POSITION(frag_color_clamp, 0x4EA); ASSERT_REG_POSITION(screen_y_control, 0x4EB); +ASSERT_REG_POSITION(line_width_smooth, 0x4EC); +ASSERT_REG_POSITION(line_width_aliased, 0x4ED); ASSERT_REG_POSITION(vb_element_base, 0x50D); ASSERT_REG_POSITION(vb_base_instance, 0x50E); ASSERT_REG_POSITION(clip_distance_enabled, 0x544); @@ -1545,11 +1607,13 @@ ASSERT_REG_POSITION(samplecnt_enable, 0x545); ASSERT_REG_POSITION(point_size, 0x546); ASSERT_REG_POSITION(point_sprite_enable, 0x548); ASSERT_REG_POSITION(counter_reset, 0x54C); +ASSERT_REG_POSITION(multisample_enable, 0x54D); ASSERT_REG_POSITION(zeta_enable, 0x54E); ASSERT_REG_POSITION(multisample_control, 0x54F); ASSERT_REG_POSITION(condition, 0x554); ASSERT_REG_POSITION(tsc, 0x557); -ASSERT_REG_POSITION(polygon_offset_factor, 0x55b); +ASSERT_REG_POSITION(polygon_offset_factor, 0x55B); +ASSERT_REG_POSITION(line_smooth_enable, 0x55C); ASSERT_REG_POSITION(tic, 0x55D); ASSERT_REG_POSITION(stencil_two_side_enable, 0x565); ASSERT_REG_POSITION(stencil_back_op_fail, 0x566); @@ -1558,6 +1622,7 @@ ASSERT_REG_POSITION(stencil_back_op_zpass, 0x568); ASSERT_REG_POSITION(stencil_back_func_func, 0x569); ASSERT_REG_POSITION(framebuffer_srgb, 0x56E); ASSERT_REG_POSITION(polygon_offset_units, 0x56F); +ASSERT_REG_POSITION(multisample_mode, 0x574); ASSERT_REG_POSITION(point_coord_replace, 0x581); ASSERT_REG_POSITION(code_address, 0x582); ASSERT_REG_POSITION(draw, 0x585); diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h index a31947ef3..5e9cfba22 100644 --- a/src/video_core/engines/shader_bytecode.h +++ b/src/video_core/engines/shader_bytecode.h @@ -290,6 +290,23 @@ enum class VmadShr : u64 { Shr15 = 2, }; +enum class VmnmxType : u64 { + Bits8, + Bits16, + Bits32, +}; + +enum class VmnmxOperation : u64 { + Mrg_16H = 0, + Mrg_16L = 1, + Mrg_8B0 = 2, + Mrg_8B2 = 3, + Acc = 4, + Min = 5, + Max = 6, + Nop = 7, +}; + enum class XmadMode : u64 { None = 0, CLo = 1, @@ -1657,6 +1674,42 @@ union Instruction { } vmad; union { + BitField<54, 1, u64> is_dest_signed; + BitField<48, 1, u64> is_src_a_signed; + BitField<49, 1, u64> is_src_b_signed; + BitField<37, 2, u64> src_format_a; + BitField<29, 2, u64> src_format_b; + BitField<56, 1, u64> mx; + BitField<55, 1, u64> sat; + BitField<36, 2, u64> selector_a; + BitField<28, 2, u64> selector_b; + BitField<50, 1, u64> is_op_b_register; + BitField<51, 3, VmnmxOperation> operation; + + VmnmxType SourceFormatA() const { + switch (src_format_a) { + case 0b11: + return VmnmxType::Bits32; + case 0b10: + return VmnmxType::Bits16; + default: + return VmnmxType::Bits8; + } + } + + VmnmxType SourceFormatB() const { + switch (src_format_b) { + case 0b11: + return VmnmxType::Bits32; + case 0b10: + return VmnmxType::Bits16; + default: + return VmnmxType::Bits8; + } + } + } vmnmx; + + union { BitField<20, 16, u64> imm20_16; BitField<35, 1, u64> high_b_rr; // used on RR BitField<36, 1, u64> product_shift_left; @@ -1718,6 +1771,7 @@ public: BRK, DEPBAR, VOTE, + VOTE_VTG, SHFL, FSWZADD, BFE_C, @@ -1765,9 +1819,11 @@ public: IPA, OUT_R, // Emit vertex/primitive ISBERD, + BAR, MEMBAR, VMAD, VSETP, + VMNMX, FFMA_IMM, // Fused Multiply and Add FFMA_CR, FFMA_RC, @@ -1822,7 +1878,8 @@ public: ICMP_R, ICMP_CR, ICMP_IMM, - FCMP_R, + FCMP_RR, + FCMP_RC, MUFU, // Multi-Function Operator RRO_C, // Range Reduction Operator RRO_R, @@ -1849,7 +1906,7 @@ public: MOV_C, MOV_R, MOV_IMM, - MOV_SYS, + S2R, MOV32_IMM, SHL_C, SHL_R, @@ -2033,6 +2090,7 @@ private: INST("111000110000----", Id::EXIT, Type::Flow, "EXIT"), INST("1111000011110---", Id::DEPBAR, Type::Synch, "DEPBAR"), INST("0101000011011---", Id::VOTE, Type::Warp, "VOTE"), + INST("0101000011100---", Id::VOTE_VTG, Type::Warp, "VOTE_VTG"), INST("1110111100010---", Id::SHFL, Type::Warp, "SHFL"), INST("0101000011111---", Id::FSWZADD, Type::Warp, "FSWZADD"), INST("1110111111011---", Id::LD_A, Type::Memory, "LD_A"), @@ -2071,9 +2129,11 @@ private: INST("11100000--------", Id::IPA, Type::Trivial, "IPA"), INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"), INST("1110111111010---", Id::ISBERD, Type::Trivial, "ISBERD"), + INST("1111000010101---", Id::BAR, Type::Trivial, "BAR"), INST("1110111110011---", Id::MEMBAR, Type::Trivial, "MEMBAR"), INST("01011111--------", Id::VMAD, Type::Video, "VMAD"), INST("0101000011110---", Id::VSETP, Type::Video, "VSETP"), + INST("0011101---------", Id::VMNMX, Type::Video, "VMNMX"), INST("0011001-1-------", Id::FFMA_IMM, Type::Ffma, "FFMA_IMM"), INST("010010011-------", Id::FFMA_CR, Type::Ffma, "FFMA_CR"), INST("010100011-------", Id::FFMA_RC, Type::Ffma, "FFMA_RC"), @@ -2128,7 +2188,8 @@ private: INST("0101110100100---", Id::HSETP2_R, Type::HalfSetPredicate, "HSETP2_R"), INST("0111111-0-------", Id::HSETP2_IMM, Type::HalfSetPredicate, "HSETP2_IMM"), INST("0101110100011---", Id::HSET2_R, Type::HalfSet, "HSET2_R"), - INST("010110111010----", Id::FCMP_R, Type::Arithmetic, "FCMP_R"), + INST("010110111010----", Id::FCMP_RR, Type::Arithmetic, "FCMP_RR"), + INST("010010111010----", Id::FCMP_RC, Type::Arithmetic, "FCMP_RC"), INST("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"), INST("0100110010010---", Id::RRO_C, Type::Arithmetic, "RRO_C"), INST("0101110010010---", Id::RRO_R, Type::Arithmetic, "RRO_R"), @@ -2142,7 +2203,7 @@ private: INST("0100110010011---", Id::MOV_C, Type::Arithmetic, "MOV_C"), INST("0101110010011---", Id::MOV_R, Type::Arithmetic, "MOV_R"), INST("0011100-10011---", Id::MOV_IMM, Type::Arithmetic, "MOV_IMM"), - INST("1111000011001---", Id::MOV_SYS, Type::Trivial, "MOV_SYS"), + INST("1111000011001---", Id::S2R, Type::Trivial, "S2R"), INST("000000010000----", Id::MOV32_IMM, Type::ArithmeticImmediate, "MOV32_IMM"), INST("0100110001100---", Id::FMNMX_C, Type::Arithmetic, "FMNMX_C"), INST("0101110001100---", Id::FMNMX_R, Type::Arithmetic, "FMNMX_R"), @@ -2174,7 +2235,7 @@ private: INST("0011011-11111---", Id::SHF_LEFT_IMM, Type::Shift, "SHF_LEFT_IMM"), INST("0100110011100---", Id::I2I_C, Type::Conversion, "I2I_C"), INST("0101110011100---", Id::I2I_R, Type::Conversion, "I2I_R"), - INST("0011101-11100---", Id::I2I_IMM, Type::Conversion, "I2I_IMM"), + INST("0011100-11100---", Id::I2I_IMM, Type::Conversion, "I2I_IMM"), INST("0100110010111---", Id::I2F_C, Type::Conversion, "I2F_C"), INST("0101110010111---", Id::I2F_R, Type::Conversion, "I2F_R"), INST("0011100-10111---", Id::I2F_IMM, Type::Conversion, "I2F_IMM"), diff --git a/src/video_core/engines/shader_header.h b/src/video_core/engines/shader_header.h index bc80661d8..72e2a33d5 100644 --- a/src/video_core/engines/shader_header.h +++ b/src/video_core/engines/shader_header.h @@ -4,6 +4,9 @@ #pragma once +#include <array> +#include <optional> + #include "common/bit_field.h" #include "common/common_funcs.h" #include "common/common_types.h" @@ -16,7 +19,7 @@ enum class OutputTopology : u32 { TriangleStrip = 7, }; -enum class AttributeUse : u8 { +enum class PixelImap : u8 { Unused = 0, Constant = 1, Perspective = 2, @@ -24,7 +27,7 @@ enum class AttributeUse : u8 { }; // Documentation in: -// http://download.nvidia.com/open-gpu-doc/Shader-Program-Header/1/Shader-Program-Header.html#ImapTexture +// http://download.nvidia.com/open-gpu-doc/Shader-Program-Header/1/Shader-Program-Header.html struct Header { union { BitField<0, 5, u32> sph_type; @@ -59,8 +62,8 @@ struct Header { union { BitField<0, 12, u32> max_output_vertices; BitField<12, 8, u32> store_req_start; // NOTE: not used by geometry shaders. - BitField<24, 4, u32> reserved; - BitField<12, 8, u32> store_req_end; // NOTE: not used by geometry shaders. + BitField<20, 4, u32> reserved; + BitField<24, 8, u32> store_req_end; // NOTE: not used by geometry shaders. } common4{}; union { @@ -93,17 +96,20 @@ struct Header { struct { INSERT_UNION_PADDING_BYTES(3); // ImapSystemValuesA INSERT_UNION_PADDING_BYTES(1); // ImapSystemValuesB + union { - BitField<0, 2, AttributeUse> x; - BitField<2, 2, AttributeUse> y; - BitField<4, 2, AttributeUse> w; - BitField<6, 2, AttributeUse> z; + BitField<0, 2, PixelImap> x; + BitField<2, 2, PixelImap> y; + BitField<4, 2, PixelImap> z; + BitField<6, 2, PixelImap> w; u8 raw; } imap_generic_vector[32]; + INSERT_UNION_PADDING_BYTES(2); // ImapColor INSERT_UNION_PADDING_BYTES(2); // ImapSystemValuesC INSERT_UNION_PADDING_BYTES(10); // ImapFixedFncTexture[10] INSERT_UNION_PADDING_BYTES(2); // ImapReserved + struct { u32 target; union { @@ -112,31 +118,30 @@ struct Header { BitField<2, 30, u32> reserved; }; } omap; + bool IsColorComponentOutputEnabled(u32 render_target, u32 component) const { const u32 bit = render_target * 4 + component; return omap.target & (1 << bit); } - AttributeUse GetAttributeIndexUse(u32 attribute, u32 index) const { - return static_cast<AttributeUse>( - (imap_generic_vector[attribute].raw >> (index * 2)) & 0x03); - } - AttributeUse GetAttributeUse(u32 attribute) const { - AttributeUse result = AttributeUse::Unused; - for (u32 i = 0; i < 4; i++) { - const auto index = GetAttributeIndexUse(attribute, i); - if (index == AttributeUse::Unused) { - continue; - } - if (result == AttributeUse::Unused || result == index) { - result = index; + + PixelImap GetPixelImap(u32 attribute) const { + const auto get_index = [this, attribute](u32 index) { + return static_cast<PixelImap>( + (imap_generic_vector[attribute].raw >> (index * 2)) & 3); + }; + + std::optional<PixelImap> result; + for (u32 component = 0; component < 4; ++component) { + const PixelImap index = get_index(component); + if (index == PixelImap::Unused) { continue; } - LOG_CRITICAL(HW_GPU, "Generic Attribute Conflict in Interpolation Mode"); - if (index == AttributeUse::Perspective) { - result = index; + if (result && result != index) { + LOG_CRITICAL(HW_GPU, "Generic attribute conflict in interpolation mode"); } + result = index; } - return result; + return result.value_or(PixelImap::Unused); } } ps; diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h index ced9d7e28..1a2d747be 100644 --- a/src/video_core/gpu.h +++ b/src/video_core/gpu.h @@ -270,13 +270,13 @@ public: virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0; /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory - virtual void FlushRegion(CacheAddr addr, u64 size) = 0; + virtual void FlushRegion(VAddr addr, u64 size) = 0; /// Notify rasterizer that any caches of the specified region should be invalidated - virtual void InvalidateRegion(CacheAddr addr, u64 size) = 0; + virtual void InvalidateRegion(VAddr addr, u64 size) = 0; /// Notify rasterizer that any caches of the specified region should be flushed and invalidated - virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0; + virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0; protected: virtual void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const = 0; diff --git a/src/video_core/gpu_asynch.cpp b/src/video_core/gpu_asynch.cpp index 925be8d7b..20e73a37e 100644 --- a/src/video_core/gpu_asynch.cpp +++ b/src/video_core/gpu_asynch.cpp @@ -12,8 +12,9 @@ namespace VideoCommon { GPUAsynch::GPUAsynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer_, std::unique_ptr<Core::Frontend::GraphicsContext>&& context) - : GPU(system, std::move(renderer_), true), gpu_thread{system}, gpu_context(std::move(context)), - cpu_context(renderer->GetRenderWindow().CreateSharedContext()) {} + : GPU(system, std::move(renderer_), true), gpu_thread{system}, + cpu_context(renderer->GetRenderWindow().CreateSharedContext()), + gpu_context(std::move(context)) {} GPUAsynch::~GPUAsynch() = default; @@ -30,15 +31,15 @@ void GPUAsynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { gpu_thread.SwapBuffers(framebuffer); } -void GPUAsynch::FlushRegion(CacheAddr addr, u64 size) { +void GPUAsynch::FlushRegion(VAddr addr, u64 size) { gpu_thread.FlushRegion(addr, size); } -void GPUAsynch::InvalidateRegion(CacheAddr addr, u64 size) { +void GPUAsynch::InvalidateRegion(VAddr addr, u64 size) { gpu_thread.InvalidateRegion(addr, size); } -void GPUAsynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) { +void GPUAsynch::FlushAndInvalidateRegion(VAddr addr, u64 size) { gpu_thread.FlushAndInvalidateRegion(addr, size); } diff --git a/src/video_core/gpu_asynch.h b/src/video_core/gpu_asynch.h index 265c62758..03fd0eef0 100644 --- a/src/video_core/gpu_asynch.h +++ b/src/video_core/gpu_asynch.h @@ -27,9 +27,9 @@ public: void Start() override; void PushGPUEntries(Tegra::CommandList&& entries) override; void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override; - void FlushRegion(CacheAddr addr, u64 size) override; - void InvalidateRegion(CacheAddr addr, u64 size) override; - void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override; + void FlushRegion(VAddr addr, u64 size) override; + void InvalidateRegion(VAddr addr, u64 size) override; + void FlushAndInvalidateRegion(VAddr addr, u64 size) override; void WaitIdle() const override; protected: diff --git a/src/video_core/gpu_synch.cpp b/src/video_core/gpu_synch.cpp index bd5278a5c..6f38a672a 100644 --- a/src/video_core/gpu_synch.cpp +++ b/src/video_core/gpu_synch.cpp @@ -26,15 +26,15 @@ void GPUSynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { renderer->SwapBuffers(framebuffer); } -void GPUSynch::FlushRegion(CacheAddr addr, u64 size) { +void GPUSynch::FlushRegion(VAddr addr, u64 size) { renderer->Rasterizer().FlushRegion(addr, size); } -void GPUSynch::InvalidateRegion(CacheAddr addr, u64 size) { +void GPUSynch::InvalidateRegion(VAddr addr, u64 size) { renderer->Rasterizer().InvalidateRegion(addr, size); } -void GPUSynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) { +void GPUSynch::FlushAndInvalidateRegion(VAddr addr, u64 size) { renderer->Rasterizer().FlushAndInvalidateRegion(addr, size); } diff --git a/src/video_core/gpu_synch.h b/src/video_core/gpu_synch.h index 866a94c8c..4a6e9a01d 100644 --- a/src/video_core/gpu_synch.h +++ b/src/video_core/gpu_synch.h @@ -26,9 +26,9 @@ public: void Start() override; void PushGPUEntries(Tegra::CommandList&& entries) override; void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override; - void FlushRegion(CacheAddr addr, u64 size) override; - void InvalidateRegion(CacheAddr addr, u64 size) override; - void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override; + void FlushRegion(VAddr addr, u64 size) override; + void InvalidateRegion(VAddr addr, u64 size) override; + void FlushAndInvalidateRegion(VAddr addr, u64 size) override; void WaitIdle() const override {} protected: diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index 270c7ae0d..10cda686b 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp @@ -77,15 +77,15 @@ void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { PushCommand(SwapBuffersCommand(framebuffer ? std::make_optional(*framebuffer) : std::nullopt)); } -void ThreadManager::FlushRegion(CacheAddr addr, u64 size) { +void ThreadManager::FlushRegion(VAddr addr, u64 size) { PushCommand(FlushRegionCommand(addr, size)); } -void ThreadManager::InvalidateRegion(CacheAddr addr, u64 size) { +void ThreadManager::InvalidateRegion(VAddr addr, u64 size) { system.Renderer().Rasterizer().InvalidateRegion(addr, size); } -void ThreadManager::FlushAndInvalidateRegion(CacheAddr addr, u64 size) { +void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) { // Skip flush on asynch mode, as FlushAndInvalidateRegion is not used for anything too important InvalidateRegion(addr, size); } diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h index be36c580e..cd74ad330 100644 --- a/src/video_core/gpu_thread.h +++ b/src/video_core/gpu_thread.h @@ -47,26 +47,26 @@ struct SwapBuffersCommand final { /// Command to signal to the GPU thread to flush a region struct FlushRegionCommand final { - explicit constexpr FlushRegionCommand(CacheAddr addr, u64 size) : addr{addr}, size{size} {} + explicit constexpr FlushRegionCommand(VAddr addr, u64 size) : addr{addr}, size{size} {} - CacheAddr addr; + VAddr addr; u64 size; }; /// Command to signal to the GPU thread to invalidate a region struct InvalidateRegionCommand final { - explicit constexpr InvalidateRegionCommand(CacheAddr addr, u64 size) : addr{addr}, size{size} {} + explicit constexpr InvalidateRegionCommand(VAddr addr, u64 size) : addr{addr}, size{size} {} - CacheAddr addr; + VAddr addr; u64 size; }; /// Command to signal to the GPU thread to flush and invalidate a region struct FlushAndInvalidateRegionCommand final { - explicit constexpr FlushAndInvalidateRegionCommand(CacheAddr addr, u64 size) + explicit constexpr FlushAndInvalidateRegionCommand(VAddr addr, u64 size) : addr{addr}, size{size} {} - CacheAddr addr; + VAddr addr; u64 size; }; @@ -111,13 +111,13 @@ public: void SwapBuffers(const Tegra::FramebufferConfig* framebuffer); /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory - void FlushRegion(CacheAddr addr, u64 size); + void FlushRegion(VAddr addr, u64 size); /// Notify rasterizer that any caches of the specified region should be invalidated - void InvalidateRegion(CacheAddr addr, u64 size); + void InvalidateRegion(VAddr addr, u64 size); /// Notify rasterizer that any caches of the specified region should be flushed and invalidated - void FlushAndInvalidateRegion(CacheAddr addr, u64 size); + void FlushAndInvalidateRegion(VAddr addr, u64 size); // Wait until the gpu thread is idle. void WaitIdle() const; diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index f5d33f27a..a3389d0d2 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -81,12 +81,11 @@ GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) { ASSERT((gpu_addr & page_mask) == 0); const u64 aligned_size{Common::AlignUp(size, page_size)}; - const CacheAddr cache_addr{ToCacheAddr(GetPointer(gpu_addr))}; const auto cpu_addr = GpuToCpuAddress(gpu_addr); ASSERT(cpu_addr); // Flush and invalidate through the GPU interface, to be asynchronous if possible. - system.GPU().FlushAndInvalidateRegion(cache_addr, aligned_size); + system.GPU().FlushAndInvalidateRegion(*cpu_addr, aligned_size); UnmapRange(gpu_addr, aligned_size); ASSERT(system.CurrentProcess() @@ -140,11 +139,11 @@ T MemoryManager::Read(GPUVAddr addr) const { return {}; } - const u8* page_pointer{page_table.pointers[addr >> page_bits]}; + const u8* page_pointer{GetPointer(addr)}; if (page_pointer) { // NOTE: Avoid adding any extra logic to this fast-path block T value; - std::memcpy(&value, &page_pointer[addr & page_mask], sizeof(T)); + std::memcpy(&value, page_pointer, sizeof(T)); return value; } @@ -167,10 +166,10 @@ void MemoryManager::Write(GPUVAddr addr, T data) { return; } - u8* page_pointer{page_table.pointers[addr >> page_bits]}; + u8* page_pointer{GetPointer(addr)}; if (page_pointer) { // NOTE: Avoid adding any extra logic to this fast-path block - std::memcpy(&page_pointer[addr & page_mask], &data, sizeof(T)); + std::memcpy(page_pointer, &data, sizeof(T)); return; } @@ -201,9 +200,12 @@ u8* MemoryManager::GetPointer(GPUVAddr addr) { return {}; } - u8* const page_pointer{page_table.pointers[addr >> page_bits]}; - if (page_pointer != nullptr) { - return page_pointer + (addr & page_mask); + auto& memory = system.Memory(); + + const VAddr page_addr{page_table.backing_addr[addr >> page_bits]}; + + if (page_addr != 0) { + return memory.GetPointer(page_addr + (addr & page_mask)); } LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr); @@ -215,9 +217,12 @@ const u8* MemoryManager::GetPointer(GPUVAddr addr) const { return {}; } - const u8* const page_pointer{page_table.pointers[addr >> page_bits]}; - if (page_pointer != nullptr) { - return page_pointer + (addr & page_mask); + const auto& memory = system.Memory(); + + const VAddr page_addr{page_table.backing_addr[addr >> page_bits]}; + + if (page_addr != 0) { + return memory.GetPointer(page_addr + (addr & page_mask)); } LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr); @@ -238,17 +243,19 @@ void MemoryManager::ReadBlock(GPUVAddr src_addr, void* dest_buffer, const std::s std::size_t page_index{src_addr >> page_bits}; std::size_t page_offset{src_addr & page_mask}; + auto& memory = system.Memory(); + while (remaining_size > 0) { const std::size_t copy_amount{ std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)}; switch (page_table.attributes[page_index]) { case Common::PageType::Memory: { - const u8* src_ptr{page_table.pointers[page_index] + page_offset}; + const VAddr src_addr{page_table.backing_addr[page_index] + page_offset}; // Flush must happen on the rasterizer interface, such that memory is always synchronous // when it is read (even when in asynchronous GPU mode). Fixes Dead Cells title menu. - rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount); - std::memcpy(dest_buffer, src_ptr, copy_amount); + rasterizer.FlushRegion(src_addr, copy_amount); + memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); break; } default: @@ -268,13 +275,15 @@ void MemoryManager::ReadBlockUnsafe(GPUVAddr src_addr, void* dest_buffer, std::size_t page_index{src_addr >> page_bits}; std::size_t page_offset{src_addr & page_mask}; + auto& memory = system.Memory(); + while (remaining_size > 0) { const std::size_t copy_amount{ std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)}; const u8* page_pointer = page_table.pointers[page_index]; if (page_pointer) { - const u8* src_ptr{page_pointer + page_offset}; - std::memcpy(dest_buffer, src_ptr, copy_amount); + const VAddr src_addr{page_table.backing_addr[page_index] + page_offset}; + memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); } else { std::memset(dest_buffer, 0, copy_amount); } @@ -290,17 +299,19 @@ void MemoryManager::WriteBlock(GPUVAddr dest_addr, const void* src_buffer, const std::size_t page_index{dest_addr >> page_bits}; std::size_t page_offset{dest_addr & page_mask}; + auto& memory = system.Memory(); + while (remaining_size > 0) { const std::size_t copy_amount{ std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)}; switch (page_table.attributes[page_index]) { case Common::PageType::Memory: { - u8* dest_ptr{page_table.pointers[page_index] + page_offset}; + const VAddr dest_addr{page_table.backing_addr[page_index] + page_offset}; // Invalidate must happen on the rasterizer interface, such that memory is always // synchronous when it is written (even when in asynchronous GPU mode). - rasterizer.InvalidateRegion(ToCacheAddr(dest_ptr), copy_amount); - std::memcpy(dest_ptr, src_buffer, copy_amount); + rasterizer.InvalidateRegion(dest_addr, copy_amount); + memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); break; } default: @@ -320,13 +331,15 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer, std::size_t page_index{dest_addr >> page_bits}; std::size_t page_offset{dest_addr & page_mask}; + auto& memory = system.Memory(); + while (remaining_size > 0) { const std::size_t copy_amount{ std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)}; u8* page_pointer = page_table.pointers[page_index]; if (page_pointer) { - u8* dest_ptr{page_pointer + page_offset}; - std::memcpy(dest_ptr, src_buffer, copy_amount); + const VAddr dest_addr{page_table.backing_addr[page_index] + page_offset}; + memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); } page_index++; page_offset = 0; @@ -336,33 +349,9 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer, } void MemoryManager::CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, const std::size_t size) { - std::size_t remaining_size{size}; - std::size_t page_index{src_addr >> page_bits}; - std::size_t page_offset{src_addr & page_mask}; - - while (remaining_size > 0) { - const std::size_t copy_amount{ - std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)}; - - switch (page_table.attributes[page_index]) { - case Common::PageType::Memory: { - // Flush must happen on the rasterizer interface, such that memory is always synchronous - // when it is copied (even when in asynchronous GPU mode). - const u8* src_ptr{page_table.pointers[page_index] + page_offset}; - rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount); - WriteBlock(dest_addr, src_ptr, copy_amount); - break; - } - default: - UNREACHABLE(); - } - - page_index++; - page_offset = 0; - dest_addr += static_cast<VAddr>(copy_amount); - src_addr += static_cast<VAddr>(copy_amount); - remaining_size -= copy_amount; - } + std::vector<u8> tmp_buffer(size); + ReadBlock(src_addr, tmp_buffer.data(), size); + WriteBlock(dest_addr, tmp_buffer.data(), size); } void MemoryManager::CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, const std::size_t size) { @@ -371,6 +360,12 @@ void MemoryManager::CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, const WriteBlockUnsafe(dest_addr, tmp_buffer.data(), size); } +bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) { + const VAddr addr = page_table.backing_addr[gpu_addr >> page_bits]; + const std::size_t page = (addr & Memory::PAGE_MASK) + size; + return page <= Memory::PAGE_SIZE; +} + void MemoryManager::MapPages(GPUVAddr base, u64 size, u8* memory, Common::PageType type, VAddr backing_addr) { LOG_DEBUG(HW_GPU, "Mapping {} onto {:016X}-{:016X}", fmt::ptr(memory), base * page_size, diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index 073bdb491..0d9468535 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -97,6 +97,11 @@ public: void WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer, std::size_t size); void CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size); + /** + * IsGranularRange checks if a gpu region can be simply read with a pointer + */ + bool IsGranularRange(GPUVAddr gpu_addr, std::size_t size); + private: using VMAMap = std::map<GPUVAddr, VirtualMemoryArea>; using VMAHandle = VMAMap::const_iterator; diff --git a/src/video_core/query_cache.h b/src/video_core/query_cache.h index e66054ed0..5ea2b01f2 100644 --- a/src/video_core/query_cache.h +++ b/src/video_core/query_cache.h @@ -98,12 +98,12 @@ public: static_cast<QueryCache&>(*this), VideoCore::QueryType::SamplesPassed}}} {} - void InvalidateRegion(CacheAddr addr, std::size_t size) { + void InvalidateRegion(VAddr addr, std::size_t size) { std::unique_lock lock{mutex}; FlushAndRemoveRegion(addr, size); } - void FlushRegion(CacheAddr addr, std::size_t size) { + void FlushRegion(VAddr addr, std::size_t size) { std::unique_lock lock{mutex}; FlushAndRemoveRegion(addr, size); } @@ -117,14 +117,16 @@ public: void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) { std::unique_lock lock{mutex}; auto& memory_manager = system.GPU().MemoryManager(); - const auto host_ptr = memory_manager.GetPointer(gpu_addr); + const std::optional<VAddr> cpu_addr_opt = memory_manager.GpuToCpuAddress(gpu_addr); + ASSERT(cpu_addr_opt); + VAddr cpu_addr = *cpu_addr_opt; - CachedQuery* query = TryGet(ToCacheAddr(host_ptr)); + CachedQuery* query = TryGet(cpu_addr); if (!query) { - const auto cpu_addr = memory_manager.GpuToCpuAddress(gpu_addr); - ASSERT_OR_EXECUTE(cpu_addr, return;); + ASSERT_OR_EXECUTE(cpu_addr_opt, return;); + const auto host_ptr = memory_manager.GetPointer(gpu_addr); - query = Register(type, *cpu_addr, host_ptr, timestamp.has_value()); + query = Register(type, cpu_addr, host_ptr, timestamp.has_value()); } query->BindCounter(Stream(type).Current(), timestamp); @@ -173,11 +175,11 @@ protected: private: /// Flushes a memory range to guest memory and removes it from the cache. - void FlushAndRemoveRegion(CacheAddr addr, std::size_t size) { + void FlushAndRemoveRegion(VAddr addr, std::size_t size) { const u64 addr_begin = static_cast<u64>(addr); const u64 addr_end = addr_begin + static_cast<u64>(size); const auto in_range = [addr_begin, addr_end](CachedQuery& query) { - const u64 cache_begin = query.GetCacheAddr(); + const u64 cache_begin = query.GetCpuAddr(); const u64 cache_end = cache_begin + query.SizeInBytes(); return cache_begin < addr_end && addr_begin < cache_end; }; @@ -193,7 +195,7 @@ private: if (!in_range(query)) { continue; } - rasterizer.UpdatePagesCachedCount(query.CpuAddr(), query.SizeInBytes(), -1); + rasterizer.UpdatePagesCachedCount(query.GetCpuAddr(), query.SizeInBytes(), -1); query.Flush(); } contents.erase(std::remove_if(std::begin(contents), std::end(contents), in_range), @@ -204,22 +206,21 @@ private: /// Registers the passed parameters as cached and returns a pointer to the stored cached query. CachedQuery* Register(VideoCore::QueryType type, VAddr cpu_addr, u8* host_ptr, bool timestamp) { rasterizer.UpdatePagesCachedCount(cpu_addr, CachedQuery::SizeInBytes(timestamp), 1); - const u64 page = static_cast<u64>(ToCacheAddr(host_ptr)) >> PAGE_SHIFT; + const u64 page = static_cast<u64>(cpu_addr) >> PAGE_SHIFT; return &cached_queries[page].emplace_back(static_cast<QueryCache&>(*this), type, cpu_addr, host_ptr); } /// Tries to a get a cached query. Returns nullptr on failure. - CachedQuery* TryGet(CacheAddr addr) { + CachedQuery* TryGet(VAddr addr) { const u64 page = static_cast<u64>(addr) >> PAGE_SHIFT; const auto it = cached_queries.find(page); if (it == std::end(cached_queries)) { return nullptr; } auto& contents = it->second; - const auto found = - std::find_if(std::begin(contents), std::end(contents), - [addr](auto& query) { return query.GetCacheAddr() == addr; }); + const auto found = std::find_if(std::begin(contents), std::end(contents), + [addr](auto& query) { return query.GetCpuAddr() == addr; }); return found != std::end(contents) ? &*found : nullptr; } @@ -323,14 +324,10 @@ public: timestamp = timestamp_; } - VAddr CpuAddr() const noexcept { + VAddr GetCpuAddr() const noexcept { return cpu_addr; } - CacheAddr GetCacheAddr() const noexcept { - return ToCacheAddr(host_ptr); - } - u64 SizeInBytes() const noexcept { return SizeInBytes(timestamp.has_value()); } diff --git a/src/video_core/rasterizer_cache.h b/src/video_core/rasterizer_cache.h index 6de1597a2..22987751e 100644 --- a/src/video_core/rasterizer_cache.h +++ b/src/video_core/rasterizer_cache.h @@ -18,22 +18,14 @@ class RasterizerCacheObject { public: - explicit RasterizerCacheObject(const u8* host_ptr) - : host_ptr{host_ptr}, cache_addr{ToCacheAddr(host_ptr)} {} + explicit RasterizerCacheObject(const VAddr cpu_addr) : cpu_addr{cpu_addr} {} virtual ~RasterizerCacheObject(); - CacheAddr GetCacheAddr() const { - return cache_addr; + VAddr GetCpuAddr() const { + return cpu_addr; } - const u8* GetHostPtr() const { - return host_ptr; - } - - /// Gets the address of the shader in guest memory, required for cache management - virtual VAddr GetCpuAddr() const = 0; - /// Gets the size of the shader in guest memory, required for cache management virtual std::size_t GetSizeInBytes() const = 0; @@ -68,8 +60,7 @@ private: bool is_registered{}; ///< Whether the object is currently registered with the cache bool is_dirty{}; ///< Whether the object is dirty (out of sync with guest memory) u64 last_modified_ticks{}; ///< When the object was last modified, used for in-order flushing - const u8* host_ptr{}; ///< Pointer to the memory backing this cached region - CacheAddr cache_addr{}; ///< Cache address memory, unique from emulated virtual address space + VAddr cpu_addr{}; ///< Cpu address memory, unique from emulated virtual address space }; template <class T> @@ -80,7 +71,7 @@ public: explicit RasterizerCache(VideoCore::RasterizerInterface& rasterizer) : rasterizer{rasterizer} {} /// Write any cached resources overlapping the specified region back to memory - void FlushRegion(CacheAddr addr, std::size_t size) { + void FlushRegion(VAddr addr, std::size_t size) { std::lock_guard lock{mutex}; const auto& objects{GetSortedObjectsFromRegion(addr, size)}; @@ -90,7 +81,7 @@ public: } /// Mark the specified region as being invalidated - void InvalidateRegion(CacheAddr addr, u64 size) { + void InvalidateRegion(VAddr addr, u64 size) { std::lock_guard lock{mutex}; const auto& objects{GetSortedObjectsFromRegion(addr, size)}; @@ -114,27 +105,20 @@ public: protected: /// Tries to get an object from the cache with the specified cache address - T TryGet(CacheAddr addr) const { + T TryGet(VAddr addr) const { const auto iter = map_cache.find(addr); if (iter != map_cache.end()) return iter->second; return nullptr; } - T TryGet(const void* addr) const { - const auto iter = map_cache.find(ToCacheAddr(addr)); - if (iter != map_cache.end()) - return iter->second; - return nullptr; - } - /// Register an object into the cache virtual void Register(const T& object) { std::lock_guard lock{mutex}; object->SetIsRegistered(true); interval_cache.add({GetInterval(object), ObjectSet{object}}); - map_cache.insert({object->GetCacheAddr(), object}); + map_cache.insert({object->GetCpuAddr(), object}); rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), 1); } @@ -144,7 +128,7 @@ protected: object->SetIsRegistered(false); rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), -1); - const CacheAddr addr = object->GetCacheAddr(); + const VAddr addr = object->GetCpuAddr(); interval_cache.subtract({GetInterval(object), ObjectSet{object}}); map_cache.erase(addr); } @@ -173,7 +157,7 @@ protected: private: /// Returns a list of cached objects from the specified memory region, ordered by access time - std::vector<T> GetSortedObjectsFromRegion(CacheAddr addr, u64 size) { + std::vector<T> GetSortedObjectsFromRegion(VAddr addr, u64 size) { if (size == 0) { return {}; } @@ -197,13 +181,13 @@ private: } using ObjectSet = std::set<T>; - using ObjectCache = std::unordered_map<CacheAddr, T>; - using IntervalCache = boost::icl::interval_map<CacheAddr, ObjectSet>; + using ObjectCache = std::unordered_map<VAddr, T>; + using IntervalCache = boost::icl::interval_map<VAddr, ObjectSet>; using ObjectInterval = typename IntervalCache::interval_type; static auto GetInterval(const T& object) { - return ObjectInterval::right_open(object->GetCacheAddr(), - object->GetCacheAddr() + object->GetSizeInBytes()); + return ObjectInterval::right_open(object->GetCpuAddr(), + object->GetCpuAddr() + object->GetSizeInBytes()); } ObjectCache map_cache; diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h index 1a68e3caa..8ae5b9c4e 100644 --- a/src/video_core/rasterizer_interface.h +++ b/src/video_core/rasterizer_interface.h @@ -53,14 +53,14 @@ public: virtual void FlushAll() = 0; /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory - virtual void FlushRegion(CacheAddr addr, u64 size) = 0; + virtual void FlushRegion(VAddr addr, u64 size) = 0; /// Notify rasterizer that any caches of the specified region should be invalidated - virtual void InvalidateRegion(CacheAddr addr, u64 size) = 0; + virtual void InvalidateRegion(VAddr addr, u64 size) = 0; /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory /// and invalidated - virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0; + virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0; /// Notify the rasterizer to send all written commands to the host GPU. virtual void FlushCommands() = 0; diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp index 0375fca17..4eb37a96c 100644 --- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp @@ -21,8 +21,8 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs; MICROPROFILE_DEFINE(OpenGL_Buffer_Download, "OpenGL", "Buffer Download", MP_RGB(192, 192, 128)); -CachedBufferBlock::CachedBufferBlock(CacheAddr cache_addr, const std::size_t size) - : VideoCommon::BufferBlock{cache_addr, size} { +CachedBufferBlock::CachedBufferBlock(VAddr cpu_addr, const std::size_t size) + : VideoCommon::BufferBlock{cpu_addr, size} { gl_buffer.Create(); glNamedBufferData(gl_buffer.handle, static_cast<GLsizeiptr>(size), nullptr, GL_DYNAMIC_DRAW); } @@ -47,8 +47,8 @@ OGLBufferCache::~OGLBufferCache() { glDeleteBuffers(static_cast<GLsizei>(std::size(cbufs)), std::data(cbufs)); } -Buffer OGLBufferCache::CreateBlock(CacheAddr cache_addr, std::size_t size) { - return std::make_shared<CachedBufferBlock>(cache_addr, size); +Buffer OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) { + return std::make_shared<CachedBufferBlock>(cpu_addr, size); } void OGLBufferCache::WriteBarrier() { diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h index 8c7145443..d94a11252 100644 --- a/src/video_core/renderer_opengl/gl_buffer_cache.h +++ b/src/video_core/renderer_opengl/gl_buffer_cache.h @@ -31,7 +31,7 @@ using GenericBufferCache = VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuf class CachedBufferBlock : public VideoCommon::BufferBlock { public: - explicit CachedBufferBlock(CacheAddr cache_addr, const std::size_t size); + explicit CachedBufferBlock(VAddr cpu_addr, const std::size_t size); ~CachedBufferBlock(); const GLuint* GetHandle() const { @@ -55,7 +55,7 @@ public: } protected: - Buffer CreateBlock(CacheAddr cache_addr, std::size_t size) override; + Buffer CreateBlock(VAddr cpu_addr, std::size_t size) override; void WriteBarrier() override; diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp index 1a2e2a9f7..c286502ba 100644 --- a/src/video_core/renderer_opengl/gl_device.cpp +++ b/src/video_core/renderer_opengl/gl_device.cpp @@ -131,6 +131,31 @@ std::array<Device::BaseBindings, Tegra::Engines::MaxShaderTypes> BuildBaseBindin return bindings; } +bool IsASTCSupported() { + static constexpr std::array formats = { + GL_COMPRESSED_RGBA_ASTC_4x4_KHR, GL_COMPRESSED_RGBA_ASTC_5x4_KHR, + GL_COMPRESSED_RGBA_ASTC_5x5_KHR, GL_COMPRESSED_RGBA_ASTC_6x5_KHR, + GL_COMPRESSED_RGBA_ASTC_6x6_KHR, GL_COMPRESSED_RGBA_ASTC_8x5_KHR, + GL_COMPRESSED_RGBA_ASTC_8x6_KHR, GL_COMPRESSED_RGBA_ASTC_8x8_KHR, + GL_COMPRESSED_RGBA_ASTC_10x5_KHR, GL_COMPRESSED_RGBA_ASTC_10x6_KHR, + GL_COMPRESSED_RGBA_ASTC_10x8_KHR, GL_COMPRESSED_RGBA_ASTC_10x10_KHR, + GL_COMPRESSED_RGBA_ASTC_12x10_KHR, GL_COMPRESSED_RGBA_ASTC_12x12_KHR, + GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x4_KHR, + GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x5_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x5_KHR, + GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x6_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x5_KHR, + GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x6_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x8_KHR, + GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x5_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x6_KHR, + GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x8_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x10_KHR, + GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x10_KHR, GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x12_KHR, + }; + return std::find_if_not(formats.begin(), formats.end(), [](GLenum format) { + GLint supported; + glGetInternalformativ(GL_TEXTURE_2D, format, GL_INTERNALFORMAT_SUPPORTED, 1, + &supported); + return supported == GL_TRUE; + }) == formats.end(); +} + } // Anonymous namespace Device::Device() : base_bindings{BuildBaseBindings()} { @@ -152,6 +177,7 @@ Device::Device() : base_bindings{BuildBaseBindings()} { has_shader_ballot = GLAD_GL_ARB_shader_ballot; has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array; has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted"); + has_astc = IsASTCSupported(); has_variable_aoffi = TestVariableAoffi(); has_component_indexing_bug = is_amd; has_precise_bug = TestPreciseBug(); diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h index d73b099d0..a55050cb5 100644 --- a/src/video_core/renderer_opengl/gl_device.h +++ b/src/video_core/renderer_opengl/gl_device.h @@ -64,6 +64,10 @@ public: return has_image_load_formatted; } + bool HasASTC() const { + return has_astc; + } + bool HasVariableAoffi() const { return has_variable_aoffi; } @@ -97,6 +101,7 @@ private: bool has_shader_ballot{}; bool has_vertex_viewport_layer{}; bool has_image_load_formatted{}; + bool has_astc{}; bool has_variable_aoffi{}; bool has_component_indexing_bug{}; bool has_precise_bug{}; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 31add708f..f4598fbf7 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -140,8 +140,8 @@ void RasterizerOpenGL::SetupVertexFormat() { const auto attrib = gpu.regs.vertex_attrib_format[index]; const auto gl_index = static_cast<GLuint>(index); - // Ignore invalid attributes. - if (!attrib.IsValid()) { + // Disable constant attributes. + if (attrib.IsConstant()) { glDisableVertexAttribArray(gl_index); continue; } @@ -345,7 +345,7 @@ void RasterizerOpenGL::ConfigureFramebuffers() { texture_cache.GuardRenderTargets(true); - View depth_surface = texture_cache.GetDepthBufferSurface(true); + View depth_surface = texture_cache.GetDepthBufferSurface(); const auto& regs = gpu.regs; UNIMPLEMENTED_IF(regs.rt_separate_frag_data == 0); @@ -354,7 +354,7 @@ void RasterizerOpenGL::ConfigureFramebuffers() { FramebufferCacheKey key; const auto colors_count = static_cast<std::size_t>(regs.rt_control.count); for (std::size_t index = 0; index < colors_count; ++index) { - View color_surface{texture_cache.GetColorBufferSurface(index, true)}; + View color_surface{texture_cache.GetColorBufferSurface(index)}; if (!color_surface) { continue; } @@ -386,11 +386,14 @@ void RasterizerOpenGL::ConfigureClearFramebuffer(bool using_color_fb, bool using texture_cache.GuardRenderTargets(true); View color_surface; if (using_color_fb) { - color_surface = texture_cache.GetColorBufferSurface(regs.clear_buffers.RT, false); + const std::size_t index = regs.clear_buffers.RT; + color_surface = texture_cache.GetColorBufferSurface(index); + texture_cache.MarkColorBufferInUse(index); } View depth_surface; if (using_depth_fb || using_stencil_fb) { - depth_surface = texture_cache.GetDepthBufferSurface(false); + depth_surface = texture_cache.GetDepthBufferSurface(); + texture_cache.MarkDepthBufferInUse(); } texture_cache.GuardRenderTargets(false); @@ -493,6 +496,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) { SyncPrimitiveRestart(); SyncScissorTest(); SyncPointState(); + SyncLineState(); SyncPolygonOffset(); SyncAlphaTest(); SyncFramebufferSRGB(); @@ -653,9 +657,9 @@ void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCore::QueryType type, void RasterizerOpenGL::FlushAll() {} -void RasterizerOpenGL::FlushRegion(CacheAddr addr, u64 size) { +void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) { MICROPROFILE_SCOPE(OpenGL_CacheManagement); - if (!addr || !size) { + if (addr == 0 || size == 0) { return; } texture_cache.FlushRegion(addr, size); @@ -663,9 +667,9 @@ void RasterizerOpenGL::FlushRegion(CacheAddr addr, u64 size) { query_cache.FlushRegion(addr, size); } -void RasterizerOpenGL::InvalidateRegion(CacheAddr addr, u64 size) { +void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) { MICROPROFILE_SCOPE(OpenGL_CacheManagement); - if (!addr || !size) { + if (addr == 0 || size == 0) { return; } texture_cache.InvalidateRegion(addr, size); @@ -674,7 +678,7 @@ void RasterizerOpenGL::InvalidateRegion(CacheAddr addr, u64 size) { query_cache.InvalidateRegion(addr, size); } -void RasterizerOpenGL::FlushAndInvalidateRegion(CacheAddr addr, u64 size) { +void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) { if (Settings::values.use_accurate_gpu_emulation) { FlushRegion(addr, size); } @@ -713,8 +717,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config, MICROPROFILE_SCOPE(OpenGL_CacheManagement); - const auto surface{ - texture_cache.TryFindFramebufferSurface(system.Memory().GetPointer(framebuffer_addr))}; + const auto surface{texture_cache.TryFindFramebufferSurface(framebuffer_addr)}; if (!surface) { return {}; } @@ -1309,6 +1312,19 @@ void RasterizerOpenGL::SyncPointState() { glDisable(GL_PROGRAM_POINT_SIZE); } +void RasterizerOpenGL::SyncLineState() { + auto& gpu = system.GPU().Maxwell3D(); + auto& flags = gpu.dirty.flags; + if (!flags[Dirty::LineWidth]) { + return; + } + flags[Dirty::LineWidth] = false; + + const auto& regs = gpu.regs; + oglEnable(GL_LINE_SMOOTH, regs.line_smooth_enable); + glLineWidth(regs.line_smooth_enable ? regs.line_width_smooth : regs.line_width_aliased); +} + void RasterizerOpenGL::SyncPolygonOffset() { auto& gpu = system.GPU().Maxwell3D(); auto& flags = gpu.dirty.flags; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 2d3be2437..435da4425 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -65,9 +65,9 @@ public: void ResetCounter(VideoCore::QueryType type) override; void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override; void FlushAll() override; - void FlushRegion(CacheAddr addr, u64 size) override; - void InvalidateRegion(CacheAddr addr, u64 size) override; - void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override; + void FlushRegion(VAddr addr, u64 size) override; + void InvalidateRegion(VAddr addr, u64 size) override; + void FlushAndInvalidateRegion(VAddr addr, u64 size) override; void FlushCommands() override; void TickFrame() override; bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src, @@ -171,6 +171,9 @@ private: /// Syncs the point state to match the guest state void SyncPointState(); + /// Syncs the line state to match the guest state + void SyncLineState(); + /// Syncs the rasterizer enable state to match the guest state void SyncRasterizeEnable(); diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index 046ee55a5..12c6dcfde 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -34,6 +34,8 @@ namespace OpenGL { using Tegra::Engines::ShaderType; +using VideoCommon::Shader::CompileDepth; +using VideoCommon::Shader::CompilerSettings; using VideoCommon::Shader::ProgramCode; using VideoCommon::Shader::Registry; using VideoCommon::Shader::ShaderIR; @@ -43,7 +45,7 @@ namespace { constexpr u32 STAGE_MAIN_OFFSET = 10; constexpr u32 KERNEL_MAIN_OFFSET = 0; -constexpr VideoCommon::Shader::CompilerSettings COMPILER_SETTINGS{}; +constexpr CompilerSettings COMPILER_SETTINGS{CompileDepth::FullDecompile}; /// Gets the address for the specified shader stage program GPUVAddr GetShaderAddress(Core::System& system, Maxwell::ShaderProgram program) { @@ -214,11 +216,11 @@ std::unordered_set<GLenum> GetSupportedFormats() { } // Anonymous namespace -CachedShader::CachedShader(const u8* host_ptr, VAddr cpu_addr, std::size_t size_in_bytes, +CachedShader::CachedShader(VAddr cpu_addr, std::size_t size_in_bytes, std::shared_ptr<VideoCommon::Shader::Registry> registry, ShaderEntries entries, std::shared_ptr<OGLProgram> program) - : RasterizerCacheObject{host_ptr}, registry{std::move(registry)}, entries{std::move(entries)}, - cpu_addr{cpu_addr}, size_in_bytes{size_in_bytes}, program{std::move(program)} {} + : RasterizerCacheObject{cpu_addr}, registry{std::move(registry)}, entries{std::move(entries)}, + size_in_bytes{size_in_bytes}, program{std::move(program)} {} CachedShader::~CachedShader() = default; @@ -254,9 +256,8 @@ Shader CachedShader::CreateStageFromMemory(const ShaderParameters& params, entry.bindless_samplers = registry->GetBindlessSamplers(); params.disk_cache.SaveEntry(std::move(entry)); - return std::shared_ptr<CachedShader>(new CachedShader(params.host_ptr, params.cpu_addr, - size_in_bytes, std::move(registry), - MakeEntries(ir), std::move(program))); + return std::shared_ptr<CachedShader>(new CachedShader( + params.cpu_addr, size_in_bytes, std::move(registry), MakeEntries(ir), std::move(program))); } Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, ProgramCode code) { @@ -279,17 +280,16 @@ Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, Prog entry.bindless_samplers = registry->GetBindlessSamplers(); params.disk_cache.SaveEntry(std::move(entry)); - return std::shared_ptr<CachedShader>(new CachedShader(params.host_ptr, params.cpu_addr, - size_in_bytes, std::move(registry), - MakeEntries(ir), std::move(program))); + return std::shared_ptr<CachedShader>(new CachedShader( + params.cpu_addr, size_in_bytes, std::move(registry), MakeEntries(ir), std::move(program))); } Shader CachedShader::CreateFromCache(const ShaderParameters& params, const PrecompiledShader& precompiled_shader, std::size_t size_in_bytes) { - return std::shared_ptr<CachedShader>(new CachedShader( - params.host_ptr, params.cpu_addr, size_in_bytes, precompiled_shader.registry, - precompiled_shader.entries, precompiled_shader.program)); + return std::shared_ptr<CachedShader>( + new CachedShader(params.cpu_addr, size_in_bytes, precompiled_shader.registry, + precompiled_shader.entries, precompiled_shader.program)); } ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system, @@ -449,12 +449,14 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) { const GPUVAddr address{GetShaderAddress(system, program)}; // Look up shader in the cache based on address - const auto host_ptr{memory_manager.GetPointer(address)}; - Shader shader{TryGet(host_ptr)}; + const auto cpu_addr{memory_manager.GpuToCpuAddress(address)}; + Shader shader{cpu_addr ? TryGet(*cpu_addr) : nullptr}; if (shader) { return last_shaders[static_cast<std::size_t>(program)] = shader; } + const auto host_ptr{memory_manager.GetPointer(address)}; + // No shader found - create a new one ProgramCode code{GetShaderCode(memory_manager, address, host_ptr)}; ProgramCode code_b; @@ -465,9 +467,9 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) { const auto unique_identifier = GetUniqueIdentifier( GetShaderType(program), program == Maxwell::ShaderProgram::VertexA, code, code_b); - const auto cpu_addr{*memory_manager.GpuToCpuAddress(address)}; - const ShaderParameters params{system, disk_cache, device, - cpu_addr, host_ptr, unique_identifier}; + + const ShaderParameters params{system, disk_cache, device, + *cpu_addr, host_ptr, unique_identifier}; const auto found = runtime_cache.find(unique_identifier); if (found == runtime_cache.end()) { @@ -484,18 +486,20 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) { Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) { auto& memory_manager{system.GPU().MemoryManager()}; - const auto host_ptr{memory_manager.GetPointer(code_addr)}; - auto kernel = TryGet(host_ptr); + const auto cpu_addr{memory_manager.GpuToCpuAddress(code_addr)}; + + auto kernel = cpu_addr ? TryGet(*cpu_addr) : nullptr; if (kernel) { return kernel; } + const auto host_ptr{memory_manager.GetPointer(code_addr)}; // No kernel found, create a new one auto code{GetShaderCode(memory_manager, code_addr, host_ptr)}; const auto unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)}; - const auto cpu_addr{*memory_manager.GpuToCpuAddress(code_addr)}; - const ShaderParameters params{system, disk_cache, device, - cpu_addr, host_ptr, unique_identifier}; + + const ShaderParameters params{system, disk_cache, device, + *cpu_addr, host_ptr, unique_identifier}; const auto found = runtime_cache.find(unique_identifier); if (found == runtime_cache.end()) { diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h index 4935019fc..c836df5bd 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.h +++ b/src/video_core/renderer_opengl/gl_shader_cache.h @@ -65,11 +65,6 @@ public: /// Gets the GL program handle for the shader GLuint GetHandle() const; - /// Returns the guest CPU address of the shader - VAddr GetCpuAddr() const override { - return cpu_addr; - } - /// Returns the size in bytes of the shader std::size_t GetSizeInBytes() const override { return size_in_bytes; @@ -90,13 +85,12 @@ public: std::size_t size_in_bytes); private: - explicit CachedShader(const u8* host_ptr, VAddr cpu_addr, std::size_t size_in_bytes, + explicit CachedShader(VAddr cpu_addr, std::size_t size_in_bytes, std::shared_ptr<VideoCommon::Shader::Registry> registry, ShaderEntries entries, std::shared_ptr<OGLProgram> program); std::shared_ptr<VideoCommon::Shader::Registry> registry; ShaderEntries entries; - VAddr cpu_addr = 0; std::size_t size_in_bytes = 0; std::shared_ptr<OGLProgram> program; }; diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index a25280a47..b1804e9ea 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -31,11 +31,11 @@ namespace { using Tegra::Engines::ShaderType; using Tegra::Shader::Attribute; -using Tegra::Shader::AttributeUse; using Tegra::Shader::Header; using Tegra::Shader::IpaInterpMode; using Tegra::Shader::IpaMode; using Tegra::Shader::IpaSampleMode; +using Tegra::Shader::PixelImap; using Tegra::Shader::Register; using VideoCommon::Shader::BuildTransformFeedback; using VideoCommon::Shader::Registry; @@ -702,20 +702,19 @@ private: code.AddNewLine(); } - std::string GetInputFlags(AttributeUse attribute) { + const char* GetInputFlags(PixelImap attribute) { switch (attribute) { - case AttributeUse::Perspective: - // Default, Smooth - return {}; - case AttributeUse::Constant: - return "flat "; - case AttributeUse::ScreenLinear: - return "noperspective "; - default: - case AttributeUse::Unused: - UNIMPLEMENTED_MSG("Unknown attribute usage index={}", static_cast<u32>(attribute)); - return {}; + case PixelImap::Perspective: + return "smooth"; + case PixelImap::Constant: + return "flat"; + case PixelImap::ScreenLinear: + return "noperspective"; + case PixelImap::Unused: + break; } + UNIMPLEMENTED_MSG("Unknown attribute usage index={}", static_cast<int>(attribute)); + return {}; } void DeclareInputAttributes() { @@ -749,8 +748,8 @@ private: std::string suffix; if (stage == ShaderType::Fragment) { - const auto input_mode{header.ps.GetAttributeUse(location)}; - if (skip_unused && input_mode == AttributeUse::Unused) { + const auto input_mode{header.ps.GetPixelImap(location)}; + if (input_mode == PixelImap::Unused) { return; } suffix = GetInputFlags(input_mode); @@ -927,7 +926,7 @@ private: const u32 address{generic_base + index * generic_stride + element * element_stride}; const bool declared = stage != ShaderType::Fragment || - header.ps.GetAttributeUse(index) != AttributeUse::Unused; + header.ps.GetPixelImap(index) != PixelImap::Unused; const std::string value = declared ? ReadAttribute(attribute, element).AsFloat() : "0.0f"; code.AddLine("case 0x{:X}U: return {};", address, value); @@ -1142,8 +1141,7 @@ private: GetSwizzle(element)), Type::Float}; case ShaderType::Fragment: - return {element == 3 ? "1.0f" : ("gl_FragCoord"s + GetSwizzle(element)), - Type::Float}; + return {"gl_FragCoord"s + GetSwizzle(element), Type::Float}; default: UNREACHABLE(); } @@ -1821,15 +1819,17 @@ private: } Expression HMergeH0(Operation operation) { - std::string dest = VisitOperand(operation, 0).AsUint(); - std::string src = VisitOperand(operation, 1).AsUint(); - return {fmt::format("(({} & 0x0000FFFFU) | ({} & 0xFFFF0000U))", src, dest), Type::Uint}; + const std::string dest = VisitOperand(operation, 0).AsUint(); + const std::string src = VisitOperand(operation, 1).AsUint(); + return {fmt::format("vec2(unpackHalf2x16({}).x, unpackHalf2x16({}).y)", src, dest), + Type::HalfFloat}; } Expression HMergeH1(Operation operation) { - std::string dest = VisitOperand(operation, 0).AsUint(); - std::string src = VisitOperand(operation, 1).AsUint(); - return {fmt::format("(({} & 0x0000FFFFU) | ({} & 0xFFFF0000U))", dest, src), Type::Uint}; + const std::string dest = VisitOperand(operation, 0).AsUint(); + const std::string src = VisitOperand(operation, 1).AsUint(); + return {fmt::format("vec2(unpackHalf2x16({}).x, unpackHalf2x16({}).y)", dest, src), + Type::HalfFloat}; } Expression HPack2(Operation operation) { diff --git a/src/video_core/renderer_opengl/gl_state_tracker.cpp b/src/video_core/renderer_opengl/gl_state_tracker.cpp index 255ac3147..d24fad3de 100644 --- a/src/video_core/renderer_opengl/gl_state_tracker.cpp +++ b/src/video_core/renderer_opengl/gl_state_tracker.cpp @@ -185,6 +185,12 @@ void SetupDirtyPointSize(Tables& tables) { tables[0][OFF(point_sprite_enable)] = PointSize; } +void SetupDirtyLineWidth(Tables& tables) { + tables[0][OFF(line_width_smooth)] = LineWidth; + tables[0][OFF(line_width_aliased)] = LineWidth; + tables[0][OFF(line_smooth_enable)] = LineWidth; +} + void SetupDirtyClipControl(Tables& tables) { auto& table = tables[0]; table[OFF(screen_y_control)] = ClipControl; @@ -233,6 +239,7 @@ void StateTracker::Initialize() { SetupDirtyLogicOp(tables); SetupDirtyFragmentClampColor(tables); SetupDirtyPointSize(tables); + SetupDirtyLineWidth(tables); SetupDirtyClipControl(tables); SetupDirtyDepthClampEnabled(tables); SetupDirtyMisc(tables); diff --git a/src/video_core/renderer_opengl/gl_state_tracker.h b/src/video_core/renderer_opengl/gl_state_tracker.h index b882d75c3..0f823288e 100644 --- a/src/video_core/renderer_opengl/gl_state_tracker.h +++ b/src/video_core/renderer_opengl/gl_state_tracker.h @@ -78,6 +78,7 @@ enum : u8 { LogicOp, FragmentClampColor, PointSize, + LineWidth, ClipControl, DepthClampEnabled, diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index f424e3000..2729d1265 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -24,7 +24,6 @@ using Tegra::Texture::SwizzleSource; using VideoCore::MortonSwizzleMode; using VideoCore::Surface::PixelFormat; -using VideoCore::Surface::SurfaceCompression; using VideoCore::Surface::SurfaceTarget; using VideoCore::Surface::SurfaceType; @@ -37,102 +36,100 @@ namespace { struct FormatTuple { GLint internal_format; - GLenum format; - GLenum type; - bool compressed; + GLenum format = GL_NONE; + GLenum type = GL_NONE; }; constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format_tuples = {{ - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false}, // ABGR8U - {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE, false}, // ABGR8S - {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE, false}, // ABGR8UI - {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, false}, // B5G6R5U - {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, false}, // A2B10G10R10U - {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, false}, // A1B5G5R5U - {GL_R8, GL_RED, GL_UNSIGNED_BYTE, false}, // R8U - {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE, false}, // R8UI - {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, false}, // RGBA16F - {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT, false}, // RGBA16U - {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT, false}, // RGBA16S - {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT, false}, // RGBA16UI - {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, false}, // R11FG11FB10F - {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, false}, // RGBA32UI - {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT1 - {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT23 - {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT45 - {GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, true}, // DXN1 - {GL_COMPRESSED_RG_RGTC2, GL_RG, GL_UNSIGNED_INT_8_8_8_8, true}, // DXN2UNORM - {GL_COMPRESSED_SIGNED_RG_RGTC2, GL_RG, GL_INT, true}, // DXN2SNORM - {GL_COMPRESSED_RGBA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // BC7U - {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, true}, // BC6H_UF16 - {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, true}, // BC6H_SF16 - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_4X4 - {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE, false}, // BGRA8 - {GL_RGBA32F, GL_RGBA, GL_FLOAT, false}, // RGBA32F - {GL_RG32F, GL_RG, GL_FLOAT, false}, // RG32F - {GL_R32F, GL_RED, GL_FLOAT, false}, // R32F - {GL_R16F, GL_RED, GL_HALF_FLOAT, false}, // R16F - {GL_R16, GL_RED, GL_UNSIGNED_SHORT, false}, // R16U - {GL_R16_SNORM, GL_RED, GL_SHORT, false}, // R16S - {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT, false}, // R16UI - {GL_R16I, GL_RED_INTEGER, GL_SHORT, false}, // R16I - {GL_RG16, GL_RG, GL_UNSIGNED_SHORT, false}, // RG16 - {GL_RG16F, GL_RG, GL_HALF_FLOAT, false}, // RG16F - {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT, false}, // RG16UI - {GL_RG16I, GL_RG_INTEGER, GL_SHORT, false}, // RG16I - {GL_RG16_SNORM, GL_RG, GL_SHORT, false}, // RG16S - {GL_RGB32F, GL_RGB, GL_FLOAT, false}, // RGB32F - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false}, // RGBA8_SRGB - {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, false}, // RG8U - {GL_RG8_SNORM, GL_RG, GL_BYTE, false}, // RG8S - {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT, false}, // RG32UI - {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT, false}, // RGBX16F - {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT, false}, // R32UI - {GL_R32I, GL_RED_INTEGER, GL_INT, false}, // R32I - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X8 - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X5 - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_5X4 - {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE, false}, // BGRA8 + {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // ABGR8U + {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE}, // ABGR8S + {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE}, // ABGR8UI + {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV}, // B5G6R5U + {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV}, // A2B10G10R10U + {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV}, // A1B5G5R5U + {GL_R8, GL_RED, GL_UNSIGNED_BYTE}, // R8U + {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE}, // R8UI + {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT}, // RGBA16F + {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT}, // RGBA16U + {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT}, // RGBA16S + {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT}, // RGBA16UI + {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV}, // R11FG11FB10F + {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT}, // RGBA32UI + {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT}, // DXT1 + {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT}, // DXT23 + {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT}, // DXT45 + {GL_COMPRESSED_RED_RGTC1}, // DXN1 + {GL_COMPRESSED_RG_RGTC2}, // DXN2UNORM + {GL_COMPRESSED_SIGNED_RG_RGTC2}, // DXN2SNORM + {GL_COMPRESSED_RGBA_BPTC_UNORM}, // BC7U + {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT}, // BC6H_UF16 + {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT}, // BC6H_SF16 + {GL_COMPRESSED_RGBA_ASTC_4x4_KHR}, // ASTC_2D_4X4 + {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE}, // BGRA8 + {GL_RGBA32F, GL_RGBA, GL_FLOAT}, // RGBA32F + {GL_RG32F, GL_RG, GL_FLOAT}, // RG32F + {GL_R32F, GL_RED, GL_FLOAT}, // R32F + {GL_R16F, GL_RED, GL_HALF_FLOAT}, // R16F + {GL_R16, GL_RED, GL_UNSIGNED_SHORT}, // R16U + {GL_R16_SNORM, GL_RED, GL_SHORT}, // R16S + {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT}, // R16UI + {GL_R16I, GL_RED_INTEGER, GL_SHORT}, // R16I + {GL_RG16, GL_RG, GL_UNSIGNED_SHORT}, // RG16 + {GL_RG16F, GL_RG, GL_HALF_FLOAT}, // RG16F + {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT}, // RG16UI + {GL_RG16I, GL_RG_INTEGER, GL_SHORT}, // RG16I + {GL_RG16_SNORM, GL_RG, GL_SHORT}, // RG16S + {GL_RGB32F, GL_RGB, GL_FLOAT}, // RGB32F + {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // RGBA8_SRGB + {GL_RG8, GL_RG, GL_UNSIGNED_BYTE}, // RG8U + {GL_RG8_SNORM, GL_RG, GL_BYTE}, // RG8S + {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT}, // RG32UI + {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT}, // RGBX16F + {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT}, // R32UI + {GL_R32I, GL_RED_INTEGER, GL_INT}, // R32I + {GL_COMPRESSED_RGBA_ASTC_8x8_KHR}, // ASTC_2D_8X8 + {GL_COMPRESSED_RGBA_ASTC_8x5_KHR}, // ASTC_2D_8X5 + {GL_COMPRESSED_RGBA_ASTC_5x4_KHR}, // ASTC_2D_5X4 + {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE}, // BGRA8 // Compressed sRGB formats - {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT1_SRGB - {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT23_SRGB - {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT45_SRGB - {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // BC7U_SRGB - {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4_REV, false}, // R4G4B4A4U - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_4X4_SRGB - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X8_SRGB - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X5_SRGB - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_5X4_SRGB - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_5X5 - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_5X5_SRGB - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_10X8 - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_10X8_SRGB - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_6X6 - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_6X6_SRGB - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_10X10 - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_10X10_SRGB - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_12X12 - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_12X12_SRGB - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X6 - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_8X6_SRGB - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_6X5 - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_6X5_SRGB - {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV, false}, // E5B9G9R9F + {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT}, // DXT1_SRGB + {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT}, // DXT23_SRGB + {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT}, // DXT45_SRGB + {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM}, // BC7U_SRGB + {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4_REV}, // R4G4B4A4U + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR}, // ASTC_2D_4X4_SRGB + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x8_KHR}, // ASTC_2D_8X8_SRGB + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x5_KHR}, // ASTC_2D_8X5_SRGB + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x4_KHR}, // ASTC_2D_5X4_SRGB + {GL_COMPRESSED_RGBA_ASTC_5x5_KHR}, // ASTC_2D_5X5 + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x5_KHR}, // ASTC_2D_5X5_SRGB + {GL_COMPRESSED_RGBA_ASTC_10x8_KHR}, // ASTC_2D_10X8 + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x8_KHR}, // ASTC_2D_10X8_SRGB + {GL_COMPRESSED_RGBA_ASTC_6x6_KHR}, // ASTC_2D_6X6 + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x6_KHR}, // ASTC_2D_6X6_SRGB + {GL_COMPRESSED_RGBA_ASTC_10x10_KHR}, // ASTC_2D_10X10 + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x10_KHR}, // ASTC_2D_10X10_SRGB + {GL_COMPRESSED_RGBA_ASTC_12x12_KHR}, // ASTC_2D_12X12 + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x12_KHR}, // ASTC_2D_12X12_SRGB + {GL_COMPRESSED_RGBA_ASTC_8x6_KHR}, // ASTC_2D_8X6 + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x6_KHR}, // ASTC_2D_8X6_SRGB + {GL_COMPRESSED_RGBA_ASTC_6x5_KHR}, // ASTC_2D_6X5 + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x5_KHR}, // ASTC_2D_6X5_SRGB + {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV}, // E5B9G9R9F // Depth formats - {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT, false}, // Z32F - {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT, false}, // Z16 + {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT}, // Z32F + {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // Z16 // DepthStencil formats - {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, false}, // Z24S8 - {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, false}, // S8Z24 - {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV, false}, // Z32FS8 + {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // Z24S8 + {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // S8Z24 + {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV}, // Z32FS8 }}; const FormatTuple& GetFormatTuple(PixelFormat pixel_format) { ASSERT(static_cast<std::size_t>(pixel_format) < tex_format_tuples.size()); - const auto& format{tex_format_tuples[static_cast<std::size_t>(pixel_format)]}; - return format; + return tex_format_tuples[static_cast<std::size_t>(pixel_format)]; } GLenum GetTextureTarget(const SurfaceTarget& target) { @@ -242,13 +239,20 @@ OGLTexture CreateTexture(const SurfaceParams& params, GLenum target, GLenum inte } // Anonymous namespace -CachedSurface::CachedSurface(const GPUVAddr gpu_addr, const SurfaceParams& params) - : VideoCommon::SurfaceBase<View>(gpu_addr, params) { - const auto& tuple{GetFormatTuple(params.pixel_format)}; - internal_format = tuple.internal_format; - format = tuple.format; - type = tuple.type; - is_compressed = tuple.compressed; +CachedSurface::CachedSurface(const GPUVAddr gpu_addr, const SurfaceParams& params, + bool is_astc_supported) + : VideoCommon::SurfaceBase<View>(gpu_addr, params, is_astc_supported) { + if (is_converted) { + internal_format = params.srgb_conversion ? GL_SRGB8_ALPHA8 : GL_RGBA8; + format = GL_RGBA; + type = GL_UNSIGNED_BYTE; + } else { + const auto& tuple{GetFormatTuple(params.pixel_format)}; + internal_format = tuple.internal_format; + format = tuple.format; + type = tuple.type; + is_compressed = params.IsCompressed(); + } target = GetTextureTarget(params.target); texture = CreateTexture(params, target, internal_format, texture_buffer); DecorateSurfaceName(); @@ -264,7 +268,7 @@ void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) { if (params.IsBuffer()) { glGetNamedBufferSubData(texture_buffer.handle, 0, - static_cast<GLsizeiptr>(params.GetHostSizeInBytes()), + static_cast<GLsizeiptr>(params.GetHostSizeInBytes(false)), staging_buffer.data()); return; } @@ -272,9 +276,10 @@ void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) { SCOPE_EXIT({ glPixelStorei(GL_PACK_ROW_LENGTH, 0); }); for (u32 level = 0; level < params.emulated_levels; ++level) { - glPixelStorei(GL_PACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level))); + glPixelStorei(GL_PACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level, is_converted))); glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(params.GetMipWidth(level))); - const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level); + const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted); + u8* const mip_data = staging_buffer.data() + mip_offset; const GLsizei size = static_cast<GLsizei>(params.GetHostMipmapSize(level)); if (is_compressed) { @@ -294,14 +299,10 @@ void CachedSurface::UploadTexture(const std::vector<u8>& staging_buffer) { } void CachedSurface::UploadTextureMipmap(u32 level, const std::vector<u8>& staging_buffer) { - glPixelStorei(GL_UNPACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level))); + glPixelStorei(GL_UNPACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level, is_converted))); glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(params.GetMipWidth(level))); - auto compression_type = params.GetCompressionType(); - - const std::size_t mip_offset = compression_type == SurfaceCompression::Converted - ? params.GetConvertedMipmapOffset(level) - : params.GetHostMipmapLevelOffset(level); + const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted); const u8* buffer{staging_buffer.data() + mip_offset}; if (is_compressed) { const auto image_size{static_cast<GLsizei>(params.GetHostMipmapSize(level))}; @@ -410,14 +411,13 @@ CachedSurfaceView::~CachedSurfaceView() = default; void CachedSurfaceView::Attach(GLenum attachment, GLenum target) const { ASSERT(params.num_levels == 1); - const GLuint texture = surface.GetTexture(); if (params.num_layers > 1) { // Layered framebuffer attachments UNIMPLEMENTED_IF(params.base_layer != 0); switch (params.target) { case SurfaceTarget::Texture2DArray: - glFramebufferTexture(target, attachment, texture, params.base_level); + glFramebufferTexture(target, attachment, GetTexture(), 0); break; default: UNIMPLEMENTED(); @@ -426,6 +426,7 @@ void CachedSurfaceView::Attach(GLenum attachment, GLenum target) const { } const GLenum view_target = surface.GetTarget(); + const GLuint texture = surface.GetTexture(); switch (surface.GetSurfaceParams().target) { case SurfaceTarget::Texture1D: glFramebufferTexture1D(target, attachment, view_target, texture, params.base_level); @@ -482,7 +483,7 @@ OGLTextureView CachedSurfaceView::CreateTextureView() const { TextureCacheOpenGL::TextureCacheOpenGL(Core::System& system, VideoCore::RasterizerInterface& rasterizer, const Device& device, StateTracker& state_tracker) - : TextureCacheBase{system, rasterizer}, state_tracker{state_tracker} { + : TextureCacheBase{system, rasterizer, device.HasASTC()}, state_tracker{state_tracker} { src_framebuffer.Create(); dst_framebuffer.Create(); } @@ -490,7 +491,7 @@ TextureCacheOpenGL::TextureCacheOpenGL(Core::System& system, TextureCacheOpenGL::~TextureCacheOpenGL() = default; Surface TextureCacheOpenGL::CreateSurface(GPUVAddr gpu_addr, const SurfaceParams& params) { - return std::make_shared<CachedSurface>(gpu_addr, params); + return std::make_shared<CachedSurface>(gpu_addr, params, is_astc_supported); } void TextureCacheOpenGL::ImageCopy(Surface& src_surface, Surface& dst_surface, @@ -596,7 +597,7 @@ void TextureCacheOpenGL::BufferCopy(Surface& src_surface, Surface& dst_surface) glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle); - if (source_format.compressed) { + if (src_surface->IsCompressed()) { glGetCompressedTextureImage(src_surface->GetTexture(), 0, static_cast<GLsizei>(source_size), nullptr); } else { @@ -610,7 +611,7 @@ void TextureCacheOpenGL::BufferCopy(Surface& src_surface, Surface& dst_surface) const GLsizei width = static_cast<GLsizei>(dst_params.width); const GLsizei height = static_cast<GLsizei>(dst_params.height); const GLsizei depth = static_cast<GLsizei>(dst_params.depth); - if (dest_format.compressed) { + if (dst_surface->IsCompressed()) { LOG_CRITICAL(HW_GPU, "Compressed buffer copy is unimplemented!"); UNREACHABLE(); } else { diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h index 6658c6ffd..02d9981a1 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.h +++ b/src/video_core/renderer_opengl/gl_texture_cache.h @@ -37,7 +37,7 @@ class CachedSurface final : public VideoCommon::SurfaceBase<View> { friend CachedSurfaceView; public: - explicit CachedSurface(GPUVAddr gpu_addr, const SurfaceParams& params); + explicit CachedSurface(GPUVAddr gpu_addr, const SurfaceParams& params, bool is_astc_supported); ~CachedSurface(); void UploadTexture(const std::vector<u8>& staging_buffer) override; @@ -51,6 +51,10 @@ public: return texture.handle; } + bool IsCompressed() const { + return is_compressed; + } + protected: void DecorateSurfaceName() override; diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index f1a28cc21..b2a179746 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -315,8 +315,8 @@ public: RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system, Core::Frontend::GraphicsContext& context) - : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system}, - frame_mailbox{}, context{context}, has_debug_tool{HasDebugTool()} {} + : RendererBase{emu_window}, emu_window{emu_window}, system{system}, context{context}, + has_debug_tool{HasDebugTool()} {} RendererOpenGL::~RendererOpenGL() = default; diff --git a/src/video_core/renderer_vulkan/declarations.h b/src/video_core/renderer_vulkan/declarations.h deleted file mode 100644 index 323bf6b39..000000000 --- a/src/video_core/renderer_vulkan/declarations.h +++ /dev/null @@ -1,58 +0,0 @@ -// Copyright 2019 yuzu Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -#pragma once - -namespace vk { -class DispatchLoaderDynamic; -} - -namespace Vulkan { -constexpr vk::DispatchLoaderDynamic* dont_use_me_dld = nullptr; -} - -#define VULKAN_HPP_DEFAULT_DISPATCHER (*::Vulkan::dont_use_me_dld) -#define VULKAN_HPP_ENABLE_DYNAMIC_LOADER_TOOL 0 -#define VULKAN_HPP_DISPATCH_LOADER_DYNAMIC 1 -#include <vulkan/vulkan.hpp> - -namespace Vulkan { - -// vulkan.hpp unique handlers use DispatchLoaderStatic -template <typename T> -using UniqueHandle = vk::UniqueHandle<T, vk::DispatchLoaderDynamic>; - -using UniqueAccelerationStructureNV = UniqueHandle<vk::AccelerationStructureNV>; -using UniqueBuffer = UniqueHandle<vk::Buffer>; -using UniqueBufferView = UniqueHandle<vk::BufferView>; -using UniqueCommandBuffer = UniqueHandle<vk::CommandBuffer>; -using UniqueCommandPool = UniqueHandle<vk::CommandPool>; -using UniqueDescriptorPool = UniqueHandle<vk::DescriptorPool>; -using UniqueDescriptorSet = UniqueHandle<vk::DescriptorSet>; -using UniqueDescriptorSetLayout = UniqueHandle<vk::DescriptorSetLayout>; -using UniqueDescriptorUpdateTemplate = UniqueHandle<vk::DescriptorUpdateTemplate>; -using UniqueDevice = UniqueHandle<vk::Device>; -using UniqueDeviceMemory = UniqueHandle<vk::DeviceMemory>; -using UniqueEvent = UniqueHandle<vk::Event>; -using UniqueFence = UniqueHandle<vk::Fence>; -using UniqueFramebuffer = UniqueHandle<vk::Framebuffer>; -using UniqueImage = UniqueHandle<vk::Image>; -using UniqueImageView = UniqueHandle<vk::ImageView>; -using UniqueIndirectCommandsLayoutNVX = UniqueHandle<vk::IndirectCommandsLayoutNVX>; -using UniqueObjectTableNVX = UniqueHandle<vk::ObjectTableNVX>; -using UniquePipeline = UniqueHandle<vk::Pipeline>; -using UniquePipelineCache = UniqueHandle<vk::PipelineCache>; -using UniquePipelineLayout = UniqueHandle<vk::PipelineLayout>; -using UniqueQueryPool = UniqueHandle<vk::QueryPool>; -using UniqueRenderPass = UniqueHandle<vk::RenderPass>; -using UniqueSampler = UniqueHandle<vk::Sampler>; -using UniqueSamplerYcbcrConversion = UniqueHandle<vk::SamplerYcbcrConversion>; -using UniqueSemaphore = UniqueHandle<vk::Semaphore>; -using UniqueShaderModule = UniqueHandle<vk::ShaderModule>; -using UniqueSwapchainKHR = UniqueHandle<vk::SwapchainKHR>; -using UniqueValidationCacheEXT = UniqueHandle<vk::ValidationCacheEXT>; -using UniqueDebugReportCallbackEXT = UniqueHandle<vk::DebugReportCallbackEXT>; -using UniqueDebugUtilsMessengerEXT = UniqueHandle<vk::DebugUtilsMessengerEXT>; - -} // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp index 7480cb7c3..8681b821f 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp @@ -2,13 +2,15 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <iterator> + #include "common/assert.h" #include "common/common_types.h" #include "common/logging/log.h" #include "video_core/engines/maxwell_3d.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/vk_device.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/surface.h" namespace Vulkan::MaxwellToVK { @@ -17,88 +19,89 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs; namespace Sampler { -vk::Filter Filter(Tegra::Texture::TextureFilter filter) { +VkFilter Filter(Tegra::Texture::TextureFilter filter) { switch (filter) { case Tegra::Texture::TextureFilter::Linear: - return vk::Filter::eLinear; + return VK_FILTER_LINEAR; case Tegra::Texture::TextureFilter::Nearest: - return vk::Filter::eNearest; + return VK_FILTER_NEAREST; } UNIMPLEMENTED_MSG("Unimplemented sampler filter={}", static_cast<u32>(filter)); return {}; } -vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter) { +VkSamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter) { switch (mipmap_filter) { case Tegra::Texture::TextureMipmapFilter::None: // TODO(Rodrigo): None seems to be mapped to OpenGL's mag and min filters without mipmapping // (e.g. GL_NEAREST and GL_LINEAR). Vulkan doesn't have such a thing, find out if we have to // use an image view with a single mipmap level to emulate this. - return vk::SamplerMipmapMode::eLinear; + return VK_SAMPLER_MIPMAP_MODE_LINEAR; + ; case Tegra::Texture::TextureMipmapFilter::Linear: - return vk::SamplerMipmapMode::eLinear; + return VK_SAMPLER_MIPMAP_MODE_LINEAR; case Tegra::Texture::TextureMipmapFilter::Nearest: - return vk::SamplerMipmapMode::eNearest; + return VK_SAMPLER_MIPMAP_MODE_NEAREST; } UNIMPLEMENTED_MSG("Unimplemented sampler mipmap mode={}", static_cast<u32>(mipmap_filter)); return {}; } -vk::SamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode, - Tegra::Texture::TextureFilter filter) { +VkSamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode, + Tegra::Texture::TextureFilter filter) { switch (wrap_mode) { case Tegra::Texture::WrapMode::Wrap: - return vk::SamplerAddressMode::eRepeat; + return VK_SAMPLER_ADDRESS_MODE_REPEAT; case Tegra::Texture::WrapMode::Mirror: - return vk::SamplerAddressMode::eMirroredRepeat; + return VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT; case Tegra::Texture::WrapMode::ClampToEdge: - return vk::SamplerAddressMode::eClampToEdge; + return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE; case Tegra::Texture::WrapMode::Border: - return vk::SamplerAddressMode::eClampToBorder; + return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; case Tegra::Texture::WrapMode::Clamp: - if (device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) { + if (device.GetDriverID() == VK_DRIVER_ID_NVIDIA_PROPRIETARY_KHR) { // Nvidia's Vulkan driver defaults to GL_CLAMP on invalid enumerations, we can hack this // by sending an invalid enumeration. - return static_cast<vk::SamplerAddressMode>(0xcafe); + return static_cast<VkSamplerAddressMode>(0xcafe); } // TODO(Rodrigo): Emulate GL_CLAMP properly on other vendors switch (filter) { case Tegra::Texture::TextureFilter::Nearest: - return vk::SamplerAddressMode::eClampToEdge; + return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE; case Tegra::Texture::TextureFilter::Linear: - return vk::SamplerAddressMode::eClampToBorder; + return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; } UNREACHABLE(); - return vk::SamplerAddressMode::eClampToEdge; + return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE; case Tegra::Texture::WrapMode::MirrorOnceClampToEdge: - return vk::SamplerAddressMode::eMirrorClampToEdge; + return VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE; case Tegra::Texture::WrapMode::MirrorOnceBorder: UNIMPLEMENTED(); - return vk::SamplerAddressMode::eMirrorClampToEdge; + return VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE; default: UNIMPLEMENTED_MSG("Unimplemented wrap mode={}", static_cast<u32>(wrap_mode)); return {}; } } -vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func) { +VkCompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func) { switch (depth_compare_func) { case Tegra::Texture::DepthCompareFunc::Never: - return vk::CompareOp::eNever; + return VK_COMPARE_OP_NEVER; case Tegra::Texture::DepthCompareFunc::Less: - return vk::CompareOp::eLess; + return VK_COMPARE_OP_LESS; case Tegra::Texture::DepthCompareFunc::LessEqual: - return vk::CompareOp::eLessOrEqual; + return VK_COMPARE_OP_LESS_OR_EQUAL; case Tegra::Texture::DepthCompareFunc::Equal: - return vk::CompareOp::eEqual; + return VK_COMPARE_OP_EQUAL; case Tegra::Texture::DepthCompareFunc::NotEqual: - return vk::CompareOp::eNotEqual; + return VK_COMPARE_OP_NOT_EQUAL; case Tegra::Texture::DepthCompareFunc::Greater: - return vk::CompareOp::eGreater; + return VK_COMPARE_OP_GREATER; case Tegra::Texture::DepthCompareFunc::GreaterEqual: - return vk::CompareOp::eGreaterOrEqual; + return VK_COMPARE_OP_GREATER_OR_EQUAL; case Tegra::Texture::DepthCompareFunc::Always: - return vk::CompareOp::eAlways; + return VK_COMPARE_OP_ALWAYS; } UNIMPLEMENTED_MSG("Unimplemented sampler depth compare function={}", static_cast<u32>(depth_compare_func)); @@ -112,92 +115,92 @@ namespace { enum : u32 { Attachable = 1, Storage = 2 }; struct FormatTuple { - vk::Format format; ///< Vulkan format - int usage; ///< Describes image format usage + VkFormat format; ///< Vulkan format + int usage = 0; ///< Describes image format usage } constexpr tex_format_tuples[] = { - {vk::Format::eA8B8G8R8UnormPack32, Attachable | Storage}, // ABGR8U - {vk::Format::eA8B8G8R8SnormPack32, Attachable | Storage}, // ABGR8S - {vk::Format::eA8B8G8R8UintPack32, Attachable | Storage}, // ABGR8UI - {vk::Format::eB5G6R5UnormPack16, {}}, // B5G6R5U - {vk::Format::eA2B10G10R10UnormPack32, Attachable | Storage}, // A2B10G10R10U - {vk::Format::eA1R5G5B5UnormPack16, Attachable}, // A1B5G5R5U (flipped with swizzle) - {vk::Format::eR8Unorm, Attachable | Storage}, // R8U - {vk::Format::eR8Uint, Attachable | Storage}, // R8UI - {vk::Format::eR16G16B16A16Sfloat, Attachable | Storage}, // RGBA16F - {vk::Format::eR16G16B16A16Unorm, Attachable | Storage}, // RGBA16U - {vk::Format::eR16G16B16A16Snorm, Attachable | Storage}, // RGBA16S - {vk::Format::eR16G16B16A16Uint, Attachable | Storage}, // RGBA16UI - {vk::Format::eB10G11R11UfloatPack32, Attachable | Storage}, // R11FG11FB10F - {vk::Format::eR32G32B32A32Uint, Attachable | Storage}, // RGBA32UI - {vk::Format::eBc1RgbaUnormBlock, {}}, // DXT1 - {vk::Format::eBc2UnormBlock, {}}, // DXT23 - {vk::Format::eBc3UnormBlock, {}}, // DXT45 - {vk::Format::eBc4UnormBlock, {}}, // DXN1 - {vk::Format::eBc5UnormBlock, {}}, // DXN2UNORM - {vk::Format::eBc5SnormBlock, {}}, // DXN2SNORM - {vk::Format::eBc7UnormBlock, {}}, // BC7U - {vk::Format::eBc6HUfloatBlock, {}}, // BC6H_UF16 - {vk::Format::eBc6HSfloatBlock, {}}, // BC6H_SF16 - {vk::Format::eAstc4x4UnormBlock, {}}, // ASTC_2D_4X4 - {vk::Format::eB8G8R8A8Unorm, {}}, // BGRA8 - {vk::Format::eR32G32B32A32Sfloat, Attachable | Storage}, // RGBA32F - {vk::Format::eR32G32Sfloat, Attachable | Storage}, // RG32F - {vk::Format::eR32Sfloat, Attachable | Storage}, // R32F - {vk::Format::eR16Sfloat, Attachable | Storage}, // R16F - {vk::Format::eR16Unorm, Attachable | Storage}, // R16U - {vk::Format::eUndefined, {}}, // R16S - {vk::Format::eUndefined, {}}, // R16UI - {vk::Format::eUndefined, {}}, // R16I - {vk::Format::eR16G16Unorm, Attachable | Storage}, // RG16 - {vk::Format::eR16G16Sfloat, Attachable | Storage}, // RG16F - {vk::Format::eUndefined, {}}, // RG16UI - {vk::Format::eUndefined, {}}, // RG16I - {vk::Format::eR16G16Snorm, Attachable | Storage}, // RG16S - {vk::Format::eUndefined, {}}, // RGB32F - {vk::Format::eR8G8B8A8Srgb, Attachable}, // RGBA8_SRGB - {vk::Format::eR8G8Unorm, Attachable | Storage}, // RG8U - {vk::Format::eR8G8Snorm, Attachable | Storage}, // RG8S - {vk::Format::eR32G32Uint, Attachable | Storage}, // RG32UI - {vk::Format::eUndefined, {}}, // RGBX16F - {vk::Format::eR32Uint, Attachable | Storage}, // R32UI - {vk::Format::eR32Sint, Attachable | Storage}, // R32I - {vk::Format::eAstc8x8UnormBlock, {}}, // ASTC_2D_8X8 - {vk::Format::eUndefined, {}}, // ASTC_2D_8X5 - {vk::Format::eUndefined, {}}, // ASTC_2D_5X4 - {vk::Format::eUndefined, {}}, // BGRA8_SRGB - {vk::Format::eBc1RgbaSrgbBlock, {}}, // DXT1_SRGB - {vk::Format::eBc2SrgbBlock, {}}, // DXT23_SRGB - {vk::Format::eBc3SrgbBlock, {}}, // DXT45_SRGB - {vk::Format::eBc7SrgbBlock, {}}, // BC7U_SRGB - {vk::Format::eR4G4B4A4UnormPack16, Attachable}, // R4G4B4A4U - {vk::Format::eAstc4x4SrgbBlock, {}}, // ASTC_2D_4X4_SRGB - {vk::Format::eAstc8x8SrgbBlock, {}}, // ASTC_2D_8X8_SRGB - {vk::Format::eAstc8x5SrgbBlock, {}}, // ASTC_2D_8X5_SRGB - {vk::Format::eAstc5x4SrgbBlock, {}}, // ASTC_2D_5X4_SRGB - {vk::Format::eAstc5x5UnormBlock, {}}, // ASTC_2D_5X5 - {vk::Format::eAstc5x5SrgbBlock, {}}, // ASTC_2D_5X5_SRGB - {vk::Format::eAstc10x8UnormBlock, {}}, // ASTC_2D_10X8 - {vk::Format::eAstc10x8SrgbBlock, {}}, // ASTC_2D_10X8_SRGB - {vk::Format::eAstc6x6UnormBlock, {}}, // ASTC_2D_6X6 - {vk::Format::eAstc6x6SrgbBlock, {}}, // ASTC_2D_6X6_SRGB - {vk::Format::eAstc10x10UnormBlock, {}}, // ASTC_2D_10X10 - {vk::Format::eAstc10x10SrgbBlock, {}}, // ASTC_2D_10X10_SRGB - {vk::Format::eAstc12x12UnormBlock, {}}, // ASTC_2D_12X12 - {vk::Format::eAstc12x12SrgbBlock, {}}, // ASTC_2D_12X12_SRGB - {vk::Format::eAstc8x6UnormBlock, {}}, // ASTC_2D_8X6 - {vk::Format::eAstc8x6SrgbBlock, {}}, // ASTC_2D_8X6_SRGB - {vk::Format::eAstc6x5UnormBlock, {}}, // ASTC_2D_6X5 - {vk::Format::eAstc6x5SrgbBlock, {}}, // ASTC_2D_6X5_SRGB - {vk::Format::eE5B9G9R9UfloatPack32, {}}, // E5B9G9R9F + {VK_FORMAT_A8B8G8R8_UNORM_PACK32, Attachable | Storage}, // ABGR8U + {VK_FORMAT_A8B8G8R8_SNORM_PACK32, Attachable | Storage}, // ABGR8S + {VK_FORMAT_A8B8G8R8_UINT_PACK32, Attachable | Storage}, // ABGR8UI + {VK_FORMAT_B5G6R5_UNORM_PACK16}, // B5G6R5U + {VK_FORMAT_A2B10G10R10_UNORM_PACK32, Attachable | Storage}, // A2B10G10R10U + {VK_FORMAT_A1R5G5B5_UNORM_PACK16, Attachable}, // A1B5G5R5U (flipped with swizzle) + {VK_FORMAT_R8_UNORM, Attachable | Storage}, // R8U + {VK_FORMAT_R8_UINT, Attachable | Storage}, // R8UI + {VK_FORMAT_R16G16B16A16_SFLOAT, Attachable | Storage}, // RGBA16F + {VK_FORMAT_R16G16B16A16_UNORM, Attachable | Storage}, // RGBA16U + {VK_FORMAT_R16G16B16A16_SNORM, Attachable | Storage}, // RGBA16S + {VK_FORMAT_R16G16B16A16_UINT, Attachable | Storage}, // RGBA16UI + {VK_FORMAT_B10G11R11_UFLOAT_PACK32, Attachable | Storage}, // R11FG11FB10F + {VK_FORMAT_R32G32B32A32_UINT, Attachable | Storage}, // RGBA32UI + {VK_FORMAT_BC1_RGBA_UNORM_BLOCK}, // DXT1 + {VK_FORMAT_BC2_UNORM_BLOCK}, // DXT23 + {VK_FORMAT_BC3_UNORM_BLOCK}, // DXT45 + {VK_FORMAT_BC4_UNORM_BLOCK}, // DXN1 + {VK_FORMAT_BC5_UNORM_BLOCK}, // DXN2UNORM + {VK_FORMAT_BC5_SNORM_BLOCK}, // DXN2SNORM + {VK_FORMAT_BC7_UNORM_BLOCK}, // BC7U + {VK_FORMAT_BC6H_UFLOAT_BLOCK}, // BC6H_UF16 + {VK_FORMAT_BC6H_SFLOAT_BLOCK}, // BC6H_SF16 + {VK_FORMAT_ASTC_4x4_UNORM_BLOCK}, // ASTC_2D_4X4 + {VK_FORMAT_B8G8R8A8_UNORM}, // BGRA8 + {VK_FORMAT_R32G32B32A32_SFLOAT, Attachable | Storage}, // RGBA32F + {VK_FORMAT_R32G32_SFLOAT, Attachable | Storage}, // RG32F + {VK_FORMAT_R32_SFLOAT, Attachable | Storage}, // R32F + {VK_FORMAT_R16_SFLOAT, Attachable | Storage}, // R16F + {VK_FORMAT_R16_UNORM, Attachable | Storage}, // R16U + {VK_FORMAT_UNDEFINED}, // R16S + {VK_FORMAT_UNDEFINED}, // R16UI + {VK_FORMAT_UNDEFINED}, // R16I + {VK_FORMAT_R16G16_UNORM, Attachable | Storage}, // RG16 + {VK_FORMAT_R16G16_SFLOAT, Attachable | Storage}, // RG16F + {VK_FORMAT_UNDEFINED}, // RG16UI + {VK_FORMAT_UNDEFINED}, // RG16I + {VK_FORMAT_R16G16_SNORM, Attachable | Storage}, // RG16S + {VK_FORMAT_UNDEFINED}, // RGB32F + {VK_FORMAT_R8G8B8A8_SRGB, Attachable}, // RGBA8_SRGB + {VK_FORMAT_R8G8_UNORM, Attachable | Storage}, // RG8U + {VK_FORMAT_R8G8_SNORM, Attachable | Storage}, // RG8S + {VK_FORMAT_R32G32_UINT, Attachable | Storage}, // RG32UI + {VK_FORMAT_UNDEFINED}, // RGBX16F + {VK_FORMAT_R32_UINT, Attachable | Storage}, // R32UI + {VK_FORMAT_R32_SINT, Attachable | Storage}, // R32I + {VK_FORMAT_ASTC_8x8_UNORM_BLOCK}, // ASTC_2D_8X8 + {VK_FORMAT_UNDEFINED}, // ASTC_2D_8X5 + {VK_FORMAT_UNDEFINED}, // ASTC_2D_5X4 + {VK_FORMAT_UNDEFINED}, // BGRA8_SRGB + {VK_FORMAT_BC1_RGBA_SRGB_BLOCK}, // DXT1_SRGB + {VK_FORMAT_BC2_SRGB_BLOCK}, // DXT23_SRGB + {VK_FORMAT_BC3_SRGB_BLOCK}, // DXT45_SRGB + {VK_FORMAT_BC7_SRGB_BLOCK}, // BC7U_SRGB + {VK_FORMAT_R4G4B4A4_UNORM_PACK16, Attachable}, // R4G4B4A4U + {VK_FORMAT_ASTC_4x4_SRGB_BLOCK}, // ASTC_2D_4X4_SRGB + {VK_FORMAT_ASTC_8x8_SRGB_BLOCK}, // ASTC_2D_8X8_SRGB + {VK_FORMAT_ASTC_8x5_SRGB_BLOCK}, // ASTC_2D_8X5_SRGB + {VK_FORMAT_ASTC_5x4_SRGB_BLOCK}, // ASTC_2D_5X4_SRGB + {VK_FORMAT_ASTC_5x5_UNORM_BLOCK}, // ASTC_2D_5X5 + {VK_FORMAT_ASTC_5x5_SRGB_BLOCK}, // ASTC_2D_5X5_SRGB + {VK_FORMAT_ASTC_10x8_UNORM_BLOCK}, // ASTC_2D_10X8 + {VK_FORMAT_ASTC_10x8_SRGB_BLOCK}, // ASTC_2D_10X8_SRGB + {VK_FORMAT_ASTC_6x6_UNORM_BLOCK}, // ASTC_2D_6X6 + {VK_FORMAT_ASTC_6x6_SRGB_BLOCK}, // ASTC_2D_6X6_SRGB + {VK_FORMAT_ASTC_10x10_UNORM_BLOCK}, // ASTC_2D_10X10 + {VK_FORMAT_ASTC_10x10_SRGB_BLOCK}, // ASTC_2D_10X10_SRGB + {VK_FORMAT_ASTC_12x12_UNORM_BLOCK}, // ASTC_2D_12X12 + {VK_FORMAT_ASTC_12x12_SRGB_BLOCK}, // ASTC_2D_12X12_SRGB + {VK_FORMAT_ASTC_8x6_UNORM_BLOCK}, // ASTC_2D_8X6 + {VK_FORMAT_ASTC_8x6_SRGB_BLOCK}, // ASTC_2D_8X6_SRGB + {VK_FORMAT_ASTC_6x5_UNORM_BLOCK}, // ASTC_2D_6X5 + {VK_FORMAT_ASTC_6x5_SRGB_BLOCK}, // ASTC_2D_6X5_SRGB + {VK_FORMAT_E5B9G9R9_UFLOAT_PACK32}, // E5B9G9R9F // Depth formats - {vk::Format::eD32Sfloat, Attachable}, // Z32F - {vk::Format::eD16Unorm, Attachable}, // Z16 + {VK_FORMAT_D32_SFLOAT, Attachable}, // Z32F + {VK_FORMAT_D16_UNORM, Attachable}, // Z16 // DepthStencil formats - {vk::Format::eD24UnormS8Uint, Attachable}, // Z24S8 - {vk::Format::eD24UnormS8Uint, Attachable}, // S8Z24 (emulated) - {vk::Format::eD32SfloatS8Uint, Attachable}, // Z32FS8 + {VK_FORMAT_D24_UNORM_S8_UINT, Attachable}, // Z24S8 + {VK_FORMAT_D24_UNORM_S8_UINT, Attachable}, // S8Z24 (emulated) + {VK_FORMAT_D32_SFLOAT_S8_UINT, Attachable}, // Z32FS8 }; static_assert(std::size(tex_format_tuples) == VideoCore::Surface::MaxPixelFormat); @@ -212,106 +215,106 @@ FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFo ASSERT(static_cast<std::size_t>(pixel_format) < std::size(tex_format_tuples)); auto tuple = tex_format_tuples[static_cast<std::size_t>(pixel_format)]; - if (tuple.format == vk::Format::eUndefined) { + if (tuple.format == VK_FORMAT_UNDEFINED) { UNIMPLEMENTED_MSG("Unimplemented texture format with pixel format={}", static_cast<u32>(pixel_format)); - return {vk::Format::eA8B8G8R8UnormPack32, true, true}; + return {VK_FORMAT_A8B8G8R8_UNORM_PACK32, true, true}; } // Use ABGR8 on hardware that doesn't support ASTC natively if (!device.IsOptimalAstcSupported() && VideoCore::Surface::IsPixelFormatASTC(pixel_format)) { tuple.format = VideoCore::Surface::IsPixelFormatSRGB(pixel_format) - ? vk::Format::eA8B8G8R8SrgbPack32 - : vk::Format::eA8B8G8R8UnormPack32; + ? VK_FORMAT_A8B8G8R8_SRGB_PACK32 + : VK_FORMAT_A8B8G8R8_UNORM_PACK32; } const bool attachable = tuple.usage & Attachable; const bool storage = tuple.usage & Storage; - vk::FormatFeatureFlags usage; + VkFormatFeatureFlags usage; if (format_type == FormatType::Buffer) { - usage = vk::FormatFeatureFlagBits::eStorageTexelBuffer | - vk::FormatFeatureFlagBits::eUniformTexelBuffer; + usage = + VK_FORMAT_FEATURE_STORAGE_TEXEL_BUFFER_BIT | VK_FORMAT_FEATURE_UNIFORM_TEXEL_BUFFER_BIT; } else { - usage = vk::FormatFeatureFlagBits::eSampledImage | vk::FormatFeatureFlagBits::eTransferDst | - vk::FormatFeatureFlagBits::eTransferSrc; + usage = VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_TRANSFER_DST_BIT | + VK_FORMAT_FEATURE_TRANSFER_SRC_BIT; if (attachable) { - usage |= IsZetaFormat(pixel_format) ? vk::FormatFeatureFlagBits::eDepthStencilAttachment - : vk::FormatFeatureFlagBits::eColorAttachment; + usage |= IsZetaFormat(pixel_format) ? VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT + : VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT; } if (storage) { - usage |= vk::FormatFeatureFlagBits::eStorageImage; + usage |= VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT; } } return {device.GetSupportedFormat(tuple.format, usage, format_type), attachable, storage}; } -vk::ShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage) { +VkShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage) { switch (stage) { case Tegra::Engines::ShaderType::Vertex: - return vk::ShaderStageFlagBits::eVertex; + return VK_SHADER_STAGE_VERTEX_BIT; case Tegra::Engines::ShaderType::TesselationControl: - return vk::ShaderStageFlagBits::eTessellationControl; + return VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT; case Tegra::Engines::ShaderType::TesselationEval: - return vk::ShaderStageFlagBits::eTessellationEvaluation; + return VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT; case Tegra::Engines::ShaderType::Geometry: - return vk::ShaderStageFlagBits::eGeometry; + return VK_SHADER_STAGE_GEOMETRY_BIT; case Tegra::Engines::ShaderType::Fragment: - return vk::ShaderStageFlagBits::eFragment; + return VK_SHADER_STAGE_FRAGMENT_BIT; case Tegra::Engines::ShaderType::Compute: - return vk::ShaderStageFlagBits::eCompute; + return VK_SHADER_STAGE_COMPUTE_BIT; } UNIMPLEMENTED_MSG("Unimplemented shader stage={}", static_cast<u32>(stage)); return {}; } -vk::PrimitiveTopology PrimitiveTopology([[maybe_unused]] const VKDevice& device, - Maxwell::PrimitiveTopology topology) { +VkPrimitiveTopology PrimitiveTopology([[maybe_unused]] const VKDevice& device, + Maxwell::PrimitiveTopology topology) { switch (topology) { case Maxwell::PrimitiveTopology::Points: - return vk::PrimitiveTopology::ePointList; + return VK_PRIMITIVE_TOPOLOGY_POINT_LIST; case Maxwell::PrimitiveTopology::Lines: - return vk::PrimitiveTopology::eLineList; + return VK_PRIMITIVE_TOPOLOGY_LINE_LIST; case Maxwell::PrimitiveTopology::LineStrip: - return vk::PrimitiveTopology::eLineStrip; + return VK_PRIMITIVE_TOPOLOGY_LINE_STRIP; case Maxwell::PrimitiveTopology::Triangles: - return vk::PrimitiveTopology::eTriangleList; + return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST; case Maxwell::PrimitiveTopology::TriangleStrip: - return vk::PrimitiveTopology::eTriangleStrip; + return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP; case Maxwell::PrimitiveTopology::TriangleFan: - return vk::PrimitiveTopology::eTriangleFan; + return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN; case Maxwell::PrimitiveTopology::Quads: // TODO(Rodrigo): Use VK_PRIMITIVE_TOPOLOGY_QUAD_LIST_EXT whenever it releases - return vk::PrimitiveTopology::eTriangleList; + return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST; case Maxwell::PrimitiveTopology::Patches: - return vk::PrimitiveTopology::ePatchList; + return VK_PRIMITIVE_TOPOLOGY_PATCH_LIST; default: UNIMPLEMENTED_MSG("Unimplemented topology={}", static_cast<u32>(topology)); return {}; } } -vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) { +VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) { switch (type) { case Maxwell::VertexAttribute::Type::SignedNorm: switch (size) { case Maxwell::VertexAttribute::Size::Size_8: - return vk::Format::eR8Snorm; + return VK_FORMAT_R8_SNORM; case Maxwell::VertexAttribute::Size::Size_8_8: - return vk::Format::eR8G8Snorm; + return VK_FORMAT_R8G8_SNORM; case Maxwell::VertexAttribute::Size::Size_8_8_8: - return vk::Format::eR8G8B8Snorm; + return VK_FORMAT_R8G8B8_SNORM; case Maxwell::VertexAttribute::Size::Size_8_8_8_8: - return vk::Format::eR8G8B8A8Snorm; + return VK_FORMAT_R8G8B8A8_SNORM; case Maxwell::VertexAttribute::Size::Size_16: - return vk::Format::eR16Snorm; + return VK_FORMAT_R16_SNORM; case Maxwell::VertexAttribute::Size::Size_16_16: - return vk::Format::eR16G16Snorm; + return VK_FORMAT_R16G16_SNORM; case Maxwell::VertexAttribute::Size::Size_16_16_16: - return vk::Format::eR16G16B16Snorm; + return VK_FORMAT_R16G16B16_SNORM; case Maxwell::VertexAttribute::Size::Size_16_16_16_16: - return vk::Format::eR16G16B16A16Snorm; + return VK_FORMAT_R16G16B16A16_SNORM; case Maxwell::VertexAttribute::Size::Size_10_10_10_2: - return vk::Format::eA2B10G10R10SnormPack32; + return VK_FORMAT_A2B10G10R10_SNORM_PACK32; default: break; } @@ -319,23 +322,23 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr case Maxwell::VertexAttribute::Type::UnsignedNorm: switch (size) { case Maxwell::VertexAttribute::Size::Size_8: - return vk::Format::eR8Unorm; + return VK_FORMAT_R8_UNORM; case Maxwell::VertexAttribute::Size::Size_8_8: - return vk::Format::eR8G8Unorm; + return VK_FORMAT_R8G8_UNORM; case Maxwell::VertexAttribute::Size::Size_8_8_8: - return vk::Format::eR8G8B8Unorm; + return VK_FORMAT_R8G8B8_UNORM; case Maxwell::VertexAttribute::Size::Size_8_8_8_8: - return vk::Format::eR8G8B8A8Unorm; + return VK_FORMAT_R8G8B8A8_UNORM; case Maxwell::VertexAttribute::Size::Size_16: - return vk::Format::eR16Unorm; + return VK_FORMAT_R16_UNORM; case Maxwell::VertexAttribute::Size::Size_16_16: - return vk::Format::eR16G16Unorm; + return VK_FORMAT_R16G16_UNORM; case Maxwell::VertexAttribute::Size::Size_16_16_16: - return vk::Format::eR16G16B16Unorm; + return VK_FORMAT_R16G16B16_UNORM; case Maxwell::VertexAttribute::Size::Size_16_16_16_16: - return vk::Format::eR16G16B16A16Unorm; + return VK_FORMAT_R16G16B16A16_UNORM; case Maxwell::VertexAttribute::Size::Size_10_10_10_2: - return vk::Format::eA2B10G10R10UnormPack32; + return VK_FORMAT_A2B10G10R10_UNORM_PACK32; default: break; } @@ -343,59 +346,69 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr case Maxwell::VertexAttribute::Type::SignedInt: switch (size) { case Maxwell::VertexAttribute::Size::Size_16_16_16_16: - return vk::Format::eR16G16B16A16Sint; + return VK_FORMAT_R16G16B16A16_SINT; case Maxwell::VertexAttribute::Size::Size_8: - return vk::Format::eR8Sint; + return VK_FORMAT_R8_SINT; case Maxwell::VertexAttribute::Size::Size_8_8: - return vk::Format::eR8G8Sint; + return VK_FORMAT_R8G8_SINT; case Maxwell::VertexAttribute::Size::Size_8_8_8: - return vk::Format::eR8G8B8Sint; + return VK_FORMAT_R8G8B8_SINT; case Maxwell::VertexAttribute::Size::Size_8_8_8_8: - return vk::Format::eR8G8B8A8Sint; + return VK_FORMAT_R8G8B8A8_SINT; case Maxwell::VertexAttribute::Size::Size_32: - return vk::Format::eR32Sint; + return VK_FORMAT_R32_SINT; default: break; } + break; case Maxwell::VertexAttribute::Type::UnsignedInt: switch (size) { case Maxwell::VertexAttribute::Size::Size_8: - return vk::Format::eR8Uint; + return VK_FORMAT_R8_UINT; case Maxwell::VertexAttribute::Size::Size_8_8: - return vk::Format::eR8G8Uint; + return VK_FORMAT_R8G8_UINT; case Maxwell::VertexAttribute::Size::Size_8_8_8: - return vk::Format::eR8G8B8Uint; + return VK_FORMAT_R8G8B8_UINT; case Maxwell::VertexAttribute::Size::Size_8_8_8_8: - return vk::Format::eR8G8B8A8Uint; + return VK_FORMAT_R8G8B8A8_UINT; + case Maxwell::VertexAttribute::Size::Size_16: + return VK_FORMAT_R16_UINT; + case Maxwell::VertexAttribute::Size::Size_16_16: + return VK_FORMAT_R16G16_UINT; + case Maxwell::VertexAttribute::Size::Size_16_16_16: + return VK_FORMAT_R16G16B16_UINT; + case Maxwell::VertexAttribute::Size::Size_16_16_16_16: + return VK_FORMAT_R16G16B16A16_UINT; case Maxwell::VertexAttribute::Size::Size_32: - return vk::Format::eR32Uint; + return VK_FORMAT_R32_UINT; case Maxwell::VertexAttribute::Size::Size_32_32: - return vk::Format::eR32G32Uint; + return VK_FORMAT_R32G32_UINT; case Maxwell::VertexAttribute::Size::Size_32_32_32: - return vk::Format::eR32G32B32Uint; + return VK_FORMAT_R32G32B32_UINT; case Maxwell::VertexAttribute::Size::Size_32_32_32_32: - return vk::Format::eR32G32B32A32Uint; + return VK_FORMAT_R32G32B32A32_UINT; default: break; } + break; case Maxwell::VertexAttribute::Type::UnsignedScaled: switch (size) { case Maxwell::VertexAttribute::Size::Size_8: - return vk::Format::eR8Uscaled; + return VK_FORMAT_R8_USCALED; case Maxwell::VertexAttribute::Size::Size_8_8: - return vk::Format::eR8G8Uscaled; + return VK_FORMAT_R8G8_USCALED; case Maxwell::VertexAttribute::Size::Size_8_8_8: - return vk::Format::eR8G8B8Uscaled; + return VK_FORMAT_R8G8B8_USCALED; case Maxwell::VertexAttribute::Size::Size_8_8_8_8: - return vk::Format::eR8G8B8A8Uscaled; + return VK_FORMAT_R8G8B8A8_USCALED; case Maxwell::VertexAttribute::Size::Size_16: - return vk::Format::eR16Uscaled; + return VK_FORMAT_R16_USCALED; case Maxwell::VertexAttribute::Size::Size_16_16: - return vk::Format::eR16G16Uscaled; + return VK_FORMAT_R16G16_USCALED; case Maxwell::VertexAttribute::Size::Size_16_16_16: - return vk::Format::eR16G16B16Uscaled; + return VK_FORMAT_R16G16B16_USCALED; case Maxwell::VertexAttribute::Size::Size_16_16_16_16: - return vk::Format::eR16G16B16A16Uscaled; + return VK_FORMAT_R16G16B16A16_USCALED; default: break; } @@ -403,21 +416,21 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr case Maxwell::VertexAttribute::Type::SignedScaled: switch (size) { case Maxwell::VertexAttribute::Size::Size_8: - return vk::Format::eR8Sscaled; + return VK_FORMAT_R8_SSCALED; case Maxwell::VertexAttribute::Size::Size_8_8: - return vk::Format::eR8G8Sscaled; + return VK_FORMAT_R8G8_SSCALED; case Maxwell::VertexAttribute::Size::Size_8_8_8: - return vk::Format::eR8G8B8Sscaled; + return VK_FORMAT_R8G8B8_SSCALED; case Maxwell::VertexAttribute::Size::Size_8_8_8_8: - return vk::Format::eR8G8B8A8Sscaled; + return VK_FORMAT_R8G8B8A8_SSCALED; case Maxwell::VertexAttribute::Size::Size_16: - return vk::Format::eR16Sscaled; + return VK_FORMAT_R16_SSCALED; case Maxwell::VertexAttribute::Size::Size_16_16: - return vk::Format::eR16G16Sscaled; + return VK_FORMAT_R16G16_SSCALED; case Maxwell::VertexAttribute::Size::Size_16_16_16: - return vk::Format::eR16G16B16Sscaled; + return VK_FORMAT_R16G16B16_SSCALED; case Maxwell::VertexAttribute::Size::Size_16_16_16_16: - return vk::Format::eR16G16B16A16Sscaled; + return VK_FORMAT_R16G16B16A16_SSCALED; default: break; } @@ -425,21 +438,21 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr case Maxwell::VertexAttribute::Type::Float: switch (size) { case Maxwell::VertexAttribute::Size::Size_32: - return vk::Format::eR32Sfloat; + return VK_FORMAT_R32_SFLOAT; case Maxwell::VertexAttribute::Size::Size_32_32: - return vk::Format::eR32G32Sfloat; + return VK_FORMAT_R32G32_SFLOAT; case Maxwell::VertexAttribute::Size::Size_32_32_32: - return vk::Format::eR32G32B32Sfloat; + return VK_FORMAT_R32G32B32_SFLOAT; case Maxwell::VertexAttribute::Size::Size_32_32_32_32: - return vk::Format::eR32G32B32A32Sfloat; + return VK_FORMAT_R32G32B32A32_SFLOAT; case Maxwell::VertexAttribute::Size::Size_16: - return vk::Format::eR16Sfloat; + return VK_FORMAT_R16_SFLOAT; case Maxwell::VertexAttribute::Size::Size_16_16: - return vk::Format::eR16G16Sfloat; + return VK_FORMAT_R16G16_SFLOAT; case Maxwell::VertexAttribute::Size::Size_16_16_16: - return vk::Format::eR16G16B16Sfloat; + return VK_FORMAT_R16G16B16_SFLOAT; case Maxwell::VertexAttribute::Size::Size_16_16_16_16: - return vk::Format::eR16G16B16A16Sfloat; + return VK_FORMAT_R16G16B16A16_SFLOAT; default: break; } @@ -450,210 +463,210 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr return {}; } -vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison) { +VkCompareOp ComparisonOp(Maxwell::ComparisonOp comparison) { switch (comparison) { case Maxwell::ComparisonOp::Never: case Maxwell::ComparisonOp::NeverOld: - return vk::CompareOp::eNever; + return VK_COMPARE_OP_NEVER; case Maxwell::ComparisonOp::Less: case Maxwell::ComparisonOp::LessOld: - return vk::CompareOp::eLess; + return VK_COMPARE_OP_LESS; case Maxwell::ComparisonOp::Equal: case Maxwell::ComparisonOp::EqualOld: - return vk::CompareOp::eEqual; + return VK_COMPARE_OP_EQUAL; case Maxwell::ComparisonOp::LessEqual: case Maxwell::ComparisonOp::LessEqualOld: - return vk::CompareOp::eLessOrEqual; + return VK_COMPARE_OP_LESS_OR_EQUAL; case Maxwell::ComparisonOp::Greater: case Maxwell::ComparisonOp::GreaterOld: - return vk::CompareOp::eGreater; + return VK_COMPARE_OP_GREATER; case Maxwell::ComparisonOp::NotEqual: case Maxwell::ComparisonOp::NotEqualOld: - return vk::CompareOp::eNotEqual; + return VK_COMPARE_OP_NOT_EQUAL; case Maxwell::ComparisonOp::GreaterEqual: case Maxwell::ComparisonOp::GreaterEqualOld: - return vk::CompareOp::eGreaterOrEqual; + return VK_COMPARE_OP_GREATER_OR_EQUAL; case Maxwell::ComparisonOp::Always: case Maxwell::ComparisonOp::AlwaysOld: - return vk::CompareOp::eAlways; + return VK_COMPARE_OP_ALWAYS; } UNIMPLEMENTED_MSG("Unimplemented comparison op={}", static_cast<u32>(comparison)); return {}; } -vk::IndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format) { +VkIndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format) { switch (index_format) { case Maxwell::IndexFormat::UnsignedByte: if (!device.IsExtIndexTypeUint8Supported()) { UNIMPLEMENTED_MSG("Native uint8 indices are not supported on this device"); - return vk::IndexType::eUint16; + return VK_INDEX_TYPE_UINT16; } - return vk::IndexType::eUint8EXT; + return VK_INDEX_TYPE_UINT8_EXT; case Maxwell::IndexFormat::UnsignedShort: - return vk::IndexType::eUint16; + return VK_INDEX_TYPE_UINT16; case Maxwell::IndexFormat::UnsignedInt: - return vk::IndexType::eUint32; + return VK_INDEX_TYPE_UINT32; } UNIMPLEMENTED_MSG("Unimplemented index_format={}", static_cast<u32>(index_format)); return {}; } -vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op) { +VkStencilOp StencilOp(Maxwell::StencilOp stencil_op) { switch (stencil_op) { case Maxwell::StencilOp::Keep: case Maxwell::StencilOp::KeepOGL: - return vk::StencilOp::eKeep; + return VK_STENCIL_OP_KEEP; case Maxwell::StencilOp::Zero: case Maxwell::StencilOp::ZeroOGL: - return vk::StencilOp::eZero; + return VK_STENCIL_OP_ZERO; case Maxwell::StencilOp::Replace: case Maxwell::StencilOp::ReplaceOGL: - return vk::StencilOp::eReplace; + return VK_STENCIL_OP_REPLACE; case Maxwell::StencilOp::Incr: case Maxwell::StencilOp::IncrOGL: - return vk::StencilOp::eIncrementAndClamp; + return VK_STENCIL_OP_INCREMENT_AND_CLAMP; case Maxwell::StencilOp::Decr: case Maxwell::StencilOp::DecrOGL: - return vk::StencilOp::eDecrementAndClamp; + return VK_STENCIL_OP_DECREMENT_AND_CLAMP; case Maxwell::StencilOp::Invert: case Maxwell::StencilOp::InvertOGL: - return vk::StencilOp::eInvert; + return VK_STENCIL_OP_INVERT; case Maxwell::StencilOp::IncrWrap: case Maxwell::StencilOp::IncrWrapOGL: - return vk::StencilOp::eIncrementAndWrap; + return VK_STENCIL_OP_INCREMENT_AND_WRAP; case Maxwell::StencilOp::DecrWrap: case Maxwell::StencilOp::DecrWrapOGL: - return vk::StencilOp::eDecrementAndWrap; + return VK_STENCIL_OP_DECREMENT_AND_WRAP; } UNIMPLEMENTED_MSG("Unimplemented stencil op={}", static_cast<u32>(stencil_op)); return {}; } -vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation) { +VkBlendOp BlendEquation(Maxwell::Blend::Equation equation) { switch (equation) { case Maxwell::Blend::Equation::Add: case Maxwell::Blend::Equation::AddGL: - return vk::BlendOp::eAdd; + return VK_BLEND_OP_ADD; case Maxwell::Blend::Equation::Subtract: case Maxwell::Blend::Equation::SubtractGL: - return vk::BlendOp::eSubtract; + return VK_BLEND_OP_SUBTRACT; case Maxwell::Blend::Equation::ReverseSubtract: case Maxwell::Blend::Equation::ReverseSubtractGL: - return vk::BlendOp::eReverseSubtract; + return VK_BLEND_OP_REVERSE_SUBTRACT; case Maxwell::Blend::Equation::Min: case Maxwell::Blend::Equation::MinGL: - return vk::BlendOp::eMin; + return VK_BLEND_OP_MIN; case Maxwell::Blend::Equation::Max: case Maxwell::Blend::Equation::MaxGL: - return vk::BlendOp::eMax; + return VK_BLEND_OP_MAX; } UNIMPLEMENTED_MSG("Unimplemented blend equation={}", static_cast<u32>(equation)); return {}; } -vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor) { +VkBlendFactor BlendFactor(Maxwell::Blend::Factor factor) { switch (factor) { case Maxwell::Blend::Factor::Zero: case Maxwell::Blend::Factor::ZeroGL: - return vk::BlendFactor::eZero; + return VK_BLEND_FACTOR_ZERO; case Maxwell::Blend::Factor::One: case Maxwell::Blend::Factor::OneGL: - return vk::BlendFactor::eOne; + return VK_BLEND_FACTOR_ONE; case Maxwell::Blend::Factor::SourceColor: case Maxwell::Blend::Factor::SourceColorGL: - return vk::BlendFactor::eSrcColor; + return VK_BLEND_FACTOR_SRC_COLOR; case Maxwell::Blend::Factor::OneMinusSourceColor: case Maxwell::Blend::Factor::OneMinusSourceColorGL: - return vk::BlendFactor::eOneMinusSrcColor; + return VK_BLEND_FACTOR_ONE_MINUS_SRC_COLOR; case Maxwell::Blend::Factor::SourceAlpha: case Maxwell::Blend::Factor::SourceAlphaGL: - return vk::BlendFactor::eSrcAlpha; + return VK_BLEND_FACTOR_SRC_ALPHA; case Maxwell::Blend::Factor::OneMinusSourceAlpha: case Maxwell::Blend::Factor::OneMinusSourceAlphaGL: - return vk::BlendFactor::eOneMinusSrcAlpha; + return VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA; case Maxwell::Blend::Factor::DestAlpha: case Maxwell::Blend::Factor::DestAlphaGL: - return vk::BlendFactor::eDstAlpha; + return VK_BLEND_FACTOR_DST_ALPHA; case Maxwell::Blend::Factor::OneMinusDestAlpha: case Maxwell::Blend::Factor::OneMinusDestAlphaGL: - return vk::BlendFactor::eOneMinusDstAlpha; + return VK_BLEND_FACTOR_ONE_MINUS_DST_ALPHA; case Maxwell::Blend::Factor::DestColor: case Maxwell::Blend::Factor::DestColorGL: - return vk::BlendFactor::eDstColor; + return VK_BLEND_FACTOR_DST_COLOR; case Maxwell::Blend::Factor::OneMinusDestColor: case Maxwell::Blend::Factor::OneMinusDestColorGL: - return vk::BlendFactor::eOneMinusDstColor; + return VK_BLEND_FACTOR_ONE_MINUS_DST_COLOR; case Maxwell::Blend::Factor::SourceAlphaSaturate: case Maxwell::Blend::Factor::SourceAlphaSaturateGL: - return vk::BlendFactor::eSrcAlphaSaturate; + return VK_BLEND_FACTOR_SRC_ALPHA_SATURATE; case Maxwell::Blend::Factor::Source1Color: case Maxwell::Blend::Factor::Source1ColorGL: - return vk::BlendFactor::eSrc1Color; + return VK_BLEND_FACTOR_SRC1_COLOR; case Maxwell::Blend::Factor::OneMinusSource1Color: case Maxwell::Blend::Factor::OneMinusSource1ColorGL: - return vk::BlendFactor::eOneMinusSrc1Color; + return VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR; case Maxwell::Blend::Factor::Source1Alpha: case Maxwell::Blend::Factor::Source1AlphaGL: - return vk::BlendFactor::eSrc1Alpha; + return VK_BLEND_FACTOR_SRC1_ALPHA; case Maxwell::Blend::Factor::OneMinusSource1Alpha: case Maxwell::Blend::Factor::OneMinusSource1AlphaGL: - return vk::BlendFactor::eOneMinusSrc1Alpha; + return VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA; case Maxwell::Blend::Factor::ConstantColor: case Maxwell::Blend::Factor::ConstantColorGL: - return vk::BlendFactor::eConstantColor; + return VK_BLEND_FACTOR_CONSTANT_COLOR; case Maxwell::Blend::Factor::OneMinusConstantColor: case Maxwell::Blend::Factor::OneMinusConstantColorGL: - return vk::BlendFactor::eOneMinusConstantColor; + return VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_COLOR; case Maxwell::Blend::Factor::ConstantAlpha: case Maxwell::Blend::Factor::ConstantAlphaGL: - return vk::BlendFactor::eConstantAlpha; + return VK_BLEND_FACTOR_CONSTANT_ALPHA; case Maxwell::Blend::Factor::OneMinusConstantAlpha: case Maxwell::Blend::Factor::OneMinusConstantAlphaGL: - return vk::BlendFactor::eOneMinusConstantAlpha; + return VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_ALPHA; } UNIMPLEMENTED_MSG("Unimplemented blend factor={}", static_cast<u32>(factor)); return {}; } -vk::FrontFace FrontFace(Maxwell::FrontFace front_face) { +VkFrontFace FrontFace(Maxwell::FrontFace front_face) { switch (front_face) { case Maxwell::FrontFace::ClockWise: - return vk::FrontFace::eClockwise; + return VK_FRONT_FACE_CLOCKWISE; case Maxwell::FrontFace::CounterClockWise: - return vk::FrontFace::eCounterClockwise; + return VK_FRONT_FACE_COUNTER_CLOCKWISE; } UNIMPLEMENTED_MSG("Unimplemented front face={}", static_cast<u32>(front_face)); return {}; } -vk::CullModeFlags CullFace(Maxwell::CullFace cull_face) { +VkCullModeFlags CullFace(Maxwell::CullFace cull_face) { switch (cull_face) { case Maxwell::CullFace::Front: - return vk::CullModeFlagBits::eFront; + return VK_CULL_MODE_FRONT_BIT; case Maxwell::CullFace::Back: - return vk::CullModeFlagBits::eBack; + return VK_CULL_MODE_BACK_BIT; case Maxwell::CullFace::FrontAndBack: - return vk::CullModeFlagBits::eFrontAndBack; + return VK_CULL_MODE_FRONT_AND_BACK; } UNIMPLEMENTED_MSG("Unimplemented cull face={}", static_cast<u32>(cull_face)); return {}; } -vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle) { +VkComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle) { switch (swizzle) { case Tegra::Texture::SwizzleSource::Zero: - return vk::ComponentSwizzle::eZero; + return VK_COMPONENT_SWIZZLE_ZERO; case Tegra::Texture::SwizzleSource::R: - return vk::ComponentSwizzle::eR; + return VK_COMPONENT_SWIZZLE_R; case Tegra::Texture::SwizzleSource::G: - return vk::ComponentSwizzle::eG; + return VK_COMPONENT_SWIZZLE_G; case Tegra::Texture::SwizzleSource::B: - return vk::ComponentSwizzle::eB; + return VK_COMPONENT_SWIZZLE_B; case Tegra::Texture::SwizzleSource::A: - return vk::ComponentSwizzle::eA; + return VK_COMPONENT_SWIZZLE_A; case Tegra::Texture::SwizzleSource::OneInt: case Tegra::Texture::SwizzleSource::OneFloat: - return vk::ComponentSwizzle::eOne; + return VK_COMPONENT_SWIZZLE_ONE; } UNIMPLEMENTED_MSG("Unimplemented swizzle source={}", static_cast<u32>(swizzle)); return {}; diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.h b/src/video_core/renderer_vulkan/maxwell_to_vk.h index 24f6ab544..81bce4c6c 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.h +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.h @@ -6,8 +6,8 @@ #include "common/common_types.h" #include "video_core/engines/maxwell_3d.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/surface.h" #include "video_core/textures/texture.h" @@ -18,46 +18,45 @@ using PixelFormat = VideoCore::Surface::PixelFormat; namespace Sampler { -vk::Filter Filter(Tegra::Texture::TextureFilter filter); +VkFilter Filter(Tegra::Texture::TextureFilter filter); -vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter); +VkSamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter); -vk::SamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode, - Tegra::Texture::TextureFilter filter); +VkSamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode, + Tegra::Texture::TextureFilter filter); -vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func); +VkCompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func); } // namespace Sampler struct FormatInfo { - vk::Format format; + VkFormat format; bool attachable; bool storage; }; FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFormat pixel_format); -vk::ShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage); +VkShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage); -vk::PrimitiveTopology PrimitiveTopology(const VKDevice& device, - Maxwell::PrimitiveTopology topology); +VkPrimitiveTopology PrimitiveTopology(const VKDevice& device, Maxwell::PrimitiveTopology topology); -vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size); +VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size); -vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison); +VkCompareOp ComparisonOp(Maxwell::ComparisonOp comparison); -vk::IndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format); +VkIndexType IndexFormat(const VKDevice& device, Maxwell::IndexFormat index_format); -vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op); +VkStencilOp StencilOp(Maxwell::StencilOp stencil_op); -vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation); +VkBlendOp BlendEquation(Maxwell::Blend::Equation equation); -vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor); +VkBlendFactor BlendFactor(Maxwell::Blend::Factor factor); -vk::FrontFace FrontFace(Maxwell::FrontFace front_face); +VkFrontFace FrontFace(Maxwell::FrontFace front_face); -vk::CullModeFlags CullFace(Maxwell::CullFace cull_face); +VkCullModeFlags CullFace(Maxwell::CullFace cull_face); -vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle); +VkComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle); } // namespace Vulkan::MaxwellToVK diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp index 6953aaafe..dd590c38b 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp @@ -2,13 +2,18 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <algorithm> +#include <array> +#include <cstring> #include <memory> #include <optional> +#include <string> #include <vector> #include <fmt/format.h> #include "common/assert.h" +#include "common/dynamic_library.h" #include "common/logging/log.h" #include "common/telemetry.h" #include "core/core.h" @@ -19,7 +24,6 @@ #include "core/settings.h" #include "core/telemetry_session.h" #include "video_core/gpu.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/renderer_vulkan.h" #include "video_core/renderer_vulkan/vk_blit_screen.h" #include "video_core/renderer_vulkan/vk_device.h" @@ -29,30 +33,145 @@ #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_state_tracker.h" #include "video_core/renderer_vulkan/vk_swapchain.h" +#include "video_core/renderer_vulkan/wrapper.h" + +// Include these late to avoid polluting previous headers +#ifdef _WIN32 +#include <windows.h> +// ensure include order +#include <vulkan/vulkan_win32.h> +#endif + +#ifdef __linux__ +#include <X11/Xlib.h> +#include <vulkan/vulkan_wayland.h> +#include <vulkan/vulkan_xlib.h> +#endif namespace Vulkan { namespace { -VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity_, +using Core::Frontend::WindowSystemType; + +VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity, VkDebugUtilsMessageTypeFlagsEXT type, const VkDebugUtilsMessengerCallbackDataEXT* data, [[maybe_unused]] void* user_data) { - const vk::DebugUtilsMessageSeverityFlagBitsEXT severity{severity_}; const char* message{data->pMessage}; - if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eError) { + if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT) { LOG_CRITICAL(Render_Vulkan, "{}", message); - } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eWarning) { + } else if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT) { LOG_WARNING(Render_Vulkan, "{}", message); - } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eInfo) { + } else if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT) { LOG_INFO(Render_Vulkan, "{}", message); - } else if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eVerbose) { + } else if (severity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT) { LOG_DEBUG(Render_Vulkan, "{}", message); } return VK_FALSE; } +Common::DynamicLibrary OpenVulkanLibrary() { + Common::DynamicLibrary library; +#ifdef __APPLE__ + // Check if a path to a specific Vulkan library has been specified. + char* libvulkan_env = getenv("LIBVULKAN_PATH"); + if (!libvulkan_env || !library.Open(libvulkan_env)) { + // Use the libvulkan.dylib from the application bundle. + std::string filename = File::GetBundleDirectory() + "/Contents/Frameworks/libvulkan.dylib"; + library.Open(filename.c_str()); + } +#else + std::string filename = Common::DynamicLibrary::GetVersionedFilename("vulkan", 1); + if (!library.Open(filename.c_str())) { + // Android devices may not have libvulkan.so.1, only libvulkan.so. + filename = Common::DynamicLibrary::GetVersionedFilename("vulkan"); + library.Open(filename.c_str()); + } +#endif + return library; +} + +vk::Instance CreateInstance(Common::DynamicLibrary& library, vk::InstanceDispatch& dld, + WindowSystemType window_type = WindowSystemType::Headless, + bool enable_layers = false) { + if (!library.IsOpen()) { + LOG_ERROR(Render_Vulkan, "Vulkan library not available"); + return {}; + } + if (!library.GetSymbol("vkGetInstanceProcAddr", &dld.vkGetInstanceProcAddr)) { + LOG_ERROR(Render_Vulkan, "vkGetInstanceProcAddr not present in Vulkan"); + return {}; + } + if (!vk::Load(dld)) { + LOG_ERROR(Render_Vulkan, "Failed to load Vulkan function pointers"); + return {}; + } + + std::vector<const char*> extensions; + extensions.reserve(6); + switch (window_type) { + case Core::Frontend::WindowSystemType::Headless: + break; +#ifdef _WIN32 + case Core::Frontend::WindowSystemType::Windows: + extensions.push_back(VK_KHR_WIN32_SURFACE_EXTENSION_NAME); + break; +#endif +#ifdef __linux__ + case Core::Frontend::WindowSystemType::X11: + extensions.push_back(VK_KHR_XLIB_SURFACE_EXTENSION_NAME); + break; + case Core::Frontend::WindowSystemType::Wayland: + extensions.push_back(VK_KHR_WAYLAND_SURFACE_EXTENSION_NAME); + break; +#endif + default: + LOG_ERROR(Render_Vulkan, "Presentation not supported on this platform"); + break; + } + if (window_type != Core::Frontend::WindowSystemType::Headless) { + extensions.push_back(VK_KHR_SURFACE_EXTENSION_NAME); + } + if (enable_layers) { + extensions.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME); + } + extensions.push_back(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME); + + const std::optional properties = vk::EnumerateInstanceExtensionProperties(dld); + if (!properties) { + LOG_ERROR(Render_Vulkan, "Failed to query extension properties"); + return {}; + } + + for (const char* extension : extensions) { + const auto it = + std::find_if(properties->begin(), properties->end(), [extension](const auto& prop) { + return !std::strcmp(extension, prop.extensionName); + }); + if (it == properties->end()) { + LOG_ERROR(Render_Vulkan, "Required instance extension {} is not available", extension); + return {}; + } + } + + static constexpr std::array layers_data{"VK_LAYER_LUNARG_standard_validation"}; + vk::Span<const char*> layers = layers_data; + if (!enable_layers) { + layers = {}; + } + vk::Instance instance = vk::Instance::Create(layers, extensions, dld); + if (!instance) { + LOG_ERROR(Render_Vulkan, "Failed to create Vulkan instance"); + return {}; + } + if (!vk::Load(*instance, dld)) { + LOG_ERROR(Render_Vulkan, "Failed to load Vulkan instance function pointers"); + } + return instance; +} + std::string GetReadableVersion(u32 version) { return fmt::format("{}.{}.{}", VK_VERSION_MAJOR(version), VK_VERSION_MINOR(version), VK_VERSION_PATCH(version)); @@ -63,14 +182,14 @@ std::string GetDriverVersion(const VKDevice& device) { // https://github.com/SaschaWillems/vulkan.gpuinfo.org/blob/5dddea46ea1120b0df14eef8f15ff8e318e35462/functions.php#L308-L314 const u32 version = device.GetDriverVersion(); - if (device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) { + if (device.GetDriverID() == VK_DRIVER_ID_NVIDIA_PROPRIETARY_KHR) { const u32 major = (version >> 22) & 0x3ff; const u32 minor = (version >> 14) & 0x0ff; const u32 secondary = (version >> 6) & 0x0ff; const u32 tertiary = version & 0x003f; return fmt::format("{}.{}.{}.{}", major, minor, secondary, tertiary); } - if (device.GetDriverID() == vk::DriverIdKHR::eIntelProprietaryWindows) { + if (device.GetDriverID() == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS_KHR) { const u32 major = version >> 14; const u32 minor = version & 0x3fff; return fmt::format("{}.{}", major, minor); @@ -147,27 +266,12 @@ bool RendererVulkan::TryPresent(int /*timeout_ms*/) { } bool RendererVulkan::Init() { - PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr{}; - render_window.RetrieveVulkanHandlers(&vkGetInstanceProcAddr, &instance, &surface); - const vk::DispatchLoaderDynamic dldi(instance, vkGetInstanceProcAddr); - - std::optional<vk::DebugUtilsMessengerEXT> callback; - if (Settings::values.renderer_debug && dldi.vkCreateDebugUtilsMessengerEXT) { - callback = CreateDebugCallback(dldi); - if (!callback) { - return false; - } - } - - if (!PickDevices(dldi)) { - if (callback) { - instance.destroy(*callback, nullptr, dldi); - } + library = OpenVulkanLibrary(); + instance = CreateInstance(library, dld, render_window.GetWindowInfo().type, + Settings::values.renderer_debug); + if (!instance || !CreateDebugCallback() || !CreateSurface() || !PickDevices()) { return false; } - debug_callback = UniqueDebugUtilsMessengerEXT( - *callback, vk::ObjectDestroy<vk::Instance, vk::DispatchLoaderDynamic>( - instance, nullptr, device->GetDispatchLoader())); Report(); @@ -176,7 +280,7 @@ bool RendererVulkan::Init() { resource_manager = std::make_unique<VKResourceManager>(*device); const auto& framebuffer = render_window.GetFramebufferLayout(); - swapchain = std::make_unique<VKSwapchain>(surface, *device); + swapchain = std::make_unique<VKSwapchain>(*surface, *device); swapchain->Create(framebuffer.width, framebuffer.height, false); state_tracker = std::make_unique<StateTracker>(system); @@ -198,10 +302,8 @@ void RendererVulkan::ShutDown() { if (!device) { return; } - const auto dev = device->GetLogical(); - const auto& dld = device->GetDispatchLoader(); - if (dev && dld.vkDeviceWaitIdle) { - dev.waitIdle(dld); + if (const auto& dev = device->GetLogical()) { + dev.WaitIdle(); } rasterizer.reset(); @@ -213,44 +315,94 @@ void RendererVulkan::ShutDown() { device.reset(); } -std::optional<vk::DebugUtilsMessengerEXT> RendererVulkan::CreateDebugCallback( - const vk::DispatchLoaderDynamic& dldi) { - const vk::DebugUtilsMessengerCreateInfoEXT callback_ci( - {}, - vk::DebugUtilsMessageSeverityFlagBitsEXT::eError | - vk::DebugUtilsMessageSeverityFlagBitsEXT::eWarning | - vk::DebugUtilsMessageSeverityFlagBitsEXT::eInfo | - vk::DebugUtilsMessageSeverityFlagBitsEXT::eVerbose, - vk::DebugUtilsMessageTypeFlagBitsEXT::eGeneral | - vk::DebugUtilsMessageTypeFlagBitsEXT::eValidation | - vk::DebugUtilsMessageTypeFlagBitsEXT::ePerformance, - &DebugCallback, nullptr); - vk::DebugUtilsMessengerEXT callback; - if (instance.createDebugUtilsMessengerEXT(&callback_ci, nullptr, &callback, dldi) != - vk::Result::eSuccess) { +bool RendererVulkan::CreateDebugCallback() { + if (!Settings::values.renderer_debug) { + return true; + } + debug_callback = instance.TryCreateDebugCallback(DebugCallback); + if (!debug_callback) { LOG_ERROR(Render_Vulkan, "Failed to create debug callback"); - return {}; + return false; } - return callback; + return true; } -bool RendererVulkan::PickDevices(const vk::DispatchLoaderDynamic& dldi) { - const auto devices = instance.enumeratePhysicalDevices(dldi); +bool RendererVulkan::CreateSurface() { + [[maybe_unused]] const auto& window_info = render_window.GetWindowInfo(); + VkSurfaceKHR unsafe_surface = nullptr; + +#ifdef _WIN32 + if (window_info.type == Core::Frontend::WindowSystemType::Windows) { + const HWND hWnd = static_cast<HWND>(window_info.render_surface); + const VkWin32SurfaceCreateInfoKHR win32_ci{VK_STRUCTURE_TYPE_WIN32_SURFACE_CREATE_INFO_KHR, + nullptr, 0, nullptr, hWnd}; + const auto vkCreateWin32SurfaceKHR = reinterpret_cast<PFN_vkCreateWin32SurfaceKHR>( + dld.vkGetInstanceProcAddr(*instance, "vkCreateWin32SurfaceKHR")); + if (!vkCreateWin32SurfaceKHR || + vkCreateWin32SurfaceKHR(*instance, &win32_ci, nullptr, &unsafe_surface) != VK_SUCCESS) { + LOG_ERROR(Render_Vulkan, "Failed to initialize Win32 surface"); + return false; + } + } +#endif +#ifdef __linux__ + if (window_info.type == Core::Frontend::WindowSystemType::X11) { + const VkXlibSurfaceCreateInfoKHR xlib_ci{ + VK_STRUCTURE_TYPE_XLIB_SURFACE_CREATE_INFO_KHR, nullptr, 0, + static_cast<Display*>(window_info.display_connection), + reinterpret_cast<Window>(window_info.render_surface)}; + const auto vkCreateXlibSurfaceKHR = reinterpret_cast<PFN_vkCreateXlibSurfaceKHR>( + dld.vkGetInstanceProcAddr(*instance, "vkCreateXlibSurfaceKHR")); + if (!vkCreateXlibSurfaceKHR || + vkCreateXlibSurfaceKHR(*instance, &xlib_ci, nullptr, &unsafe_surface) != VK_SUCCESS) { + LOG_ERROR(Render_Vulkan, "Failed to initialize Xlib surface"); + return false; + } + } + if (window_info.type == Core::Frontend::WindowSystemType::Wayland) { + const VkWaylandSurfaceCreateInfoKHR wayland_ci{ + VK_STRUCTURE_TYPE_WAYLAND_SURFACE_CREATE_INFO_KHR, nullptr, 0, + static_cast<wl_display*>(window_info.display_connection), + static_cast<wl_surface*>(window_info.render_surface)}; + const auto vkCreateWaylandSurfaceKHR = reinterpret_cast<PFN_vkCreateWaylandSurfaceKHR>( + dld.vkGetInstanceProcAddr(*instance, "vkCreateWaylandSurfaceKHR")); + if (!vkCreateWaylandSurfaceKHR || + vkCreateWaylandSurfaceKHR(*instance, &wayland_ci, nullptr, &unsafe_surface) != + VK_SUCCESS) { + LOG_ERROR(Render_Vulkan, "Failed to initialize Wayland surface"); + return false; + } + } +#endif + if (!unsafe_surface) { + LOG_ERROR(Render_Vulkan, "Presentation not supported on this platform"); + return false; + } + + surface = vk::SurfaceKHR(unsafe_surface, *instance, dld); + return true; +} + +bool RendererVulkan::PickDevices() { + const auto devices = instance.EnumeratePhysicalDevices(); + if (!devices) { + LOG_ERROR(Render_Vulkan, "Failed to enumerate physical devices"); + return false; + } - // TODO(Rodrigo): Choose device from config file const s32 device_index = Settings::values.vulkan_device; - if (device_index < 0 || device_index >= static_cast<s32>(devices.size())) { + if (device_index < 0 || device_index >= static_cast<s32>(devices->size())) { LOG_ERROR(Render_Vulkan, "Invalid device index {}!", device_index); return false; } - const vk::PhysicalDevice physical_device = devices[device_index]; - - if (!VKDevice::IsSuitable(dldi, physical_device, surface)) { + const vk::PhysicalDevice physical_device((*devices)[static_cast<std::size_t>(device_index)], + dld); + if (!VKDevice::IsSuitable(physical_device, *surface)) { return false; } - device = std::make_unique<VKDevice>(dldi, physical_device, surface); - return device->Create(dldi, instance); + device = std::make_unique<VKDevice>(*instance, physical_device, *surface, dld); + return device->Create(); } void RendererVulkan::Report() const { @@ -276,4 +428,25 @@ void RendererVulkan::Report() const { telemetry_session.AddField(field, "GPU_Vulkan_Extensions", extensions); } +std::vector<std::string> RendererVulkan::EnumerateDevices() { + vk::InstanceDispatch dld; + Common::DynamicLibrary library = OpenVulkanLibrary(); + vk::Instance instance = CreateInstance(library, dld); + if (!instance) { + return {}; + } + + const std::optional physical_devices = instance.EnumeratePhysicalDevices(); + if (!physical_devices) { + return {}; + } + + std::vector<std::string> names; + names.reserve(physical_devices->size()); + for (const auto& device : *physical_devices) { + names.push_back(vk::PhysicalDevice(device, dld).GetProperties().deviceName); + } + return names; +} + } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h index d14384e79..18270909b 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.h +++ b/src/video_core/renderer_vulkan/renderer_vulkan.h @@ -6,10 +6,13 @@ #include <memory> #include <optional> +#include <string> #include <vector> +#include "common/dynamic_library.h" + #include "video_core/renderer_base.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Core { class System; @@ -44,22 +47,28 @@ public: void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override; bool TryPresent(int timeout_ms) override; + static std::vector<std::string> EnumerateDevices(); + private: - std::optional<vk::DebugUtilsMessengerEXT> CreateDebugCallback( - const vk::DispatchLoaderDynamic& dldi); + bool CreateDebugCallback(); - bool PickDevices(const vk::DispatchLoaderDynamic& dldi); + bool CreateSurface(); + + bool PickDevices(); void Report() const; Core::System& system; + Common::DynamicLibrary library; + vk::InstanceDispatch dld; + vk::Instance instance; vk::SurfaceKHR surface; VKScreenInfo screen_info; - UniqueDebugUtilsMessengerEXT debug_callback; + vk::DebugCallback debug_callback; std::unique_ptr<VKDevice> device; std::unique_ptr<VKSwapchain> swapchain; std::unique_ptr<VKMemoryManager> memory_manager; diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp index 855cfc883..fbd406f2b 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp +++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp @@ -20,7 +20,6 @@ #include "video_core/gpu.h" #include "video_core/morton.h" #include "video_core/rasterizer_interface.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/renderer_vulkan.h" #include "video_core/renderer_vulkan/vk_blit_screen.h" #include "video_core/renderer_vulkan/vk_device.h" @@ -30,6 +29,7 @@ #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_shader_util.h" #include "video_core/renderer_vulkan/vk_swapchain.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/surface.h" namespace Vulkan { @@ -140,16 +140,25 @@ struct ScreenRectVertex { std::array<f32, 2> position; std::array<f32, 2> tex_coord; - static vk::VertexInputBindingDescription GetDescription() { - return vk::VertexInputBindingDescription(0, sizeof(ScreenRectVertex), - vk::VertexInputRate::eVertex); + static VkVertexInputBindingDescription GetDescription() { + VkVertexInputBindingDescription description; + description.binding = 0; + description.stride = sizeof(ScreenRectVertex); + description.inputRate = VK_VERTEX_INPUT_RATE_VERTEX; + return description; } - static std::array<vk::VertexInputAttributeDescription, 2> GetAttributes() { - return {vk::VertexInputAttributeDescription(0, 0, vk::Format::eR32G32Sfloat, - offsetof(ScreenRectVertex, position)), - vk::VertexInputAttributeDescription(1, 0, vk::Format::eR32G32Sfloat, - offsetof(ScreenRectVertex, tex_coord))}; + static std::array<VkVertexInputAttributeDescription, 2> GetAttributes() { + std::array<VkVertexInputAttributeDescription, 2> attributes; + attributes[0].location = 0; + attributes[0].binding = 0; + attributes[0].format = VK_FORMAT_R32G32_SFLOAT; + attributes[0].offset = offsetof(ScreenRectVertex, position); + attributes[1].location = 1; + attributes[1].binding = 0; + attributes[1].format = VK_FORMAT_R32G32_SFLOAT; + attributes[1].offset = offsetof(ScreenRectVertex, tex_coord); + return attributes; } }; @@ -172,16 +181,16 @@ std::size_t GetSizeInBytes(const Tegra::FramebufferConfig& framebuffer) { static_cast<std::size_t>(framebuffer.height) * GetBytesPerPixel(framebuffer); } -vk::Format GetFormat(const Tegra::FramebufferConfig& framebuffer) { +VkFormat GetFormat(const Tegra::FramebufferConfig& framebuffer) { switch (framebuffer.pixel_format) { case Tegra::FramebufferConfig::PixelFormat::ABGR8: - return vk::Format::eA8B8G8R8UnormPack32; + return VK_FORMAT_A8B8G8R8_UNORM_PACK32; case Tegra::FramebufferConfig::PixelFormat::RGB565: - return vk::Format::eR5G6B5UnormPack16; + return VK_FORMAT_R5G6B5_UNORM_PACK16; default: UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", static_cast<u32>(framebuffer.pixel_format)); - return vk::Format::eA8B8G8R8UnormPack32; + return VK_FORMAT_A8B8G8R8_UNORM_PACK32; } } @@ -219,8 +228,8 @@ void VKBlitScreen::Recreate() { CreateDynamicResources(); } -std::tuple<VKFence&, vk::Semaphore> VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer, - bool use_accelerated) { +std::tuple<VKFence&, VkSemaphore> VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer, + bool use_accelerated) { RefreshResources(framebuffer); // Finish any pending renderpass @@ -255,46 +264,76 @@ std::tuple<VKFence&, vk::Semaphore> VKBlitScreen::Draw(const Tegra::FramebufferC framebuffer.stride, block_height_log2, framebuffer.height, 0, 1, 1, map.GetAddress() + image_offset, host_ptr); - blit_image->Transition(0, 1, 0, 1, vk::PipelineStageFlagBits::eTransfer, - vk::AccessFlagBits::eTransferWrite, - vk::ImageLayout::eTransferDstOptimal); - - const vk::BufferImageCopy copy(image_offset, 0, 0, - {vk::ImageAspectFlagBits::eColor, 0, 0, 1}, {0, 0, 0}, - {framebuffer.width, framebuffer.height, 1}); - scheduler.Record([buffer_handle = *buffer, image = blit_image->GetHandle(), - copy](auto cmdbuf, auto& dld) { - cmdbuf.copyBufferToImage(buffer_handle, image, vk::ImageLayout::eTransferDstOptimal, - {copy}, dld); - }); + blit_image->Transition(0, 1, 0, 1, VK_PIPELINE_STAGE_TRANSFER_BIT, + VK_ACCESS_TRANSFER_WRITE_BIT, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL); + + VkBufferImageCopy copy; + copy.bufferOffset = image_offset; + copy.bufferRowLength = 0; + copy.bufferImageHeight = 0; + copy.imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; + copy.imageSubresource.mipLevel = 0; + copy.imageSubresource.baseArrayLayer = 0; + copy.imageSubresource.layerCount = 1; + copy.imageOffset.x = 0; + copy.imageOffset.y = 0; + copy.imageOffset.z = 0; + copy.imageExtent.width = framebuffer.width; + copy.imageExtent.height = framebuffer.height; + copy.imageExtent.depth = 1; + scheduler.Record( + [buffer = *buffer, image = *blit_image->GetHandle(), copy](vk::CommandBuffer cmdbuf) { + cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy); + }); } map.Release(); - blit_image->Transition(0, 1, 0, 1, vk::PipelineStageFlagBits::eFragmentShader, - vk::AccessFlagBits::eShaderRead, - vk::ImageLayout::eShaderReadOnlyOptimal); + blit_image->Transition(0, 1, 0, 1, VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, + VK_ACCESS_SHADER_READ_BIT, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL); scheduler.Record([renderpass = *renderpass, framebuffer = *framebuffers[image_index], descriptor_set = descriptor_sets[image_index], buffer = *buffer, size = swapchain.GetSize(), pipeline = *pipeline, - layout = *pipeline_layout](auto cmdbuf, auto& dld) { - const vk::ClearValue clear_color{std::array{0.0f, 0.0f, 0.0f, 1.0f}}; - const vk::RenderPassBeginInfo renderpass_bi(renderpass, framebuffer, {{0, 0}, size}, 1, - &clear_color); - - cmdbuf.beginRenderPass(renderpass_bi, vk::SubpassContents::eInline, dld); - cmdbuf.bindPipeline(vk::PipelineBindPoint::eGraphics, pipeline, dld); - cmdbuf.setViewport( - 0, - {{0.0f, 0.0f, static_cast<f32>(size.width), static_cast<f32>(size.height), 0.0f, 1.0f}}, - dld); - cmdbuf.setScissor(0, {{{0, 0}, size}}, dld); - - cmdbuf.bindVertexBuffers(0, {buffer}, {offsetof(BufferData, vertices)}, dld); - cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eGraphics, layout, 0, {descriptor_set}, {}, - dld); - cmdbuf.draw(4, 1, 0, 0, dld); - cmdbuf.endRenderPass(dld); + layout = *pipeline_layout](vk::CommandBuffer cmdbuf) { + VkClearValue clear_color; + clear_color.color.float32[0] = 0.0f; + clear_color.color.float32[1] = 0.0f; + clear_color.color.float32[2] = 0.0f; + clear_color.color.float32[3] = 0.0f; + + VkRenderPassBeginInfo renderpass_bi; + renderpass_bi.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO; + renderpass_bi.pNext = nullptr; + renderpass_bi.renderPass = renderpass; + renderpass_bi.framebuffer = framebuffer; + renderpass_bi.renderArea.offset.x = 0; + renderpass_bi.renderArea.offset.y = 0; + renderpass_bi.renderArea.extent = size; + renderpass_bi.clearValueCount = 1; + renderpass_bi.pClearValues = &clear_color; + + VkViewport viewport; + viewport.x = 0.0f; + viewport.y = 0.0f; + viewport.width = static_cast<float>(size.width); + viewport.height = static_cast<float>(size.height); + viewport.minDepth = 0.0f; + viewport.maxDepth = 1.0f; + + VkRect2D scissor; + scissor.offset.x = 0; + scissor.offset.y = 0; + scissor.extent = size; + + cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE); + cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline); + cmdbuf.SetViewport(0, viewport); + cmdbuf.SetScissor(0, scissor); + + cmdbuf.BindVertexBuffer(0, buffer, offsetof(BufferData, vertices)); + cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set, {}); + cmdbuf.Draw(4, 1, 0, 0); + cmdbuf.EndRenderPass(); }); return {scheduler.GetFence(), *semaphores[image_index]}; @@ -334,165 +373,297 @@ void VKBlitScreen::CreateShaders() { } void VKBlitScreen::CreateSemaphores() { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - semaphores.resize(image_count); - for (std::size_t i = 0; i < image_count; ++i) { - semaphores[i] = dev.createSemaphoreUnique({}, nullptr, dld); - } + std::generate(semaphores.begin(), semaphores.end(), + [this] { return device.GetLogical().CreateSemaphore(); }); } void VKBlitScreen::CreateDescriptorPool() { - const std::array<vk::DescriptorPoolSize, 2> pool_sizes{ - vk::DescriptorPoolSize{vk::DescriptorType::eUniformBuffer, static_cast<u32>(image_count)}, - vk::DescriptorPoolSize{vk::DescriptorType::eCombinedImageSampler, - static_cast<u32>(image_count)}}; - const vk::DescriptorPoolCreateInfo pool_ci( - {}, static_cast<u32>(image_count), static_cast<u32>(pool_sizes.size()), pool_sizes.data()); - const auto dev = device.GetLogical(); - descriptor_pool = dev.createDescriptorPoolUnique(pool_ci, nullptr, device.GetDispatchLoader()); + std::array<VkDescriptorPoolSize, 2> pool_sizes; + pool_sizes[0].type = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; + pool_sizes[0].descriptorCount = static_cast<u32>(image_count); + pool_sizes[1].type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; + pool_sizes[1].descriptorCount = static_cast<u32>(image_count); + + VkDescriptorPoolCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT; + ci.maxSets = static_cast<u32>(image_count); + ci.poolSizeCount = static_cast<u32>(pool_sizes.size()); + ci.pPoolSizes = pool_sizes.data(); + descriptor_pool = device.GetLogical().CreateDescriptorPool(ci); } void VKBlitScreen::CreateRenderPass() { - const vk::AttachmentDescription color_attachment( - {}, swapchain.GetImageFormat(), vk::SampleCountFlagBits::e1, vk::AttachmentLoadOp::eClear, - vk::AttachmentStoreOp::eStore, vk::AttachmentLoadOp::eDontCare, - vk::AttachmentStoreOp::eDontCare, vk::ImageLayout::eUndefined, - vk::ImageLayout::ePresentSrcKHR); - - const vk::AttachmentReference color_attachment_ref(0, vk::ImageLayout::eColorAttachmentOptimal); - - const vk::SubpassDescription subpass_description({}, vk::PipelineBindPoint::eGraphics, 0, - nullptr, 1, &color_attachment_ref, nullptr, - nullptr, 0, nullptr); - - const vk::SubpassDependency dependency( - VK_SUBPASS_EXTERNAL, 0, vk::PipelineStageFlagBits::eColorAttachmentOutput, - vk::PipelineStageFlagBits::eColorAttachmentOutput, {}, - vk::AccessFlagBits::eColorAttachmentRead | vk::AccessFlagBits::eColorAttachmentWrite, {}); - - const vk::RenderPassCreateInfo renderpass_ci({}, 1, &color_attachment, 1, &subpass_description, - 1, &dependency); - - const auto dev = device.GetLogical(); - renderpass = dev.createRenderPassUnique(renderpass_ci, nullptr, device.GetDispatchLoader()); + VkAttachmentDescription color_attachment; + color_attachment.flags = 0; + color_attachment.format = swapchain.GetImageFormat(); + color_attachment.samples = VK_SAMPLE_COUNT_1_BIT; + color_attachment.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR; + color_attachment.storeOp = VK_ATTACHMENT_STORE_OP_STORE; + color_attachment.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; + color_attachment.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; + color_attachment.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; + color_attachment.finalLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR; + + VkAttachmentReference color_attachment_ref; + color_attachment_ref.attachment = 0; + color_attachment_ref.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; + + VkSubpassDescription subpass_description; + subpass_description.flags = 0; + subpass_description.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; + subpass_description.inputAttachmentCount = 0; + subpass_description.pInputAttachments = nullptr; + subpass_description.colorAttachmentCount = 1; + subpass_description.pColorAttachments = &color_attachment_ref; + subpass_description.pResolveAttachments = nullptr; + subpass_description.pDepthStencilAttachment = nullptr; + subpass_description.preserveAttachmentCount = 0; + subpass_description.pPreserveAttachments = nullptr; + + VkSubpassDependency dependency; + dependency.srcSubpass = VK_SUBPASS_EXTERNAL; + dependency.dstSubpass = 0; + dependency.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT; + dependency.dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT; + dependency.srcAccessMask = 0; + dependency.dstAccessMask = + VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT; + dependency.dependencyFlags = 0; + + VkRenderPassCreateInfo renderpass_ci; + renderpass_ci.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO; + renderpass_ci.pNext = nullptr; + renderpass_ci.flags = 0; + renderpass_ci.attachmentCount = 1; + renderpass_ci.pAttachments = &color_attachment; + renderpass_ci.subpassCount = 1; + renderpass_ci.pSubpasses = &subpass_description; + renderpass_ci.dependencyCount = 1; + renderpass_ci.pDependencies = &dependency; + + renderpass = device.GetLogical().CreateRenderPass(renderpass_ci); } void VKBlitScreen::CreateDescriptorSetLayout() { - const std::array<vk::DescriptorSetLayoutBinding, 2> layout_bindings{ - vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eUniformBuffer, 1, - vk::ShaderStageFlagBits::eVertex, nullptr), - vk::DescriptorSetLayoutBinding(1, vk::DescriptorType::eCombinedImageSampler, 1, - vk::ShaderStageFlagBits::eFragment, nullptr)}; - const vk::DescriptorSetLayoutCreateInfo descriptor_layout_ci( - {}, static_cast<u32>(layout_bindings.size()), layout_bindings.data()); - - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - descriptor_set_layout = dev.createDescriptorSetLayoutUnique(descriptor_layout_ci, nullptr, dld); + std::array<VkDescriptorSetLayoutBinding, 2> layout_bindings; + layout_bindings[0].binding = 0; + layout_bindings[0].descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; + layout_bindings[0].descriptorCount = 1; + layout_bindings[0].stageFlags = VK_SHADER_STAGE_VERTEX_BIT; + layout_bindings[0].pImmutableSamplers = nullptr; + layout_bindings[1].binding = 1; + layout_bindings[1].descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; + layout_bindings[1].descriptorCount = 1; + layout_bindings[1].stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT; + layout_bindings[1].pImmutableSamplers = nullptr; + + VkDescriptorSetLayoutCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.bindingCount = static_cast<u32>(layout_bindings.size()); + ci.pBindings = layout_bindings.data(); + + descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout(ci); } void VKBlitScreen::CreateDescriptorSets() { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - - descriptor_sets.resize(image_count); - for (std::size_t i = 0; i < image_count; ++i) { - const vk::DescriptorSetLayout layout = *descriptor_set_layout; - const vk::DescriptorSetAllocateInfo descriptor_set_ai(*descriptor_pool, 1, &layout); - const vk::Result result = - dev.allocateDescriptorSets(&descriptor_set_ai, &descriptor_sets[i], dld); - ASSERT(result == vk::Result::eSuccess); - } + const std::vector layouts(image_count, *descriptor_set_layout); + + VkDescriptorSetAllocateInfo ai; + ai.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO; + ai.pNext = nullptr; + ai.descriptorPool = *descriptor_pool; + ai.descriptorSetCount = static_cast<u32>(image_count); + ai.pSetLayouts = layouts.data(); + descriptor_sets = descriptor_pool.Allocate(ai); } void VKBlitScreen::CreatePipelineLayout() { - const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &descriptor_set_layout.get(), 0, - nullptr); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - pipeline_layout = dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld); + VkPipelineLayoutCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.setLayoutCount = 1; + ci.pSetLayouts = descriptor_set_layout.address(); + ci.pushConstantRangeCount = 0; + ci.pPushConstantRanges = nullptr; + pipeline_layout = device.GetLogical().CreatePipelineLayout(ci); } void VKBlitScreen::CreateGraphicsPipeline() { - const std::array shader_stages = { - vk::PipelineShaderStageCreateInfo({}, vk::ShaderStageFlagBits::eVertex, *vertex_shader, - "main", nullptr), - vk::PipelineShaderStageCreateInfo({}, vk::ShaderStageFlagBits::eFragment, *fragment_shader, - "main", nullptr)}; + std::array<VkPipelineShaderStageCreateInfo, 2> shader_stages; + shader_stages[0].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; + shader_stages[0].pNext = nullptr; + shader_stages[0].flags = 0; + shader_stages[0].stage = VK_SHADER_STAGE_VERTEX_BIT; + shader_stages[0].module = *vertex_shader; + shader_stages[0].pName = "main"; + shader_stages[0].pSpecializationInfo = nullptr; + shader_stages[1].sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; + shader_stages[1].pNext = nullptr; + shader_stages[1].flags = 0; + shader_stages[1].stage = VK_SHADER_STAGE_FRAGMENT_BIT; + shader_stages[1].module = *fragment_shader; + shader_stages[1].pName = "main"; + shader_stages[1].pSpecializationInfo = nullptr; const auto vertex_binding_description = ScreenRectVertex::GetDescription(); const auto vertex_attrs_description = ScreenRectVertex::GetAttributes(); - const vk::PipelineVertexInputStateCreateInfo vertex_input( - {}, 1, &vertex_binding_description, static_cast<u32>(vertex_attrs_description.size()), - vertex_attrs_description.data()); - - const vk::PipelineInputAssemblyStateCreateInfo input_assembly( - {}, vk::PrimitiveTopology::eTriangleStrip, false); - - // Set a dummy viewport, it's going to be replaced by dynamic states. - const vk::Viewport viewport(0.0f, 0.0f, 1.0f, 1.0f, 0.0f, 1.0f); - const vk::Rect2D scissor({0, 0}, {1, 1}); - const vk::PipelineViewportStateCreateInfo viewport_state({}, 1, &viewport, 1, &scissor); - - const vk::PipelineRasterizationStateCreateInfo rasterizer( - {}, false, false, vk::PolygonMode::eFill, vk::CullModeFlagBits::eNone, - vk::FrontFace::eClockwise, false, 0.0f, 0.0f, 0.0f, 1.0f); - - const vk::PipelineMultisampleStateCreateInfo multisampling({}, vk::SampleCountFlagBits::e1, - false, 0.0f, nullptr, false, false); - - const vk::PipelineColorBlendAttachmentState color_blend_attachment( - false, vk::BlendFactor::eZero, vk::BlendFactor::eZero, vk::BlendOp::eAdd, - vk::BlendFactor::eZero, vk::BlendFactor::eZero, vk::BlendOp::eAdd, - vk::ColorComponentFlagBits::eR | vk::ColorComponentFlagBits::eG | - vk::ColorComponentFlagBits::eB | vk::ColorComponentFlagBits::eA); - - const vk::PipelineColorBlendStateCreateInfo color_blending( - {}, false, vk::LogicOp::eCopy, 1, &color_blend_attachment, {0.0f, 0.0f, 0.0f, 0.0f}); - - const std::array<vk::DynamicState, 2> dynamic_states = {vk::DynamicState::eViewport, - vk::DynamicState::eScissor}; - - const vk::PipelineDynamicStateCreateInfo dynamic_state( - {}, static_cast<u32>(dynamic_states.size()), dynamic_states.data()); - - const vk::GraphicsPipelineCreateInfo pipeline_ci( - {}, static_cast<u32>(shader_stages.size()), shader_stages.data(), &vertex_input, - &input_assembly, nullptr, &viewport_state, &rasterizer, &multisampling, nullptr, - &color_blending, &dynamic_state, *pipeline_layout, *renderpass, 0, nullptr, 0); - - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - pipeline = dev.createGraphicsPipelineUnique({}, pipeline_ci, nullptr, dld); + VkPipelineVertexInputStateCreateInfo vertex_input_ci; + vertex_input_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO; + vertex_input_ci.pNext = nullptr; + vertex_input_ci.flags = 0; + vertex_input_ci.vertexBindingDescriptionCount = 1; + vertex_input_ci.pVertexBindingDescriptions = &vertex_binding_description; + vertex_input_ci.vertexAttributeDescriptionCount = u32{vertex_attrs_description.size()}; + vertex_input_ci.pVertexAttributeDescriptions = vertex_attrs_description.data(); + + VkPipelineInputAssemblyStateCreateInfo input_assembly_ci; + input_assembly_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO; + input_assembly_ci.pNext = nullptr; + input_assembly_ci.flags = 0; + input_assembly_ci.topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP; + input_assembly_ci.primitiveRestartEnable = VK_FALSE; + + VkPipelineViewportStateCreateInfo viewport_state_ci; + viewport_state_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO; + viewport_state_ci.pNext = nullptr; + viewport_state_ci.flags = 0; + viewport_state_ci.viewportCount = 1; + viewport_state_ci.pViewports = nullptr; + viewport_state_ci.scissorCount = 1; + viewport_state_ci.pScissors = nullptr; + + VkPipelineRasterizationStateCreateInfo rasterization_ci; + rasterization_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO; + rasterization_ci.pNext = nullptr; + rasterization_ci.flags = 0; + rasterization_ci.depthClampEnable = VK_FALSE; + rasterization_ci.rasterizerDiscardEnable = VK_FALSE; + rasterization_ci.polygonMode = VK_POLYGON_MODE_FILL; + rasterization_ci.cullMode = VK_CULL_MODE_NONE; + rasterization_ci.frontFace = VK_FRONT_FACE_CLOCKWISE; + rasterization_ci.depthBiasEnable = VK_FALSE; + rasterization_ci.depthBiasConstantFactor = 0.0f; + rasterization_ci.depthBiasClamp = 0.0f; + rasterization_ci.depthBiasSlopeFactor = 0.0f; + rasterization_ci.lineWidth = 1.0f; + + VkPipelineMultisampleStateCreateInfo multisampling_ci; + multisampling_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO; + multisampling_ci.pNext = nullptr; + multisampling_ci.flags = 0; + multisampling_ci.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT; + multisampling_ci.sampleShadingEnable = VK_FALSE; + multisampling_ci.minSampleShading = 0.0f; + multisampling_ci.pSampleMask = nullptr; + multisampling_ci.alphaToCoverageEnable = VK_FALSE; + multisampling_ci.alphaToOneEnable = VK_FALSE; + + VkPipelineColorBlendAttachmentState color_blend_attachment; + color_blend_attachment.blendEnable = VK_FALSE; + color_blend_attachment.srcColorBlendFactor = VK_BLEND_FACTOR_ZERO; + color_blend_attachment.dstColorBlendFactor = VK_BLEND_FACTOR_ZERO; + color_blend_attachment.colorBlendOp = VK_BLEND_OP_ADD; + color_blend_attachment.srcAlphaBlendFactor = VK_BLEND_FACTOR_ZERO; + color_blend_attachment.dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO; + color_blend_attachment.alphaBlendOp = VK_BLEND_OP_ADD; + color_blend_attachment.colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT | + VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT; + + VkPipelineColorBlendStateCreateInfo color_blend_ci; + color_blend_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO; + color_blend_ci.flags = 0; + color_blend_ci.pNext = nullptr; + color_blend_ci.logicOpEnable = VK_FALSE; + color_blend_ci.logicOp = VK_LOGIC_OP_COPY; + color_blend_ci.attachmentCount = 1; + color_blend_ci.pAttachments = &color_blend_attachment; + color_blend_ci.blendConstants[0] = 0.0f; + color_blend_ci.blendConstants[1] = 0.0f; + color_blend_ci.blendConstants[2] = 0.0f; + color_blend_ci.blendConstants[3] = 0.0f; + + static constexpr std::array dynamic_states = {VK_DYNAMIC_STATE_VIEWPORT, + VK_DYNAMIC_STATE_SCISSOR}; + VkPipelineDynamicStateCreateInfo dynamic_state_ci; + dynamic_state_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO; + dynamic_state_ci.pNext = nullptr; + dynamic_state_ci.flags = 0; + dynamic_state_ci.dynamicStateCount = static_cast<u32>(dynamic_states.size()); + dynamic_state_ci.pDynamicStates = dynamic_states.data(); + + VkGraphicsPipelineCreateInfo pipeline_ci; + pipeline_ci.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO; + pipeline_ci.pNext = nullptr; + pipeline_ci.flags = 0; + pipeline_ci.stageCount = static_cast<u32>(shader_stages.size()); + pipeline_ci.pStages = shader_stages.data(); + pipeline_ci.pVertexInputState = &vertex_input_ci; + pipeline_ci.pInputAssemblyState = &input_assembly_ci; + pipeline_ci.pTessellationState = nullptr; + pipeline_ci.pViewportState = &viewport_state_ci; + pipeline_ci.pRasterizationState = &rasterization_ci; + pipeline_ci.pMultisampleState = &multisampling_ci; + pipeline_ci.pDepthStencilState = nullptr; + pipeline_ci.pColorBlendState = &color_blend_ci; + pipeline_ci.pDynamicState = &dynamic_state_ci; + pipeline_ci.layout = *pipeline_layout; + pipeline_ci.renderPass = *renderpass; + pipeline_ci.subpass = 0; + pipeline_ci.basePipelineHandle = 0; + pipeline_ci.basePipelineIndex = 0; + + pipeline = device.GetLogical().CreateGraphicsPipeline(pipeline_ci); } void VKBlitScreen::CreateSampler() { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - const vk::SamplerCreateInfo sampler_ci( - {}, vk::Filter::eLinear, vk::Filter::eLinear, vk::SamplerMipmapMode::eLinear, - vk::SamplerAddressMode::eClampToBorder, vk::SamplerAddressMode::eClampToBorder, - vk::SamplerAddressMode::eClampToBorder, 0.0f, false, 0.0f, false, vk::CompareOp::eNever, - 0.0f, 0.0f, vk::BorderColor::eFloatOpaqueBlack, false); - sampler = dev.createSamplerUnique(sampler_ci, nullptr, dld); + VkSamplerCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.magFilter = VK_FILTER_LINEAR; + ci.minFilter = VK_FILTER_NEAREST; + ci.mipmapMode = VK_SAMPLER_MIPMAP_MODE_LINEAR; + ci.addressModeU = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; + ci.addressModeV = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; + ci.addressModeW = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER; + ci.mipLodBias = 0.0f; + ci.anisotropyEnable = VK_FALSE; + ci.maxAnisotropy = 0.0f; + ci.compareEnable = VK_FALSE; + ci.compareOp = VK_COMPARE_OP_NEVER; + ci.minLod = 0.0f; + ci.maxLod = 0.0f; + ci.borderColor = VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK; + ci.unnormalizedCoordinates = VK_FALSE; + + sampler = device.GetLogical().CreateSampler(ci); } void VKBlitScreen::CreateFramebuffers() { - const vk::Extent2D size{swapchain.GetSize()}; - framebuffers.clear(); + const VkExtent2D size{swapchain.GetSize()}; framebuffers.resize(image_count); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); + VkFramebufferCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.renderPass = *renderpass; + ci.attachmentCount = 1; + ci.width = size.width; + ci.height = size.height; + ci.layers = 1; for (std::size_t i = 0; i < image_count; ++i) { - const vk::ImageView image_view{swapchain.GetImageViewIndex(i)}; - const vk::FramebufferCreateInfo framebuffer_ci({}, *renderpass, 1, &image_view, size.width, - size.height, 1); - framebuffers[i] = dev.createFramebufferUnique(framebuffer_ci, nullptr, dld); + const VkImageView image_view{swapchain.GetImageViewIndex(i)}; + ci.pAttachments = &image_view; + framebuffers[i] = device.GetLogical().CreateFramebuffer(ci); } } @@ -507,54 +678,86 @@ void VKBlitScreen::ReleaseRawImages() { } void VKBlitScreen::CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer) { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - - const vk::BufferCreateInfo buffer_ci({}, CalculateBufferSize(framebuffer), - vk::BufferUsageFlagBits::eTransferSrc | - vk::BufferUsageFlagBits::eVertexBuffer | - vk::BufferUsageFlagBits::eUniformBuffer, - vk::SharingMode::eExclusive, 0, nullptr); - buffer = dev.createBufferUnique(buffer_ci, nullptr, dld); - buffer_commit = memory_manager.Commit(*buffer, true); + VkBufferCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.size = CalculateBufferSize(framebuffer); + ci.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | + VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT; + ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; + ci.queueFamilyIndexCount = 0; + ci.pQueueFamilyIndices = nullptr; + + buffer = device.GetLogical().CreateBuffer(ci); + buffer_commit = memory_manager.Commit(buffer, true); } void VKBlitScreen::CreateRawImages(const Tegra::FramebufferConfig& framebuffer) { raw_images.resize(image_count); raw_buffer_commits.resize(image_count); - const auto format = GetFormat(framebuffer); + VkImageCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.imageType = VK_IMAGE_TYPE_2D; + ci.format = GetFormat(framebuffer); + ci.extent.width = framebuffer.width; + ci.extent.height = framebuffer.height; + ci.extent.depth = 1; + ci.mipLevels = 1; + ci.arrayLayers = 1; + ci.samples = VK_SAMPLE_COUNT_1_BIT; + ci.tiling = VK_IMAGE_TILING_LINEAR; + ci.usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT; + ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; + ci.queueFamilyIndexCount = 0; + ci.pQueueFamilyIndices = nullptr; + ci.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; + for (std::size_t i = 0; i < image_count; ++i) { - const vk::ImageCreateInfo image_ci( - {}, vk::ImageType::e2D, format, {framebuffer.width, framebuffer.height, 1}, 1, 1, - vk::SampleCountFlagBits::e1, vk::ImageTiling::eOptimal, - vk::ImageUsageFlagBits::eTransferDst | vk::ImageUsageFlagBits::eSampled, - vk::SharingMode::eExclusive, 0, nullptr, vk::ImageLayout::eUndefined); - - raw_images[i] = - std::make_unique<VKImage>(device, scheduler, image_ci, vk::ImageAspectFlagBits::eColor); + raw_images[i] = std::make_unique<VKImage>(device, scheduler, ci, VK_IMAGE_ASPECT_COLOR_BIT); raw_buffer_commits[i] = memory_manager.Commit(raw_images[i]->GetHandle(), false); } } -void VKBlitScreen::UpdateDescriptorSet(std::size_t image_index, vk::ImageView image_view) const { - const vk::DescriptorSet descriptor_set = descriptor_sets[image_index]; - - const vk::DescriptorBufferInfo buffer_info(*buffer, offsetof(BufferData, uniform), - sizeof(BufferData::uniform)); - const vk::WriteDescriptorSet ubo_write(descriptor_set, 0, 0, 1, - vk::DescriptorType::eUniformBuffer, nullptr, - &buffer_info, nullptr); - - const vk::DescriptorImageInfo image_info(*sampler, image_view, - vk::ImageLayout::eShaderReadOnlyOptimal); - const vk::WriteDescriptorSet sampler_write(descriptor_set, 1, 0, 1, - vk::DescriptorType::eCombinedImageSampler, - &image_info, nullptr, nullptr); - - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - dev.updateDescriptorSets({ubo_write, sampler_write}, {}, dld); +void VKBlitScreen::UpdateDescriptorSet(std::size_t image_index, VkImageView image_view) const { + VkDescriptorBufferInfo buffer_info; + buffer_info.buffer = *buffer; + buffer_info.offset = offsetof(BufferData, uniform); + buffer_info.range = sizeof(BufferData::uniform); + + VkWriteDescriptorSet ubo_write; + ubo_write.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET; + ubo_write.pNext = nullptr; + ubo_write.dstSet = descriptor_sets[image_index]; + ubo_write.dstBinding = 0; + ubo_write.dstArrayElement = 0; + ubo_write.descriptorCount = 1; + ubo_write.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; + ubo_write.pImageInfo = nullptr; + ubo_write.pBufferInfo = &buffer_info; + ubo_write.pTexelBufferView = nullptr; + + VkDescriptorImageInfo image_info; + image_info.sampler = *sampler; + image_info.imageView = image_view; + image_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; + + VkWriteDescriptorSet sampler_write; + sampler_write.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET; + sampler_write.pNext = nullptr; + sampler_write.dstSet = descriptor_sets[image_index]; + sampler_write.dstBinding = 1; + sampler_write.dstArrayElement = 0; + sampler_write.descriptorCount = 1; + sampler_write.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; + sampler_write.pImageInfo = &image_info; + sampler_write.pBufferInfo = nullptr; + sampler_write.pTexelBufferView = nullptr; + + device.GetLogical().UpdateDescriptorSets(std::array{ubo_write, sampler_write}, {}); } void VKBlitScreen::SetUniformData(BufferData& data, diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.h b/src/video_core/renderer_vulkan/vk_blit_screen.h index ea680b3f5..5eb544aea 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.h +++ b/src/video_core/renderer_vulkan/vk_blit_screen.h @@ -8,9 +8,9 @@ #include <memory> #include <tuple> -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_memory_manager.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Core { class System; @@ -49,8 +49,8 @@ public: void Recreate(); - std::tuple<VKFence&, vk::Semaphore> Draw(const Tegra::FramebufferConfig& framebuffer, - bool use_accelerated); + std::tuple<VKFence&, VkSemaphore> Draw(const Tegra::FramebufferConfig& framebuffer, + bool use_accelerated); private: struct BufferData; @@ -74,7 +74,7 @@ private: void CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer); void CreateRawImages(const Tegra::FramebufferConfig& framebuffer); - void UpdateDescriptorSet(std::size_t image_index, vk::ImageView image_view) const; + void UpdateDescriptorSet(std::size_t image_index, VkImageView image_view) const; void SetUniformData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const; void SetVertexData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const; @@ -93,23 +93,23 @@ private: const std::size_t image_count; const VKScreenInfo& screen_info; - UniqueShaderModule vertex_shader; - UniqueShaderModule fragment_shader; - UniqueDescriptorPool descriptor_pool; - UniqueDescriptorSetLayout descriptor_set_layout; - UniquePipelineLayout pipeline_layout; - UniquePipeline pipeline; - UniqueRenderPass renderpass; - std::vector<UniqueFramebuffer> framebuffers; - std::vector<vk::DescriptorSet> descriptor_sets; - UniqueSampler sampler; - - UniqueBuffer buffer; + vk::ShaderModule vertex_shader; + vk::ShaderModule fragment_shader; + vk::DescriptorPool descriptor_pool; + vk::DescriptorSetLayout descriptor_set_layout; + vk::PipelineLayout pipeline_layout; + vk::Pipeline pipeline; + vk::RenderPass renderpass; + std::vector<vk::Framebuffer> framebuffers; + vk::DescriptorSets descriptor_sets; + vk::Sampler sampler; + + vk::Buffer buffer; VKMemoryCommit buffer_commit; std::vector<std::unique_ptr<VKFenceWatch>> watches; - std::vector<UniqueSemaphore> semaphores; + std::vector<vk::Semaphore> semaphores; std::vector<std::unique_ptr<VKImage>> raw_images; std::vector<VKMemoryCommit> raw_buffer_commits; u32 raw_width = 0; diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp index 1ba544943..0d167afbd 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp @@ -11,48 +11,50 @@ #include "common/assert.h" #include "common/bit_util.h" #include "core/core.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_buffer_cache.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_stream_buffer.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { namespace { -const auto BufferUsage = - vk::BufferUsageFlagBits::eVertexBuffer | vk::BufferUsageFlagBits::eIndexBuffer | - vk::BufferUsageFlagBits::eUniformBuffer | vk::BufferUsageFlagBits::eStorageBuffer; +constexpr VkBufferUsageFlags BUFFER_USAGE = + VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_INDEX_BUFFER_BIT | + VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT; -const auto UploadPipelineStage = - vk::PipelineStageFlagBits::eTransfer | vk::PipelineStageFlagBits::eVertexInput | - vk::PipelineStageFlagBits::eVertexShader | vk::PipelineStageFlagBits::eFragmentShader | - vk::PipelineStageFlagBits::eComputeShader; +constexpr VkPipelineStageFlags UPLOAD_PIPELINE_STAGE = + VK_PIPELINE_STAGE_TRANSFER_BIT | VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | + VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | + VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; -const auto UploadAccessBarriers = - vk::AccessFlagBits::eTransferRead | vk::AccessFlagBits::eShaderRead | - vk::AccessFlagBits::eUniformRead | vk::AccessFlagBits::eVertexAttributeRead | - vk::AccessFlagBits::eIndexRead; +constexpr VkAccessFlags UPLOAD_ACCESS_BARRIERS = + VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_UNIFORM_READ_BIT | + VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDEX_READ_BIT; -auto CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) { - return std::make_unique<VKStreamBuffer>(device, scheduler, BufferUsage); +std::unique_ptr<VKStreamBuffer> CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) { + return std::make_unique<VKStreamBuffer>(device, scheduler, BUFFER_USAGE); } } // Anonymous namespace CachedBufferBlock::CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager, - CacheAddr cache_addr, std::size_t size) - : VideoCommon::BufferBlock{cache_addr, size} { - const vk::BufferCreateInfo buffer_ci({}, static_cast<vk::DeviceSize>(size), - BufferUsage | vk::BufferUsageFlagBits::eTransferSrc | - vk::BufferUsageFlagBits::eTransferDst, - vk::SharingMode::eExclusive, 0, nullptr); - - const auto& dld{device.GetDispatchLoader()}; - const auto dev{device.GetLogical()}; - buffer.handle = dev.createBufferUnique(buffer_ci, nullptr, dld); - buffer.commit = memory_manager.Commit(*buffer.handle, false); + VAddr cpu_addr, std::size_t size) + : VideoCommon::BufferBlock{cpu_addr, size} { + VkBufferCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.size = static_cast<VkDeviceSize>(size); + ci.usage = BUFFER_USAGE | VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; + ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; + ci.queueFamilyIndexCount = 0; + ci.pQueueFamilyIndices = nullptr; + + buffer.handle = device.GetLogical().CreateBuffer(ci); + buffer.commit = memory_manager.Commit(buffer.handle, false); } CachedBufferBlock::~CachedBufferBlock() = default; @@ -60,30 +62,30 @@ CachedBufferBlock::~CachedBufferBlock() = default; VKBufferCache::VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system, const VKDevice& device, VKMemoryManager& memory_manager, VKScheduler& scheduler, VKStagingBufferPool& staging_pool) - : VideoCommon::BufferCache<Buffer, vk::Buffer, VKStreamBuffer>{rasterizer, system, - CreateStreamBuffer(device, - scheduler)}, + : VideoCommon::BufferCache<Buffer, VkBuffer, VKStreamBuffer>{rasterizer, system, + CreateStreamBuffer(device, + scheduler)}, device{device}, memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{ staging_pool} {} VKBufferCache::~VKBufferCache() = default; -Buffer VKBufferCache::CreateBlock(CacheAddr cache_addr, std::size_t size) { - return std::make_shared<CachedBufferBlock>(device, memory_manager, cache_addr, size); +Buffer VKBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) { + return std::make_shared<CachedBufferBlock>(device, memory_manager, cpu_addr, size); } -const vk::Buffer* VKBufferCache::ToHandle(const Buffer& buffer) { +const VkBuffer* VKBufferCache::ToHandle(const Buffer& buffer) { return buffer->GetHandle(); } -const vk::Buffer* VKBufferCache::GetEmptyBuffer(std::size_t size) { +const VkBuffer* VKBufferCache::GetEmptyBuffer(std::size_t size) { size = std::max(size, std::size_t(4)); const auto& empty = staging_pool.GetUnusedBuffer(size, false); scheduler.RequestOutsideRenderPassOperationContext(); - scheduler.Record([size, buffer = *empty.handle](vk::CommandBuffer cmdbuf, auto& dld) { - cmdbuf.fillBuffer(buffer, 0, size, 0, dld); + scheduler.Record([size, buffer = *empty.handle](vk::CommandBuffer cmdbuf) { + cmdbuf.FillBuffer(buffer, 0, size, 0); }); - return &*empty.handle; + return empty.handle.address(); } void VKBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size, @@ -93,14 +95,21 @@ void VKBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, st scheduler.RequestOutsideRenderPassOperationContext(); scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset, - size](auto cmdbuf, auto& dld) { - cmdbuf.copyBuffer(staging, buffer, {{0, offset, size}}, dld); - cmdbuf.pipelineBarrier( - vk::PipelineStageFlagBits::eTransfer, UploadPipelineStage, {}, {}, - {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, UploadAccessBarriers, - VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, - offset, size)}, - {}, dld); + size](vk::CommandBuffer cmdbuf) { + cmdbuf.CopyBuffer(staging, buffer, VkBufferCopy{0, offset, size}); + + VkBufferMemoryBarrier barrier; + barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; + barrier.pNext = nullptr; + barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT; + barrier.dstAccessMask = UPLOAD_ACCESS_BARRIERS; + barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.buffer = buffer; + barrier.offset = offset; + barrier.size = size; + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, UPLOAD_PIPELINE_STAGE, 0, {}, + barrier, {}); }); } @@ -109,16 +118,23 @@ void VKBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset, const auto& staging = staging_pool.GetUnusedBuffer(size, true); scheduler.RequestOutsideRenderPassOperationContext(); scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset, - size](auto cmdbuf, auto& dld) { - cmdbuf.pipelineBarrier( - vk::PipelineStageFlagBits::eVertexShader | vk::PipelineStageFlagBits::eFragmentShader | - vk::PipelineStageFlagBits::eComputeShader, - vk::PipelineStageFlagBits::eTransfer, {}, {}, - {vk::BufferMemoryBarrier(vk::AccessFlagBits::eShaderWrite, - vk::AccessFlagBits::eTransferRead, VK_QUEUE_FAMILY_IGNORED, - VK_QUEUE_FAMILY_IGNORED, buffer, offset, size)}, - {}, dld); - cmdbuf.copyBuffer(buffer, staging, {{offset, 0, size}}, dld); + size](vk::CommandBuffer cmdbuf) { + VkBufferMemoryBarrier barrier; + barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; + barrier.pNext = nullptr; + barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT; + barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT; + barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.buffer = buffer; + barrier.offset = offset; + barrier.size = size; + + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | + VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | + VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, + VK_PIPELINE_STAGE_TRANSFER_BIT, 0, {}, barrier, {}); + cmdbuf.CopyBuffer(buffer, staging, VkBufferCopy{offset, 0, size}); }); scheduler.Finish(); @@ -129,17 +145,30 @@ void VKBufferCache::CopyBlock(const Buffer& src, const Buffer& dst, std::size_t std::size_t dst_offset, std::size_t size) { scheduler.RequestOutsideRenderPassOperationContext(); scheduler.Record([src_buffer = *src->GetHandle(), dst_buffer = *dst->GetHandle(), src_offset, - dst_offset, size](auto cmdbuf, auto& dld) { - cmdbuf.copyBuffer(src_buffer, dst_buffer, {{src_offset, dst_offset, size}}, dld); - cmdbuf.pipelineBarrier( - vk::PipelineStageFlagBits::eTransfer, UploadPipelineStage, {}, {}, - {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferRead, - vk::AccessFlagBits::eShaderWrite, VK_QUEUE_FAMILY_IGNORED, - VK_QUEUE_FAMILY_IGNORED, src_buffer, src_offset, size), - vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, UploadAccessBarriers, - VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, dst_buffer, - dst_offset, size)}, - {}, dld); + dst_offset, size](vk::CommandBuffer cmdbuf) { + cmdbuf.CopyBuffer(src_buffer, dst_buffer, VkBufferCopy{src_offset, dst_offset, size}); + + std::array<VkBufferMemoryBarrier, 2> barriers; + barriers[0].sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; + barriers[0].pNext = nullptr; + barriers[0].srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT; + barriers[0].dstAccessMask = VK_ACCESS_SHADER_WRITE_BIT; + barriers[0].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barriers[0].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barriers[0].buffer = src_buffer; + barriers[0].offset = src_offset; + barriers[0].size = size; + barriers[1].sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; + barriers[1].pNext = nullptr; + barriers[1].srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT; + barriers[1].dstAccessMask = UPLOAD_ACCESS_BARRIERS; + barriers[1].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barriers[1].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barriers[1].buffer = dst_buffer; + barriers[1].offset = dst_offset; + barriers[1].size = size; + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, UPLOAD_PIPELINE_STAGE, 0, {}, + barriers, {}); }); } diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h index 3f38eed0c..d3c23da98 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.h +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h @@ -11,11 +11,11 @@ #include "common/common_types.h" #include "video_core/buffer_cache/buffer_cache.h" #include "video_core/rasterizer_cache.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_memory_manager.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" #include "video_core/renderer_vulkan/vk_stream_buffer.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Core { class System; @@ -30,11 +30,11 @@ class VKScheduler; class CachedBufferBlock final : public VideoCommon::BufferBlock { public: explicit CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager, - CacheAddr cache_addr, std::size_t size); + VAddr cpu_addr, std::size_t size); ~CachedBufferBlock(); - const vk::Buffer* GetHandle() const { - return &*buffer.handle; + const VkBuffer* GetHandle() const { + return buffer.handle.address(); } private: @@ -43,21 +43,21 @@ private: using Buffer = std::shared_ptr<CachedBufferBlock>; -class VKBufferCache final : public VideoCommon::BufferCache<Buffer, vk::Buffer, VKStreamBuffer> { +class VKBufferCache final : public VideoCommon::BufferCache<Buffer, VkBuffer, VKStreamBuffer> { public: explicit VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system, const VKDevice& device, VKMemoryManager& memory_manager, VKScheduler& scheduler, VKStagingBufferPool& staging_pool); ~VKBufferCache(); - const vk::Buffer* GetEmptyBuffer(std::size_t size) override; + const VkBuffer* GetEmptyBuffer(std::size_t size) override; protected: void WriteBarrier() override {} - Buffer CreateBlock(CacheAddr cache_addr, std::size_t size) override; + Buffer CreateBlock(VAddr cpu_addr, std::size_t size) override; - const vk::Buffer* ToHandle(const Buffer& buffer) override; + const VkBuffer* ToHandle(const Buffer& buffer) override; void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size, const u8* data) override; diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.cpp b/src/video_core/renderer_vulkan/vk_compute_pass.cpp index 7bdda3d79..9d92305f4 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pass.cpp @@ -10,13 +10,13 @@ #include "common/alignment.h" #include "common/assert.h" #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_compute_pass.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -114,6 +114,35 @@ constexpr u8 quad_array[] = { 0xf9, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x4b, 0x00, 0x00, 0x00, 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00}; +VkDescriptorSetLayoutBinding BuildQuadArrayPassDescriptorSetLayoutBinding() { + VkDescriptorSetLayoutBinding binding; + binding.binding = 0; + binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; + binding.descriptorCount = 1; + binding.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; + binding.pImmutableSamplers = nullptr; + return binding; +} + +VkDescriptorUpdateTemplateEntryKHR BuildQuadArrayPassDescriptorUpdateTemplateEntry() { + VkDescriptorUpdateTemplateEntryKHR entry; + entry.dstBinding = 0; + entry.dstArrayElement = 0; + entry.descriptorCount = 1; + entry.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; + entry.offset = 0; + entry.stride = sizeof(DescriptorUpdateEntry); + return entry; +} + +VkPushConstantRange BuildQuadArrayPassPushConstantRange() { + VkPushConstantRange range; + range.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; + range.offset = 0; + range.size = sizeof(u32); + return range; +} + // Uint8 SPIR-V module. Generated from the "shaders/" directory. constexpr u8 uint8_pass[] = { 0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x2f, 0x00, 0x00, 0x00, @@ -191,53 +220,111 @@ constexpr u8 uint8_pass[] = { 0xf9, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00}; +std::array<VkDescriptorSetLayoutBinding, 2> BuildUint8PassDescriptorSetBindings() { + std::array<VkDescriptorSetLayoutBinding, 2> bindings; + bindings[0].binding = 0; + bindings[0].descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; + bindings[0].descriptorCount = 1; + bindings[0].stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; + bindings[0].pImmutableSamplers = nullptr; + bindings[1].binding = 1; + bindings[1].descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; + bindings[1].descriptorCount = 1; + bindings[1].stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; + bindings[1].pImmutableSamplers = nullptr; + return bindings; +} + +VkDescriptorUpdateTemplateEntryKHR BuildUint8PassDescriptorUpdateTemplateEntry() { + VkDescriptorUpdateTemplateEntryKHR entry; + entry.dstBinding = 0; + entry.dstArrayElement = 0; + entry.descriptorCount = 2; + entry.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; + entry.offset = 0; + entry.stride = sizeof(DescriptorUpdateEntry); + return entry; +} + } // Anonymous namespace VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool, - const std::vector<vk::DescriptorSetLayoutBinding>& bindings, - const std::vector<vk::DescriptorUpdateTemplateEntry>& templates, - const std::vector<vk::PushConstantRange> push_constants, - std::size_t code_size, const u8* code) { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - - const vk::DescriptorSetLayoutCreateInfo descriptor_layout_ci( - {}, static_cast<u32>(bindings.size()), bindings.data()); - descriptor_set_layout = dev.createDescriptorSetLayoutUnique(descriptor_layout_ci, nullptr, dld); - - const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &*descriptor_set_layout, - static_cast<u32>(push_constants.size()), - push_constants.data()); - layout = dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld); + vk::Span<VkDescriptorSetLayoutBinding> bindings, + vk::Span<VkDescriptorUpdateTemplateEntryKHR> templates, + vk::Span<VkPushConstantRange> push_constants, std::size_t code_size, + const u8* code) { + VkDescriptorSetLayoutCreateInfo descriptor_layout_ci; + descriptor_layout_ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; + descriptor_layout_ci.pNext = nullptr; + descriptor_layout_ci.flags = 0; + descriptor_layout_ci.bindingCount = bindings.size(); + descriptor_layout_ci.pBindings = bindings.data(); + descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout(descriptor_layout_ci); + + VkPipelineLayoutCreateInfo pipeline_layout_ci; + pipeline_layout_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; + pipeline_layout_ci.pNext = nullptr; + pipeline_layout_ci.flags = 0; + pipeline_layout_ci.setLayoutCount = 1; + pipeline_layout_ci.pSetLayouts = descriptor_set_layout.address(); + pipeline_layout_ci.pushConstantRangeCount = push_constants.size(); + pipeline_layout_ci.pPushConstantRanges = push_constants.data(); + layout = device.GetLogical().CreatePipelineLayout(pipeline_layout_ci); if (!templates.empty()) { - const vk::DescriptorUpdateTemplateCreateInfo template_ci( - {}, static_cast<u32>(templates.size()), templates.data(), - vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout, - vk::PipelineBindPoint::eGraphics, *layout, 0); - descriptor_template = dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld); + VkDescriptorUpdateTemplateCreateInfoKHR template_ci; + template_ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR; + template_ci.pNext = nullptr; + template_ci.flags = 0; + template_ci.descriptorUpdateEntryCount = templates.size(); + template_ci.pDescriptorUpdateEntries = templates.data(); + template_ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR; + template_ci.descriptorSetLayout = *descriptor_set_layout; + template_ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; + template_ci.pipelineLayout = *layout; + template_ci.set = 0; + descriptor_template = device.GetLogical().CreateDescriptorUpdateTemplateKHR(template_ci); descriptor_allocator.emplace(descriptor_pool, *descriptor_set_layout); } auto code_copy = std::make_unique<u32[]>(code_size / sizeof(u32) + 1); std::memcpy(code_copy.get(), code, code_size); - const vk::ShaderModuleCreateInfo module_ci({}, code_size, code_copy.get()); - module = dev.createShaderModuleUnique(module_ci, nullptr, dld); - const vk::PipelineShaderStageCreateInfo stage_ci({}, vk::ShaderStageFlagBits::eCompute, *module, - "main", nullptr); + VkShaderModuleCreateInfo module_ci; + module_ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; + module_ci.pNext = nullptr; + module_ci.flags = 0; + module_ci.codeSize = code_size; + module_ci.pCode = code_copy.get(); + module = device.GetLogical().CreateShaderModule(module_ci); + + VkComputePipelineCreateInfo pipeline_ci; + pipeline_ci.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO; + pipeline_ci.pNext = nullptr; + pipeline_ci.flags = 0; + pipeline_ci.layout = *layout; + pipeline_ci.basePipelineHandle = nullptr; + pipeline_ci.basePipelineIndex = 0; - const vk::ComputePipelineCreateInfo pipeline_ci({}, stage_ci, *layout, nullptr, 0); - pipeline = dev.createComputePipelineUnique(nullptr, pipeline_ci, nullptr, dld); + VkPipelineShaderStageCreateInfo& stage_ci = pipeline_ci.stage; + stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; + stage_ci.pNext = nullptr; + stage_ci.flags = 0; + stage_ci.stage = VK_SHADER_STAGE_COMPUTE_BIT; + stage_ci.module = *module; + stage_ci.pName = "main"; + stage_ci.pSpecializationInfo = nullptr; + + pipeline = device.GetLogical().CreateComputePipeline(pipeline_ci); } VKComputePass::~VKComputePass() = default; -vk::DescriptorSet VKComputePass::CommitDescriptorSet( - VKUpdateDescriptorQueue& update_descriptor_queue, VKFence& fence) { +VkDescriptorSet VKComputePass::CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue, + VKFence& fence) { if (!descriptor_template) { - return {}; + return nullptr; } const auto set = descriptor_allocator->Commit(fence); update_descriptor_queue.Send(*descriptor_template, set); @@ -248,25 +335,21 @@ QuadArrayPass::QuadArrayPass(const VKDevice& device, VKScheduler& scheduler, VKDescriptorPool& descriptor_pool, VKStagingBufferPool& staging_buffer_pool, VKUpdateDescriptorQueue& update_descriptor_queue) - : VKComputePass(device, descriptor_pool, - {vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eStorageBuffer, 1, - vk::ShaderStageFlagBits::eCompute, nullptr)}, - {vk::DescriptorUpdateTemplateEntry(0, 0, 1, vk::DescriptorType::eStorageBuffer, - 0, sizeof(DescriptorUpdateEntry))}, - {vk::PushConstantRange(vk::ShaderStageFlagBits::eCompute, 0, sizeof(u32))}, - std::size(quad_array), quad_array), + : VKComputePass(device, descriptor_pool, BuildQuadArrayPassDescriptorSetLayoutBinding(), + BuildQuadArrayPassDescriptorUpdateTemplateEntry(), + BuildQuadArrayPassPushConstantRange(), std::size(quad_array), quad_array), scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool}, update_descriptor_queue{update_descriptor_queue} {} QuadArrayPass::~QuadArrayPass() = default; -std::pair<const vk::Buffer&, vk::DeviceSize> QuadArrayPass::Assemble(u32 num_vertices, u32 first) { +std::pair<const VkBuffer*, VkDeviceSize> QuadArrayPass::Assemble(u32 num_vertices, u32 first) { const u32 num_triangle_vertices = num_vertices * 6 / 4; const std::size_t staging_size = num_triangle_vertices * sizeof(u32); auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false); update_descriptor_queue.Acquire(); - update_descriptor_queue.AddBuffer(&*buffer.handle, 0, staging_size); + update_descriptor_queue.AddBuffer(buffer.handle.address(), 0, staging_size); const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence()); scheduler.RequestOutsideRenderPassOperationContext(); @@ -274,66 +357,72 @@ std::pair<const vk::Buffer&, vk::DeviceSize> QuadArrayPass::Assemble(u32 num_ver ASSERT(num_vertices % 4 == 0); const u32 num_quads = num_vertices / 4; scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, num_quads, - first, set](auto cmdbuf, auto& dld) { + first, set](vk::CommandBuffer cmdbuf) { constexpr u32 dispatch_size = 1024; - cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline, dld); - cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, 0, {set}, {}, dld); - cmdbuf.pushConstants(layout, vk::ShaderStageFlagBits::eCompute, 0, sizeof(first), &first, - dld); - cmdbuf.dispatch(Common::AlignUp(num_quads, dispatch_size) / dispatch_size, 1, 1, dld); - - const vk::BufferMemoryBarrier barrier( - vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eVertexAttributeRead, - VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, 0, - static_cast<vk::DeviceSize>(num_quads) * 6 * sizeof(u32)); - cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader, - vk::PipelineStageFlagBits::eVertexInput, {}, {}, {barrier}, {}, dld); + cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline); + cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, 0, set, {}); + cmdbuf.PushConstants(layout, VK_SHADER_STAGE_COMPUTE_BIT, 0, sizeof(first), &first); + cmdbuf.Dispatch(Common::AlignUp(num_quads, dispatch_size) / dispatch_size, 1, 1); + + VkBufferMemoryBarrier barrier; + barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; + barrier.pNext = nullptr; + barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT; + barrier.dstAccessMask = VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT; + barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.buffer = buffer; + barrier.offset = 0; + barrier.size = static_cast<VkDeviceSize>(num_quads) * 6 * sizeof(u32); + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, + VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, {}, {barrier}, {}); }); - return {*buffer.handle, 0}; + return {buffer.handle.address(), 0}; } Uint8Pass::Uint8Pass(const VKDevice& device, VKScheduler& scheduler, VKDescriptorPool& descriptor_pool, VKStagingBufferPool& staging_buffer_pool, VKUpdateDescriptorQueue& update_descriptor_queue) - : VKComputePass(device, descriptor_pool, - {vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eStorageBuffer, 1, - vk::ShaderStageFlagBits::eCompute, nullptr), - vk::DescriptorSetLayoutBinding(1, vk::DescriptorType::eStorageBuffer, 1, - vk::ShaderStageFlagBits::eCompute, nullptr)}, - {vk::DescriptorUpdateTemplateEntry(0, 0, 2, vk::DescriptorType::eStorageBuffer, - 0, sizeof(DescriptorUpdateEntry))}, - {}, std::size(uint8_pass), uint8_pass), + : VKComputePass(device, descriptor_pool, BuildUint8PassDescriptorSetBindings(), + BuildUint8PassDescriptorUpdateTemplateEntry(), {}, std::size(uint8_pass), + uint8_pass), scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool}, update_descriptor_queue{update_descriptor_queue} {} Uint8Pass::~Uint8Pass() = default; -std::pair<const vk::Buffer*, u64> Uint8Pass::Assemble(u32 num_vertices, vk::Buffer src_buffer, - u64 src_offset) { +std::pair<const VkBuffer*, u64> Uint8Pass::Assemble(u32 num_vertices, VkBuffer src_buffer, + u64 src_offset) { const auto staging_size = static_cast<u32>(num_vertices * sizeof(u16)); auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false); update_descriptor_queue.Acquire(); update_descriptor_queue.AddBuffer(&src_buffer, src_offset, num_vertices); - update_descriptor_queue.AddBuffer(&*buffer.handle, 0, staging_size); + update_descriptor_queue.AddBuffer(buffer.handle.address(), 0, staging_size); const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence()); scheduler.RequestOutsideRenderPassOperationContext(); scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, set, - num_vertices](auto cmdbuf, auto& dld) { + num_vertices](vk::CommandBuffer cmdbuf) { constexpr u32 dispatch_size = 1024; - cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline, dld); - cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, 0, {set}, {}, dld); - cmdbuf.dispatch(Common::AlignUp(num_vertices, dispatch_size) / dispatch_size, 1, 1, dld); - - const vk::BufferMemoryBarrier barrier( - vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eVertexAttributeRead, - VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, 0, - static_cast<vk::DeviceSize>(num_vertices) * sizeof(u16)); - cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader, - vk::PipelineStageFlagBits::eVertexInput, {}, {}, {barrier}, {}, dld); + cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline); + cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, 0, set, {}); + cmdbuf.Dispatch(Common::AlignUp(num_vertices, dispatch_size) / dispatch_size, 1, 1); + + VkBufferMemoryBarrier barrier; + barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; + barrier.pNext = nullptr; + barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT; + barrier.dstAccessMask = VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT; + barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.buffer = buffer; + barrier.offset = 0; + barrier.size = static_cast<VkDeviceSize>(num_vertices * sizeof(u16)); + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, + VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, {}, barrier, {}); }); - return {&*buffer.handle, 0}; + return {buffer.handle.address(), 0}; } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.h b/src/video_core/renderer_vulkan/vk_compute_pass.h index 7057eb837..c62516bff 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.h +++ b/src/video_core/renderer_vulkan/vk_compute_pass.h @@ -8,8 +8,8 @@ #include <utility> #include <vector> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -22,24 +22,24 @@ class VKUpdateDescriptorQueue; class VKComputePass { public: explicit VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool, - const std::vector<vk::DescriptorSetLayoutBinding>& bindings, - const std::vector<vk::DescriptorUpdateTemplateEntry>& templates, - const std::vector<vk::PushConstantRange> push_constants, - std::size_t code_size, const u8* code); + vk::Span<VkDescriptorSetLayoutBinding> bindings, + vk::Span<VkDescriptorUpdateTemplateEntryKHR> templates, + vk::Span<VkPushConstantRange> push_constants, std::size_t code_size, + const u8* code); ~VKComputePass(); protected: - vk::DescriptorSet CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue, - VKFence& fence); + VkDescriptorSet CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue, + VKFence& fence); - UniqueDescriptorUpdateTemplate descriptor_template; - UniquePipelineLayout layout; - UniquePipeline pipeline; + vk::DescriptorUpdateTemplateKHR descriptor_template; + vk::PipelineLayout layout; + vk::Pipeline pipeline; private: - UniqueDescriptorSetLayout descriptor_set_layout; + vk::DescriptorSetLayout descriptor_set_layout; std::optional<DescriptorAllocator> descriptor_allocator; - UniqueShaderModule module; + vk::ShaderModule module; }; class QuadArrayPass final : public VKComputePass { @@ -50,7 +50,7 @@ public: VKUpdateDescriptorQueue& update_descriptor_queue); ~QuadArrayPass(); - std::pair<const vk::Buffer&, vk::DeviceSize> Assemble(u32 num_vertices, u32 first); + std::pair<const VkBuffer*, VkDeviceSize> Assemble(u32 num_vertices, u32 first); private: VKScheduler& scheduler; @@ -65,8 +65,7 @@ public: VKUpdateDescriptorQueue& update_descriptor_queue); ~Uint8Pass(); - std::pair<const vk::Buffer*, u64> Assemble(u32 num_vertices, vk::Buffer src_buffer, - u64 src_offset); + std::pair<const VkBuffer*, u64> Assemble(u32 num_vertices, VkBuffer src_buffer, u64 src_offset); private: VKScheduler& scheduler; diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp index 60f57d83e..23beafa4f 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp @@ -5,7 +5,6 @@ #include <memory> #include <vector> -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_compute_pipeline.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_device.h" @@ -14,6 +13,7 @@ #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_shader_decompiler.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -30,7 +30,7 @@ VKComputePipeline::VKComputePipeline(const VKDevice& device, VKScheduler& schedu VKComputePipeline::~VKComputePipeline() = default; -vk::DescriptorSet VKComputePipeline::CommitDescriptorSet() { +VkDescriptorSet VKComputePipeline::CommitDescriptorSet() { if (!descriptor_template) { return {}; } @@ -39,74 +39,109 @@ vk::DescriptorSet VKComputePipeline::CommitDescriptorSet() { return set; } -UniqueDescriptorSetLayout VKComputePipeline::CreateDescriptorSetLayout() const { - std::vector<vk::DescriptorSetLayoutBinding> bindings; +vk::DescriptorSetLayout VKComputePipeline::CreateDescriptorSetLayout() const { + std::vector<VkDescriptorSetLayoutBinding> bindings; u32 binding = 0; - const auto AddBindings = [&](vk::DescriptorType descriptor_type, std::size_t num_entries) { + const auto add_bindings = [&](VkDescriptorType descriptor_type, std::size_t num_entries) { // TODO(Rodrigo): Maybe make individual bindings here? for (u32 bindpoint = 0; bindpoint < static_cast<u32>(num_entries); ++bindpoint) { - bindings.emplace_back(binding++, descriptor_type, 1, vk::ShaderStageFlagBits::eCompute, - nullptr); + VkDescriptorSetLayoutBinding& entry = bindings.emplace_back(); + entry.binding = binding++; + entry.descriptorType = descriptor_type; + entry.descriptorCount = 1; + entry.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT; + entry.pImmutableSamplers = nullptr; } }; - AddBindings(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size()); - AddBindings(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size()); - AddBindings(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size()); - AddBindings(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size()); - AddBindings(vk::DescriptorType::eStorageImage, entries.images.size()); - - const vk::DescriptorSetLayoutCreateInfo descriptor_set_layout_ci( - {}, static_cast<u32>(bindings.size()), bindings.data()); - - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - return dev.createDescriptorSetLayoutUnique(descriptor_set_layout_ci, nullptr, dld); + add_bindings(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, entries.const_buffers.size()); + add_bindings(VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, entries.global_buffers.size()); + add_bindings(VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER, entries.texel_buffers.size()); + add_bindings(VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, entries.samplers.size()); + add_bindings(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, entries.images.size()); + + VkDescriptorSetLayoutCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.bindingCount = static_cast<u32>(bindings.size()); + ci.pBindings = bindings.data(); + return device.GetLogical().CreateDescriptorSetLayout(ci); } -UniquePipelineLayout VKComputePipeline::CreatePipelineLayout() const { - const vk::PipelineLayoutCreateInfo layout_ci({}, 1, &*descriptor_set_layout, 0, nullptr); - const auto dev = device.GetLogical(); - return dev.createPipelineLayoutUnique(layout_ci, nullptr, device.GetDispatchLoader()); +vk::PipelineLayout VKComputePipeline::CreatePipelineLayout() const { + VkPipelineLayoutCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.setLayoutCount = 1; + ci.pSetLayouts = descriptor_set_layout.address(); + ci.pushConstantRangeCount = 0; + ci.pPushConstantRanges = nullptr; + return device.GetLogical().CreatePipelineLayout(ci); } -UniqueDescriptorUpdateTemplate VKComputePipeline::CreateDescriptorUpdateTemplate() const { - std::vector<vk::DescriptorUpdateTemplateEntry> template_entries; +vk::DescriptorUpdateTemplateKHR VKComputePipeline::CreateDescriptorUpdateTemplate() const { + std::vector<VkDescriptorUpdateTemplateEntryKHR> template_entries; u32 binding = 0; u32 offset = 0; FillDescriptorUpdateTemplateEntries(entries, binding, offset, template_entries); if (template_entries.empty()) { // If the shader doesn't use descriptor sets, skip template creation. - return UniqueDescriptorUpdateTemplate{}; + return {}; } - const vk::DescriptorUpdateTemplateCreateInfo template_ci( - {}, static_cast<u32>(template_entries.size()), template_entries.data(), - vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout, - vk::PipelineBindPoint::eGraphics, *layout, DESCRIPTOR_SET); - - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - return dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld); + VkDescriptorUpdateTemplateCreateInfoKHR ci; + ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR; + ci.pNext = nullptr; + ci.flags = 0; + ci.descriptorUpdateEntryCount = static_cast<u32>(template_entries.size()); + ci.pDescriptorUpdateEntries = template_entries.data(); + ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR; + ci.descriptorSetLayout = *descriptor_set_layout; + ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; + ci.pipelineLayout = *layout; + ci.set = DESCRIPTOR_SET; + return device.GetLogical().CreateDescriptorUpdateTemplateKHR(ci); } -UniqueShaderModule VKComputePipeline::CreateShaderModule(const std::vector<u32>& code) const { - const vk::ShaderModuleCreateInfo module_ci({}, code.size() * sizeof(u32), code.data()); - const auto dev = device.GetLogical(); - return dev.createShaderModuleUnique(module_ci, nullptr, device.GetDispatchLoader()); +vk::ShaderModule VKComputePipeline::CreateShaderModule(const std::vector<u32>& code) const { + VkShaderModuleCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.codeSize = code.size() * sizeof(u32); + ci.pCode = code.data(); + return device.GetLogical().CreateShaderModule(ci); } -UniquePipeline VKComputePipeline::CreatePipeline() const { - vk::PipelineShaderStageCreateInfo shader_stage_ci({}, vk::ShaderStageFlagBits::eCompute, - *shader_module, "main", nullptr); - vk::PipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci; +vk::Pipeline VKComputePipeline::CreatePipeline() const { + VkComputePipelineCreateInfo ci; + VkPipelineShaderStageCreateInfo& stage_ci = ci.stage; + stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; + stage_ci.pNext = nullptr; + stage_ci.flags = 0; + stage_ci.stage = VK_SHADER_STAGE_COMPUTE_BIT; + stage_ci.module = *shader_module; + stage_ci.pName = "main"; + stage_ci.pSpecializationInfo = nullptr; + + VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci; + subgroup_size_ci.sType = + VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_REQUIRED_SUBGROUP_SIZE_CREATE_INFO_EXT; + subgroup_size_ci.pNext = nullptr; subgroup_size_ci.requiredSubgroupSize = GuestWarpSize; - if (entries.uses_warps && device.IsGuestWarpSizeSupported(vk::ShaderStageFlagBits::eCompute)) { - shader_stage_ci.pNext = &subgroup_size_ci; + + if (entries.uses_warps && device.IsGuestWarpSizeSupported(VK_SHADER_STAGE_COMPUTE_BIT)) { + stage_ci.pNext = &subgroup_size_ci; } - const vk::ComputePipelineCreateInfo create_info({}, shader_stage_ci, *layout, {}, 0); - const auto dev = device.GetLogical(); - return dev.createComputePipelineUnique({}, create_info, nullptr, device.GetDispatchLoader()); + ci.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.layout = *layout; + ci.basePipelineHandle = nullptr; + ci.basePipelineIndex = 0; + return device.GetLogical().CreateComputePipeline(ci); } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.h b/src/video_core/renderer_vulkan/vk_compute_pipeline.h index 22235c6c9..33b9af29e 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pipeline.h +++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.h @@ -7,9 +7,9 @@ #include <memory> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_shader_decompiler.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -25,42 +25,42 @@ public: const SPIRVShader& shader); ~VKComputePipeline(); - vk::DescriptorSet CommitDescriptorSet(); + VkDescriptorSet CommitDescriptorSet(); - vk::Pipeline GetHandle() const { + VkPipeline GetHandle() const { return *pipeline; } - vk::PipelineLayout GetLayout() const { + VkPipelineLayout GetLayout() const { return *layout; } - const ShaderEntries& GetEntries() { + const ShaderEntries& GetEntries() const { return entries; } private: - UniqueDescriptorSetLayout CreateDescriptorSetLayout() const; + vk::DescriptorSetLayout CreateDescriptorSetLayout() const; - UniquePipelineLayout CreatePipelineLayout() const; + vk::PipelineLayout CreatePipelineLayout() const; - UniqueDescriptorUpdateTemplate CreateDescriptorUpdateTemplate() const; + vk::DescriptorUpdateTemplateKHR CreateDescriptorUpdateTemplate() const; - UniqueShaderModule CreateShaderModule(const std::vector<u32>& code) const; + vk::ShaderModule CreateShaderModule(const std::vector<u32>& code) const; - UniquePipeline CreatePipeline() const; + vk::Pipeline CreatePipeline() const; const VKDevice& device; VKScheduler& scheduler; ShaderEntries entries; - UniqueDescriptorSetLayout descriptor_set_layout; + vk::DescriptorSetLayout descriptor_set_layout; DescriptorAllocator descriptor_allocator; VKUpdateDescriptorQueue& update_descriptor_queue; - UniquePipelineLayout layout; - UniqueDescriptorUpdateTemplate descriptor_template; - UniqueShaderModule shader_module; - UniquePipeline pipeline; + vk::PipelineLayout layout; + vk::DescriptorUpdateTemplateKHR descriptor_template; + vk::ShaderModule shader_module; + vk::Pipeline pipeline; }; } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp b/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp index cc7c281a0..e9d528aa6 100644 --- a/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp +++ b/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp @@ -6,10 +6,10 @@ #include <vector> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -17,19 +17,18 @@ namespace Vulkan { constexpr std::size_t SETS_GROW_RATE = 0x20; DescriptorAllocator::DescriptorAllocator(VKDescriptorPool& descriptor_pool, - vk::DescriptorSetLayout layout) + VkDescriptorSetLayout layout) : VKFencedPool{SETS_GROW_RATE}, descriptor_pool{descriptor_pool}, layout{layout} {} DescriptorAllocator::~DescriptorAllocator() = default; -vk::DescriptorSet DescriptorAllocator::Commit(VKFence& fence) { - return *descriptors[CommitResource(fence)]; +VkDescriptorSet DescriptorAllocator::Commit(VKFence& fence) { + const std::size_t index = CommitResource(fence); + return descriptors_allocations[index / SETS_GROW_RATE][index % SETS_GROW_RATE]; } void DescriptorAllocator::Allocate(std::size_t begin, std::size_t end) { - auto new_sets = descriptor_pool.AllocateDescriptors(layout, end - begin); - descriptors.insert(descriptors.end(), std::make_move_iterator(new_sets.begin()), - std::make_move_iterator(new_sets.end())); + descriptors_allocations.push_back(descriptor_pool.AllocateDescriptors(layout, end - begin)); } VKDescriptorPool::VKDescriptorPool(const VKDevice& device) @@ -37,53 +36,50 @@ VKDescriptorPool::VKDescriptorPool(const VKDevice& device) VKDescriptorPool::~VKDescriptorPool() = default; -vk::DescriptorPool VKDescriptorPool::AllocateNewPool() { +vk::DescriptorPool* VKDescriptorPool::AllocateNewPool() { static constexpr u32 num_sets = 0x20000; - static constexpr vk::DescriptorPoolSize pool_sizes[] = { - {vk::DescriptorType::eUniformBuffer, num_sets * 90}, - {vk::DescriptorType::eStorageBuffer, num_sets * 60}, - {vk::DescriptorType::eUniformTexelBuffer, num_sets * 64}, - {vk::DescriptorType::eCombinedImageSampler, num_sets * 64}, - {vk::DescriptorType::eStorageImage, num_sets * 40}}; - - const vk::DescriptorPoolCreateInfo create_info( - vk::DescriptorPoolCreateFlagBits::eFreeDescriptorSet, num_sets, - static_cast<u32>(std::size(pool_sizes)), std::data(pool_sizes)); - const auto dev = device.GetLogical(); - return *pools.emplace_back( - dev.createDescriptorPoolUnique(create_info, nullptr, device.GetDispatchLoader())); + static constexpr VkDescriptorPoolSize pool_sizes[] = { + {VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, num_sets * 90}, + {VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, num_sets * 60}, + {VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER, num_sets * 64}, + {VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, num_sets * 64}, + {VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, num_sets * 40}}; + + VkDescriptorPoolCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT; + ci.maxSets = num_sets; + ci.poolSizeCount = static_cast<u32>(std::size(pool_sizes)); + ci.pPoolSizes = std::data(pool_sizes); + return &pools.emplace_back(device.GetLogical().CreateDescriptorPool(ci)); } -std::vector<UniqueDescriptorSet> VKDescriptorPool::AllocateDescriptors( - vk::DescriptorSetLayout layout, std::size_t count) { - std::vector layout_copies(count, layout); - vk::DescriptorSetAllocateInfo allocate_info(active_pool, static_cast<u32>(count), - layout_copies.data()); - - std::vector<vk::DescriptorSet> sets(count); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - switch (const auto result = dev.allocateDescriptorSets(&allocate_info, sets.data(), dld)) { - case vk::Result::eSuccess: - break; - case vk::Result::eErrorOutOfPoolMemory: - active_pool = AllocateNewPool(); - allocate_info.descriptorPool = active_pool; - if (dev.allocateDescriptorSets(&allocate_info, sets.data(), dld) == vk::Result::eSuccess) { - break; - } - [[fallthrough]]; - default: - vk::throwResultException(result, "vk::Device::allocateDescriptorSetsUnique"); +vk::DescriptorSets VKDescriptorPool::AllocateDescriptors(VkDescriptorSetLayout layout, + std::size_t count) { + const std::vector layout_copies(count, layout); + VkDescriptorSetAllocateInfo ai; + ai.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO; + ai.pNext = nullptr; + ai.descriptorPool = **active_pool; + ai.descriptorSetCount = static_cast<u32>(count); + ai.pSetLayouts = layout_copies.data(); + + vk::DescriptorSets sets = active_pool->Allocate(ai); + if (!sets.IsOutOfPoolMemory()) { + return sets; } - vk::PoolFree deleter(dev, active_pool, dld); - std::vector<UniqueDescriptorSet> unique_sets; - unique_sets.reserve(count); - for (const auto set : sets) { - unique_sets.push_back(UniqueDescriptorSet{set, deleter}); + // Our current pool is out of memory. Allocate a new one and retry + active_pool = AllocateNewPool(); + ai.descriptorPool = **active_pool; + sets = active_pool->Allocate(ai); + if (!sets.IsOutOfPoolMemory()) { + return sets; } - return unique_sets; + + // After allocating a new pool, we are out of memory again. We can't handle this from here. + throw vk::Exception(VK_ERROR_OUT_OF_POOL_MEMORY); } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_descriptor_pool.h b/src/video_core/renderer_vulkan/vk_descriptor_pool.h index a441dbc0f..ab40c70f0 100644 --- a/src/video_core/renderer_vulkan/vk_descriptor_pool.h +++ b/src/video_core/renderer_vulkan/vk_descriptor_pool.h @@ -8,8 +8,8 @@ #include <vector> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -17,21 +17,21 @@ class VKDescriptorPool; class DescriptorAllocator final : public VKFencedPool { public: - explicit DescriptorAllocator(VKDescriptorPool& descriptor_pool, vk::DescriptorSetLayout layout); + explicit DescriptorAllocator(VKDescriptorPool& descriptor_pool, VkDescriptorSetLayout layout); ~DescriptorAllocator() override; DescriptorAllocator(const DescriptorAllocator&) = delete; - vk::DescriptorSet Commit(VKFence& fence); + VkDescriptorSet Commit(VKFence& fence); protected: void Allocate(std::size_t begin, std::size_t end) override; private: VKDescriptorPool& descriptor_pool; - const vk::DescriptorSetLayout layout; + const VkDescriptorSetLayout layout; - std::vector<UniqueDescriptorSet> descriptors; + std::vector<vk::DescriptorSets> descriptors_allocations; }; class VKDescriptorPool final { @@ -42,15 +42,14 @@ public: ~VKDescriptorPool(); private: - vk::DescriptorPool AllocateNewPool(); + vk::DescriptorPool* AllocateNewPool(); - std::vector<UniqueDescriptorSet> AllocateDescriptors(vk::DescriptorSetLayout layout, - std::size_t count); + vk::DescriptorSets AllocateDescriptors(VkDescriptorSetLayout layout, std::size_t count); const VKDevice& device; - std::vector<UniqueDescriptorPool> pools; - vk::DescriptorPool active_pool; + std::vector<vk::DescriptorPool> pools; + vk::DescriptorPool* active_pool; }; } // namespace Vulkan
\ No newline at end of file diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp index 28d2fbc4f..52d29e49d 100644 --- a/src/video_core/renderer_vulkan/vk_device.cpp +++ b/src/video_core/renderer_vulkan/vk_device.cpp @@ -6,14 +6,15 @@ #include <chrono> #include <cstdlib> #include <optional> -#include <set> #include <string_view> #include <thread> +#include <unordered_set> #include <vector> + #include "common/assert.h" #include "core/settings.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -21,49 +22,43 @@ namespace { namespace Alternatives { -constexpr std::array Depth24UnormS8Uint = {vk::Format::eD32SfloatS8Uint, - vk::Format::eD16UnormS8Uint, vk::Format{}}; -constexpr std::array Depth16UnormS8Uint = {vk::Format::eD24UnormS8Uint, - vk::Format::eD32SfloatS8Uint, vk::Format{}}; +constexpr std::array Depth24UnormS8_UINT = {VK_FORMAT_D32_SFLOAT_S8_UINT, + VK_FORMAT_D16_UNORM_S8_UINT, VkFormat{}}; +constexpr std::array Depth16UnormS8_UINT = {VK_FORMAT_D24_UNORM_S8_UINT, + VK_FORMAT_D32_SFLOAT_S8_UINT, VkFormat{}}; } // namespace Alternatives +constexpr std::array REQUIRED_EXTENSIONS = { + VK_KHR_SWAPCHAIN_EXTENSION_NAME, + VK_KHR_16BIT_STORAGE_EXTENSION_NAME, + VK_KHR_8BIT_STORAGE_EXTENSION_NAME, + VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME, + VK_KHR_DESCRIPTOR_UPDATE_TEMPLATE_EXTENSION_NAME, + VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME, + VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME, + VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME, + VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME, +}; + template <typename T> void SetNext(void**& next, T& data) { *next = &data; next = &data.pNext; } -template <typename T> -T GetFeatures(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dldi) { - vk::PhysicalDeviceFeatures2 features; - T extension_features; - features.pNext = &extension_features; - physical.getFeatures2(&features, dldi); - return extension_features; -} - -template <typename T> -T GetProperties(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dldi) { - vk::PhysicalDeviceProperties2 properties; - T extension_properties; - properties.pNext = &extension_properties; - physical.getProperties2(&properties, dldi); - return extension_properties; -} - -constexpr const vk::Format* GetFormatAlternatives(vk::Format format) { +constexpr const VkFormat* GetFormatAlternatives(VkFormat format) { switch (format) { - case vk::Format::eD24UnormS8Uint: - return Alternatives::Depth24UnormS8Uint.data(); - case vk::Format::eD16UnormS8Uint: - return Alternatives::Depth16UnormS8Uint.data(); + case VK_FORMAT_D24_UNORM_S8_UINT: + return Alternatives::Depth24UnormS8_UINT.data(); + case VK_FORMAT_D16_UNORM_S8_UINT: + return Alternatives::Depth16UnormS8_UINT.data(); default: return nullptr; } } -vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties, FormatType format_type) { +VkFormatFeatureFlags GetFormatFeatures(VkFormatProperties properties, FormatType format_type) { switch (format_type) { case FormatType::Linear: return properties.linearTilingFeatures; @@ -76,79 +71,220 @@ vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties, Format } } +std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties( + vk::PhysicalDevice physical, const vk::InstanceDispatch& dld) { + static constexpr std::array formats{VK_FORMAT_A8B8G8R8_UNORM_PACK32, + VK_FORMAT_A8B8G8R8_UINT_PACK32, + VK_FORMAT_A8B8G8R8_SNORM_PACK32, + VK_FORMAT_A8B8G8R8_SRGB_PACK32, + VK_FORMAT_B5G6R5_UNORM_PACK16, + VK_FORMAT_A2B10G10R10_UNORM_PACK32, + VK_FORMAT_A1R5G5B5_UNORM_PACK16, + VK_FORMAT_R32G32B32A32_SFLOAT, + VK_FORMAT_R32G32B32A32_UINT, + VK_FORMAT_R32G32_SFLOAT, + VK_FORMAT_R32G32_UINT, + VK_FORMAT_R16G16B16A16_UINT, + VK_FORMAT_R16G16B16A16_SNORM, + VK_FORMAT_R16G16B16A16_UNORM, + VK_FORMAT_R16G16_UNORM, + VK_FORMAT_R16G16_SNORM, + VK_FORMAT_R16G16_SFLOAT, + VK_FORMAT_R16_UNORM, + VK_FORMAT_R8G8B8A8_SRGB, + VK_FORMAT_R8G8_UNORM, + VK_FORMAT_R8G8_SNORM, + VK_FORMAT_R8_UNORM, + VK_FORMAT_R8_UINT, + VK_FORMAT_B10G11R11_UFLOAT_PACK32, + VK_FORMAT_R32_SFLOAT, + VK_FORMAT_R32_UINT, + VK_FORMAT_R32_SINT, + VK_FORMAT_R16_SFLOAT, + VK_FORMAT_R16G16B16A16_SFLOAT, + VK_FORMAT_B8G8R8A8_UNORM, + VK_FORMAT_R4G4B4A4_UNORM_PACK16, + VK_FORMAT_D32_SFLOAT, + VK_FORMAT_D16_UNORM, + VK_FORMAT_D16_UNORM_S8_UINT, + VK_FORMAT_D24_UNORM_S8_UINT, + VK_FORMAT_D32_SFLOAT_S8_UINT, + VK_FORMAT_BC1_RGBA_UNORM_BLOCK, + VK_FORMAT_BC2_UNORM_BLOCK, + VK_FORMAT_BC3_UNORM_BLOCK, + VK_FORMAT_BC4_UNORM_BLOCK, + VK_FORMAT_BC5_UNORM_BLOCK, + VK_FORMAT_BC5_SNORM_BLOCK, + VK_FORMAT_BC7_UNORM_BLOCK, + VK_FORMAT_BC6H_UFLOAT_BLOCK, + VK_FORMAT_BC6H_SFLOAT_BLOCK, + VK_FORMAT_BC1_RGBA_SRGB_BLOCK, + VK_FORMAT_BC2_SRGB_BLOCK, + VK_FORMAT_BC3_SRGB_BLOCK, + VK_FORMAT_BC7_SRGB_BLOCK, + VK_FORMAT_ASTC_4x4_SRGB_BLOCK, + VK_FORMAT_ASTC_8x8_SRGB_BLOCK, + VK_FORMAT_ASTC_8x5_SRGB_BLOCK, + VK_FORMAT_ASTC_5x4_SRGB_BLOCK, + VK_FORMAT_ASTC_5x5_UNORM_BLOCK, + VK_FORMAT_ASTC_5x5_SRGB_BLOCK, + VK_FORMAT_ASTC_10x8_UNORM_BLOCK, + VK_FORMAT_ASTC_10x8_SRGB_BLOCK, + VK_FORMAT_ASTC_6x6_UNORM_BLOCK, + VK_FORMAT_ASTC_6x6_SRGB_BLOCK, + VK_FORMAT_ASTC_10x10_UNORM_BLOCK, + VK_FORMAT_ASTC_10x10_SRGB_BLOCK, + VK_FORMAT_ASTC_12x12_UNORM_BLOCK, + VK_FORMAT_ASTC_12x12_SRGB_BLOCK, + VK_FORMAT_ASTC_8x6_UNORM_BLOCK, + VK_FORMAT_ASTC_8x6_SRGB_BLOCK, + VK_FORMAT_ASTC_6x5_UNORM_BLOCK, + VK_FORMAT_ASTC_6x5_SRGB_BLOCK, + VK_FORMAT_E5B9G9R9_UFLOAT_PACK32}; + std::unordered_map<VkFormat, VkFormatProperties> format_properties; + for (const auto format : formats) { + format_properties.emplace(format, physical.GetFormatProperties(format)); + } + return format_properties; +} + } // Anonymous namespace -VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical, - vk::SurfaceKHR surface) - : physical{physical}, properties{physical.getProperties(dldi)}, - format_properties{GetFormatProperties(dldi, physical)} { - SetupFamilies(dldi, surface); - SetupFeatures(dldi); +VKDevice::VKDevice(VkInstance instance, vk::PhysicalDevice physical, VkSurfaceKHR surface, + const vk::InstanceDispatch& dld) + : dld{dld}, physical{physical}, properties{physical.GetProperties()}, + format_properties{GetFormatProperties(physical, dld)} { + SetupFamilies(surface); + SetupFeatures(); } VKDevice::~VKDevice() = default; -bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance) { +bool VKDevice::Create() { const auto queue_cis = GetDeviceQueueCreateInfos(); - const std::vector extensions = LoadExtensions(dldi); + const std::vector extensions = LoadExtensions(); - vk::PhysicalDeviceFeatures2 features2; + VkPhysicalDeviceFeatures2 features2; + features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; + features2.pNext = nullptr; void** next = &features2.pNext; + auto& features = features2.features; - features.vertexPipelineStoresAndAtomics = true; + features.robustBufferAccess = false; + features.fullDrawIndexUint32 = false; + features.imageCubeArray = false; features.independentBlend = true; + features.geometryShader = true; + features.tessellationShader = true; + features.sampleRateShading = false; + features.dualSrcBlend = false; + features.logicOp = false; + features.multiDrawIndirect = false; + features.drawIndirectFirstInstance = false; features.depthClamp = true; - features.samplerAnisotropy = true; + features.depthBiasClamp = true; + features.fillModeNonSolid = false; + features.depthBounds = false; + features.wideLines = false; features.largePoints = true; + features.alphaToOne = false; features.multiViewport = true; - features.depthBiasClamp = true; - features.geometryShader = true; - features.tessellationShader = true; + features.samplerAnisotropy = true; + features.textureCompressionETC2 = false; + features.textureCompressionASTC_LDR = is_optimal_astc_supported; + features.textureCompressionBC = false; features.occlusionQueryPrecise = true; + features.pipelineStatisticsQuery = false; + features.vertexPipelineStoresAndAtomics = true; features.fragmentStoresAndAtomics = true; + features.shaderTessellationAndGeometryPointSize = false; features.shaderImageGatherExtended = true; + features.shaderStorageImageExtendedFormats = false; + features.shaderStorageImageMultisample = false; features.shaderStorageImageReadWithoutFormat = is_formatless_image_load_supported; features.shaderStorageImageWriteWithoutFormat = true; - features.textureCompressionASTC_LDR = is_optimal_astc_supported; - - vk::PhysicalDevice16BitStorageFeaturesKHR bit16_storage; + features.shaderUniformBufferArrayDynamicIndexing = false; + features.shaderSampledImageArrayDynamicIndexing = false; + features.shaderStorageBufferArrayDynamicIndexing = false; + features.shaderStorageImageArrayDynamicIndexing = false; + features.shaderClipDistance = false; + features.shaderCullDistance = false; + features.shaderFloat64 = false; + features.shaderInt64 = false; + features.shaderInt16 = false; + features.shaderResourceResidency = false; + features.shaderResourceMinLod = false; + features.sparseBinding = false; + features.sparseResidencyBuffer = false; + features.sparseResidencyImage2D = false; + features.sparseResidencyImage3D = false; + features.sparseResidency2Samples = false; + features.sparseResidency4Samples = false; + features.sparseResidency8Samples = false; + features.sparseResidency16Samples = false; + features.sparseResidencyAliased = false; + features.variableMultisampleRate = false; + features.inheritedQueries = false; + + VkPhysicalDevice16BitStorageFeaturesKHR bit16_storage; + bit16_storage.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES_KHR; + bit16_storage.pNext = nullptr; + bit16_storage.storageBuffer16BitAccess = false; bit16_storage.uniformAndStorageBuffer16BitAccess = true; + bit16_storage.storagePushConstant16 = false; + bit16_storage.storageInputOutput16 = false; SetNext(next, bit16_storage); - vk::PhysicalDevice8BitStorageFeaturesKHR bit8_storage; + VkPhysicalDevice8BitStorageFeaturesKHR bit8_storage; + bit8_storage.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_8BIT_STORAGE_FEATURES_KHR; + bit8_storage.pNext = nullptr; + bit8_storage.storageBuffer8BitAccess = false; bit8_storage.uniformAndStorageBuffer8BitAccess = true; + bit8_storage.storagePushConstant8 = false; SetNext(next, bit8_storage); - vk::PhysicalDeviceHostQueryResetFeaturesEXT host_query_reset; + VkPhysicalDeviceHostQueryResetFeaturesEXT host_query_reset; + host_query_reset.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_HOST_QUERY_RESET_FEATURES_EXT; host_query_reset.hostQueryReset = true; SetNext(next, host_query_reset); - vk::PhysicalDeviceFloat16Int8FeaturesKHR float16_int8; + VkPhysicalDeviceFloat16Int8FeaturesKHR float16_int8; if (is_float16_supported) { + float16_int8.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT16_INT8_FEATURES_KHR; + float16_int8.pNext = nullptr; float16_int8.shaderFloat16 = true; + float16_int8.shaderInt8 = false; SetNext(next, float16_int8); } else { LOG_INFO(Render_Vulkan, "Device doesn't support float16 natively"); } - vk::PhysicalDeviceUniformBufferStandardLayoutFeaturesKHR std430_layout; + VkPhysicalDeviceUniformBufferStandardLayoutFeaturesKHR std430_layout; if (khr_uniform_buffer_standard_layout) { + std430_layout.sType = + VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_UNIFORM_BUFFER_STANDARD_LAYOUT_FEATURES_KHR; + std430_layout.pNext = nullptr; std430_layout.uniformBufferStandardLayout = true; SetNext(next, std430_layout); } else { LOG_INFO(Render_Vulkan, "Device doesn't support packed UBOs"); } - vk::PhysicalDeviceIndexTypeUint8FeaturesEXT index_type_uint8; + VkPhysicalDeviceIndexTypeUint8FeaturesEXT index_type_uint8; if (ext_index_type_uint8) { + index_type_uint8.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_INDEX_TYPE_UINT8_FEATURES_EXT; + index_type_uint8.pNext = nullptr; index_type_uint8.indexTypeUint8 = true; SetNext(next, index_type_uint8); } else { LOG_INFO(Render_Vulkan, "Device doesn't support uint8 indexes"); } - vk::PhysicalDeviceTransformFeedbackFeaturesEXT transform_feedback; + VkPhysicalDeviceTransformFeedbackFeaturesEXT transform_feedback; if (ext_transform_feedback) { + transform_feedback.sType = + VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT; + transform_feedback.pNext = nullptr; transform_feedback.transformFeedback = true; transform_feedback.geometryStreams = true; SetNext(next, transform_feedback); @@ -160,62 +296,48 @@ bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instan LOG_INFO(Render_Vulkan, "Device doesn't support depth range unrestricted"); } - vk::DeviceCreateInfo device_ci({}, static_cast<u32>(queue_cis.size()), queue_cis.data(), 0, - nullptr, static_cast<u32>(extensions.size()), extensions.data(), - nullptr); - device_ci.pNext = &features2; - - vk::Device dummy_logical; - if (physical.createDevice(&device_ci, nullptr, &dummy_logical, dldi) != vk::Result::eSuccess) { - LOG_CRITICAL(Render_Vulkan, "Logical device failed to be created!"); + logical = vk::Device::Create(physical, queue_cis, extensions, features2, dld); + if (!logical) { + LOG_ERROR(Render_Vulkan, "Failed to create logical device"); return false; } - dld.init(instance, dldi.vkGetInstanceProcAddr, dummy_logical, dldi.vkGetDeviceProcAddr); - logical = UniqueDevice( - dummy_logical, vk::ObjectDestroy<vk::NoParent, vk::DispatchLoaderDynamic>(nullptr, dld)); - CollectTelemetryParameters(); - graphics_queue = logical->getQueue(graphics_family, 0, dld); - present_queue = logical->getQueue(present_family, 0, dld); + graphics_queue = logical.GetQueue(graphics_family); + present_queue = logical.GetQueue(present_family); return true; } -vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format, - vk::FormatFeatureFlags wanted_usage, - FormatType format_type) const { +VkFormat VKDevice::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage, + FormatType format_type) const { if (IsFormatSupported(wanted_format, wanted_usage, format_type)) { return wanted_format; } // The wanted format is not supported by hardware, search for alternatives - const vk::Format* alternatives = GetFormatAlternatives(wanted_format); + const VkFormat* alternatives = GetFormatAlternatives(wanted_format); if (alternatives == nullptr) { UNREACHABLE_MSG("Format={} with usage={} and type={} has no defined alternatives and host " "hardware does not support it", - vk::to_string(wanted_format), vk::to_string(wanted_usage), - static_cast<u32>(format_type)); + wanted_format, wanted_usage, format_type); return wanted_format; } std::size_t i = 0; - for (vk::Format alternative = alternatives[0]; alternative != vk::Format{}; - alternative = alternatives[++i]) { + for (VkFormat alternative = *alternatives; alternative; alternative = alternatives[++i]) { if (!IsFormatSupported(alternative, wanted_usage, format_type)) { continue; } LOG_WARNING(Render_Vulkan, "Emulating format={} with alternative format={} with usage={} and type={}", - static_cast<u32>(wanted_format), static_cast<u32>(alternative), - static_cast<u32>(wanted_usage), static_cast<u32>(format_type)); + wanted_format, alternative, wanted_usage, format_type); return alternative; } // No alternatives found, panic UNREACHABLE_MSG("Format={} with usage={} and type={} is not supported by the host hardware and " "doesn't support any of the alternatives", - static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage), - static_cast<u32>(format_type)); + wanted_format, wanted_usage, format_type); return wanted_format; } @@ -229,35 +351,39 @@ void VKDevice::ReportLoss() const { return; } - [[maybe_unused]] const std::vector data = graphics_queue.getCheckpointDataNV(dld); + [[maybe_unused]] const std::vector data = graphics_queue.GetCheckpointDataNV(dld); // Catch here in debug builds (or with optimizations disabled) the last graphics pipeline to be // executed. It can be done on a debugger by evaluating the expression: // *(VKGraphicsPipeline*)data[0] } -bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features, - const vk::DispatchLoaderDynamic& dldi) const { +bool VKDevice::IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const { // Disable for now to avoid converting ASTC twice. - return false; static constexpr std::array astc_formats = { - vk::Format::eAstc4x4SrgbBlock, vk::Format::eAstc8x8SrgbBlock, - vk::Format::eAstc8x5SrgbBlock, vk::Format::eAstc5x4SrgbBlock, - vk::Format::eAstc5x5UnormBlock, vk::Format::eAstc5x5SrgbBlock, - vk::Format::eAstc10x8UnormBlock, vk::Format::eAstc10x8SrgbBlock, - vk::Format::eAstc6x6UnormBlock, vk::Format::eAstc6x6SrgbBlock, - vk::Format::eAstc10x10UnormBlock, vk::Format::eAstc10x10SrgbBlock, - vk::Format::eAstc12x12UnormBlock, vk::Format::eAstc12x12SrgbBlock, - vk::Format::eAstc8x6UnormBlock, vk::Format::eAstc8x6SrgbBlock, - vk::Format::eAstc6x5UnormBlock, vk::Format::eAstc6x5SrgbBlock}; + VK_FORMAT_ASTC_4x4_UNORM_BLOCK, VK_FORMAT_ASTC_4x4_SRGB_BLOCK, + VK_FORMAT_ASTC_5x4_UNORM_BLOCK, VK_FORMAT_ASTC_5x4_SRGB_BLOCK, + VK_FORMAT_ASTC_5x5_UNORM_BLOCK, VK_FORMAT_ASTC_5x5_SRGB_BLOCK, + VK_FORMAT_ASTC_6x5_UNORM_BLOCK, VK_FORMAT_ASTC_6x5_SRGB_BLOCK, + VK_FORMAT_ASTC_6x6_UNORM_BLOCK, VK_FORMAT_ASTC_6x6_SRGB_BLOCK, + VK_FORMAT_ASTC_8x5_UNORM_BLOCK, VK_FORMAT_ASTC_8x5_SRGB_BLOCK, + VK_FORMAT_ASTC_8x6_UNORM_BLOCK, VK_FORMAT_ASTC_8x6_SRGB_BLOCK, + VK_FORMAT_ASTC_8x8_UNORM_BLOCK, VK_FORMAT_ASTC_8x8_SRGB_BLOCK, + VK_FORMAT_ASTC_10x5_UNORM_BLOCK, VK_FORMAT_ASTC_10x5_SRGB_BLOCK, + VK_FORMAT_ASTC_10x6_UNORM_BLOCK, VK_FORMAT_ASTC_10x6_SRGB_BLOCK, + VK_FORMAT_ASTC_10x8_UNORM_BLOCK, VK_FORMAT_ASTC_10x8_SRGB_BLOCK, + VK_FORMAT_ASTC_10x10_UNORM_BLOCK, VK_FORMAT_ASTC_10x10_SRGB_BLOCK, + 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) { return false; } const auto format_feature_usage{ - vk::FormatFeatureFlagBits::eSampledImage | vk::FormatFeatureFlagBits::eBlitSrc | - vk::FormatFeatureFlagBits::eBlitDst | vk::FormatFeatureFlagBits::eTransferSrc | - vk::FormatFeatureFlagBits::eTransferDst}; + VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_BLIT_SRC_BIT | + VK_FORMAT_FEATURE_BLIT_DST_BIT | VK_FORMAT_FEATURE_TRANSFER_SRC_BIT | + VK_FORMAT_FEATURE_TRANSFER_DST_BIT}; for (const auto format : astc_formats) { - const auto format_properties{physical.getFormatProperties(format, dldi)}; + const auto format_properties{physical.GetFormatProperties(format)}; if (!(format_properties.optimalTilingFeatures & format_feature_usage)) { return false; } @@ -265,62 +391,49 @@ bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features return true; } -bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage, +bool VKDevice::IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage, FormatType format_type) const { const auto it = format_properties.find(wanted_format); if (it == format_properties.end()) { - UNIMPLEMENTED_MSG("Unimplemented format query={}", vk::to_string(wanted_format)); + UNIMPLEMENTED_MSG("Unimplemented format query={}", wanted_format); return true; } const auto supported_usage = GetFormatFeatures(it->second, format_type); return (supported_usage & wanted_usage) == wanted_usage; } -bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical, - vk::SurfaceKHR surface) { +bool VKDevice::IsSuitable(vk::PhysicalDevice physical, VkSurfaceKHR surface) { bool is_suitable = true; + std::bitset<REQUIRED_EXTENSIONS.size()> available_extensions; - constexpr std::array required_extensions = { - VK_KHR_SWAPCHAIN_EXTENSION_NAME, - VK_KHR_16BIT_STORAGE_EXTENSION_NAME, - VK_KHR_8BIT_STORAGE_EXTENSION_NAME, - VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME, - VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME, - VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME, - VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME, - VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME, - }; - std::bitset<required_extensions.size()> available_extensions{}; - - for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) { - for (std::size_t i = 0; i < required_extensions.size(); ++i) { + for (const auto& prop : physical.EnumerateDeviceExtensionProperties()) { + for (std::size_t i = 0; i < REQUIRED_EXTENSIONS.size(); ++i) { if (available_extensions[i]) { continue; } - available_extensions[i] = - required_extensions[i] == std::string_view{prop.extensionName}; + const std::string_view name{prop.extensionName}; + available_extensions[i] = name == REQUIRED_EXTENSIONS[i]; } } if (!available_extensions.all()) { - for (std::size_t i = 0; i < required_extensions.size(); ++i) { + for (std::size_t i = 0; i < REQUIRED_EXTENSIONS.size(); ++i) { if (available_extensions[i]) { continue; } - LOG_ERROR(Render_Vulkan, "Missing required extension: {}", required_extensions[i]); + LOG_ERROR(Render_Vulkan, "Missing required extension: {}", REQUIRED_EXTENSIONS[i]); is_suitable = false; } } bool has_graphics{}, has_present{}; - const auto queue_family_properties = physical.getQueueFamilyProperties(dldi); + const std::vector queue_family_properties = physical.GetQueueFamilyProperties(); for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) { const auto& family = queue_family_properties[i]; if (family.queueCount == 0) { continue; } - has_graphics |= - (family.queueFlags & vk::QueueFlagBits::eGraphics) != static_cast<vk::QueueFlagBits>(0); - has_present |= physical.getSurfaceSupportKHR(i, surface, dldi) != 0; + has_graphics |= family.queueFlags & VK_QUEUE_GRAPHICS_BIT; + has_present |= physical.GetSurfaceSupportKHR(i, surface); } if (!has_graphics || !has_present) { LOG_ERROR(Render_Vulkan, "Device lacks a graphics and present queue"); @@ -328,7 +441,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev } // TODO(Rodrigo): Check if the device matches all requeriments. - const auto properties{physical.getProperties(dldi)}; + const auto properties{physical.GetProperties()}; const auto& limits{properties.limits}; constexpr u32 required_ubo_size = 65536; @@ -345,7 +458,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev is_suitable = false; } - const auto features{physical.getFeatures(dldi)}; + const auto features{physical.GetFeatures()}; const std::array feature_report = { std::make_pair(features.vertexPipelineStoresAndAtomics, "vertexPipelineStoresAndAtomics"), std::make_pair(features.independentBlend, "independentBlend"), @@ -377,9 +490,9 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev return is_suitable; } -std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynamic& dldi) { +std::vector<const char*> VKDevice::LoadExtensions() { std::vector<const char*> extensions; - const auto Test = [&](const vk::ExtensionProperties& extension, + const auto Test = [&](const VkExtensionProperties& extension, std::optional<std::reference_wrapper<bool>> status, const char* name, bool push) { if (extension.extensionName != std::string_view(name)) { @@ -393,22 +506,13 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami } }; - extensions.reserve(15); - extensions.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME); - extensions.push_back(VK_KHR_16BIT_STORAGE_EXTENSION_NAME); - extensions.push_back(VK_KHR_8BIT_STORAGE_EXTENSION_NAME); - extensions.push_back(VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME); - extensions.push_back(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME); - extensions.push_back(VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME); - extensions.push_back(VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME); - extensions.push_back(VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME); - - [[maybe_unused]] const bool nsight = - std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED"); + extensions.reserve(7 + REQUIRED_EXTENSIONS.size()); + extensions.insert(extensions.begin(), REQUIRED_EXTENSIONS.begin(), REQUIRED_EXTENSIONS.end()); + bool has_khr_shader_float16_int8{}; bool has_ext_subgroup_size_control{}; bool has_ext_transform_feedback{}; - for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) { + for (const auto& extension : physical.EnumerateDeviceExtensionProperties()) { Test(extension, khr_uniform_buffer_standard_layout, VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true); Test(extension, has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME, @@ -428,38 +532,67 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami } } + VkPhysicalDeviceFeatures2KHR features; + features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2_KHR; + + VkPhysicalDeviceProperties2KHR properties; + properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2_KHR; + if (has_khr_shader_float16_int8) { - is_float16_supported = - GetFeatures<vk::PhysicalDeviceFloat16Int8FeaturesKHR>(physical, dldi).shaderFloat16; + VkPhysicalDeviceFloat16Int8FeaturesKHR float16_int8_features; + float16_int8_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT16_INT8_FEATURES_KHR; + float16_int8_features.pNext = nullptr; + features.pNext = &float16_int8_features; + + physical.GetFeatures2KHR(features); + is_float16_supported = float16_int8_features.shaderFloat16; extensions.push_back(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME); } if (has_ext_subgroup_size_control) { - const auto features = - GetFeatures<vk::PhysicalDeviceSubgroupSizeControlFeaturesEXT>(physical, dldi); - const auto properties = - GetProperties<vk::PhysicalDeviceSubgroupSizeControlPropertiesEXT>(physical, dldi); - - is_warp_potentially_bigger = properties.maxSubgroupSize > GuestWarpSize; - - if (features.subgroupSizeControl && properties.minSubgroupSize <= GuestWarpSize && - properties.maxSubgroupSize >= GuestWarpSize) { + VkPhysicalDeviceSubgroupSizeControlFeaturesEXT subgroup_features; + subgroup_features.sType = + VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_FEATURES_EXT; + subgroup_features.pNext = nullptr; + features.pNext = &subgroup_features; + physical.GetFeatures2KHR(features); + + VkPhysicalDeviceSubgroupSizeControlPropertiesEXT subgroup_properties; + subgroup_properties.sType = + VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_PROPERTIES_EXT; + subgroup_properties.pNext = nullptr; + properties.pNext = &subgroup_properties; + physical.GetProperties2KHR(properties); + + is_warp_potentially_bigger = subgroup_properties.maxSubgroupSize > GuestWarpSize; + + 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 = properties.requiredSubgroupSizeStages; + guest_warp_stages = subgroup_properties.requiredSubgroupSizeStages; } } else { is_warp_potentially_bigger = true; } if (has_ext_transform_feedback) { - const auto features = - GetFeatures<vk::PhysicalDeviceTransformFeedbackFeaturesEXT>(physical, dldi); - const auto properties = - GetProperties<vk::PhysicalDeviceTransformFeedbackPropertiesEXT>(physical, dldi); - - if (features.transformFeedback && features.geometryStreams && - properties.maxTransformFeedbackStreams >= 4 && properties.maxTransformFeedbackBuffers && - properties.transformFeedbackQueries && properties.transformFeedbackDraw) { + VkPhysicalDeviceTransformFeedbackFeaturesEXT tfb_features; + tfb_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT; + tfb_features.pNext = nullptr; + features.pNext = &tfb_features; + physical.GetFeatures2KHR(features); + + VkPhysicalDeviceTransformFeedbackPropertiesEXT tfb_properties; + tfb_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_PROPERTIES_EXT; + tfb_properties.pNext = nullptr; + properties.pNext = &tfb_properties; + physical.GetProperties2KHR(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; } @@ -468,10 +601,10 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami return extensions; } -void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface) { +void VKDevice::SetupFamilies(VkSurfaceKHR surface) { std::optional<u32> graphics_family_, present_family_; - const auto queue_family_properties = physical.getQueueFamilyProperties(dldi); + const std::vector queue_family_properties = physical.GetQueueFamilyProperties(); for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) { if (graphics_family_ && present_family_) break; @@ -480,10 +613,12 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK if (queue_family.queueCount == 0) continue; - if (queue_family.queueFlags & vk::QueueFlagBits::eGraphics) + if (queue_family.queueFlags & VK_QUEUE_GRAPHICS_BIT) { graphics_family_ = i; - if (physical.getSurfaceSupportKHR(i, surface, dldi)) + } + if (physical.GetSurfaceSupportKHR(i, surface)) { present_family_ = i; + } } ASSERT(graphics_family_ && present_family_); @@ -491,111 +626,49 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK present_family = *present_family_; } -void VKDevice::SetupFeatures(const vk::DispatchLoaderDynamic& dldi) { - const auto supported_features{physical.getFeatures(dldi)}; +void VKDevice::SetupFeatures() { + const auto supported_features{physical.GetFeatures()}; is_formatless_image_load_supported = supported_features.shaderStorageImageReadWithoutFormat; - is_optimal_astc_supported = IsOptimalAstcSupported(supported_features, dldi); + is_optimal_astc_supported = IsOptimalAstcSupported(supported_features); } void VKDevice::CollectTelemetryParameters() { - const auto driver = GetProperties<vk::PhysicalDeviceDriverPropertiesKHR>(physical, dld); + VkPhysicalDeviceDriverPropertiesKHR driver; + driver.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES_KHR; + driver.pNext = nullptr; + + VkPhysicalDeviceProperties2KHR properties; + properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2_KHR; + properties.pNext = &driver; + physical.GetProperties2KHR(properties); + driver_id = driver.driverID; vendor_name = driver.driverName; - const auto extensions = physical.enumerateDeviceExtensionProperties(nullptr, dld); + const std::vector extensions = physical.EnumerateDeviceExtensionProperties(); reported_extensions.reserve(std::size(extensions)); for (const auto& extension : extensions) { reported_extensions.push_back(extension.extensionName); } } -std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const { - static const float QUEUE_PRIORITY = 1.0f; +std::vector<VkDeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const { + static constexpr float QUEUE_PRIORITY = 1.0f; - std::set<u32> unique_queue_families = {graphics_family, present_family}; - std::vector<vk::DeviceQueueCreateInfo> queue_cis; + std::unordered_set<u32> unique_queue_families = {graphics_family, present_family}; + std::vector<VkDeviceQueueCreateInfo> queue_cis; - for (u32 queue_family : unique_queue_families) - queue_cis.push_back({{}, queue_family, 1, &QUEUE_PRIORITY}); + for (const u32 queue_family : unique_queue_families) { + VkDeviceQueueCreateInfo& ci = queue_cis.emplace_back(); + ci.sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.queueFamilyIndex = queue_family; + ci.queueCount = 1; + ci.pQueuePriorities = &QUEUE_PRIORITY; + } return queue_cis; } -std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties( - const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical) { - static constexpr std::array formats{vk::Format::eA8B8G8R8UnormPack32, - vk::Format::eA8B8G8R8UintPack32, - vk::Format::eA8B8G8R8SnormPack32, - vk::Format::eA8B8G8R8SrgbPack32, - vk::Format::eB5G6R5UnormPack16, - vk::Format::eA2B10G10R10UnormPack32, - vk::Format::eA1R5G5B5UnormPack16, - vk::Format::eR32G32B32A32Sfloat, - vk::Format::eR32G32B32A32Uint, - vk::Format::eR32G32Sfloat, - vk::Format::eR32G32Uint, - vk::Format::eR16G16B16A16Uint, - vk::Format::eR16G16B16A16Snorm, - vk::Format::eR16G16B16A16Unorm, - vk::Format::eR16G16Unorm, - vk::Format::eR16G16Snorm, - vk::Format::eR16G16Sfloat, - vk::Format::eR16Unorm, - vk::Format::eR8G8B8A8Srgb, - vk::Format::eR8G8Unorm, - vk::Format::eR8G8Snorm, - vk::Format::eR8Unorm, - vk::Format::eR8Uint, - vk::Format::eB10G11R11UfloatPack32, - vk::Format::eR32Sfloat, - vk::Format::eR32Uint, - vk::Format::eR32Sint, - vk::Format::eR16Sfloat, - vk::Format::eR16G16B16A16Sfloat, - vk::Format::eB8G8R8A8Unorm, - vk::Format::eR4G4B4A4UnormPack16, - vk::Format::eD32Sfloat, - vk::Format::eD16Unorm, - vk::Format::eD16UnormS8Uint, - vk::Format::eD24UnormS8Uint, - vk::Format::eD32SfloatS8Uint, - vk::Format::eBc1RgbaUnormBlock, - vk::Format::eBc2UnormBlock, - vk::Format::eBc3UnormBlock, - vk::Format::eBc4UnormBlock, - vk::Format::eBc5UnormBlock, - vk::Format::eBc5SnormBlock, - vk::Format::eBc7UnormBlock, - vk::Format::eBc6HUfloatBlock, - vk::Format::eBc6HSfloatBlock, - vk::Format::eBc1RgbaSrgbBlock, - vk::Format::eBc2SrgbBlock, - vk::Format::eBc3SrgbBlock, - vk::Format::eBc7SrgbBlock, - vk::Format::eAstc4x4SrgbBlock, - vk::Format::eAstc8x8SrgbBlock, - vk::Format::eAstc8x5SrgbBlock, - vk::Format::eAstc5x4SrgbBlock, - vk::Format::eAstc5x5UnormBlock, - vk::Format::eAstc5x5SrgbBlock, - vk::Format::eAstc10x8UnormBlock, - vk::Format::eAstc10x8SrgbBlock, - vk::Format::eAstc6x6UnormBlock, - vk::Format::eAstc6x6SrgbBlock, - vk::Format::eAstc10x10UnormBlock, - vk::Format::eAstc10x10SrgbBlock, - vk::Format::eAstc12x12UnormBlock, - vk::Format::eAstc12x12SrgbBlock, - vk::Format::eAstc8x6UnormBlock, - vk::Format::eAstc8x6SrgbBlock, - vk::Format::eAstc6x5UnormBlock, - vk::Format::eAstc6x5SrgbBlock, - vk::Format::eE5B9G9R9UfloatPack32}; - std::unordered_map<vk::Format, vk::FormatProperties> format_properties; - for (const auto format : formats) { - format_properties.emplace(format, physical.getFormatProperties(format, dldi)); - } - return format_properties; -} - } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h index 6e656517f..60d64572a 100644 --- a/src/video_core/renderer_vulkan/vk_device.h +++ b/src/video_core/renderer_vulkan/vk_device.h @@ -8,8 +8,9 @@ #include <string_view> #include <unordered_map> #include <vector> + #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -22,12 +23,12 @@ const u32 GuestWarpSize = 32; /// Handles data specific to a physical device. class VKDevice final { public: - explicit VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical, - vk::SurfaceKHR surface); + explicit VKDevice(VkInstance instance, vk::PhysicalDevice physical, VkSurfaceKHR surface, + const vk::InstanceDispatch& dld); ~VKDevice(); /// Initializes the device. Returns true on success. - bool Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance); + bool Create(); /** * Returns a format supported by the device for the passed requeriments. @@ -36,20 +37,20 @@ public: * @param format_type Format type usage. * @returns A format supported by the device. */ - vk::Format GetSupportedFormat(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage, - FormatType format_type) const; + VkFormat GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage, + FormatType format_type) const; /// Reports a device loss. void ReportLoss() const; /// Returns the dispatch loader with direct function pointers of the device. - const vk::DispatchLoaderDynamic& GetDispatchLoader() const { + const vk::DeviceDispatch& GetDispatchLoader() const { return dld; } /// Returns the logical device. - vk::Device GetLogical() const { - return logical.get(); + const vk::Device& GetLogical() const { + return logical; } /// Returns the physical device. @@ -79,7 +80,7 @@ public: /// Returns true if the device is integrated with the host CPU. bool IsIntegrated() const { - return properties.deviceType == vk::PhysicalDeviceType::eIntegratedGpu; + return properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU; } /// Returns the current Vulkan API version provided in Vulkan-formatted version numbers. @@ -98,27 +99,27 @@ public: } /// Returns the driver ID. - vk::DriverIdKHR GetDriverID() const { + VkDriverIdKHR GetDriverID() const { return driver_id; } /// Returns uniform buffer alignment requeriment. - vk::DeviceSize GetUniformBufferAlignment() const { + VkDeviceSize GetUniformBufferAlignment() const { return properties.limits.minUniformBufferOffsetAlignment; } /// Returns storage alignment requeriment. - vk::DeviceSize GetStorageBufferAlignment() const { + VkDeviceSize GetStorageBufferAlignment() const { return properties.limits.minStorageBufferOffsetAlignment; } /// Returns the maximum range for storage buffers. - vk::DeviceSize GetMaxStorageBufferRange() const { + VkDeviceSize GetMaxStorageBufferRange() const { return properties.limits.maxStorageBufferRange; } /// Returns the maximum size for push constants. - vk::DeviceSize GetMaxPushConstantsSize() const { + VkDeviceSize GetMaxPushConstantsSize() const { return properties.limits.maxPushConstantsSize; } @@ -138,8 +139,8 @@ public: } /// Returns true if the device can be forced to use the guest warp size. - bool IsGuestWarpSizeSupported(vk::ShaderStageFlagBits stage) const { - return (guest_warp_stages & stage) != vk::ShaderStageFlags{}; + bool IsGuestWarpSizeSupported(VkShaderStageFlagBits stage) const { + return guest_warp_stages & stage; } /// Returns true if formatless image load is supported. @@ -188,50 +189,44 @@ public: } /// Checks if the physical device is suitable. - static bool IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical, - vk::SurfaceKHR surface); + static bool IsSuitable(vk::PhysicalDevice physical, VkSurfaceKHR surface); private: /// Loads extensions into a vector and stores available ones in this object. - std::vector<const char*> LoadExtensions(const vk::DispatchLoaderDynamic& dldi); + std::vector<const char*> LoadExtensions(); /// Sets up queue families. - void SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface); + void SetupFamilies(VkSurfaceKHR surface); /// Sets up device features. - void SetupFeatures(const vk::DispatchLoaderDynamic& dldi); + void SetupFeatures(); /// Collects telemetry information from the device. void CollectTelemetryParameters(); /// Returns a list of queue initialization descriptors. - std::vector<vk::DeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const; + std::vector<VkDeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const; /// Returns true if ASTC textures are natively supported. - bool IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features, - const vk::DispatchLoaderDynamic& dldi) const; + bool IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const; /// Returns true if a format is supported. - bool IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage, + bool IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage, FormatType format_type) const; - /// Returns the device properties for Vulkan formats. - static std::unordered_map<vk::Format, vk::FormatProperties> GetFormatProperties( - const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical); - - const vk::PhysicalDevice physical; ///< Physical device. - vk::DispatchLoaderDynamic dld; ///< Device function pointers. - vk::PhysicalDeviceProperties properties; ///< Device properties. - UniqueDevice logical; ///< Logical device. - vk::Queue graphics_queue; ///< Main graphics queue. - vk::Queue present_queue; ///< Main present queue. - u32 graphics_family{}; ///< Main graphics queue family index. - u32 present_family{}; ///< Main present queue family index. - vk::DriverIdKHR driver_id{}; ///< Driver ID. - vk::ShaderStageFlags guest_warp_stages{}; ///< Stages where the guest warp size can be forced.ed - bool is_optimal_astc_supported{}; ///< Support for native ASTC. - bool is_float16_supported{}; ///< Support for float16 arithmetics. - bool is_warp_potentially_bigger{}; ///< Host warp size can be bigger than guest. + vk::DeviceDispatch dld; ///< Device function pointers. + vk::PhysicalDevice physical; ///< Physical device. + VkPhysicalDeviceProperties properties; ///< Device properties. + vk::Device logical; ///< Logical device. + vk::Queue graphics_queue; ///< Main graphics queue. + vk::Queue present_queue; ///< Main present queue. + 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.ed + bool is_optimal_astc_supported{}; ///< Support for native ASTC. + bool is_float16_supported{}; ///< Support for float16 arithmetics. + bool is_warp_potentially_bigger{}; ///< Host warp size can be bigger than guest. bool is_formatless_image_load_supported{}; ///< Support for shader image read without format. bool khr_uniform_buffer_standard_layout{}; ///< Support for std430 on UBOs. bool ext_index_type_uint8{}; ///< Support for VK_EXT_index_type_uint8. @@ -245,7 +240,7 @@ private: std::vector<std::string> reported_extensions; ///< Reported Vulkan extensions. /// Format properties dictionary. - std::unordered_map<vk::Format, vk::FormatProperties> format_properties; + std::unordered_map<VkFormat, VkFormatProperties> format_properties; }; } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp index 6a02403c1..b540b838d 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp @@ -2,11 +2,13 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <array> +#include <cstring> #include <vector> + #include "common/assert.h" #include "common/common_types.h" #include "common/microprofile.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" @@ -16,6 +18,7 @@ #include "video_core/renderer_vulkan/vk_renderpass_cache.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -23,21 +26,26 @@ MICROPROFILE_DECLARE(Vulkan_PipelineCache); namespace { -vk::StencilOpState GetStencilFaceState(const FixedPipelineState::StencilFace& face) { - return vk::StencilOpState(MaxwellToVK::StencilOp(face.action_stencil_fail), - MaxwellToVK::StencilOp(face.action_depth_pass), - MaxwellToVK::StencilOp(face.action_depth_fail), - MaxwellToVK::ComparisonOp(face.test_func), 0, 0, 0); +VkStencilOpState GetStencilFaceState(const FixedPipelineState::StencilFace& face) { + VkStencilOpState state; + state.failOp = MaxwellToVK::StencilOp(face.action_stencil_fail); + state.passOp = MaxwellToVK::StencilOp(face.action_depth_pass); + state.depthFailOp = MaxwellToVK::StencilOp(face.action_depth_fail); + state.compareOp = MaxwellToVK::ComparisonOp(face.test_func); + state.compareMask = 0; + state.writeMask = 0; + state.reference = 0; + return state; } -bool SupportsPrimitiveRestart(vk::PrimitiveTopology topology) { +bool SupportsPrimitiveRestart(VkPrimitiveTopology topology) { static constexpr std::array unsupported_topologies = { - vk::PrimitiveTopology::ePointList, - vk::PrimitiveTopology::eLineList, - vk::PrimitiveTopology::eTriangleList, - vk::PrimitiveTopology::eLineListWithAdjacency, - vk::PrimitiveTopology::eTriangleListWithAdjacency, - vk::PrimitiveTopology::ePatchList}; + VK_PRIMITIVE_TOPOLOGY_POINT_LIST, + VK_PRIMITIVE_TOPOLOGY_LINE_LIST, + VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST, + VK_PRIMITIVE_TOPOLOGY_LINE_LIST_WITH_ADJACENCY, + VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_WITH_ADJACENCY, + VK_PRIMITIVE_TOPOLOGY_PATCH_LIST}; return std::find(std::begin(unsupported_topologies), std::end(unsupported_topologies), topology) == std::end(unsupported_topologies); } @@ -49,7 +57,7 @@ VKGraphicsPipeline::VKGraphicsPipeline(const VKDevice& device, VKScheduler& sche VKUpdateDescriptorQueue& update_descriptor_queue, VKRenderPassCache& renderpass_cache, const GraphicsPipelineCacheKey& key, - const std::vector<vk::DescriptorSetLayoutBinding>& bindings, + vk::Span<VkDescriptorSetLayoutBinding> bindings, const SPIRVProgram& program) : device{device}, scheduler{scheduler}, fixed_state{key.fixed_state}, hash{key.Hash()}, descriptor_set_layout{CreateDescriptorSetLayout(bindings)}, @@ -63,7 +71,7 @@ VKGraphicsPipeline::VKGraphicsPipeline(const VKDevice& device, VKScheduler& sche VKGraphicsPipeline::~VKGraphicsPipeline() = default; -vk::DescriptorSet VKGraphicsPipeline::CommitDescriptorSet() { +VkDescriptorSet VKGraphicsPipeline::CommitDescriptorSet() { if (!descriptor_template) { return {}; } @@ -72,27 +80,32 @@ vk::DescriptorSet VKGraphicsPipeline::CommitDescriptorSet() { return set; } -UniqueDescriptorSetLayout VKGraphicsPipeline::CreateDescriptorSetLayout( - const std::vector<vk::DescriptorSetLayoutBinding>& bindings) const { - const vk::DescriptorSetLayoutCreateInfo descriptor_set_layout_ci( - {}, static_cast<u32>(bindings.size()), bindings.data()); - - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - return dev.createDescriptorSetLayoutUnique(descriptor_set_layout_ci, nullptr, dld); +vk::DescriptorSetLayout VKGraphicsPipeline::CreateDescriptorSetLayout( + vk::Span<VkDescriptorSetLayoutBinding> bindings) const { + VkDescriptorSetLayoutCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.bindingCount = bindings.size(); + ci.pBindings = bindings.data(); + return device.GetLogical().CreateDescriptorSetLayout(ci); } -UniquePipelineLayout VKGraphicsPipeline::CreatePipelineLayout() const { - const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &*descriptor_set_layout, 0, - nullptr); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - return dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld); +vk::PipelineLayout VKGraphicsPipeline::CreatePipelineLayout() const { + VkPipelineLayoutCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.setLayoutCount = 1; + ci.pSetLayouts = descriptor_set_layout.address(); + ci.pushConstantRangeCount = 0; + ci.pPushConstantRanges = nullptr; + return device.GetLogical().CreatePipelineLayout(ci); } -UniqueDescriptorUpdateTemplate VKGraphicsPipeline::CreateDescriptorUpdateTemplate( +vk::DescriptorUpdateTemplateKHR VKGraphicsPipeline::CreateDescriptorUpdateTemplate( const SPIRVProgram& program) const { - std::vector<vk::DescriptorUpdateTemplateEntry> template_entries; + std::vector<VkDescriptorUpdateTemplateEntry> template_entries; u32 binding = 0; u32 offset = 0; for (const auto& stage : program) { @@ -102,38 +115,47 @@ UniqueDescriptorUpdateTemplate VKGraphicsPipeline::CreateDescriptorUpdateTemplat } if (template_entries.empty()) { // If the shader doesn't use descriptor sets, skip template creation. - return UniqueDescriptorUpdateTemplate{}; + return {}; } - const vk::DescriptorUpdateTemplateCreateInfo template_ci( - {}, static_cast<u32>(template_entries.size()), template_entries.data(), - vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout, - vk::PipelineBindPoint::eGraphics, *layout, DESCRIPTOR_SET); - - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - return dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld); + VkDescriptorUpdateTemplateCreateInfoKHR ci; + ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR; + ci.pNext = nullptr; + ci.flags = 0; + ci.descriptorUpdateEntryCount = static_cast<u32>(template_entries.size()); + ci.pDescriptorUpdateEntries = template_entries.data(); + ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR; + ci.descriptorSetLayout = *descriptor_set_layout; + ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; + ci.pipelineLayout = *layout; + ci.set = DESCRIPTOR_SET; + return device.GetLogical().CreateDescriptorUpdateTemplateKHR(ci); } -std::vector<UniqueShaderModule> VKGraphicsPipeline::CreateShaderModules( +std::vector<vk::ShaderModule> VKGraphicsPipeline::CreateShaderModules( const SPIRVProgram& program) const { - std::vector<UniqueShaderModule> modules; - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); + VkShaderModuleCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + + std::vector<vk::ShaderModule> modules; + modules.reserve(Maxwell::MaxShaderStage); for (std::size_t i = 0; i < Maxwell::MaxShaderStage; ++i) { const auto& stage = program[i]; if (!stage) { continue; } - const vk::ShaderModuleCreateInfo module_ci({}, stage->code.size() * sizeof(u32), - stage->code.data()); - modules.emplace_back(dev.createShaderModuleUnique(module_ci, nullptr, dld)); + + ci.codeSize = stage->code.size() * sizeof(u32); + ci.pCode = stage->code.data(); + modules.push_back(device.GetLogical().CreateShaderModule(ci)); } return modules; } -UniquePipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpass_params, - const SPIRVProgram& program) const { +vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpass_params, + const SPIRVProgram& program) const { const auto& vi = fixed_state.vertex_input; const auto& ia = fixed_state.input_assembly; const auto& ds = fixed_state.depth_stencil; @@ -141,19 +163,26 @@ UniquePipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& render const auto& ts = fixed_state.tessellation; const auto& rs = fixed_state.rasterizer; - std::vector<vk::VertexInputBindingDescription> vertex_bindings; - std::vector<vk::VertexInputBindingDivisorDescriptionEXT> vertex_binding_divisors; + std::vector<VkVertexInputBindingDescription> vertex_bindings; + std::vector<VkVertexInputBindingDivisorDescriptionEXT> vertex_binding_divisors; for (std::size_t i = 0; i < vi.num_bindings; ++i) { const auto& binding = vi.bindings[i]; const bool instanced = binding.divisor != 0; - const auto rate = instanced ? vk::VertexInputRate::eInstance : vk::VertexInputRate::eVertex; - vertex_bindings.emplace_back(binding.index, binding.stride, rate); + const auto rate = instanced ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX; + + auto& vertex_binding = vertex_bindings.emplace_back(); + vertex_binding.binding = binding.index; + vertex_binding.stride = binding.stride; + vertex_binding.inputRate = rate; + if (instanced) { - vertex_binding_divisors.emplace_back(binding.index, binding.divisor); + auto& binding_divisor = vertex_binding_divisors.emplace_back(); + binding_divisor.binding = binding.index; + binding_divisor.divisor = binding.divisor; } } - std::vector<vk::VertexInputAttributeDescription> vertex_attributes; + std::vector<VkVertexInputAttributeDescription> vertex_attributes; const auto& input_attributes = program[0]->entries.attributes; for (std::size_t i = 0; i < vi.num_attributes; ++i) { const auto& attribute = vi.attributes[i]; @@ -161,109 +190,194 @@ UniquePipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& render // Skip attributes not used by the vertex shaders. continue; } - vertex_attributes.emplace_back(attribute.index, attribute.buffer, - MaxwellToVK::VertexFormat(attribute.type, attribute.size), - attribute.offset); + auto& vertex_attribute = vertex_attributes.emplace_back(); + vertex_attribute.location = attribute.index; + vertex_attribute.binding = attribute.buffer; + vertex_attribute.format = MaxwellToVK::VertexFormat(attribute.type, attribute.size); + vertex_attribute.offset = attribute.offset; } - vk::PipelineVertexInputStateCreateInfo vertex_input_ci( - {}, static_cast<u32>(vertex_bindings.size()), vertex_bindings.data(), - static_cast<u32>(vertex_attributes.size()), vertex_attributes.data()); - - const vk::PipelineVertexInputDivisorStateCreateInfoEXT vertex_input_divisor_ci( - static_cast<u32>(vertex_binding_divisors.size()), vertex_binding_divisors.data()); + VkPipelineVertexInputStateCreateInfo vertex_input_ci; + vertex_input_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO; + vertex_input_ci.pNext = nullptr; + vertex_input_ci.flags = 0; + vertex_input_ci.vertexBindingDescriptionCount = static_cast<u32>(vertex_bindings.size()); + vertex_input_ci.pVertexBindingDescriptions = vertex_bindings.data(); + vertex_input_ci.vertexAttributeDescriptionCount = static_cast<u32>(vertex_attributes.size()); + vertex_input_ci.pVertexAttributeDescriptions = vertex_attributes.data(); + + VkPipelineVertexInputDivisorStateCreateInfoEXT input_divisor_ci; + input_divisor_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_DIVISOR_STATE_CREATE_INFO_EXT; + input_divisor_ci.pNext = nullptr; + input_divisor_ci.vertexBindingDivisorCount = static_cast<u32>(vertex_binding_divisors.size()); + input_divisor_ci.pVertexBindingDivisors = vertex_binding_divisors.data(); if (!vertex_binding_divisors.empty()) { - vertex_input_ci.pNext = &vertex_input_divisor_ci; + vertex_input_ci.pNext = &input_divisor_ci; } - const auto primitive_topology = MaxwellToVK::PrimitiveTopology(device, ia.topology); - const vk::PipelineInputAssemblyStateCreateInfo input_assembly_ci( - {}, primitive_topology, - ia.primitive_restart_enable && SupportsPrimitiveRestart(primitive_topology)); - - const vk::PipelineTessellationStateCreateInfo tessellation_ci({}, ts.patch_control_points); - - const vk::PipelineViewportStateCreateInfo viewport_ci({}, Maxwell::NumViewports, nullptr, - Maxwell::NumViewports, nullptr); - - // TODO(Rodrigo): Find out what's the default register value for front face - const vk::PipelineRasterizationStateCreateInfo rasterizer_ci( - {}, rs.depth_clamp_enable, false, vk::PolygonMode::eFill, - rs.cull_enable ? MaxwellToVK::CullFace(rs.cull_face) : vk::CullModeFlagBits::eNone, - MaxwellToVK::FrontFace(rs.front_face), rs.depth_bias_enable, 0.0f, 0.0f, 0.0f, 1.0f); - - const vk::PipelineMultisampleStateCreateInfo multisampling_ci( - {}, vk::SampleCountFlagBits::e1, false, 0.0f, nullptr, false, false); - - const vk::CompareOp depth_test_compare = ds.depth_test_enable - ? MaxwellToVK::ComparisonOp(ds.depth_test_function) - : vk::CompareOp::eAlways; - - const vk::PipelineDepthStencilStateCreateInfo depth_stencil_ci( - {}, ds.depth_test_enable, ds.depth_write_enable, depth_test_compare, ds.depth_bounds_enable, - ds.stencil_enable, GetStencilFaceState(ds.front_stencil), - GetStencilFaceState(ds.back_stencil), 0.0f, 0.0f); - - std::array<vk::PipelineColorBlendAttachmentState, Maxwell::NumRenderTargets> cb_attachments; + VkPipelineInputAssemblyStateCreateInfo input_assembly_ci; + input_assembly_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO; + input_assembly_ci.pNext = nullptr; + input_assembly_ci.flags = 0; + input_assembly_ci.topology = MaxwellToVK::PrimitiveTopology(device, ia.topology); + input_assembly_ci.primitiveRestartEnable = + ia.primitive_restart_enable && SupportsPrimitiveRestart(input_assembly_ci.topology); + + VkPipelineTessellationStateCreateInfo tessellation_ci; + tessellation_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO; + tessellation_ci.pNext = nullptr; + tessellation_ci.flags = 0; + tessellation_ci.patchControlPoints = ts.patch_control_points; + + VkPipelineViewportStateCreateInfo viewport_ci; + viewport_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO; + viewport_ci.pNext = nullptr; + viewport_ci.flags = 0; + viewport_ci.viewportCount = Maxwell::NumViewports; + viewport_ci.pViewports = nullptr; + viewport_ci.scissorCount = Maxwell::NumViewports; + viewport_ci.pScissors = nullptr; + + VkPipelineRasterizationStateCreateInfo rasterization_ci; + rasterization_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO; + rasterization_ci.pNext = nullptr; + rasterization_ci.flags = 0; + rasterization_ci.depthClampEnable = rs.depth_clamp_enable; + rasterization_ci.rasterizerDiscardEnable = VK_FALSE; + rasterization_ci.polygonMode = VK_POLYGON_MODE_FILL; + rasterization_ci.cullMode = + rs.cull_enable ? MaxwellToVK::CullFace(rs.cull_face) : VK_CULL_MODE_NONE; + rasterization_ci.frontFace = MaxwellToVK::FrontFace(rs.front_face); + rasterization_ci.depthBiasEnable = rs.depth_bias_enable; + rasterization_ci.depthBiasConstantFactor = 0.0f; + rasterization_ci.depthBiasClamp = 0.0f; + rasterization_ci.depthBiasSlopeFactor = 0.0f; + rasterization_ci.lineWidth = 1.0f; + + VkPipelineMultisampleStateCreateInfo multisample_ci; + multisample_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO; + multisample_ci.pNext = nullptr; + multisample_ci.flags = 0; + multisample_ci.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT; + multisample_ci.sampleShadingEnable = VK_FALSE; + multisample_ci.minSampleShading = 0.0f; + multisample_ci.pSampleMask = nullptr; + multisample_ci.alphaToCoverageEnable = VK_FALSE; + multisample_ci.alphaToOneEnable = VK_FALSE; + + VkPipelineDepthStencilStateCreateInfo depth_stencil_ci; + depth_stencil_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO; + depth_stencil_ci.pNext = nullptr; + depth_stencil_ci.flags = 0; + depth_stencil_ci.depthTestEnable = ds.depth_test_enable; + depth_stencil_ci.depthWriteEnable = ds.depth_write_enable; + depth_stencil_ci.depthCompareOp = ds.depth_test_enable + ? MaxwellToVK::ComparisonOp(ds.depth_test_function) + : VK_COMPARE_OP_ALWAYS; + depth_stencil_ci.depthBoundsTestEnable = ds.depth_bounds_enable; + depth_stencil_ci.stencilTestEnable = ds.stencil_enable; + depth_stencil_ci.front = GetStencilFaceState(ds.front_stencil); + depth_stencil_ci.back = GetStencilFaceState(ds.back_stencil); + depth_stencil_ci.minDepthBounds = 0.0f; + depth_stencil_ci.maxDepthBounds = 0.0f; + + std::array<VkPipelineColorBlendAttachmentState, Maxwell::NumRenderTargets> cb_attachments; const std::size_t num_attachments = std::min(cd.attachments_count, renderpass_params.color_attachments.size()); for (std::size_t i = 0; i < num_attachments; ++i) { - constexpr std::array component_table{ - vk::ColorComponentFlagBits::eR, vk::ColorComponentFlagBits::eG, - vk::ColorComponentFlagBits::eB, vk::ColorComponentFlagBits::eA}; + static constexpr std::array component_table = { + VK_COLOR_COMPONENT_R_BIT, VK_COLOR_COMPONENT_G_BIT, VK_COLOR_COMPONENT_B_BIT, + VK_COLOR_COMPONENT_A_BIT}; const auto& blend = cd.attachments[i]; - vk::ColorComponentFlags color_components{}; + VkColorComponentFlags color_components = 0; for (std::size_t j = 0; j < component_table.size(); ++j) { - if (blend.components[j]) + if (blend.components[j]) { color_components |= component_table[j]; + } } - cb_attachments[i] = vk::PipelineColorBlendAttachmentState( - blend.enable, MaxwellToVK::BlendFactor(blend.src_rgb_func), - MaxwellToVK::BlendFactor(blend.dst_rgb_func), - MaxwellToVK::BlendEquation(blend.rgb_equation), - MaxwellToVK::BlendFactor(blend.src_a_func), MaxwellToVK::BlendFactor(blend.dst_a_func), - MaxwellToVK::BlendEquation(blend.a_equation), color_components); + VkPipelineColorBlendAttachmentState& attachment = cb_attachments[i]; + attachment.blendEnable = blend.enable; + attachment.srcColorBlendFactor = MaxwellToVK::BlendFactor(blend.src_rgb_func); + attachment.dstColorBlendFactor = MaxwellToVK::BlendFactor(blend.dst_rgb_func); + attachment.colorBlendOp = MaxwellToVK::BlendEquation(blend.rgb_equation); + attachment.srcAlphaBlendFactor = MaxwellToVK::BlendFactor(blend.src_a_func); + attachment.dstAlphaBlendFactor = MaxwellToVK::BlendFactor(blend.dst_a_func); + attachment.alphaBlendOp = MaxwellToVK::BlendEquation(blend.a_equation); + attachment.colorWriteMask = color_components; } - const vk::PipelineColorBlendStateCreateInfo color_blending_ci({}, false, vk::LogicOp::eCopy, - static_cast<u32>(num_attachments), - cb_attachments.data(), {}); - - constexpr std::array dynamic_states = { - vk::DynamicState::eViewport, vk::DynamicState::eScissor, - vk::DynamicState::eDepthBias, vk::DynamicState::eBlendConstants, - vk::DynamicState::eDepthBounds, vk::DynamicState::eStencilCompareMask, - vk::DynamicState::eStencilWriteMask, vk::DynamicState::eStencilReference}; - const vk::PipelineDynamicStateCreateInfo dynamic_state_ci( - {}, static_cast<u32>(dynamic_states.size()), dynamic_states.data()); - - vk::PipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci; + + VkPipelineColorBlendStateCreateInfo color_blend_ci; + color_blend_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO; + color_blend_ci.pNext = nullptr; + color_blend_ci.flags = 0; + color_blend_ci.logicOpEnable = VK_FALSE; + color_blend_ci.logicOp = VK_LOGIC_OP_COPY; + color_blend_ci.attachmentCount = static_cast<u32>(num_attachments); + color_blend_ci.pAttachments = cb_attachments.data(); + std::memset(color_blend_ci.blendConstants, 0, sizeof(color_blend_ci.blendConstants)); + + static constexpr std::array dynamic_states = { + VK_DYNAMIC_STATE_VIEWPORT, VK_DYNAMIC_STATE_SCISSOR, + VK_DYNAMIC_STATE_DEPTH_BIAS, VK_DYNAMIC_STATE_BLEND_CONSTANTS, + VK_DYNAMIC_STATE_DEPTH_BOUNDS, VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK, + VK_DYNAMIC_STATE_STENCIL_WRITE_MASK, VK_DYNAMIC_STATE_STENCIL_REFERENCE}; + + VkPipelineDynamicStateCreateInfo dynamic_state_ci; + dynamic_state_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO; + dynamic_state_ci.pNext = nullptr; + dynamic_state_ci.flags = 0; + dynamic_state_ci.dynamicStateCount = static_cast<u32>(dynamic_states.size()); + dynamic_state_ci.pDynamicStates = dynamic_states.data(); + + VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci; + subgroup_size_ci.sType = + VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_REQUIRED_SUBGROUP_SIZE_CREATE_INFO_EXT; + subgroup_size_ci.pNext = nullptr; subgroup_size_ci.requiredSubgroupSize = GuestWarpSize; - std::vector<vk::PipelineShaderStageCreateInfo> shader_stages; + std::vector<VkPipelineShaderStageCreateInfo> shader_stages; std::size_t module_index = 0; for (std::size_t stage = 0; stage < Maxwell::MaxShaderStage; ++stage) { if (!program[stage]) { continue; } - const auto stage_enum = static_cast<Tegra::Engines::ShaderType>(stage); - const auto vk_stage = MaxwellToVK::ShaderStage(stage_enum); - auto& stage_ci = shader_stages.emplace_back(vk::PipelineShaderStageCreateFlags{}, vk_stage, - *modules[module_index++], "main", nullptr); - if (program[stage]->entries.uses_warps && device.IsGuestWarpSizeSupported(vk_stage)) { + VkPipelineShaderStageCreateInfo& stage_ci = shader_stages.emplace_back(); + stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; + stage_ci.pNext = nullptr; + stage_ci.flags = 0; + stage_ci.stage = MaxwellToVK::ShaderStage(static_cast<Tegra::Engines::ShaderType>(stage)); + stage_ci.module = *modules[module_index++]; + stage_ci.pName = "main"; + stage_ci.pSpecializationInfo = nullptr; + + if (program[stage]->entries.uses_warps && device.IsGuestWarpSizeSupported(stage_ci.stage)) { stage_ci.pNext = &subgroup_size_ci; } } - const vk::GraphicsPipelineCreateInfo create_info( - {}, static_cast<u32>(shader_stages.size()), shader_stages.data(), &vertex_input_ci, - &input_assembly_ci, &tessellation_ci, &viewport_ci, &rasterizer_ci, &multisampling_ci, - &depth_stencil_ci, &color_blending_ci, &dynamic_state_ci, *layout, renderpass, 0, {}, 0); - - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - return dev.createGraphicsPipelineUnique(nullptr, create_info, nullptr, dld); + VkGraphicsPipelineCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.stageCount = static_cast<u32>(shader_stages.size()); + ci.pStages = shader_stages.data(); + ci.pVertexInputState = &vertex_input_ci; + ci.pInputAssemblyState = &input_assembly_ci; + ci.pTessellationState = &tessellation_ci; + ci.pViewportState = &viewport_ci; + ci.pRasterizationState = &rasterization_ci; + ci.pMultisampleState = &multisample_ci; + ci.pDepthStencilState = &depth_stencil_ci; + ci.pColorBlendState = &color_blend_ci; + ci.pDynamicState = &dynamic_state_ci; + ci.layout = *layout; + ci.renderPass = renderpass; + ci.subpass = 0; + ci.basePipelineHandle = nullptr; + ci.basePipelineIndex = 0; + return device.GetLogical().CreateGraphicsPipeline(ci); } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h index 4f5e4ea2d..7aba70960 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h @@ -11,12 +11,12 @@ #include <vector> #include "video_core/engines/maxwell_3d.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_renderpass_cache.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_shader_decompiler.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -39,52 +39,52 @@ public: VKUpdateDescriptorQueue& update_descriptor_queue, VKRenderPassCache& renderpass_cache, const GraphicsPipelineCacheKey& key, - const std::vector<vk::DescriptorSetLayoutBinding>& bindings, + vk::Span<VkDescriptorSetLayoutBinding> bindings, const SPIRVProgram& program); ~VKGraphicsPipeline(); - vk::DescriptorSet CommitDescriptorSet(); + VkDescriptorSet CommitDescriptorSet(); - vk::Pipeline GetHandle() const { + VkPipeline GetHandle() const { return *pipeline; } - vk::PipelineLayout GetLayout() const { + VkPipelineLayout GetLayout() const { return *layout; } - vk::RenderPass GetRenderPass() const { + VkRenderPass GetRenderPass() const { return renderpass; } private: - UniqueDescriptorSetLayout CreateDescriptorSetLayout( - const std::vector<vk::DescriptorSetLayoutBinding>& bindings) const; + vk::DescriptorSetLayout CreateDescriptorSetLayout( + vk::Span<VkDescriptorSetLayoutBinding> bindings) const; - UniquePipelineLayout CreatePipelineLayout() const; + vk::PipelineLayout CreatePipelineLayout() const; - UniqueDescriptorUpdateTemplate CreateDescriptorUpdateTemplate( + vk::DescriptorUpdateTemplateKHR CreateDescriptorUpdateTemplate( const SPIRVProgram& program) const; - std::vector<UniqueShaderModule> CreateShaderModules(const SPIRVProgram& program) const; + std::vector<vk::ShaderModule> CreateShaderModules(const SPIRVProgram& program) const; - UniquePipeline CreatePipeline(const RenderPassParams& renderpass_params, - const SPIRVProgram& program) const; + vk::Pipeline CreatePipeline(const RenderPassParams& renderpass_params, + const SPIRVProgram& program) const; const VKDevice& device; VKScheduler& scheduler; const FixedPipelineState fixed_state; const u64 hash; - UniqueDescriptorSetLayout descriptor_set_layout; + vk::DescriptorSetLayout descriptor_set_layout; DescriptorAllocator descriptor_allocator; VKUpdateDescriptorQueue& update_descriptor_queue; - UniquePipelineLayout layout; - UniqueDescriptorUpdateTemplate descriptor_template; - std::vector<UniqueShaderModule> modules; + vk::PipelineLayout layout; + vk::DescriptorUpdateTemplateKHR descriptor_template; + std::vector<vk::ShaderModule> modules; - vk::RenderPass renderpass; - UniquePipeline pipeline; + VkRenderPass renderpass; + vk::Pipeline pipeline; }; } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_image.cpp b/src/video_core/renderer_vulkan/vk_image.cpp index 4bcbef959..9bceb3861 100644 --- a/src/video_core/renderer_vulkan/vk_image.cpp +++ b/src/video_core/renderer_vulkan/vk_image.cpp @@ -6,22 +6,21 @@ #include <vector> #include "common/assert.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_image.h" #include "video_core/renderer_vulkan/vk_scheduler.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { -VKImage::VKImage(const VKDevice& device, VKScheduler& scheduler, - const vk::ImageCreateInfo& image_ci, vk::ImageAspectFlags aspect_mask) +VKImage::VKImage(const VKDevice& device, VKScheduler& scheduler, const VkImageCreateInfo& image_ci, + VkImageAspectFlags aspect_mask) : device{device}, scheduler{scheduler}, format{image_ci.format}, aspect_mask{aspect_mask}, image_num_layers{image_ci.arrayLayers}, image_num_levels{image_ci.mipLevels} { UNIMPLEMENTED_IF_MSG(image_ci.queueFamilyIndexCount != 0, "Queue family tracking is not implemented"); - const auto dev = device.GetLogical(); - image = dev.createImageUnique(image_ci, nullptr, device.GetDispatchLoader()); + image = device.GetLogical().CreateImage(image_ci); const u32 num_ranges = image_num_layers * image_num_levels; barriers.resize(num_ranges); @@ -31,8 +30,8 @@ VKImage::VKImage(const VKDevice& device, VKScheduler& scheduler, VKImage::~VKImage() = default; void VKImage::Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, - vk::PipelineStageFlags new_stage_mask, vk::AccessFlags new_access, - vk::ImageLayout new_layout) { + VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access, + VkImageLayout new_layout) { if (!HasChanged(base_layer, num_layers, base_level, num_levels, new_access, new_layout)) { return; } @@ -43,9 +42,21 @@ void VKImage::Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num const u32 layer = base_layer + layer_it; const u32 level = base_level + level_it; auto& state = GetSubrangeState(layer, level); - barriers[cursor] = vk::ImageMemoryBarrier( - state.access, new_access, state.layout, new_layout, VK_QUEUE_FAMILY_IGNORED, - VK_QUEUE_FAMILY_IGNORED, *image, {aspect_mask, level, 1, layer, 1}); + auto& barrier = barriers[cursor]; + barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; + barrier.pNext = nullptr; + barrier.srcAccessMask = state.access; + barrier.dstAccessMask = new_access; + barrier.oldLayout = state.layout; + barrier.newLayout = new_layout; + barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; + barrier.image = *image; + barrier.subresourceRange.aspectMask = aspect_mask; + barrier.subresourceRange.baseMipLevel = level; + barrier.subresourceRange.levelCount = 1; + barrier.subresourceRange.baseArrayLayer = layer; + barrier.subresourceRange.layerCount = 1; state.access = new_access; state.layout = new_layout; } @@ -53,16 +64,16 @@ void VKImage::Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num scheduler.RequestOutsideRenderPassOperationContext(); - scheduler.Record([barriers = barriers, cursor](auto cmdbuf, auto& dld) { + scheduler.Record([barriers = barriers, cursor](vk::CommandBuffer cmdbuf) { // TODO(Rodrigo): Implement a way to use the latest stage across subresources. - constexpr auto stage_stub = vk::PipelineStageFlagBits::eAllCommands; - cmdbuf.pipelineBarrier(stage_stub, stage_stub, {}, 0, nullptr, 0, nullptr, - static_cast<u32>(cursor), barriers.data(), dld); + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, + VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, {}, {}, + vk::Span(barriers.data(), cursor)); }); } bool VKImage::HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, - vk::AccessFlags new_access, vk::ImageLayout new_layout) noexcept { + VkAccessFlags new_access, VkImageLayout new_layout) noexcept { const bool is_full_range = base_layer == 0 && num_layers == image_num_layers && base_level == 0 && num_levels == image_num_levels; if (!is_full_range) { @@ -91,11 +102,21 @@ bool VKImage::HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num void VKImage::CreatePresentView() { // Image type has to be 2D to be presented. - const vk::ImageViewCreateInfo image_view_ci({}, *image, vk::ImageViewType::e2D, format, {}, - {aspect_mask, 0, 1, 0, 1}); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - present_view = dev.createImageViewUnique(image_view_ci, nullptr, dld); + VkImageViewCreateInfo image_view_ci; + image_view_ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; + image_view_ci.pNext = nullptr; + image_view_ci.flags = 0; + image_view_ci.image = *image; + image_view_ci.viewType = VK_IMAGE_VIEW_TYPE_2D; + image_view_ci.format = format; + image_view_ci.components = {VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY, + VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY}; + image_view_ci.subresourceRange.aspectMask = aspect_mask; + image_view_ci.subresourceRange.baseMipLevel = 0; + image_view_ci.subresourceRange.levelCount = 1; + image_view_ci.subresourceRange.baseArrayLayer = 0; + image_view_ci.subresourceRange.layerCount = 1; + present_view = device.GetLogical().CreateImageView(image_view_ci); } VKImage::SubrangeState& VKImage::GetSubrangeState(u32 layer, u32 level) noexcept { diff --git a/src/video_core/renderer_vulkan/vk_image.h b/src/video_core/renderer_vulkan/vk_image.h index b78242512..b4d7229e5 100644 --- a/src/video_core/renderer_vulkan/vk_image.h +++ b/src/video_core/renderer_vulkan/vk_image.h @@ -8,7 +8,7 @@ #include <vector> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -18,16 +18,16 @@ class VKScheduler; class VKImage { public: explicit VKImage(const VKDevice& device, VKScheduler& scheduler, - const vk::ImageCreateInfo& image_ci, vk::ImageAspectFlags aspect_mask); + const VkImageCreateInfo& image_ci, VkImageAspectFlags aspect_mask); ~VKImage(); /// Records in the passed command buffer an image transition and updates the state of the image. void Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, - vk::PipelineStageFlags new_stage_mask, vk::AccessFlags new_access, - vk::ImageLayout new_layout); + VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access, + VkImageLayout new_layout); /// Returns a view compatible with presentation, the image has to be 2D. - vk::ImageView GetPresentView() { + VkImageView GetPresentView() { if (!present_view) { CreatePresentView(); } @@ -35,28 +35,28 @@ public: } /// Returns the Vulkan image handler. - vk::Image GetHandle() const { - return *image; + const vk::Image& GetHandle() const { + return image; } /// Returns the Vulkan format for this image. - vk::Format GetFormat() const { + VkFormat GetFormat() const { return format; } /// Returns the Vulkan aspect mask. - vk::ImageAspectFlags GetAspectMask() const { + VkImageAspectFlags GetAspectMask() const { return aspect_mask; } private: struct SubrangeState final { - vk::AccessFlags access{}; ///< Current access bits. - vk::ImageLayout layout = vk::ImageLayout::eUndefined; ///< Current image layout. + VkAccessFlags access = 0; ///< Current access bits. + VkImageLayout layout = VK_IMAGE_LAYOUT_UNDEFINED; ///< Current image layout. }; bool HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, - vk::AccessFlags new_access, vk::ImageLayout new_layout) noexcept; + VkAccessFlags new_access, VkImageLayout new_layout) noexcept; /// Creates a presentation view. void CreatePresentView(); @@ -67,16 +67,16 @@ private: const VKDevice& device; ///< Device handler. VKScheduler& scheduler; ///< Device scheduler. - const vk::Format format; ///< Vulkan format. - const vk::ImageAspectFlags aspect_mask; ///< Vulkan aspect mask. - const u32 image_num_layers; ///< Number of layers. - const u32 image_num_levels; ///< Number of mipmap levels. + const VkFormat format; ///< Vulkan format. + const VkImageAspectFlags aspect_mask; ///< Vulkan aspect mask. + const u32 image_num_layers; ///< Number of layers. + const u32 image_num_levels; ///< Number of mipmap levels. - UniqueImage image; ///< Image handle. - UniqueImageView present_view; ///< Image view compatible with presentation. + vk::Image image; ///< Image handle. + vk::ImageView present_view; ///< Image view compatible with presentation. - std::vector<vk::ImageMemoryBarrier> barriers; ///< Pool of barriers. - std::vector<SubrangeState> subrange_states; ///< Current subrange state. + std::vector<VkImageMemoryBarrier> barriers; ///< Pool of barriers. + std::vector<SubrangeState> subrange_states; ///< Current subrange state. bool state_diverged = false; ///< True when subresources mismatch in layout. }; diff --git a/src/video_core/renderer_vulkan/vk_memory_manager.cpp b/src/video_core/renderer_vulkan/vk_memory_manager.cpp index 9cc9979d0..6a9e658bf 100644 --- a/src/video_core/renderer_vulkan/vk_memory_manager.cpp +++ b/src/video_core/renderer_vulkan/vk_memory_manager.cpp @@ -11,9 +11,9 @@ #include "common/assert.h" #include "common/common_types.h" #include "common/logging/log.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_memory_manager.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -30,17 +30,11 @@ u64 GetAllocationChunkSize(u64 required_size) { class VKMemoryAllocation final { public: explicit VKMemoryAllocation(const VKDevice& device, vk::DeviceMemory memory, - vk::MemoryPropertyFlags properties, u64 allocation_size, u32 type) - : device{device}, memory{memory}, properties{properties}, allocation_size{allocation_size}, - shifted_type{ShiftType(type)} {} - - ~VKMemoryAllocation() { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - dev.free(memory, nullptr, dld); - } + VkMemoryPropertyFlags properties, u64 allocation_size, u32 type) + : device{device}, memory{std::move(memory)}, properties{properties}, + allocation_size{allocation_size}, shifted_type{ShiftType(type)} {} - VKMemoryCommit Commit(vk::DeviceSize commit_size, vk::DeviceSize alignment) { + VKMemoryCommit Commit(VkDeviceSize commit_size, VkDeviceSize alignment) { auto found = TryFindFreeSection(free_iterator, allocation_size, static_cast<u64>(commit_size), static_cast<u64>(alignment)); if (!found) { @@ -73,9 +67,8 @@ public: } /// Returns whether this allocation is compatible with the arguments. - bool IsCompatible(vk::MemoryPropertyFlags wanted_properties, u32 type_mask) const { - return (wanted_properties & properties) != vk::MemoryPropertyFlagBits(0) && - (type_mask & shifted_type) != 0; + bool IsCompatible(VkMemoryPropertyFlags wanted_properties, u32 type_mask) const { + return (wanted_properties & properties) && (type_mask & shifted_type) != 0; } private: @@ -111,11 +104,11 @@ private: return std::nullopt; } - const VKDevice& device; ///< Vulkan device. - const vk::DeviceMemory memory; ///< Vulkan memory allocation handler. - const vk::MemoryPropertyFlags properties; ///< Vulkan properties. - const u64 allocation_size; ///< Size of this allocation. - const u32 shifted_type; ///< Stored Vulkan type of this allocation, shifted. + const VKDevice& device; ///< Vulkan device. + const vk::DeviceMemory memory; ///< Vulkan memory allocation handler. + const VkMemoryPropertyFlags properties; ///< Vulkan properties. + const u64 allocation_size; ///< Size of this allocation. + const u32 shifted_type; ///< Stored Vulkan type of this allocation, shifted. /// Hints where the next free region is likely going to be. u64 free_iterator{}; @@ -125,22 +118,20 @@ private: }; VKMemoryManager::VKMemoryManager(const VKDevice& device) - : device{device}, properties{device.GetPhysical().getMemoryProperties( - device.GetDispatchLoader())}, + : device{device}, properties{device.GetPhysical().GetMemoryProperties()}, is_memory_unified{GetMemoryUnified(properties)} {} VKMemoryManager::~VKMemoryManager() = default; -VKMemoryCommit VKMemoryManager::Commit(const vk::MemoryRequirements& requirements, +VKMemoryCommit VKMemoryManager::Commit(const VkMemoryRequirements& requirements, bool host_visible) { const u64 chunk_size = GetAllocationChunkSize(requirements.size); // When a host visible commit is asked, search for host visible and coherent, otherwise search // for a fast device local type. - const vk::MemoryPropertyFlags wanted_properties = - host_visible - ? vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent - : vk::MemoryPropertyFlagBits::eDeviceLocal; + const VkMemoryPropertyFlags wanted_properties = + host_visible ? VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT + : VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT; if (auto commit = TryAllocCommit(requirements, wanted_properties)) { return commit; @@ -161,23 +152,19 @@ VKMemoryCommit VKMemoryManager::Commit(const vk::MemoryRequirements& requirement return commit; } -VKMemoryCommit VKMemoryManager::Commit(vk::Buffer buffer, bool host_visible) { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - auto commit = Commit(dev.getBufferMemoryRequirements(buffer, dld), host_visible); - dev.bindBufferMemory(buffer, commit->GetMemory(), commit->GetOffset(), dld); +VKMemoryCommit VKMemoryManager::Commit(const vk::Buffer& buffer, bool host_visible) { + auto commit = Commit(device.GetLogical().GetBufferMemoryRequirements(*buffer), host_visible); + buffer.BindMemory(commit->GetMemory(), commit->GetOffset()); return commit; } -VKMemoryCommit VKMemoryManager::Commit(vk::Image image, bool host_visible) { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - auto commit = Commit(dev.getImageMemoryRequirements(image, dld), host_visible); - dev.bindImageMemory(image, commit->GetMemory(), commit->GetOffset(), dld); +VKMemoryCommit VKMemoryManager::Commit(const vk::Image& image, bool host_visible) { + auto commit = Commit(device.GetLogical().GetImageMemoryRequirements(*image), host_visible); + image.BindMemory(commit->GetMemory(), commit->GetOffset()); return commit; } -bool VKMemoryManager::AllocMemory(vk::MemoryPropertyFlags wanted_properties, u32 type_mask, +bool VKMemoryManager::AllocMemory(VkMemoryPropertyFlags wanted_properties, u32 type_mask, u64 size) { const u32 type = [&] { for (u32 type_index = 0; type_index < properties.memoryTypeCount; ++type_index) { @@ -191,24 +178,26 @@ bool VKMemoryManager::AllocMemory(vk::MemoryPropertyFlags wanted_properties, u32 return 0U; }(); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - // Try to allocate found type. - const vk::MemoryAllocateInfo memory_ai(size, type); - vk::DeviceMemory memory; - if (const auto res = dev.allocateMemory(&memory_ai, nullptr, &memory, dld); - res != vk::Result::eSuccess) { - LOG_CRITICAL(Render_Vulkan, "Device allocation failed with code {}!", vk::to_string(res)); + VkMemoryAllocateInfo memory_ai; + memory_ai.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO; + memory_ai.pNext = nullptr; + memory_ai.allocationSize = size; + memory_ai.memoryTypeIndex = type; + + vk::DeviceMemory memory = device.GetLogical().TryAllocateMemory(memory_ai); + if (!memory) { + LOG_CRITICAL(Render_Vulkan, "Device allocation failed!"); return false; } - allocations.push_back( - std::make_unique<VKMemoryAllocation>(device, memory, wanted_properties, size, type)); + + allocations.push_back(std::make_unique<VKMemoryAllocation>(device, std::move(memory), + wanted_properties, size, type)); return true; } -VKMemoryCommit VKMemoryManager::TryAllocCommit(const vk::MemoryRequirements& requirements, - vk::MemoryPropertyFlags wanted_properties) { +VKMemoryCommit VKMemoryManager::TryAllocCommit(const VkMemoryRequirements& requirements, + VkMemoryPropertyFlags wanted_properties) { for (auto& allocation : allocations) { if (!allocation->IsCompatible(wanted_properties, requirements.memoryTypeBits)) { continue; @@ -220,10 +209,9 @@ VKMemoryCommit VKMemoryManager::TryAllocCommit(const vk::MemoryRequirements& req return {}; } -/*static*/ bool VKMemoryManager::GetMemoryUnified( - const vk::PhysicalDeviceMemoryProperties& properties) { +bool VKMemoryManager::GetMemoryUnified(const VkPhysicalDeviceMemoryProperties& properties) { for (u32 heap_index = 0; heap_index < properties.memoryHeapCount; ++heap_index) { - if (!(properties.memoryHeaps[heap_index].flags & vk::MemoryHeapFlagBits::eDeviceLocal)) { + if (!(properties.memoryHeaps[heap_index].flags & VK_MEMORY_HEAP_DEVICE_LOCAL_BIT)) { // Memory is considered unified when heaps are device local only. return false; } @@ -232,23 +220,19 @@ VKMemoryCommit VKMemoryManager::TryAllocCommit(const vk::MemoryRequirements& req } VKMemoryCommitImpl::VKMemoryCommitImpl(const VKDevice& device, VKMemoryAllocation* allocation, - vk::DeviceMemory memory, u64 begin, u64 end) - : device{device}, interval{begin, end}, memory{memory}, allocation{allocation} {} + const vk::DeviceMemory& memory, u64 begin, u64 end) + : device{device}, memory{memory}, interval{begin, end}, allocation{allocation} {} VKMemoryCommitImpl::~VKMemoryCommitImpl() { allocation->Free(this); } MemoryMap VKMemoryCommitImpl::Map(u64 size, u64 offset_) const { - const auto dev = device.GetLogical(); - const auto address = reinterpret_cast<u8*>( - dev.mapMemory(memory, interval.first + offset_, size, {}, device.GetDispatchLoader())); - return MemoryMap{this, address}; + return MemoryMap{this, memory.Map(interval.first + offset_, size)}; } void VKMemoryCommitImpl::Unmap() const { - const auto dev = device.GetLogical(); - dev.unmapMemory(memory, device.GetDispatchLoader()); + memory.Unmap(); } MemoryMap VKMemoryCommitImpl::Map() const { diff --git a/src/video_core/renderer_vulkan/vk_memory_manager.h b/src/video_core/renderer_vulkan/vk_memory_manager.h index cd00bb91b..35ee54d30 100644 --- a/src/video_core/renderer_vulkan/vk_memory_manager.h +++ b/src/video_core/renderer_vulkan/vk_memory_manager.h @@ -8,7 +8,7 @@ #include <utility> #include <vector> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -32,13 +32,13 @@ public: * memory. When passing false, it will try to allocate device local memory. * @returns A memory commit. */ - VKMemoryCommit Commit(const vk::MemoryRequirements& reqs, bool host_visible); + VKMemoryCommit Commit(const VkMemoryRequirements& reqs, bool host_visible); /// Commits memory required by the buffer and binds it. - VKMemoryCommit Commit(vk::Buffer buffer, bool host_visible); + VKMemoryCommit Commit(const vk::Buffer& buffer, bool host_visible); /// Commits memory required by the image and binds it. - VKMemoryCommit Commit(vk::Image image, bool host_visible); + VKMemoryCommit Commit(const vk::Image& image, bool host_visible); /// Returns true if the memory allocations are done always in host visible and coherent memory. bool IsMemoryUnified() const { @@ -47,18 +47,18 @@ public: private: /// Allocates a chunk of memory. - bool AllocMemory(vk::MemoryPropertyFlags wanted_properties, u32 type_mask, u64 size); + bool AllocMemory(VkMemoryPropertyFlags wanted_properties, u32 type_mask, u64 size); /// Tries to allocate a memory commit. - VKMemoryCommit TryAllocCommit(const vk::MemoryRequirements& requirements, - vk::MemoryPropertyFlags wanted_properties); + VKMemoryCommit TryAllocCommit(const VkMemoryRequirements& requirements, + VkMemoryPropertyFlags wanted_properties); /// Returns true if the device uses an unified memory model. - static bool GetMemoryUnified(const vk::PhysicalDeviceMemoryProperties& properties); + static bool GetMemoryUnified(const VkPhysicalDeviceMemoryProperties& properties); - const VKDevice& device; ///< Device handler. - const vk::PhysicalDeviceMemoryProperties properties; ///< Physical device properties. - const bool is_memory_unified; ///< True if memory model is unified. + const VKDevice& device; ///< Device handler. + const VkPhysicalDeviceMemoryProperties properties; ///< Physical device properties. + const bool is_memory_unified; ///< True if memory model is unified. std::vector<std::unique_ptr<VKMemoryAllocation>> allocations; ///< Current allocations. }; @@ -68,7 +68,7 @@ class VKMemoryCommitImpl final { public: explicit VKMemoryCommitImpl(const VKDevice& device, VKMemoryAllocation* allocation, - vk::DeviceMemory memory, u64 begin, u64 end); + const vk::DeviceMemory& memory, u64 begin, u64 end); ~VKMemoryCommitImpl(); /// Maps a memory region and returns a pointer to it. @@ -80,13 +80,13 @@ public: MemoryMap Map() const; /// Returns the Vulkan memory handler. - vk::DeviceMemory GetMemory() const { - return memory; + VkDeviceMemory GetMemory() const { + return *memory; } /// Returns the start position of the commit relative to the allocation. - vk::DeviceSize GetOffset() const { - return static_cast<vk::DeviceSize>(interval.first); + VkDeviceSize GetOffset() const { + return static_cast<VkDeviceSize>(interval.first); } private: @@ -94,8 +94,8 @@ private: void Unmap() const; const VKDevice& device; ///< Vulkan device. + const vk::DeviceMemory& memory; ///< Vulkan device memory handler. std::pair<u64, u64> interval{}; ///< Interval where the commit exists. - vk::DeviceMemory memory; ///< Vulkan device memory handler. VKMemoryAllocation* allocation{}; ///< Pointer to the large memory allocation. }; diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp index 557b9d662..90e3a8edd 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp @@ -13,7 +13,6 @@ #include "video_core/engines/kepler_compute.h" #include "video_core/engines/maxwell_3d.h" #include "video_core/memory_manager.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/vk_compute_pipeline.h" @@ -26,6 +25,7 @@ #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/shader/compiler_settings.h" namespace Vulkan { @@ -36,12 +36,11 @@ using Tegra::Engines::ShaderType; namespace { -// C++20's using enum -constexpr auto eUniformBuffer = vk::DescriptorType::eUniformBuffer; -constexpr auto eStorageBuffer = vk::DescriptorType::eStorageBuffer; -constexpr auto eUniformTexelBuffer = vk::DescriptorType::eUniformTexelBuffer; -constexpr auto eCombinedImageSampler = vk::DescriptorType::eCombinedImageSampler; -constexpr auto eStorageImage = vk::DescriptorType::eStorageImage; +constexpr VkDescriptorType UNIFORM_BUFFER = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; +constexpr VkDescriptorType STORAGE_BUFFER = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; +constexpr VkDescriptorType UNIFORM_TEXEL_BUFFER = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER; +constexpr VkDescriptorType COMBINED_IMAGE_SAMPLER = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; +constexpr VkDescriptorType STORAGE_IMAGE = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE; constexpr VideoCommon::Shader::CompilerSettings compiler_settings{ VideoCommon::Shader::CompileDepth::FullDecompile}; @@ -126,43 +125,48 @@ ShaderType GetShaderType(Maxwell::ShaderProgram program) { } } -template <vk::DescriptorType descriptor_type, class Container> -void AddBindings(std::vector<vk::DescriptorSetLayoutBinding>& bindings, u32& binding, - vk::ShaderStageFlags stage_flags, const Container& container) { +template <VkDescriptorType descriptor_type, class Container> +void AddBindings(std::vector<VkDescriptorSetLayoutBinding>& bindings, u32& binding, + VkShaderStageFlags stage_flags, const Container& container) { const u32 num_entries = static_cast<u32>(std::size(container)); for (std::size_t i = 0; i < num_entries; ++i) { u32 count = 1; - if constexpr (descriptor_type == eCombinedImageSampler) { + if constexpr (descriptor_type == VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER) { // Combined image samplers can be arrayed. count = container[i].Size(); } - bindings.emplace_back(binding++, descriptor_type, count, stage_flags, nullptr); + VkDescriptorSetLayoutBinding& entry = bindings.emplace_back(); + entry.binding = binding++; + entry.descriptorType = descriptor_type; + entry.descriptorCount = count; + entry.stageFlags = stage_flags; + entry.pImmutableSamplers = nullptr; } } u32 FillDescriptorLayout(const ShaderEntries& entries, - std::vector<vk::DescriptorSetLayoutBinding>& bindings, + std::vector<VkDescriptorSetLayoutBinding>& bindings, Maxwell::ShaderProgram program_type, u32 base_binding) { const ShaderType stage = GetStageFromProgram(program_type); - const vk::ShaderStageFlags flags = MaxwellToVK::ShaderStage(stage); + const VkShaderStageFlags flags = MaxwellToVK::ShaderStage(stage); u32 binding = base_binding; - AddBindings<eUniformBuffer>(bindings, binding, flags, entries.const_buffers); - AddBindings<eStorageBuffer>(bindings, binding, flags, entries.global_buffers); - AddBindings<eUniformTexelBuffer>(bindings, binding, flags, entries.texel_buffers); - AddBindings<eCombinedImageSampler>(bindings, binding, flags, entries.samplers); - AddBindings<eStorageImage>(bindings, binding, flags, entries.images); + AddBindings<UNIFORM_BUFFER>(bindings, binding, flags, entries.const_buffers); + AddBindings<STORAGE_BUFFER>(bindings, binding, flags, entries.global_buffers); + AddBindings<UNIFORM_TEXEL_BUFFER>(bindings, binding, flags, entries.texel_buffers); + AddBindings<COMBINED_IMAGE_SAMPLER>(bindings, binding, flags, entries.samplers); + AddBindings<STORAGE_IMAGE>(bindings, binding, flags, entries.images); return binding; } } // Anonymous namespace CachedShader::CachedShader(Core::System& system, Tegra::Engines::ShaderType stage, - GPUVAddr gpu_addr, VAddr cpu_addr, u8* host_ptr, - ProgramCode program_code, u32 main_offset) - : RasterizerCacheObject{host_ptr}, gpu_addr{gpu_addr}, cpu_addr{cpu_addr}, - program_code{std::move(program_code)}, registry{stage, GetEngine(system, stage)}, - shader_ir{this->program_code, main_offset, compiler_settings, registry}, + GPUVAddr gpu_addr, VAddr cpu_addr, ProgramCode program_code, + u32 main_offset) + : RasterizerCacheObject{cpu_addr}, gpu_addr{gpu_addr}, program_code{std::move(program_code)}, + registry{stage, GetEngine(system, stage)}, shader_ir{this->program_code, main_offset, + compiler_settings, registry}, entries{GenerateShaderEntries(shader_ir)} {} CachedShader::~CachedShader() = default; @@ -201,19 +205,19 @@ std::array<Shader, Maxwell::MaxShaderProgram> VKPipelineCache::GetShaders() { auto& memory_manager{system.GPU().MemoryManager()}; const GPUVAddr program_addr{GetShaderAddress(system, program)}; - const auto host_ptr{memory_manager.GetPointer(program_addr)}; - auto shader = TryGet(host_ptr); + const std::optional cpu_addr = memory_manager.GpuToCpuAddress(program_addr); + ASSERT(cpu_addr); + auto shader = cpu_addr ? TryGet(*cpu_addr) : nullptr; if (!shader) { + const auto host_ptr{memory_manager.GetPointer(program_addr)}; + // No shader found - create a new one constexpr u32 stage_offset = 10; const auto stage = static_cast<Tegra::Engines::ShaderType>(index == 0 ? 0 : index - 1); auto code = GetShaderCode(memory_manager, program_addr, host_ptr, false); - const std::optional cpu_addr = memory_manager.GpuToCpuAddress(program_addr); - ASSERT(cpu_addr); - shader = std::make_shared<CachedShader>(system, stage, program_addr, *cpu_addr, - host_ptr, std::move(code), stage_offset); + std::move(code), stage_offset); Register(shader); } shaders[index] = std::move(shader); @@ -253,18 +257,19 @@ VKComputePipeline& VKPipelineCache::GetComputePipeline(const ComputePipelineCach auto& memory_manager = system.GPU().MemoryManager(); const auto program_addr = key.shader; - const auto host_ptr = memory_manager.GetPointer(program_addr); - auto shader = TryGet(host_ptr); + const auto cpu_addr = memory_manager.GpuToCpuAddress(program_addr); + ASSERT(cpu_addr); + + auto shader = cpu_addr ? TryGet(*cpu_addr) : nullptr; if (!shader) { // No shader found - create a new one - const auto cpu_addr = memory_manager.GpuToCpuAddress(program_addr); - ASSERT(cpu_addr); + const auto host_ptr = memory_manager.GetPointer(program_addr); auto code = GetShaderCode(memory_manager, program_addr, host_ptr, true); constexpr u32 kernel_main_offset = 0; shader = std::make_shared<CachedShader>(system, Tegra::Engines::ShaderType::Compute, - program_addr, *cpu_addr, host_ptr, std::move(code), + program_addr, *cpu_addr, std::move(code), kernel_main_offset); Register(shader); } @@ -317,7 +322,7 @@ void VKPipelineCache::Unregister(const Shader& shader) { RasterizerCache::Unregister(shader); } -std::pair<SPIRVProgram, std::vector<vk::DescriptorSetLayoutBinding>> +std::pair<SPIRVProgram, std::vector<VkDescriptorSetLayoutBinding>> VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) { const auto& fixed_state = key.fixed_state; auto& memory_manager = system.GPU().MemoryManager(); @@ -334,7 +339,7 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) { specialization.ndc_minus_one_to_one = fixed_state.rasterizer.ndc_minus_one_to_one; SPIRVProgram program; - std::vector<vk::DescriptorSetLayoutBinding> bindings; + std::vector<VkDescriptorSetLayoutBinding> bindings; for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) { const auto program_enum = static_cast<Maxwell::ShaderProgram>(index); @@ -345,8 +350,9 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) { } const GPUVAddr gpu_addr = GetShaderAddress(system, program_enum); - const auto host_ptr = memory_manager.GetPointer(gpu_addr); - const auto shader = TryGet(host_ptr); + const auto cpu_addr = memory_manager.GpuToCpuAddress(gpu_addr); + ASSERT(cpu_addr); + const auto shader = TryGet(*cpu_addr); ASSERT(shader); const std::size_t stage = index == 0 ? 0 : index - 1; // Stage indices are 0 - 5 @@ -369,32 +375,49 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) { return {std::move(program), std::move(bindings)}; } -template <vk::DescriptorType descriptor_type, class Container> -void AddEntry(std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries, u32& binding, +template <VkDescriptorType descriptor_type, class Container> +void AddEntry(std::vector<VkDescriptorUpdateTemplateEntry>& template_entries, u32& binding, u32& offset, const Container& container) { static constexpr u32 entry_size = static_cast<u32>(sizeof(DescriptorUpdateEntry)); const u32 count = static_cast<u32>(std::size(container)); - if constexpr (descriptor_type == eCombinedImageSampler) { + if constexpr (descriptor_type == COMBINED_IMAGE_SAMPLER) { for (u32 i = 0; i < count; ++i) { const u32 num_samplers = container[i].Size(); - template_entries.emplace_back(binding, 0, num_samplers, descriptor_type, offset, - entry_size); + VkDescriptorUpdateTemplateEntry& entry = template_entries.emplace_back(); + entry.dstBinding = binding; + entry.dstArrayElement = 0; + entry.descriptorCount = num_samplers; + entry.descriptorType = descriptor_type; + entry.offset = offset; + entry.stride = entry_size; + ++binding; offset += num_samplers * entry_size; } return; } - if constexpr (descriptor_type == eUniformTexelBuffer) { + if constexpr (descriptor_type == UNIFORM_TEXEL_BUFFER) { // Nvidia has a bug where updating multiple uniform texels at once causes the driver to // crash. for (u32 i = 0; i < count; ++i) { - template_entries.emplace_back(binding + i, 0, 1, descriptor_type, - offset + i * entry_size, entry_size); + VkDescriptorUpdateTemplateEntry& entry = template_entries.emplace_back(); + entry.dstBinding = binding + i; + entry.dstArrayElement = 0; + entry.descriptorCount = 1; + entry.descriptorType = descriptor_type; + entry.offset = offset + i * entry_size; + entry.stride = entry_size; } } else if (count > 0) { - template_entries.emplace_back(binding, 0, count, descriptor_type, offset, entry_size); + VkDescriptorUpdateTemplateEntry& entry = template_entries.emplace_back(); + entry.dstBinding = binding; + entry.dstArrayElement = 0; + entry.descriptorCount = count; + entry.descriptorType = descriptor_type; + entry.offset = offset; + entry.stride = entry_size; } offset += count * entry_size; binding += count; @@ -402,12 +425,12 @@ void AddEntry(std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries, void FillDescriptorUpdateTemplateEntries( const ShaderEntries& entries, u32& binding, u32& offset, - std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries) { - AddEntry<eUniformBuffer>(template_entries, offset, binding, entries.const_buffers); - AddEntry<eStorageBuffer>(template_entries, offset, binding, entries.global_buffers); - AddEntry<eUniformTexelBuffer>(template_entries, offset, binding, entries.texel_buffers); - AddEntry<eCombinedImageSampler>(template_entries, offset, binding, entries.samplers); - AddEntry<eStorageImage>(template_entries, offset, binding, entries.images); + std::vector<VkDescriptorUpdateTemplateEntryKHR>& template_entries) { + AddEntry<UNIFORM_BUFFER>(template_entries, offset, binding, entries.const_buffers); + AddEntry<STORAGE_BUFFER>(template_entries, offset, binding, entries.global_buffers); + AddEntry<UNIFORM_TEXEL_BUFFER>(template_entries, offset, binding, entries.texel_buffers); + AddEntry<COMBINED_IMAGE_SAMPLER>(template_entries, offset, binding, entries.samplers); + AddEntry<STORAGE_IMAGE>(template_entries, offset, binding, entries.images); } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h index c4c112290..7ccdb7083 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h @@ -19,12 +19,12 @@ #include "video_core/engines/const_buffer_engine_interface.h" #include "video_core/engines/maxwell_3d.h" #include "video_core/rasterizer_cache.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h" #include "video_core/renderer_vulkan/vk_graphics_pipeline.h" #include "video_core/renderer_vulkan/vk_renderpass_cache.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_shader_decompiler.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/shader/registry.h" #include "video_core/shader/shader_ir.h" #include "video_core/surface.h" @@ -113,17 +113,13 @@ namespace Vulkan { class CachedShader final : public RasterizerCacheObject { public: explicit CachedShader(Core::System& system, Tegra::Engines::ShaderType stage, GPUVAddr gpu_addr, - VAddr cpu_addr, u8* host_ptr, ProgramCode program_code, u32 main_offset); + VAddr cpu_addr, ProgramCode program_code, u32 main_offset); ~CachedShader(); GPUVAddr GetGpuAddr() const { return gpu_addr; } - VAddr GetCpuAddr() const override { - return cpu_addr; - } - std::size_t GetSizeInBytes() const override { return program_code.size() * sizeof(u64); } @@ -149,7 +145,6 @@ private: Tegra::Engines::ShaderType stage); GPUVAddr gpu_addr{}; - VAddr cpu_addr{}; ProgramCode program_code; VideoCommon::Shader::Registry registry; VideoCommon::Shader::ShaderIR shader_ir; @@ -177,7 +172,7 @@ protected: void FlushObjectInner(const Shader& object) override {} private: - std::pair<SPIRVProgram, std::vector<vk::DescriptorSetLayoutBinding>> DecompileShaders( + std::pair<SPIRVProgram, std::vector<VkDescriptorSetLayoutBinding>> DecompileShaders( const GraphicsPipelineCacheKey& key); Core::System& system; @@ -199,6 +194,6 @@ private: void FillDescriptorUpdateTemplateEntries( const ShaderEntries& entries, u32& binding, u32& offset, - std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries); + std::vector<VkDescriptorUpdateTemplateEntryKHR>& template_entries); } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_query_cache.cpp b/src/video_core/renderer_vulkan/vk_query_cache.cpp index ffbf60dda..0966c7ff7 100644 --- a/src/video_core/renderer_vulkan/vk_query_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_query_cache.cpp @@ -8,19 +8,19 @@ #include <utility> #include <vector> -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_query_cache.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_scheduler.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { namespace { -constexpr std::array QUERY_TARGETS = {vk::QueryType::eOcclusion}; +constexpr std::array QUERY_TARGETS = {VK_QUERY_TYPE_OCCLUSION}; -constexpr vk::QueryType GetTarget(VideoCore::QueryType type) { +constexpr VkQueryType GetTarget(VideoCore::QueryType type) { return QUERY_TARGETS[static_cast<std::size_t>(type)]; } @@ -35,29 +35,34 @@ void QueryPool::Initialize(const VKDevice& device_, VideoCore::QueryType type_) type = type_; } -std::pair<vk::QueryPool, std::uint32_t> QueryPool::Commit(VKFence& fence) { +std::pair<VkQueryPool, u32> QueryPool::Commit(VKFence& fence) { std::size_t index; do { index = CommitResource(fence); } while (usage[index]); usage[index] = true; - return {*pools[index / GROW_STEP], static_cast<std::uint32_t>(index % GROW_STEP)}; + return {*pools[index / GROW_STEP], static_cast<u32>(index % GROW_STEP)}; } void QueryPool::Allocate(std::size_t begin, std::size_t end) { usage.resize(end); - const auto dev = device->GetLogical(); - const u32 size = static_cast<u32>(end - begin); - const vk::QueryPoolCreateInfo query_pool_ci({}, GetTarget(type), size, {}); - pools.push_back(dev.createQueryPoolUnique(query_pool_ci, nullptr, device->GetDispatchLoader())); + VkQueryPoolCreateInfo query_pool_ci; + query_pool_ci.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO; + query_pool_ci.pNext = nullptr; + query_pool_ci.flags = 0; + query_pool_ci.queryType = GetTarget(type); + query_pool_ci.queryCount = static_cast<u32>(end - begin); + query_pool_ci.pipelineStatistics = 0; + pools.push_back(device->GetLogical().CreateQueryPool(query_pool_ci)); } -void QueryPool::Reserve(std::pair<vk::QueryPool, std::uint32_t> query) { +void QueryPool::Reserve(std::pair<VkQueryPool, u32> query) { const auto it = - std::find_if(std::begin(pools), std::end(pools), - [query_pool = query.first](auto& pool) { return query_pool == *pool; }); + std::find_if(pools.begin(), pools.end(), [query_pool = query.first](vk::QueryPool& pool) { + return query_pool == *pool; + }); ASSERT(it != std::end(pools)); const std::ptrdiff_t pool_index = std::distance(std::begin(pools), it); @@ -76,12 +81,11 @@ VKQueryCache::VKQueryCache(Core::System& system, VideoCore::RasterizerInterface& VKQueryCache::~VKQueryCache() = default; -std::pair<vk::QueryPool, std::uint32_t> VKQueryCache::AllocateQuery(VideoCore::QueryType type) { +std::pair<VkQueryPool, u32> VKQueryCache::AllocateQuery(VideoCore::QueryType type) { return query_pools[static_cast<std::size_t>(type)].Commit(scheduler.GetFence()); } -void VKQueryCache::Reserve(VideoCore::QueryType type, - std::pair<vk::QueryPool, std::uint32_t> query) { +void VKQueryCache::Reserve(VideoCore::QueryType type, std::pair<VkQueryPool, u32> query) { query_pools[static_cast<std::size_t>(type)].Reserve(query); } @@ -89,10 +93,10 @@ HostCounter::HostCounter(VKQueryCache& cache, std::shared_ptr<HostCounter> depen VideoCore::QueryType type) : VideoCommon::HostCounterBase<VKQueryCache, HostCounter>{std::move(dependency)}, cache{cache}, type{type}, query{cache.AllocateQuery(type)}, ticks{cache.Scheduler().Ticks()} { - const auto dev = cache.Device().GetLogical(); - cache.Scheduler().Record([dev, query = query](vk::CommandBuffer cmdbuf, auto& dld) { - dev.resetQueryPoolEXT(query.first, query.second, 1, dld); - cmdbuf.beginQuery(query.first, query.second, vk::QueryControlFlagBits::ePrecise, dld); + const vk::Device* logical = &cache.Device().GetLogical(); + cache.Scheduler().Record([logical, query = query](vk::CommandBuffer cmdbuf) { + logical->ResetQueryPoolEXT(query.first, query.second, 1); + cmdbuf.BeginQuery(query.first, query.second, VK_QUERY_CONTROL_PRECISE_BIT); }); } @@ -101,22 +105,16 @@ HostCounter::~HostCounter() { } void HostCounter::EndQuery() { - cache.Scheduler().Record([query = query](auto cmdbuf, auto& dld) { - cmdbuf.endQuery(query.first, query.second, dld); - }); + cache.Scheduler().Record( + [query = query](vk::CommandBuffer cmdbuf) { cmdbuf.EndQuery(query.first, query.second); }); } u64 HostCounter::BlockingQuery() const { if (ticks >= cache.Scheduler().Ticks()) { cache.Scheduler().Flush(); } - - const auto dev = cache.Device().GetLogical(); - const auto& dld = cache.Device().GetDispatchLoader(); - u64 value; - dev.getQueryPoolResults(query.first, query.second, 1, sizeof(value), &value, sizeof(value), - vk::QueryResultFlagBits::e64 | vk::QueryResultFlagBits::eWait, dld); - return value; + return cache.Device().GetLogical().GetQueryResult<u64>( + query.first, query.second, VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT); } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_query_cache.h b/src/video_core/renderer_vulkan/vk_query_cache.h index c3092ee96..b63784f4b 100644 --- a/src/video_core/renderer_vulkan/vk_query_cache.h +++ b/src/video_core/renderer_vulkan/vk_query_cache.h @@ -12,8 +12,8 @@ #include "common/common_types.h" #include "video_core/query_cache.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace VideoCore { class RasterizerInterface; @@ -36,9 +36,9 @@ public: void Initialize(const VKDevice& device, VideoCore::QueryType type); - std::pair<vk::QueryPool, std::uint32_t> Commit(VKFence& fence); + std::pair<VkQueryPool, u32> Commit(VKFence& fence); - void Reserve(std::pair<vk::QueryPool, std::uint32_t> query); + void Reserve(std::pair<VkQueryPool, u32> query); protected: void Allocate(std::size_t begin, std::size_t end) override; @@ -49,7 +49,7 @@ private: const VKDevice* device = nullptr; VideoCore::QueryType type = {}; - std::vector<UniqueQueryPool> pools; + std::vector<vk::QueryPool> pools; std::vector<bool> usage; }; @@ -61,9 +61,9 @@ public: const VKDevice& device, VKScheduler& scheduler); ~VKQueryCache(); - std::pair<vk::QueryPool, std::uint32_t> AllocateQuery(VideoCore::QueryType type); + std::pair<VkQueryPool, u32> AllocateQuery(VideoCore::QueryType type); - void Reserve(VideoCore::QueryType type, std::pair<vk::QueryPool, std::uint32_t> query); + void Reserve(VideoCore::QueryType type, std::pair<VkQueryPool, u32> query); const VKDevice& Device() const noexcept { return device; @@ -91,7 +91,7 @@ private: VKQueryCache& cache; const VideoCore::QueryType type; - const std::pair<vk::QueryPool, std::uint32_t> query; + const std::pair<VkQueryPool, u32> query; const u64 ticks; }; diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index 58c69b786..774ba1f26 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -19,7 +19,6 @@ #include "core/memory.h" #include "video_core/engines/kepler_compute.h" #include "video_core/engines/maxwell_3d.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/renderer_vulkan.h" @@ -39,6 +38,7 @@ #include "video_core/renderer_vulkan/vk_state_tracker.h" #include "video_core/renderer_vulkan/vk_texture_cache.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -60,32 +60,42 @@ namespace { constexpr auto ComputeShaderIndex = static_cast<std::size_t>(Tegra::Engines::ShaderType::Compute); -vk::Viewport GetViewportState(const VKDevice& device, const Maxwell& regs, std::size_t index) { - const auto& viewport = regs.viewport_transform[index]; - const float x = viewport.translate_x - viewport.scale_x; - const float y = viewport.translate_y - viewport.scale_y; - const float width = viewport.scale_x * 2.0f; - const float height = viewport.scale_y * 2.0f; +VkViewport GetViewportState(const VKDevice& device, const Maxwell& regs, std::size_t index) { + const auto& src = regs.viewport_transform[index]; + const float width = src.scale_x * 2.0f; + const float height = src.scale_y * 2.0f; - const float reduce_z = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne; - float near = viewport.translate_z - viewport.scale_z * reduce_z; - float far = viewport.translate_z + viewport.scale_z; + VkViewport viewport; + viewport.x = src.translate_x - src.scale_x; + viewport.y = src.translate_y - src.scale_y; + viewport.width = width != 0.0f ? width : 1.0f; + viewport.height = height != 0.0f ? height : 1.0f; + + const float reduce_z = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne ? 1.0f : 0.0f; + viewport.minDepth = src.translate_z - src.scale_z * reduce_z; + viewport.maxDepth = src.translate_z + src.scale_z; if (!device.IsExtDepthRangeUnrestrictedSupported()) { - near = std::clamp(near, 0.0f, 1.0f); - far = std::clamp(far, 0.0f, 1.0f); + viewport.minDepth = std::clamp(viewport.minDepth, 0.0f, 1.0f); + viewport.maxDepth = std::clamp(viewport.maxDepth, 0.0f, 1.0f); } - - return vk::Viewport(x, y, width != 0 ? width : 1.0f, height != 0 ? height : 1.0f, near, far); + return viewport; } -constexpr vk::Rect2D GetScissorState(const Maxwell& regs, std::size_t index) { - const auto& scissor = regs.scissor_test[index]; - if (!scissor.enable) { - return {{0, 0}, {INT32_MAX, INT32_MAX}}; +VkRect2D GetScissorState(const Maxwell& regs, std::size_t index) { + const auto& src = regs.scissor_test[index]; + VkRect2D scissor; + if (src.enable) { + scissor.offset.x = static_cast<s32>(src.min_x); + scissor.offset.y = static_cast<s32>(src.min_y); + scissor.extent.width = src.max_x - src.min_x; + scissor.extent.height = src.max_y - src.min_y; + } else { + scissor.offset.x = 0; + scissor.offset.y = 0; + scissor.extent.width = std::numeric_limits<s32>::max(); + scissor.extent.height = std::numeric_limits<s32>::max(); } - const u32 width = scissor.max_x - scissor.min_x; - const u32 height = scissor.max_y - scissor.min_y; - return {{static_cast<s32>(scissor.min_x), static_cast<s32>(scissor.min_y)}, {width, height}}; + return scissor; } std::array<GPUVAddr, Maxwell::MaxShaderProgram> GetShaderAddresses( @@ -97,8 +107,8 @@ std::array<GPUVAddr, Maxwell::MaxShaderProgram> GetShaderAddresses( return addresses; } -void TransitionImages(const std::vector<ImageView>& views, vk::PipelineStageFlags pipeline_stage, - vk::AccessFlags access) { +void TransitionImages(const std::vector<ImageView>& views, VkPipelineStageFlags pipeline_stage, + VkAccessFlags access) { for (auto& [view, layout] : views) { view->Transition(*layout, pipeline_stage, access); } @@ -127,13 +137,13 @@ Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry class BufferBindings final { public: - void AddVertexBinding(const vk::Buffer* buffer, vk::DeviceSize offset) { + void AddVertexBinding(const VkBuffer* buffer, VkDeviceSize offset) { vertex.buffer_ptrs[vertex.num_buffers] = buffer; vertex.offsets[vertex.num_buffers] = offset; ++vertex.num_buffers; } - void SetIndexBinding(const vk::Buffer* buffer, vk::DeviceSize offset, vk::IndexType type) { + void SetIndexBinding(const VkBuffer* buffer, VkDeviceSize offset, VkIndexType type) { index.buffer = buffer; index.offset = offset; index.type = type; @@ -217,14 +227,14 @@ private: // Some of these fields are intentionally left uninitialized to avoid initializing them twice. struct { std::size_t num_buffers = 0; - std::array<const vk::Buffer*, Maxwell::NumVertexArrays> buffer_ptrs; - std::array<vk::DeviceSize, Maxwell::NumVertexArrays> offsets; + std::array<const VkBuffer*, Maxwell::NumVertexArrays> buffer_ptrs; + std::array<VkDeviceSize, Maxwell::NumVertexArrays> offsets; } vertex; struct { - const vk::Buffer* buffer = nullptr; - vk::DeviceSize offset; - vk::IndexType type; + const VkBuffer* buffer = nullptr; + VkDeviceSize offset; + VkIndexType type; } index; template <std::size_t N> @@ -243,38 +253,35 @@ private: return; } - std::array<vk::Buffer, N> buffers; + std::array<VkBuffer, N> buffers; std::transform(vertex.buffer_ptrs.begin(), vertex.buffer_ptrs.begin() + N, buffers.begin(), [](const auto ptr) { return *ptr; }); - std::array<vk::DeviceSize, N> offsets; + std::array<VkDeviceSize, N> offsets; std::copy(vertex.offsets.begin(), vertex.offsets.begin() + N, offsets.begin()); if constexpr (is_indexed) { // Indexed draw scheduler.Record([buffers, offsets, index_buffer = *index.buffer, index_offset = index.offset, - index_type = index.type](auto cmdbuf, auto& dld) { - cmdbuf.bindIndexBuffer(index_buffer, index_offset, index_type, dld); - cmdbuf.bindVertexBuffers(0, static_cast<u32>(N), buffers.data(), offsets.data(), - dld); + index_type = index.type](vk::CommandBuffer cmdbuf) { + cmdbuf.BindIndexBuffer(index_buffer, index_offset, index_type); + cmdbuf.BindVertexBuffers(0, static_cast<u32>(N), buffers.data(), offsets.data()); }); } else { // Array draw - scheduler.Record([buffers, offsets](auto cmdbuf, auto& dld) { - cmdbuf.bindVertexBuffers(0, static_cast<u32>(N), buffers.data(), offsets.data(), - dld); + scheduler.Record([buffers, offsets](vk::CommandBuffer cmdbuf) { + cmdbuf.BindVertexBuffers(0, static_cast<u32>(N), buffers.data(), offsets.data()); }); } } }; -void RasterizerVulkan::DrawParameters::Draw(vk::CommandBuffer cmdbuf, - const vk::DispatchLoaderDynamic& dld) const { +void RasterizerVulkan::DrawParameters::Draw(vk::CommandBuffer cmdbuf) const { if (is_indexed) { - cmdbuf.drawIndexed(num_vertices, num_instances, 0, base_vertex, base_instance, dld); + cmdbuf.DrawIndexed(num_vertices, num_instances, 0, base_vertex, base_instance); } else { - cmdbuf.draw(num_vertices, num_instances, base_vertex, base_instance, dld); + cmdbuf.Draw(num_vertices, num_instances, base_vertex, base_instance); } } @@ -337,7 +344,7 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) { const auto renderpass = pipeline.GetRenderPass(); const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass); - scheduler.RequestRenderpass({renderpass, framebuffer, {{0, 0}, render_area}, 0, nullptr}); + scheduler.RequestRenderpass(renderpass, framebuffer, render_area); UpdateDynamicStates(); @@ -345,19 +352,19 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) { if (device.IsNvDeviceDiagnosticCheckpoints()) { scheduler.Record( - [&pipeline](auto cmdbuf, auto& dld) { cmdbuf.setCheckpointNV(&pipeline, dld); }); + [&pipeline](vk::CommandBuffer cmdbuf) { cmdbuf.SetCheckpointNV(&pipeline); }); } BeginTransformFeedback(); const auto pipeline_layout = pipeline.GetLayout(); const auto descriptor_set = pipeline.CommitDescriptorSet(); - scheduler.Record([pipeline_layout, descriptor_set, draw_params](auto cmdbuf, auto& dld) { + scheduler.Record([pipeline_layout, descriptor_set, draw_params](vk::CommandBuffer cmdbuf) { if (descriptor_set) { - cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eGraphics, pipeline_layout, - DESCRIPTOR_SET, 1, &descriptor_set, 0, nullptr, dld); + cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline_layout, + DESCRIPTOR_SET, descriptor_set, {}); } - draw_params.Draw(cmdbuf, dld); + draw_params.Draw(cmdbuf); }); EndTransformFeedback(); @@ -389,48 +396,54 @@ void RasterizerVulkan::Clear() { DEBUG_ASSERT(texceptions.none()); SetupImageTransitions(0, color_attachments, zeta_attachment); - const vk::RenderPass renderpass = renderpass_cache.GetRenderPass(GetRenderPassParams(0)); + const VkRenderPass renderpass = renderpass_cache.GetRenderPass(GetRenderPassParams(0)); const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass); - scheduler.RequestRenderpass({renderpass, framebuffer, {{0, 0}, render_area}, 0, nullptr}); - - const auto& scissor = regs.scissor_test[0]; - const vk::Offset2D scissor_offset(scissor.min_x, scissor.min_y); - vk::Extent2D scissor_extent{scissor.max_x - scissor.min_x, scissor.max_y - scissor.min_y}; - scissor_extent.width = std::min(scissor_extent.width, render_area.width); - scissor_extent.height = std::min(scissor_extent.height, render_area.height); + scheduler.RequestRenderpass(renderpass, framebuffer, render_area); - const u32 layer = regs.clear_buffers.layer; - const vk::ClearRect clear_rect({scissor_offset, scissor_extent}, layer, 1); + VkClearRect clear_rect; + clear_rect.baseArrayLayer = regs.clear_buffers.layer; + clear_rect.layerCount = 1; + clear_rect.rect = GetScissorState(regs, 0); + clear_rect.rect.extent.width = std::min(clear_rect.rect.extent.width, render_area.width); + clear_rect.rect.extent.height = std::min(clear_rect.rect.extent.height, render_area.height); if (use_color) { - const std::array clear_color = {regs.clear_color[0], regs.clear_color[1], - regs.clear_color[2], regs.clear_color[3]}; - const vk::ClearValue clear_value{clear_color}; + VkClearValue clear_value; + std::memcpy(clear_value.color.float32, regs.clear_color, sizeof(regs.clear_color)); + const u32 color_attachment = regs.clear_buffers.RT; - scheduler.Record([color_attachment, clear_value, clear_rect](auto cmdbuf, auto& dld) { - const vk::ClearAttachment attachment(vk::ImageAspectFlagBits::eColor, color_attachment, - clear_value); - cmdbuf.clearAttachments(1, &attachment, 1, &clear_rect, dld); + scheduler.Record([color_attachment, clear_value, clear_rect](vk::CommandBuffer cmdbuf) { + VkClearAttachment attachment; + attachment.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; + attachment.colorAttachment = color_attachment; + attachment.clearValue = clear_value; + cmdbuf.ClearAttachments(attachment, clear_rect); }); } if (!use_depth && !use_stencil) { return; } - vk::ImageAspectFlags aspect_flags; + VkImageAspectFlags aspect_flags = 0; if (use_depth) { - aspect_flags |= vk::ImageAspectFlagBits::eDepth; + aspect_flags |= VK_IMAGE_ASPECT_DEPTH_BIT; } if (use_stencil) { - aspect_flags |= vk::ImageAspectFlagBits::eStencil; + aspect_flags |= VK_IMAGE_ASPECT_STENCIL_BIT; } scheduler.Record([clear_depth = regs.clear_depth, clear_stencil = regs.clear_stencil, - clear_rect, aspect_flags](auto cmdbuf, auto& dld) { - const vk::ClearDepthStencilValue clear_zeta(clear_depth, clear_stencil); - const vk::ClearValue clear_value{clear_zeta}; - const vk::ClearAttachment attachment(aspect_flags, 0, clear_value); - cmdbuf.clearAttachments(1, &attachment, 1, &clear_rect, dld); + clear_rect, aspect_flags](vk::CommandBuffer cmdbuf) { + VkClearValue clear_value; + clear_value.depthStencil.depth = clear_depth; + clear_value.depthStencil.stencil = clear_stencil; + + VkClearAttachment attachment; + attachment.aspectMask = aspect_flags; + attachment.colorAttachment = 0; + attachment.clearValue.depthStencil.depth = clear_depth; + attachment.clearValue.depthStencil.stencil = clear_stencil; + cmdbuf.ClearAttachments(attachment, clear_rect); }); } @@ -463,24 +476,24 @@ void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) { buffer_cache.Unmap(); - TransitionImages(sampled_views, vk::PipelineStageFlagBits::eComputeShader, - vk::AccessFlagBits::eShaderRead); - TransitionImages(image_views, vk::PipelineStageFlagBits::eComputeShader, - vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite); + TransitionImages(sampled_views, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, + VK_ACCESS_SHADER_READ_BIT); + TransitionImages(image_views, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, + VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT); if (device.IsNvDeviceDiagnosticCheckpoints()) { scheduler.Record( - [&pipeline](auto cmdbuf, auto& dld) { cmdbuf.setCheckpointNV(nullptr, dld); }); + [&pipeline](vk::CommandBuffer cmdbuf) { cmdbuf.SetCheckpointNV(nullptr); }); } scheduler.Record([grid_x = launch_desc.grid_dim_x, grid_y = launch_desc.grid_dim_y, grid_z = launch_desc.grid_dim_z, pipeline_handle = pipeline.GetHandle(), layout = pipeline.GetLayout(), - descriptor_set = pipeline.CommitDescriptorSet()](auto cmdbuf, auto& dld) { - cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline_handle, dld); - cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, DESCRIPTOR_SET, 1, - &descriptor_set, 0, nullptr, dld); - cmdbuf.dispatch(grid_x, grid_y, grid_z, dld); + descriptor_set = pipeline.CommitDescriptorSet()](vk::CommandBuffer cmdbuf) { + cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline_handle); + cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, DESCRIPTOR_SET, + descriptor_set, {}); + cmdbuf.Dispatch(grid_x, grid_y, grid_z); }); } @@ -495,20 +508,26 @@ void RasterizerVulkan::Query(GPUVAddr gpu_addr, VideoCore::QueryType type, void RasterizerVulkan::FlushAll() {} -void RasterizerVulkan::FlushRegion(CacheAddr addr, u64 size) { +void RasterizerVulkan::FlushRegion(VAddr addr, u64 size) { + if (addr == 0 || size == 0) { + return; + } texture_cache.FlushRegion(addr, size); buffer_cache.FlushRegion(addr, size); query_cache.FlushRegion(addr, size); } -void RasterizerVulkan::InvalidateRegion(CacheAddr addr, u64 size) { +void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size) { + if (addr == 0 || size == 0) { + return; + } texture_cache.InvalidateRegion(addr, size); pipeline_cache.InvalidateRegion(addr, size); buffer_cache.InvalidateRegion(addr, size); query_cache.InvalidateRegion(addr, size); } -void RasterizerVulkan::FlushAndInvalidateRegion(CacheAddr addr, u64 size) { +void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size) { FlushRegion(addr, size); InvalidateRegion(addr, size); } @@ -540,8 +559,7 @@ bool RasterizerVulkan::AccelerateDisplay(const Tegra::FramebufferConfig& config, return false; } - const u8* host_ptr{system.Memory().GetPointer(framebuffer_addr)}; - const auto surface{texture_cache.TryFindFramebufferSurface(host_ptr)}; + const auto surface{texture_cache.TryFindFramebufferSurface(framebuffer_addr)}; if (!surface) { return false; } @@ -594,7 +612,7 @@ RasterizerVulkan::Texceptions RasterizerVulkan::UpdateAttachments() { Texceptions texceptions; for (std::size_t rt = 0; rt < Maxwell::NumRenderTargets; ++rt) { if (update_rendertargets) { - color_attachments[rt] = texture_cache.GetColorBufferSurface(rt, true); + color_attachments[rt] = texture_cache.GetColorBufferSurface(rt); } if (color_attachments[rt] && WalkAttachmentOverlaps(*color_attachments[rt])) { texceptions[rt] = true; @@ -602,7 +620,7 @@ RasterizerVulkan::Texceptions RasterizerVulkan::UpdateAttachments() { } if (update_rendertargets) { - zeta_attachment = texture_cache.GetDepthBufferSurface(true); + zeta_attachment = texture_cache.GetDepthBufferSurface(); } if (zeta_attachment && WalkAttachmentOverlaps(*zeta_attachment)) { texceptions[ZETA_TEXCEPTION_INDEX] = true; @@ -620,13 +638,13 @@ bool RasterizerVulkan::WalkAttachmentOverlaps(const CachedSurfaceView& attachmen continue; } overlap = true; - *layout = vk::ImageLayout::eGeneral; + *layout = VK_IMAGE_LAYOUT_GENERAL; } return overlap; } -std::tuple<vk::Framebuffer, vk::Extent2D> RasterizerVulkan::ConfigureFramebuffers( - vk::RenderPass renderpass) { +std::tuple<VkFramebuffer, VkExtent2D> RasterizerVulkan::ConfigureFramebuffers( + VkRenderPass renderpass) { FramebufferCacheKey key{renderpass, std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max()}; @@ -653,15 +671,20 @@ std::tuple<vk::Framebuffer, vk::Extent2D> RasterizerVulkan::ConfigureFramebuffer const auto [fbentry, is_cache_miss] = framebuffer_cache.try_emplace(key); auto& framebuffer = fbentry->second; if (is_cache_miss) { - const vk::FramebufferCreateInfo framebuffer_ci( - {}, key.renderpass, static_cast<u32>(key.views.size()), key.views.data(), key.width, - key.height, key.layers); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - framebuffer = dev.createFramebufferUnique(framebuffer_ci, nullptr, dld); - } - - return {*framebuffer, vk::Extent2D{key.width, key.height}}; + VkFramebufferCreateInfo framebuffer_ci; + framebuffer_ci.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; + framebuffer_ci.pNext = nullptr; + framebuffer_ci.flags = 0; + framebuffer_ci.renderPass = key.renderpass; + framebuffer_ci.attachmentCount = static_cast<u32>(key.views.size()); + framebuffer_ci.pAttachments = key.views.data(); + framebuffer_ci.width = key.width; + framebuffer_ci.height = key.height; + framebuffer_ci.layers = key.layers; + framebuffer = device.GetLogical().CreateFramebuffer(framebuffer_ci); + } + + return {*framebuffer, VkExtent2D{key.width, key.height}}; } RasterizerVulkan::DrawParameters RasterizerVulkan::SetupGeometry(FixedPipelineState& fixed_state, @@ -709,10 +732,9 @@ void RasterizerVulkan::SetupShaderDescriptors( void RasterizerVulkan::SetupImageTransitions( Texceptions texceptions, const std::array<View, Maxwell::NumRenderTargets>& color_attachments, const View& zeta_attachment) { - TransitionImages(sampled_views, vk::PipelineStageFlagBits::eAllGraphics, - vk::AccessFlagBits::eShaderRead); - TransitionImages(image_views, vk::PipelineStageFlagBits::eAllGraphics, - vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite); + TransitionImages(sampled_views, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, VK_ACCESS_SHADER_READ_BIT); + TransitionImages(image_views, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, + VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT); for (std::size_t rt = 0; rt < std::size(color_attachments); ++rt) { const auto color_attachment = color_attachments[rt]; @@ -720,19 +742,19 @@ void RasterizerVulkan::SetupImageTransitions( continue; } const auto image_layout = - texceptions[rt] ? vk::ImageLayout::eGeneral : vk::ImageLayout::eColorAttachmentOptimal; - color_attachment->Transition( - image_layout, vk::PipelineStageFlagBits::eColorAttachmentOutput, - vk::AccessFlagBits::eColorAttachmentRead | vk::AccessFlagBits::eColorAttachmentWrite); + texceptions[rt] ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; + color_attachment->Transition(image_layout, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, + VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | + VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT); } if (zeta_attachment != nullptr) { const auto image_layout = texceptions[ZETA_TEXCEPTION_INDEX] - ? vk::ImageLayout::eGeneral - : vk::ImageLayout::eDepthStencilAttachmentOptimal; - zeta_attachment->Transition(image_layout, vk::PipelineStageFlagBits::eLateFragmentTests, - vk::AccessFlagBits::eDepthStencilAttachmentRead | - vk::AccessFlagBits::eDepthStencilAttachmentWrite); + ? VK_IMAGE_LAYOUT_GENERAL + : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; + zeta_attachment->Transition(image_layout, VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT, + VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | + VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT); } } @@ -768,9 +790,9 @@ void RasterizerVulkan::BeginTransformFeedback() { const std::size_t size = binding.buffer_size; const auto [buffer, offset] = buffer_cache.UploadMemory(gpu_addr, size, 4, true); - scheduler.Record([buffer = *buffer, offset = offset, size](auto cmdbuf, auto& dld) { - cmdbuf.bindTransformFeedbackBuffersEXT(0, {buffer}, {offset}, {size}, dld); - cmdbuf.beginTransformFeedbackEXT(0, {}, {}, dld); + scheduler.Record([buffer = *buffer, offset = offset, size](vk::CommandBuffer cmdbuf) { + cmdbuf.BindTransformFeedbackBuffersEXT(0, 1, &buffer, &offset, &size); + cmdbuf.BeginTransformFeedbackEXT(0, 0, nullptr, nullptr); }); } @@ -781,7 +803,7 @@ void RasterizerVulkan::EndTransformFeedback() { } scheduler.Record( - [](auto cmdbuf, auto& dld) { cmdbuf.endTransformFeedbackEXT(0, {}, {}, dld); }); + [](vk::CommandBuffer cmdbuf) { cmdbuf.EndTransformFeedbackEXT(0, 0, nullptr, nullptr); }); } void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex_input, @@ -832,7 +854,7 @@ void RasterizerVulkan::SetupIndexBuffer(BufferBindings& buffer_bindings, DrawPar } else { const auto [buffer, offset] = quad_array_pass.Assemble(params.num_vertices, params.base_vertex); - buffer_bindings.SetIndexBinding(&buffer, offset, vk::IndexType::eUint32); + buffer_bindings.SetIndexBinding(buffer, offset, VK_INDEX_TYPE_UINT32); params.base_vertex = 0; params.num_vertices = params.num_vertices * 6 / 4; params.is_indexed = true; @@ -1017,7 +1039,7 @@ void RasterizerVulkan::SetupTexture(const Tegra::Texture::FullTextureInfo& textu update_descriptor_queue.AddSampledImage(sampler, image_view); const auto image_layout = update_descriptor_queue.GetLastImageLayout(); - *image_layout = vk::ImageLayout::eShaderReadOnlyOptimal; + *image_layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; sampled_views.push_back(ImageView{std::move(view), image_layout}); } @@ -1034,7 +1056,7 @@ void RasterizerVulkan::SetupImage(const Tegra::Texture::TICEntry& tic, const Ima update_descriptor_queue.AddImage(image_view); const auto image_layout = update_descriptor_queue.GetLastImageLayout(); - *image_layout = vk::ImageLayout::eGeneral; + *image_layout = VK_IMAGE_LAYOUT_GENERAL; image_views.push_back(ImageView{std::move(view), image_layout}); } @@ -1051,9 +1073,7 @@ void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& reg GetViewportState(device, regs, 10), GetViewportState(device, regs, 11), GetViewportState(device, regs, 12), GetViewportState(device, regs, 13), GetViewportState(device, regs, 14), GetViewportState(device, regs, 15)}; - scheduler.Record([viewports](auto cmdbuf, auto& dld) { - cmdbuf.setViewport(0, static_cast<u32>(viewports.size()), viewports.data(), dld); - }); + scheduler.Record([viewports](vk::CommandBuffer cmdbuf) { cmdbuf.SetViewport(0, viewports); }); } void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs) { @@ -1067,9 +1087,7 @@ void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs GetScissorState(regs, 9), GetScissorState(regs, 10), GetScissorState(regs, 11), GetScissorState(regs, 12), GetScissorState(regs, 13), GetScissorState(regs, 14), GetScissorState(regs, 15)}; - scheduler.Record([scissors](auto cmdbuf, auto& dld) { - cmdbuf.setScissor(0, static_cast<u32>(scissors.size()), scissors.data(), dld); - }); + scheduler.Record([scissors](vk::CommandBuffer cmdbuf) { cmdbuf.SetScissor(0, scissors); }); } void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) { @@ -1077,8 +1095,8 @@ void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) { return; } scheduler.Record([constant = regs.polygon_offset_units, clamp = regs.polygon_offset_clamp, - factor = regs.polygon_offset_factor](auto cmdbuf, auto& dld) { - cmdbuf.setDepthBias(constant, clamp, factor / 2.0f, dld); + factor = regs.polygon_offset_factor](vk::CommandBuffer cmdbuf) { + cmdbuf.SetDepthBias(constant, clamp, factor / 2.0f); }); } @@ -1088,9 +1106,8 @@ void RasterizerVulkan::UpdateBlendConstants(Tegra::Engines::Maxwell3D::Regs& reg } const std::array blend_color = {regs.blend_color.r, regs.blend_color.g, regs.blend_color.b, regs.blend_color.a}; - scheduler.Record([blend_color](auto cmdbuf, auto& dld) { - cmdbuf.setBlendConstants(blend_color.data(), dld); - }); + scheduler.Record( + [blend_color](vk::CommandBuffer cmdbuf) { cmdbuf.SetBlendConstants(blend_color.data()); }); } void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs) { @@ -1098,7 +1115,7 @@ void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs) return; } scheduler.Record([min = regs.depth_bounds[0], max = regs.depth_bounds[1]]( - auto cmdbuf, auto& dld) { cmdbuf.setDepthBounds(min, max, dld); }); + vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBounds(min, max); }); } void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) { @@ -1111,24 +1128,24 @@ void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) [front_ref = regs.stencil_front_func_ref, front_write_mask = regs.stencil_front_mask, front_test_mask = regs.stencil_front_func_mask, back_ref = regs.stencil_back_func_ref, back_write_mask = regs.stencil_back_mask, - back_test_mask = regs.stencil_back_func_mask](auto cmdbuf, auto& dld) { + back_test_mask = regs.stencil_back_func_mask](vk::CommandBuffer cmdbuf) { // Front face - cmdbuf.setStencilReference(vk::StencilFaceFlagBits::eFront, front_ref, dld); - cmdbuf.setStencilWriteMask(vk::StencilFaceFlagBits::eFront, front_write_mask, dld); - cmdbuf.setStencilCompareMask(vk::StencilFaceFlagBits::eFront, front_test_mask, dld); + cmdbuf.SetStencilReference(VK_STENCIL_FACE_FRONT_BIT, front_ref); + cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_FRONT_BIT, front_write_mask); + cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_FRONT_BIT, front_test_mask); // Back face - cmdbuf.setStencilReference(vk::StencilFaceFlagBits::eBack, back_ref, dld); - cmdbuf.setStencilWriteMask(vk::StencilFaceFlagBits::eBack, back_write_mask, dld); - cmdbuf.setStencilCompareMask(vk::StencilFaceFlagBits::eBack, back_test_mask, dld); + cmdbuf.SetStencilReference(VK_STENCIL_FACE_BACK_BIT, back_ref); + cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_BACK_BIT, back_write_mask); + cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_BACK_BIT, back_test_mask); }); } else { // Front face defines both faces scheduler.Record([ref = regs.stencil_back_func_ref, write_mask = regs.stencil_back_mask, - test_mask = regs.stencil_back_func_mask](auto cmdbuf, auto& dld) { - cmdbuf.setStencilReference(vk::StencilFaceFlagBits::eFrontAndBack, ref, dld); - cmdbuf.setStencilWriteMask(vk::StencilFaceFlagBits::eFrontAndBack, write_mask, dld); - cmdbuf.setStencilCompareMask(vk::StencilFaceFlagBits::eFrontAndBack, test_mask, dld); + test_mask = regs.stencil_back_func_mask](vk::CommandBuffer cmdbuf) { + cmdbuf.SetStencilReference(VK_STENCIL_FACE_FRONT_AND_BACK, ref); + cmdbuf.SetStencilWriteMask(VK_STENCIL_FACE_FRONT_AND_BACK, write_mask); + cmdbuf.SetStencilCompareMask(VK_STENCIL_FACE_FRONT_AND_BACK, test_mask); }); } } diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h index 3185868e9..46037860a 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.h +++ b/src/video_core/renderer_vulkan/vk_rasterizer.h @@ -17,7 +17,6 @@ #include "video_core/memory_manager.h" #include "video_core/rasterizer_accelerated.h" #include "video_core/rasterizer_interface.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/fixed_pipeline_state.h" #include "video_core/renderer_vulkan/vk_buffer_cache.h" #include "video_core/renderer_vulkan/vk_compute_pass.h" @@ -32,6 +31,7 @@ #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" #include "video_core/renderer_vulkan/vk_texture_cache.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Core { class System; @@ -49,11 +49,10 @@ namespace Vulkan { struct VKScreenInfo; -using ImageViewsPack = - boost::container::static_vector<vk::ImageView, Maxwell::NumRenderTargets + 1>; +using ImageViewsPack = boost::container::static_vector<VkImageView, Maxwell::NumRenderTargets + 1>; struct FramebufferCacheKey { - vk::RenderPass renderpass{}; + VkRenderPass renderpass{}; u32 width = 0; u32 height = 0; u32 layers = 0; @@ -101,7 +100,7 @@ class BufferBindings; struct ImageView { View view; - vk::ImageLayout* layout = nullptr; + VkImageLayout* layout = nullptr; }; class RasterizerVulkan final : public VideoCore::RasterizerAccelerated { @@ -118,9 +117,9 @@ public: void ResetCounter(VideoCore::QueryType type) override; void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override; void FlushAll() override; - void FlushRegion(CacheAddr addr, u64 size) override; - void InvalidateRegion(CacheAddr addr, u64 size) override; - void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override; + void FlushRegion(VAddr addr, u64 size) override; + void InvalidateRegion(VAddr addr, u64 size) override; + void FlushAndInvalidateRegion(VAddr addr, u64 size) override; void FlushCommands() override; void TickFrame() override; bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src, @@ -137,7 +136,7 @@ public: private: struct DrawParameters { - void Draw(vk::CommandBuffer cmdbuf, const vk::DispatchLoaderDynamic& dld) const; + void Draw(vk::CommandBuffer cmdbuf) const; u32 base_instance = 0; u32 num_instances = 0; @@ -154,7 +153,7 @@ private: Texceptions UpdateAttachments(); - std::tuple<vk::Framebuffer, vk::Extent2D> ConfigureFramebuffers(vk::RenderPass renderpass); + std::tuple<VkFramebuffer, VkExtent2D> ConfigureFramebuffers(VkRenderPass renderpass); /// Setups geometry buffers and state. DrawParameters SetupGeometry(FixedPipelineState& fixed_state, BufferBindings& buffer_bindings, @@ -272,7 +271,7 @@ private: u32 draw_counter = 0; // TODO(Rodrigo): Invalidate on image destruction - std::unordered_map<FramebufferCacheKey, UniqueFramebuffer> framebuffer_cache; + std::unordered_map<FramebufferCacheKey, vk::Framebuffer> framebuffer_cache; }; } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp b/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp index 93f5d7ba0..4e5286a69 100644 --- a/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp @@ -6,10 +6,10 @@ #include <vector> #include "video_core/engines/maxwell_3d.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_renderpass_cache.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -17,7 +17,7 @@ VKRenderPassCache::VKRenderPassCache(const VKDevice& device) : device{device} {} VKRenderPassCache::~VKRenderPassCache() = default; -vk::RenderPass VKRenderPassCache::GetRenderPass(const RenderPassParams& params) { +VkRenderPass VKRenderPassCache::GetRenderPass(const RenderPassParams& params) { const auto [pair, is_cache_miss] = cache.try_emplace(params); auto& entry = pair->second; if (is_cache_miss) { @@ -26,9 +26,9 @@ vk::RenderPass VKRenderPassCache::GetRenderPass(const RenderPassParams& params) return *entry; } -UniqueRenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& params) const { - std::vector<vk::AttachmentDescription> descriptors; - std::vector<vk::AttachmentReference> color_references; +vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& params) const { + std::vector<VkAttachmentDescription> descriptors; + std::vector<VkAttachmentReference> color_references; for (std::size_t rt = 0; rt < params.color_attachments.size(); ++rt) { const auto attachment = params.color_attachments[rt]; @@ -39,16 +39,25 @@ UniqueRenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& par // TODO(Rodrigo): Add eMayAlias when it's needed. const auto color_layout = attachment.is_texception - ? vk::ImageLayout::eGeneral - : vk::ImageLayout::eColorAttachmentOptimal; - descriptors.emplace_back(vk::AttachmentDescriptionFlagBits::eMayAlias, format.format, - vk::SampleCountFlagBits::e1, vk::AttachmentLoadOp::eLoad, - vk::AttachmentStoreOp::eStore, vk::AttachmentLoadOp::eDontCare, - vk::AttachmentStoreOp::eDontCare, color_layout, color_layout); - color_references.emplace_back(static_cast<u32>(rt), color_layout); + ? VK_IMAGE_LAYOUT_GENERAL + : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; + VkAttachmentDescription& descriptor = descriptors.emplace_back(); + descriptor.flags = VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT; + descriptor.format = format.format; + descriptor.samples = VK_SAMPLE_COUNT_1_BIT; + descriptor.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; + descriptor.storeOp = VK_ATTACHMENT_STORE_OP_STORE; + descriptor.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; + descriptor.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; + descriptor.initialLayout = color_layout; + descriptor.finalLayout = color_layout; + + VkAttachmentReference& reference = color_references.emplace_back(); + reference.attachment = static_cast<u32>(rt); + reference.layout = color_layout; } - vk::AttachmentReference zeta_attachment_ref; + VkAttachmentReference zeta_attachment_ref; if (params.has_zeta) { const auto format = MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, params.zeta_pixel_format); @@ -56,45 +65,68 @@ UniqueRenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& par static_cast<u32>(params.zeta_pixel_format)); const auto zeta_layout = params.zeta_texception - ? vk::ImageLayout::eGeneral - : vk::ImageLayout::eDepthStencilAttachmentOptimal; - descriptors.emplace_back(vk::AttachmentDescriptionFlags{}, format.format, - vk::SampleCountFlagBits::e1, vk::AttachmentLoadOp::eLoad, - vk::AttachmentStoreOp::eStore, vk::AttachmentLoadOp::eLoad, - vk::AttachmentStoreOp::eStore, zeta_layout, zeta_layout); - zeta_attachment_ref = - vk::AttachmentReference(static_cast<u32>(params.color_attachments.size()), zeta_layout); + ? VK_IMAGE_LAYOUT_GENERAL + : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; + VkAttachmentDescription& descriptor = descriptors.emplace_back(); + descriptor.flags = 0; + descriptor.format = format.format; + descriptor.samples = VK_SAMPLE_COUNT_1_BIT; + descriptor.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; + descriptor.storeOp = VK_ATTACHMENT_STORE_OP_STORE; + descriptor.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD; + descriptor.stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE; + descriptor.initialLayout = zeta_layout; + descriptor.finalLayout = zeta_layout; + + zeta_attachment_ref.attachment = static_cast<u32>(params.color_attachments.size()); + zeta_attachment_ref.layout = zeta_layout; } - const vk::SubpassDescription subpass_description( - {}, vk::PipelineBindPoint::eGraphics, 0, nullptr, static_cast<u32>(color_references.size()), - color_references.data(), nullptr, params.has_zeta ? &zeta_attachment_ref : nullptr, 0, - nullptr); - - vk::AccessFlags access; - vk::PipelineStageFlags stage; + VkSubpassDescription subpass_description; + subpass_description.flags = 0; + subpass_description.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; + subpass_description.inputAttachmentCount = 0; + subpass_description.pInputAttachments = nullptr; + subpass_description.colorAttachmentCount = static_cast<u32>(color_references.size()); + subpass_description.pColorAttachments = color_references.data(); + subpass_description.pResolveAttachments = nullptr; + subpass_description.pDepthStencilAttachment = params.has_zeta ? &zeta_attachment_ref : nullptr; + subpass_description.preserveAttachmentCount = 0; + subpass_description.pPreserveAttachments = nullptr; + + VkAccessFlags access = 0; + VkPipelineStageFlags stage = 0; if (!color_references.empty()) { - access |= - vk::AccessFlagBits::eColorAttachmentRead | vk::AccessFlagBits::eColorAttachmentWrite; - stage |= vk::PipelineStageFlagBits::eColorAttachmentOutput; + access |= VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT; + stage |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT; } if (params.has_zeta) { - access |= vk::AccessFlagBits::eDepthStencilAttachmentRead | - vk::AccessFlagBits::eDepthStencilAttachmentWrite; - stage |= vk::PipelineStageFlagBits::eLateFragmentTests; + access |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | + VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT; + stage |= VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT; } - const vk::SubpassDependency subpass_dependency(VK_SUBPASS_EXTERNAL, 0, stage, stage, {}, access, - {}); - - const vk::RenderPassCreateInfo create_info({}, static_cast<u32>(descriptors.size()), - descriptors.data(), 1, &subpass_description, 1, - &subpass_dependency); - - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - return dev.createRenderPassUnique(create_info, nullptr, dld); + VkSubpassDependency subpass_dependency; + subpass_dependency.srcSubpass = VK_SUBPASS_EXTERNAL; + subpass_dependency.dstSubpass = 0; + subpass_dependency.srcStageMask = stage; + subpass_dependency.dstStageMask = stage; + subpass_dependency.srcAccessMask = 0; + subpass_dependency.dstAccessMask = access; + subpass_dependency.dependencyFlags = 0; + + VkRenderPassCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.attachmentCount = static_cast<u32>(descriptors.size()); + ci.pAttachments = descriptors.data(); + ci.subpassCount = 1; + ci.pSubpasses = &subpass_description; + ci.dependencyCount = 1; + ci.pDependencies = &subpass_dependency; + return device.GetLogical().CreateRenderPass(ci); } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_renderpass_cache.h b/src/video_core/renderer_vulkan/vk_renderpass_cache.h index b49b2db48..921b6efb5 100644 --- a/src/video_core/renderer_vulkan/vk_renderpass_cache.h +++ b/src/video_core/renderer_vulkan/vk_renderpass_cache.h @@ -12,7 +12,7 @@ #include <boost/functional/hash.hpp> #include "video_core/engines/maxwell_3d.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/surface.h" namespace Vulkan { @@ -85,13 +85,13 @@ public: explicit VKRenderPassCache(const VKDevice& device); ~VKRenderPassCache(); - vk::RenderPass GetRenderPass(const RenderPassParams& params); + VkRenderPass GetRenderPass(const RenderPassParams& params); private: - UniqueRenderPass CreateRenderPass(const RenderPassParams& params) const; + vk::RenderPass CreateRenderPass(const RenderPassParams& params) const; const VKDevice& device; - std::unordered_map<RenderPassParams, UniqueRenderPass> cache; + std::unordered_map<RenderPassParams, vk::RenderPass> cache; }; } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_resource_manager.cpp b/src/video_core/renderer_vulkan/vk_resource_manager.cpp index 525b4bb46..dc06f545a 100644 --- a/src/video_core/renderer_vulkan/vk_resource_manager.cpp +++ b/src/video_core/renderer_vulkan/vk_resource_manager.cpp @@ -6,83 +6,83 @@ #include <optional> #include "common/assert.h" #include "common/logging/log.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { +namespace { + // TODO(Rodrigo): Fine tune these numbers. constexpr std::size_t COMMAND_BUFFER_POOL_SIZE = 0x1000; constexpr std::size_t FENCES_GROW_STEP = 0x40; +VkFenceCreateInfo BuildFenceCreateInfo() { + VkFenceCreateInfo fence_ci; + fence_ci.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO; + fence_ci.pNext = nullptr; + fence_ci.flags = 0; + return fence_ci; +} + +} // Anonymous namespace + class CommandBufferPool final : public VKFencedPool { public: CommandBufferPool(const VKDevice& device) : VKFencedPool(COMMAND_BUFFER_POOL_SIZE), device{device} {} void Allocate(std::size_t begin, std::size_t end) override { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - const u32 graphics_family = device.GetGraphicsFamily(); - - auto pool = std::make_unique<Pool>(); - // Command buffers are going to be commited, recorded, executed every single usage cycle. // They are also going to be reseted when commited. - const auto pool_flags = vk::CommandPoolCreateFlagBits::eTransient | - vk::CommandPoolCreateFlagBits::eResetCommandBuffer; - const vk::CommandPoolCreateInfo cmdbuf_pool_ci(pool_flags, graphics_family); - pool->handle = dev.createCommandPoolUnique(cmdbuf_pool_ci, nullptr, dld); - - const vk::CommandBufferAllocateInfo cmdbuf_ai(*pool->handle, - vk::CommandBufferLevel::ePrimary, - static_cast<u32>(COMMAND_BUFFER_POOL_SIZE)); - pool->cmdbufs = - dev.allocateCommandBuffersUnique<std::allocator<UniqueCommandBuffer>>(cmdbuf_ai, dld); - - pools.push_back(std::move(pool)); + VkCommandPoolCreateInfo command_pool_ci; + command_pool_ci.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO; + command_pool_ci.pNext = nullptr; + command_pool_ci.flags = + VK_COMMAND_POOL_CREATE_TRANSIENT_BIT | VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT; + command_pool_ci.queueFamilyIndex = device.GetGraphicsFamily(); + + Pool& pool = pools.emplace_back(); + pool.handle = device.GetLogical().CreateCommandPool(command_pool_ci); + pool.cmdbufs = pool.handle.Allocate(COMMAND_BUFFER_POOL_SIZE); } - vk::CommandBuffer Commit(VKFence& fence) { + VkCommandBuffer Commit(VKFence& fence) { const std::size_t index = CommitResource(fence); const auto pool_index = index / COMMAND_BUFFER_POOL_SIZE; const auto sub_index = index % COMMAND_BUFFER_POOL_SIZE; - return *pools[pool_index]->cmdbufs[sub_index]; + return pools[pool_index].cmdbufs[sub_index]; } private: struct Pool { - UniqueCommandPool handle; - std::vector<UniqueCommandBuffer> cmdbufs; + vk::CommandPool handle; + vk::CommandBuffers cmdbufs; }; const VKDevice& device; - - std::vector<std::unique_ptr<Pool>> pools; + std::vector<Pool> pools; }; VKResource::VKResource() = default; VKResource::~VKResource() = default; -VKFence::VKFence(const VKDevice& device, UniqueFence handle) - : device{device}, handle{std::move(handle)} {} +VKFence::VKFence(const VKDevice& device) + : device{device}, handle{device.GetLogical().CreateFence(BuildFenceCreateInfo())} {} VKFence::~VKFence() = default; void VKFence::Wait() { - static constexpr u64 timeout = std::numeric_limits<u64>::max(); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - switch (const auto result = dev.waitForFences(1, &*handle, true, timeout, dld)) { - case vk::Result::eSuccess: + switch (const VkResult result = handle.Wait()) { + case VK_SUCCESS: return; - case vk::Result::eErrorDeviceLost: + case VK_ERROR_DEVICE_LOST: device.ReportLoss(); [[fallthrough]]; default: - vk::throwResultException(result, "vk::waitForFences"); + throw vk::Exception(result); } } @@ -107,13 +107,11 @@ bool VKFence::Tick(bool gpu_wait, bool owner_wait) { return false; } - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); if (gpu_wait) { // Wait for the fence if it has been requested. - dev.waitForFences({*handle}, true, std::numeric_limits<u64>::max(), dld); + (void)handle.Wait(); } else { - if (dev.getFenceStatus(*handle, dld) != vk::Result::eSuccess) { + if (handle.GetStatus() != VK_SUCCESS) { // Vulkan fence is not ready, not much it can do here return false; } @@ -126,7 +124,7 @@ bool VKFence::Tick(bool gpu_wait, bool owner_wait) { protected_resources.clear(); // Prepare fence for reusage. - dev.resetFences({*handle}, dld); + handle.Reset(); is_used = false; return true; } @@ -299,21 +297,16 @@ VKFence& VKResourceManager::CommitFence() { return *found_fence; } -vk::CommandBuffer VKResourceManager::CommitCommandBuffer(VKFence& fence) { +VkCommandBuffer VKResourceManager::CommitCommandBuffer(VKFence& fence) { return command_buffer_pool->Commit(fence); } void VKResourceManager::GrowFences(std::size_t new_fences_count) { - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - const vk::FenceCreateInfo fence_ci; - const std::size_t previous_size = fences.size(); fences.resize(previous_size + new_fences_count); - std::generate(fences.begin() + previous_size, fences.end(), [&]() { - return std::make_unique<VKFence>(device, dev.createFenceUnique(fence_ci, nullptr, dld)); - }); + std::generate(fences.begin() + previous_size, fences.end(), + [this] { return std::make_unique<VKFence>(device); }); } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_resource_manager.h b/src/video_core/renderer_vulkan/vk_resource_manager.h index d4cbc95a5..f683d2276 100644 --- a/src/video_core/renderer_vulkan/vk_resource_manager.h +++ b/src/video_core/renderer_vulkan/vk_resource_manager.h @@ -7,7 +7,7 @@ #include <cstddef> #include <memory> #include <vector> -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -42,7 +42,7 @@ class VKFence { friend class VKResourceManager; public: - explicit VKFence(const VKDevice& device, UniqueFence handle); + explicit VKFence(const VKDevice& device); ~VKFence(); /** @@ -69,7 +69,7 @@ public: void RedirectProtection(VKResource* old_resource, VKResource* new_resource) noexcept; /// Retreives the fence. - operator vk::Fence() const { + operator VkFence() const { return *handle; } @@ -87,7 +87,7 @@ private: bool Tick(bool gpu_wait, bool owner_wait); const VKDevice& device; ///< Device handler - UniqueFence handle; ///< Vulkan fence + vk::Fence handle; ///< Vulkan fence std::vector<VKResource*> protected_resources; ///< List of resources protected by this fence bool is_owned = false; ///< The fence has been commited but not released yet. bool is_used = false; ///< The fence has been commited but it has not been checked to be free. @@ -181,7 +181,7 @@ public: VKFence& CommitFence(); /// Commits an unused command buffer and protects it with a fence. - vk::CommandBuffer CommitCommandBuffer(VKFence& fence); + VkCommandBuffer CommitCommandBuffer(VKFence& fence); private: /// Allocates new fences. diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp index 204b7c39c..07bbcf520 100644 --- a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp @@ -7,64 +7,64 @@ #include <unordered_map> #include "common/assert.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/vk_sampler_cache.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/textures/texture.h" namespace Vulkan { -static std::optional<vk::BorderColor> TryConvertBorderColor(std::array<float, 4> color) { +namespace { + +VkBorderColor ConvertBorderColor(std::array<float, 4> color) { // TODO(Rodrigo): Manage integer border colors if (color == std::array<float, 4>{0, 0, 0, 0}) { - return vk::BorderColor::eFloatTransparentBlack; + return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK; } else if (color == std::array<float, 4>{0, 0, 0, 1}) { - return vk::BorderColor::eFloatOpaqueBlack; + return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK; } else if (color == std::array<float, 4>{1, 1, 1, 1}) { - return vk::BorderColor::eFloatOpaqueWhite; + return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE; + } + if (color[0] + color[1] + color[2] > 1.35f) { + // If color elements are brighter than roughly 0.5 average, use white border + return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE; + } else if (color[3] > 0.5f) { + return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK; } else { - if (color[0] + color[1] + color[2] > 1.35f) { - // If color elements are brighter than roughly 0.5 average, use white border - return vk::BorderColor::eFloatOpaqueWhite; - } - if (color[3] > 0.5f) { - return vk::BorderColor::eFloatOpaqueBlack; - } - return vk::BorderColor::eFloatTransparentBlack; + return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK; } } +} // Anonymous namespace + VKSamplerCache::VKSamplerCache(const VKDevice& device) : device{device} {} VKSamplerCache::~VKSamplerCache() = default; -UniqueSampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) const { - const float max_anisotropy{tsc.GetMaxAnisotropy()}; - const bool has_anisotropy{max_anisotropy > 1.0f}; - - const auto border_color{tsc.GetBorderColor()}; - const auto vk_border_color{TryConvertBorderColor(border_color)}; - - constexpr bool unnormalized_coords{false}; - - const vk::SamplerCreateInfo sampler_ci( - {}, MaxwellToVK::Sampler::Filter(tsc.mag_filter), - MaxwellToVK::Sampler::Filter(tsc.min_filter), - MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter), - MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter), - MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter), - MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter), tsc.GetLodBias(), - has_anisotropy, max_anisotropy, tsc.depth_compare_enabled, - MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func), tsc.GetMinLod(), - tsc.GetMaxLod(), vk_border_color.value_or(vk::BorderColor::eFloatTransparentBlack), - unnormalized_coords); - - const auto& dld{device.GetDispatchLoader()}; - const auto dev{device.GetLogical()}; - return dev.createSamplerUnique(sampler_ci, nullptr, dld); +vk::Sampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) const { + VkSamplerCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.magFilter = MaxwellToVK::Sampler::Filter(tsc.mag_filter); + ci.minFilter = MaxwellToVK::Sampler::Filter(tsc.min_filter); + ci.mipmapMode = MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter); + ci.addressModeU = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter); + ci.addressModeV = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter); + ci.addressModeW = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter); + ci.mipLodBias = tsc.GetLodBias(); + ci.anisotropyEnable = tsc.GetMaxAnisotropy() > 1.0f ? VK_TRUE : VK_FALSE; + ci.maxAnisotropy = tsc.GetMaxAnisotropy(); + ci.compareEnable = tsc.depth_compare_enabled; + ci.compareOp = MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func); + ci.minLod = tsc.GetMinLod(); + ci.maxLod = tsc.GetMaxLod(); + ci.borderColor = ConvertBorderColor(tsc.GetBorderColor()); + ci.unnormalizedCoordinates = VK_FALSE; + return device.GetLogical().CreateSampler(ci); } -vk::Sampler VKSamplerCache::ToSamplerType(const UniqueSampler& sampler) const { +VkSampler VKSamplerCache::ToSamplerType(const vk::Sampler& sampler) const { return *sampler; } diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.h b/src/video_core/renderer_vulkan/vk_sampler_cache.h index 1f73b716b..a33d1c0ee 100644 --- a/src/video_core/renderer_vulkan/vk_sampler_cache.h +++ b/src/video_core/renderer_vulkan/vk_sampler_cache.h @@ -4,7 +4,7 @@ #pragma once -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/sampler_cache.h" #include "video_core/textures/texture.h" @@ -12,15 +12,15 @@ namespace Vulkan { class VKDevice; -class VKSamplerCache final : public VideoCommon::SamplerCache<vk::Sampler, UniqueSampler> { +class VKSamplerCache final : public VideoCommon::SamplerCache<VkSampler, vk::Sampler> { public: explicit VKSamplerCache(const VKDevice& device); ~VKSamplerCache(); protected: - UniqueSampler CreateSampler(const Tegra::Texture::TSCEntry& tsc) const override; + vk::Sampler CreateSampler(const Tegra::Texture::TSCEntry& tsc) const override; - vk::Sampler ToSamplerType(const UniqueSampler& sampler) const override; + VkSampler ToSamplerType(const vk::Sampler& sampler) const override; private: const VKDevice& device; diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp index b61d4fe63..900f551b3 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.cpp +++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp @@ -10,23 +10,22 @@ #include "common/assert.h" #include "common/microprofile.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_query_cache.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_state_tracker.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { MICROPROFILE_DECLARE(Vulkan_WaitForWorker); -void VKScheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf, - const vk::DispatchLoaderDynamic& dld) { +void VKScheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf) { auto command = first; while (command != nullptr) { auto next = command->GetNext(); - command->Execute(cmdbuf, dld); + command->Execute(cmdbuf); command->~Command(); command = next; } @@ -51,7 +50,7 @@ VKScheduler::~VKScheduler() { worker_thread.join(); } -void VKScheduler::Flush(bool release_fence, vk::Semaphore semaphore) { +void VKScheduler::Flush(bool release_fence, VkSemaphore semaphore) { SubmitExecution(semaphore); if (release_fence) { current_fence->Release(); @@ -59,7 +58,7 @@ void VKScheduler::Flush(bool release_fence, vk::Semaphore semaphore) { AllocateNewContext(); } -void VKScheduler::Finish(bool release_fence, vk::Semaphore semaphore) { +void VKScheduler::Finish(bool release_fence, VkSemaphore semaphore) { SubmitExecution(semaphore); current_fence->Wait(); if (release_fence) { @@ -89,17 +88,34 @@ void VKScheduler::DispatchWork() { AcquireNewChunk(); } -void VKScheduler::RequestRenderpass(const vk::RenderPassBeginInfo& renderpass_bi) { - if (state.renderpass && renderpass_bi == *state.renderpass) { +void VKScheduler::RequestRenderpass(VkRenderPass renderpass, VkFramebuffer framebuffer, + VkExtent2D render_area) { + if (renderpass == state.renderpass && framebuffer == state.framebuffer && + render_area.width == state.render_area.width && + render_area.height == state.render_area.height) { return; } - const bool end_renderpass = state.renderpass.has_value(); - state.renderpass = renderpass_bi; - Record([renderpass_bi, end_renderpass](auto cmdbuf, auto& dld) { + const bool end_renderpass = state.renderpass != nullptr; + state.renderpass = renderpass; + state.framebuffer = framebuffer; + state.render_area = render_area; + + VkRenderPassBeginInfo renderpass_bi; + renderpass_bi.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO; + renderpass_bi.pNext = nullptr; + renderpass_bi.renderPass = renderpass; + renderpass_bi.framebuffer = framebuffer; + renderpass_bi.renderArea.offset.x = 0; + renderpass_bi.renderArea.offset.y = 0; + renderpass_bi.renderArea.extent = render_area; + renderpass_bi.clearValueCount = 0; + renderpass_bi.pClearValues = nullptr; + + Record([renderpass_bi, end_renderpass](vk::CommandBuffer cmdbuf) { if (end_renderpass) { - cmdbuf.endRenderPass(dld); + cmdbuf.EndRenderPass(); } - cmdbuf.beginRenderPass(renderpass_bi, vk::SubpassContents::eInline, dld); + cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE); }); } @@ -107,13 +123,13 @@ void VKScheduler::RequestOutsideRenderPassOperationContext() { EndRenderPass(); } -void VKScheduler::BindGraphicsPipeline(vk::Pipeline pipeline) { +void VKScheduler::BindGraphicsPipeline(VkPipeline pipeline) { if (state.graphics_pipeline == pipeline) { return; } state.graphics_pipeline = pipeline; - Record([pipeline](auto cmdbuf, auto& dld) { - cmdbuf.bindPipeline(vk::PipelineBindPoint::eGraphics, pipeline, dld); + Record([pipeline](vk::CommandBuffer cmdbuf) { + cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline); }); } @@ -126,37 +142,50 @@ void VKScheduler::WorkerThread() { } auto extracted_chunk = std::move(chunk_queue.Front()); chunk_queue.Pop(); - extracted_chunk->ExecuteAll(current_cmdbuf, device.GetDispatchLoader()); + extracted_chunk->ExecuteAll(current_cmdbuf); chunk_reserve.Push(std::move(extracted_chunk)); } while (!quit); } -void VKScheduler::SubmitExecution(vk::Semaphore semaphore) { +void VKScheduler::SubmitExecution(VkSemaphore semaphore) { EndPendingOperations(); InvalidateState(); WaitWorker(); std::unique_lock lock{mutex}; - const auto queue = device.GetGraphicsQueue(); - const auto& dld = device.GetDispatchLoader(); - current_cmdbuf.end(dld); + current_cmdbuf.End(); - const vk::SubmitInfo submit_info(0, nullptr, nullptr, 1, ¤t_cmdbuf, semaphore ? 1U : 0U, - &semaphore); - queue.submit({submit_info}, static_cast<vk::Fence>(*current_fence), dld); + VkSubmitInfo submit_info; + submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO; + submit_info.pNext = nullptr; + submit_info.waitSemaphoreCount = 0; + submit_info.pWaitSemaphores = nullptr; + submit_info.pWaitDstStageMask = nullptr; + submit_info.commandBufferCount = 1; + submit_info.pCommandBuffers = current_cmdbuf.address(); + submit_info.signalSemaphoreCount = semaphore ? 1 : 0; + submit_info.pSignalSemaphores = &semaphore; + device.GetGraphicsQueue().Submit(submit_info, *current_fence); } void VKScheduler::AllocateNewContext() { ++ticks; + VkCommandBufferBeginInfo cmdbuf_bi; + cmdbuf_bi.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO; + cmdbuf_bi.pNext = nullptr; + cmdbuf_bi.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT; + cmdbuf_bi.pInheritanceInfo = nullptr; + std::unique_lock lock{mutex}; current_fence = next_fence; next_fence = &resource_manager.CommitFence(); - current_cmdbuf = resource_manager.CommitCommandBuffer(*current_fence); - current_cmdbuf.begin({vk::CommandBufferUsageFlagBits::eOneTimeSubmit}, - device.GetDispatchLoader()); + current_cmdbuf = vk::CommandBuffer(resource_manager.CommitCommandBuffer(*current_fence), + device.GetDispatchLoader()); + current_cmdbuf.Begin(cmdbuf_bi); + // Enable counters once again. These are disabled when a command buffer is finished. if (query_cache) { query_cache->UpdateCounters(); @@ -177,8 +206,8 @@ void VKScheduler::EndRenderPass() { if (!state.renderpass) { return; } - state.renderpass = std::nullopt; - Record([](auto cmdbuf, auto& dld) { cmdbuf.endRenderPass(dld); }); + state.renderpass = nullptr; + Record([](vk::CommandBuffer cmdbuf) { cmdbuf.EndRenderPass(); }); } void VKScheduler::AcquireNewChunk() { diff --git a/src/video_core/renderer_vulkan/vk_scheduler.h b/src/video_core/renderer_vulkan/vk_scheduler.h index c7cc291c3..82a8adc69 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.h +++ b/src/video_core/renderer_vulkan/vk_scheduler.h @@ -13,7 +13,7 @@ #include <utility> #include "common/common_types.h" #include "common/threadsafe_queue.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -49,10 +49,10 @@ public: ~VKScheduler(); /// Sends the current execution context to the GPU. - void Flush(bool release_fence = true, vk::Semaphore semaphore = nullptr); + void Flush(bool release_fence = true, VkSemaphore semaphore = nullptr); /// Sends the current execution context to the GPU and waits for it to complete. - void Finish(bool release_fence = true, vk::Semaphore semaphore = nullptr); + void Finish(bool release_fence = true, VkSemaphore semaphore = nullptr); /// Waits for the worker thread to finish executing everything. After this function returns it's /// safe to touch worker resources. @@ -62,14 +62,15 @@ public: void DispatchWork(); /// Requests to begin a renderpass. - void RequestRenderpass(const vk::RenderPassBeginInfo& renderpass_bi); + void RequestRenderpass(VkRenderPass renderpass, VkFramebuffer framebuffer, + VkExtent2D render_area); /// Requests the current executino context to be able to execute operations only allowed outside /// of a renderpass. void RequestOutsideRenderPassOperationContext(); /// Binds a pipeline to the current execution context. - void BindGraphicsPipeline(vk::Pipeline pipeline); + void BindGraphicsPipeline(VkPipeline pipeline); /// Assigns the query cache. void SetQueryCache(VKQueryCache& query_cache_) { @@ -101,8 +102,7 @@ private: public: virtual ~Command() = default; - virtual void Execute(vk::CommandBuffer cmdbuf, - const vk::DispatchLoaderDynamic& dld) const = 0; + virtual void Execute(vk::CommandBuffer cmdbuf) const = 0; Command* GetNext() const { return next; @@ -125,9 +125,8 @@ private: TypedCommand(TypedCommand&&) = delete; TypedCommand& operator=(TypedCommand&&) = delete; - void Execute(vk::CommandBuffer cmdbuf, - const vk::DispatchLoaderDynamic& dld) const override { - command(cmdbuf, dld); + void Execute(vk::CommandBuffer cmdbuf) const override { + command(cmdbuf); } private: @@ -136,7 +135,7 @@ private: class CommandChunk final { public: - void ExecuteAll(vk::CommandBuffer cmdbuf, const vk::DispatchLoaderDynamic& dld); + void ExecuteAll(vk::CommandBuffer cmdbuf); template <typename T> bool Record(T& command) { @@ -175,7 +174,7 @@ private: void WorkerThread(); - void SubmitExecution(vk::Semaphore semaphore); + void SubmitExecution(VkSemaphore semaphore); void AllocateNewContext(); @@ -198,8 +197,10 @@ private: VKFence* next_fence = nullptr; struct State { - std::optional<vk::RenderPassBeginInfo> renderpass; - vk::Pipeline graphics_pipeline; + VkRenderPass renderpass = nullptr; + VkFramebuffer framebuffer = nullptr; + VkExtent2D render_area = {0, 0}; + VkPipeline graphics_pipeline = nullptr; } state; std::unique_ptr<CommandChunk> chunk; diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp index 24d3ca08f..aaa138f52 100644 --- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp +++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp @@ -35,7 +35,7 @@ namespace { using Sirit::Id; using Tegra::Engines::ShaderType; using Tegra::Shader::Attribute; -using Tegra::Shader::AttributeUse; +using Tegra::Shader::PixelImap; using Tegra::Shader::Register; using namespace VideoCommon::Shader; @@ -752,16 +752,16 @@ private: if (stage != ShaderType::Fragment) { continue; } - switch (header.ps.GetAttributeUse(location)) { - case AttributeUse::Constant: + switch (header.ps.GetPixelImap(location)) { + case PixelImap::Constant: Decorate(id, spv::Decoration::Flat); break; - case AttributeUse::ScreenLinear: - Decorate(id, spv::Decoration::NoPerspective); - break; - case AttributeUse::Perspective: + case PixelImap::Perspective: // Default break; + case PixelImap::ScreenLinear: + Decorate(id, spv::Decoration::NoPerspective); + break; default: UNREACHABLE_MSG("Unused attribute being fetched"); } @@ -801,7 +801,7 @@ private: if (IsOutputAttributeArray()) { const u32 num = GetNumOutputVertices(); type = TypeArray(type, Constant(t_uint, num)); - if (device.GetDriverID() != vk::DriverIdKHR::eIntelProprietaryWindows) { + if (device.GetDriverID() != VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS_KHR) { // Intel's proprietary driver fails to setup defaults for arrayed output // attributes. varying_default = ConstantComposite(type, std::vector(num, varying_default)); @@ -1145,9 +1145,6 @@ private: switch (attribute) { case Attribute::Index::Position: { if (stage == ShaderType::Fragment) { - if (element == 3) { - return {Constant(t_float, 1.0f), Type::Float}; - } return {OpLoad(t_float, AccessElement(t_in_float, frag_coord, element)), Type::Float}; } diff --git a/src/video_core/renderer_vulkan/vk_shader_util.cpp b/src/video_core/renderer_vulkan/vk_shader_util.cpp index b97c4cb3d..784839327 100644 --- a/src/video_core/renderer_vulkan/vk_shader_util.cpp +++ b/src/video_core/renderer_vulkan/vk_shader_util.cpp @@ -8,27 +8,25 @@ #include "common/alignment.h" #include "common/assert.h" #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_shader_util.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { -UniqueShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data) { +vk::ShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data) { // Avoid undefined behavior by copying to a staging allocation ASSERT(code_size % sizeof(u32) == 0); const auto data = std::make_unique<u32[]>(code_size / sizeof(u32)); std::memcpy(data.get(), code_data, code_size); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - const vk::ShaderModuleCreateInfo shader_ci({}, code_size, data.get()); - vk::ShaderModule shader_module; - if (dev.createShaderModule(&shader_ci, nullptr, &shader_module, dld) != vk::Result::eSuccess) { - UNREACHABLE_MSG("Shader module failed to build!"); - } - - return UniqueShaderModule(shader_module, vk::ObjectDestroy(dev, nullptr, dld)); + VkShaderModuleCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.codeSize = code_size; + ci.pCode = data.get(); + return device.GetLogical().CreateShaderModule(ci); } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_shader_util.h b/src/video_core/renderer_vulkan/vk_shader_util.h index c06d65970..be38d6697 100644 --- a/src/video_core/renderer_vulkan/vk_shader_util.h +++ b/src/video_core/renderer_vulkan/vk_shader_util.h @@ -6,12 +6,12 @@ #include <vector> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { class VKDevice; -UniqueShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data); +vk::ShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data); } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp index 374959f82..94d954d7a 100644 --- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp +++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp @@ -13,6 +13,7 @@ #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -71,17 +72,23 @@ VKBuffer* VKStagingBufferPool::TryGetReservedBuffer(std::size_t size, bool host_ } VKBuffer& VKStagingBufferPool::CreateStagingBuffer(std::size_t size, bool host_visible) { - const auto usage = - vk::BufferUsageFlagBits::eTransferSrc | vk::BufferUsageFlagBits::eTransferDst | - vk::BufferUsageFlagBits::eUniformBuffer | vk::BufferUsageFlagBits::eStorageBuffer | - vk::BufferUsageFlagBits::eIndexBuffer; const u32 log2 = Common::Log2Ceil64(size); - const vk::BufferCreateInfo buffer_ci({}, 1ULL << log2, usage, vk::SharingMode::eExclusive, 0, - nullptr); - const auto dev = device.GetLogical(); + + VkBufferCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.size = 1ULL << log2; + ci.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | + VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | + VK_BUFFER_USAGE_INDEX_BUFFER_BIT; + ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; + ci.queueFamilyIndexCount = 0; + ci.pQueueFamilyIndices = nullptr; + auto buffer = std::make_unique<VKBuffer>(); - buffer->handle = dev.createBufferUnique(buffer_ci, nullptr, device.GetDispatchLoader()); - buffer->commit = memory_manager.Commit(*buffer->handle, host_visible); + buffer->handle = device.GetLogical().CreateBuffer(ci); + buffer->commit = memory_manager.Commit(buffer->handle, host_visible); auto& entries = GetCache(host_visible)[log2].entries; return *entries.emplace_back(std::move(buffer), scheduler.GetFence(), epoch).buffer; diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h index 4d9488f49..a0840ff8c 100644 --- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h +++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.h @@ -11,9 +11,9 @@ #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_memory_manager.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -22,7 +22,7 @@ class VKFenceWatch; class VKScheduler; struct VKBuffer final { - UniqueBuffer handle; + vk::Buffer handle; VKMemoryCommit commit; }; diff --git a/src/video_core/renderer_vulkan/vk_stream_buffer.cpp b/src/video_core/renderer_vulkan/vk_stream_buffer.cpp index d48d3b44c..38a93a01a 100644 --- a/src/video_core/renderer_vulkan/vk_stream_buffer.cpp +++ b/src/video_core/renderer_vulkan/vk_stream_buffer.cpp @@ -9,11 +9,11 @@ #include "common/alignment.h" #include "common/assert.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_stream_buffer.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -25,8 +25,8 @@ constexpr u64 WATCHES_RESERVE_CHUNK = 0x1000; constexpr u64 STREAM_BUFFER_SIZE = 256 * 1024 * 1024; std::optional<u32> FindMemoryType(const VKDevice& device, u32 filter, - vk::MemoryPropertyFlags wanted) { - const auto properties = device.GetPhysical().getMemoryProperties(device.GetDispatchLoader()); + VkMemoryPropertyFlags wanted) { + const auto properties = device.GetPhysical().GetMemoryProperties(); for (u32 i = 0; i < properties.memoryTypeCount; i++) { if (!(filter & (1 << i))) { continue; @@ -35,13 +35,13 @@ std::optional<u32> FindMemoryType(const VKDevice& device, u32 filter, return i; } } - return {}; + return std::nullopt; } } // Anonymous namespace VKStreamBuffer::VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler, - vk::BufferUsageFlags usage) + VkBufferUsageFlags usage) : device{device}, scheduler{scheduler} { CreateBuffers(usage); ReserveWatches(current_watches, WATCHES_INITIAL_RESERVE); @@ -78,17 +78,13 @@ std::tuple<u8*, u64, bool> VKStreamBuffer::Map(u64 size, u64 alignment) { invalidated = true; } - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - const auto pointer = reinterpret_cast<u8*>(dev.mapMemory(*memory, offset, size, {}, dld)); - return {pointer, offset, invalidated}; + return {memory.Map(offset, size), offset, invalidated}; } void VKStreamBuffer::Unmap(u64 size) { ASSERT_MSG(size <= mapped_size, "Reserved size is too small"); - const auto dev = device.GetLogical(); - dev.unmapMemory(*memory, device.GetDispatchLoader()); + memory.Unmap(); offset += size; @@ -101,30 +97,42 @@ void VKStreamBuffer::Unmap(u64 size) { watch.fence.Watch(scheduler.GetFence()); } -void VKStreamBuffer::CreateBuffers(vk::BufferUsageFlags usage) { - const vk::BufferCreateInfo buffer_ci({}, STREAM_BUFFER_SIZE, usage, vk::SharingMode::eExclusive, - 0, nullptr); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - buffer = dev.createBufferUnique(buffer_ci, nullptr, dld); +void VKStreamBuffer::CreateBuffers(VkBufferUsageFlags usage) { + VkBufferCreateInfo buffer_ci; + buffer_ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; + buffer_ci.pNext = nullptr; + buffer_ci.flags = 0; + buffer_ci.size = STREAM_BUFFER_SIZE; + buffer_ci.usage = usage; + buffer_ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; + buffer_ci.queueFamilyIndexCount = 0; + buffer_ci.pQueueFamilyIndices = nullptr; + + const auto& dev = device.GetLogical(); + buffer = dev.CreateBuffer(buffer_ci); - const auto requirements = dev.getBufferMemoryRequirements(*buffer, dld); + const auto& dld = device.GetDispatchLoader(); + const auto requirements = dev.GetBufferMemoryRequirements(*buffer); // Prefer device local host visible allocations (this should hit AMD's pinned memory). - auto type = FindMemoryType(device, requirements.memoryTypeBits, - vk::MemoryPropertyFlagBits::eHostVisible | - vk::MemoryPropertyFlagBits::eHostCoherent | - vk::MemoryPropertyFlagBits::eDeviceLocal); + auto type = + FindMemoryType(device, requirements.memoryTypeBits, + VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT | + VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT); if (!type) { // Otherwise search for a host visible allocation. type = FindMemoryType(device, requirements.memoryTypeBits, - vk::MemoryPropertyFlagBits::eHostVisible | - vk::MemoryPropertyFlagBits::eHostCoherent); + VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | + VK_MEMORY_PROPERTY_HOST_COHERENT_BIT); ASSERT_MSG(type, "No host visible and coherent memory type found"); } - const vk::MemoryAllocateInfo alloc_ci(requirements.size, *type); - memory = dev.allocateMemoryUnique(alloc_ci, nullptr, dld); - - dev.bindBufferMemory(*buffer, *memory, 0, dld); + VkMemoryAllocateInfo memory_ai; + memory_ai.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO; + memory_ai.pNext = nullptr; + memory_ai.allocationSize = requirements.size; + memory_ai.memoryTypeIndex = *type; + + memory = dev.AllocateMemory(memory_ai); + buffer.BindMemory(*memory, 0); } void VKStreamBuffer::ReserveWatches(std::vector<Watch>& watches, std::size_t grow_size) { diff --git a/src/video_core/renderer_vulkan/vk_stream_buffer.h b/src/video_core/renderer_vulkan/vk_stream_buffer.h index 187c0c612..58ce8b973 100644 --- a/src/video_core/renderer_vulkan/vk_stream_buffer.h +++ b/src/video_core/renderer_vulkan/vk_stream_buffer.h @@ -9,7 +9,7 @@ #include <vector> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -21,7 +21,7 @@ class VKScheduler; class VKStreamBuffer final { public: explicit VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler, - vk::BufferUsageFlags usage); + VkBufferUsageFlags usage); ~VKStreamBuffer(); /** @@ -35,7 +35,7 @@ public: /// Ensures that "size" bytes of memory are available to the GPU, potentially recording a copy. void Unmap(u64 size); - vk::Buffer GetHandle() const { + VkBuffer GetHandle() const { return *buffer; } @@ -46,20 +46,18 @@ private: }; /// Creates Vulkan buffer handles committing the required the required memory. - void CreateBuffers(vk::BufferUsageFlags usage); + void CreateBuffers(VkBufferUsageFlags usage); /// Increases the amount of watches available. void ReserveWatches(std::vector<Watch>& watches, std::size_t grow_size); void WaitPendingOperations(u64 requested_upper_bound); - const VKDevice& device; ///< Vulkan device manager. - VKScheduler& scheduler; ///< Command scheduler. - const vk::AccessFlags access; ///< Access usage of this stream buffer. - const vk::PipelineStageFlags pipeline_stage; ///< Pipeline usage of this stream buffer. + const VKDevice& device; ///< Vulkan device manager. + VKScheduler& scheduler; ///< Command scheduler. - UniqueBuffer buffer; ///< Mapped buffer. - UniqueDeviceMemory memory; ///< Memory allocation. + vk::Buffer buffer; ///< Mapped buffer. + vk::DeviceMemory memory; ///< Memory allocation. u64 offset{}; ///< Buffer iterator. u64 mapped_size{}; ///< Size reserved for the current copy. diff --git a/src/video_core/renderer_vulkan/vk_swapchain.cpp b/src/video_core/renderer_vulkan/vk_swapchain.cpp index 9e73fa9cd..bffd8f32a 100644 --- a/src/video_core/renderer_vulkan/vk_swapchain.cpp +++ b/src/video_core/renderer_vulkan/vk_swapchain.cpp @@ -11,69 +11,64 @@ #include "common/logging/log.h" #include "core/core.h" #include "core/frontend/framebuffer_layout.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_resource_manager.h" #include "video_core/renderer_vulkan/vk_swapchain.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { namespace { -vk::SurfaceFormatKHR ChooseSwapSurfaceFormat(const std::vector<vk::SurfaceFormatKHR>& formats, - bool srgb) { - if (formats.size() == 1 && formats[0].format == vk::Format::eUndefined) { - vk::SurfaceFormatKHR format; - format.format = vk::Format::eB8G8R8A8Unorm; - format.colorSpace = vk::ColorSpaceKHR::eSrgbNonlinear; +VkSurfaceFormatKHR ChooseSwapSurfaceFormat(vk::Span<VkSurfaceFormatKHR> formats, bool srgb) { + if (formats.size() == 1 && formats[0].format == VK_FORMAT_UNDEFINED) { + VkSurfaceFormatKHR format; + format.format = VK_FORMAT_B8G8R8A8_UNORM; + format.colorSpace = VK_COLOR_SPACE_SRGB_NONLINEAR_KHR; return format; } const auto& found = std::find_if(formats.begin(), formats.end(), [srgb](const auto& format) { - const auto request_format = srgb ? vk::Format::eB8G8R8A8Srgb : vk::Format::eB8G8R8A8Unorm; + const auto request_format = srgb ? VK_FORMAT_B8G8R8A8_SRGB : VK_FORMAT_B8G8R8A8_UNORM; return format.format == request_format && - format.colorSpace == vk::ColorSpaceKHR::eSrgbNonlinear; + format.colorSpace == VK_COLOR_SPACE_SRGB_NONLINEAR_KHR; }); return found != formats.end() ? *found : formats[0]; } -vk::PresentModeKHR ChooseSwapPresentMode(const std::vector<vk::PresentModeKHR>& modes) { +VkPresentModeKHR ChooseSwapPresentMode(vk::Span<VkPresentModeKHR> modes) { // Mailbox doesn't lock the application like fifo (vsync), prefer it - const auto& found = std::find_if(modes.begin(), modes.end(), [](const auto& mode) { - return mode == vk::PresentModeKHR::eMailbox; - }); - return found != modes.end() ? *found : vk::PresentModeKHR::eFifo; + const auto found = std::find(modes.begin(), modes.end(), VK_PRESENT_MODE_MAILBOX_KHR); + return found != modes.end() ? *found : VK_PRESENT_MODE_FIFO_KHR; } -vk::Extent2D ChooseSwapExtent(const vk::SurfaceCapabilitiesKHR& capabilities, u32 width, - u32 height) { +VkExtent2D ChooseSwapExtent(const VkSurfaceCapabilitiesKHR& capabilities, u32 width, u32 height) { constexpr auto undefined_size{std::numeric_limits<u32>::max()}; if (capabilities.currentExtent.width != undefined_size) { return capabilities.currentExtent; } - vk::Extent2D extent = {width, height}; + VkExtent2D extent; extent.width = std::max(capabilities.minImageExtent.width, - std::min(capabilities.maxImageExtent.width, extent.width)); + std::min(capabilities.maxImageExtent.width, width)); extent.height = std::max(capabilities.minImageExtent.height, - std::min(capabilities.maxImageExtent.height, extent.height)); + std::min(capabilities.maxImageExtent.height, height)); return extent; } } // Anonymous namespace -VKSwapchain::VKSwapchain(vk::SurfaceKHR surface, const VKDevice& device) +VKSwapchain::VKSwapchain(VkSurfaceKHR surface, const VKDevice& device) : surface{surface}, device{device} {} VKSwapchain::~VKSwapchain() = default; void VKSwapchain::Create(u32 width, u32 height, bool srgb) { - const auto& dld = device.GetDispatchLoader(); const auto physical_device = device.GetPhysical(); - const auto capabilities{physical_device.getSurfaceCapabilitiesKHR(surface, dld)}; + const auto capabilities{physical_device.GetSurfaceCapabilitiesKHR(surface)}; if (capabilities.maxImageExtent.width == 0 || capabilities.maxImageExtent.height == 0) { return; } - device.GetLogical().waitIdle(dld); + device.GetLogical().WaitIdle(); Destroy(); CreateSwapchain(capabilities, width, height, srgb); @@ -84,10 +79,8 @@ void VKSwapchain::Create(u32 width, u32 height, bool srgb) { } void VKSwapchain::AcquireNextImage() { - const auto dev{device.GetLogical()}; - const auto& dld{device.GetDispatchLoader()}; - dev.acquireNextImageKHR(*swapchain, std::numeric_limits<u64>::max(), - *present_semaphores[frame_index], {}, &image_index, dld); + device.GetLogical().AcquireNextImageKHR(*swapchain, std::numeric_limits<u64>::max(), + *present_semaphores[frame_index], {}, &image_index); if (auto& fence = fences[image_index]; fence) { fence->Wait(); @@ -96,29 +89,37 @@ void VKSwapchain::AcquireNextImage() { } } -bool VKSwapchain::Present(vk::Semaphore render_semaphore, VKFence& fence) { - const vk::Semaphore present_semaphore{*present_semaphores[frame_index]}; - const std::array<vk::Semaphore, 2> semaphores{present_semaphore, render_semaphore}; - const u32 wait_semaphore_count{render_semaphore ? 2U : 1U}; - const auto& dld{device.GetDispatchLoader()}; +bool VKSwapchain::Present(VkSemaphore render_semaphore, VKFence& fence) { + const VkSemaphore present_semaphore{*present_semaphores[frame_index]}; + const std::array<VkSemaphore, 2> semaphores{present_semaphore, render_semaphore}; const auto present_queue{device.GetPresentQueue()}; bool recreated = false; - const vk::PresentInfoKHR present_info(wait_semaphore_count, semaphores.data(), 1, - &swapchain.get(), &image_index, {}); - switch (const auto result = present_queue.presentKHR(&present_info, dld); result) { - case vk::Result::eSuccess: + VkPresentInfoKHR present_info; + present_info.sType = VK_STRUCTURE_TYPE_PRESENT_INFO_KHR; + present_info.pNext = nullptr; + present_info.waitSemaphoreCount = render_semaphore ? 2U : 1U; + present_info.pWaitSemaphores = semaphores.data(); + present_info.swapchainCount = 1; + present_info.pSwapchains = swapchain.address(); + present_info.pImageIndices = &image_index; + present_info.pResults = nullptr; + + switch (const VkResult result = present_queue.Present(present_info)) { + case VK_SUCCESS: + break; + case VK_SUBOPTIMAL_KHR: + LOG_DEBUG(Render_Vulkan, "Suboptimal swapchain"); break; - case vk::Result::eErrorOutOfDateKHR: + case VK_ERROR_OUT_OF_DATE_KHR: if (current_width > 0 && current_height > 0) { Create(current_width, current_height, current_srgb); recreated = true; } break; default: - LOG_CRITICAL(Render_Vulkan, "Vulkan failed to present swapchain due to {}!", - vk::to_string(result)); - UNREACHABLE(); + LOG_CRITICAL(Render_Vulkan, "Failed to present with error {}", vk::ToString(result)); + break; } ASSERT(fences[image_index] == nullptr); @@ -132,74 +133,92 @@ bool VKSwapchain::HasFramebufferChanged(const Layout::FramebufferLayout& framebu return framebuffer.width != current_width || framebuffer.height != current_height; } -void VKSwapchain::CreateSwapchain(const vk::SurfaceCapabilitiesKHR& capabilities, u32 width, +void VKSwapchain::CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities, u32 width, u32 height, bool srgb) { - const auto& dld{device.GetDispatchLoader()}; const auto physical_device{device.GetPhysical()}; - const auto formats{physical_device.getSurfaceFormatsKHR(surface, dld)}; - const auto present_modes{physical_device.getSurfacePresentModesKHR(surface, dld)}; + const auto formats{physical_device.GetSurfaceFormatsKHR(surface)}; + const auto present_modes{physical_device.GetSurfacePresentModesKHR(surface)}; - const vk::SurfaceFormatKHR surface_format{ChooseSwapSurfaceFormat(formats, srgb)}; - const vk::PresentModeKHR present_mode{ChooseSwapPresentMode(present_modes)}; + const VkSurfaceFormatKHR surface_format{ChooseSwapSurfaceFormat(formats, srgb)}; + const VkPresentModeKHR present_mode{ChooseSwapPresentMode(present_modes)}; u32 requested_image_count{capabilities.minImageCount + 1}; if (capabilities.maxImageCount > 0 && requested_image_count > capabilities.maxImageCount) { requested_image_count = capabilities.maxImageCount; } - vk::SwapchainCreateInfoKHR swapchain_ci( - {}, surface, requested_image_count, surface_format.format, surface_format.colorSpace, {}, 1, - vk::ImageUsageFlagBits::eColorAttachment, {}, {}, {}, capabilities.currentTransform, - vk::CompositeAlphaFlagBitsKHR::eOpaque, present_mode, false, {}); + VkSwapchainCreateInfoKHR swapchain_ci; + swapchain_ci.sType = VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR; + swapchain_ci.pNext = nullptr; + swapchain_ci.flags = 0; + swapchain_ci.surface = surface; + swapchain_ci.minImageCount = requested_image_count; + swapchain_ci.imageFormat = surface_format.format; + swapchain_ci.imageColorSpace = surface_format.colorSpace; + swapchain_ci.imageArrayLayers = 1; + swapchain_ci.imageUsage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; + swapchain_ci.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE; + swapchain_ci.queueFamilyIndexCount = 0; + swapchain_ci.pQueueFamilyIndices = nullptr; + swapchain_ci.preTransform = capabilities.currentTransform; + swapchain_ci.compositeAlpha = VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR; + swapchain_ci.presentMode = present_mode; + swapchain_ci.clipped = VK_FALSE; + swapchain_ci.oldSwapchain = nullptr; const u32 graphics_family{device.GetGraphicsFamily()}; const u32 present_family{device.GetPresentFamily()}; const std::array<u32, 2> queue_indices{graphics_family, present_family}; if (graphics_family != present_family) { - swapchain_ci.imageSharingMode = vk::SharingMode::eConcurrent; + swapchain_ci.imageSharingMode = VK_SHARING_MODE_CONCURRENT; swapchain_ci.queueFamilyIndexCount = static_cast<u32>(queue_indices.size()); swapchain_ci.pQueueFamilyIndices = queue_indices.data(); } else { - swapchain_ci.imageSharingMode = vk::SharingMode::eExclusive; + swapchain_ci.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE; } // Request the size again to reduce the possibility of a TOCTOU race condition. - const auto updated_capabilities = physical_device.getSurfaceCapabilitiesKHR(surface, dld); + const auto updated_capabilities = physical_device.GetSurfaceCapabilitiesKHR(surface); swapchain_ci.imageExtent = ChooseSwapExtent(updated_capabilities, width, height); // Don't add code within this and the swapchain creation. - const auto dev{device.GetLogical()}; - swapchain = dev.createSwapchainKHRUnique(swapchain_ci, nullptr, dld); + swapchain = device.GetLogical().CreateSwapchainKHR(swapchain_ci); extent = swapchain_ci.imageExtent; current_width = extent.width; current_height = extent.height; current_srgb = srgb; - images = dev.getSwapchainImagesKHR(*swapchain, dld); + images = swapchain.GetImages(); image_count = static_cast<u32>(images.size()); image_format = surface_format.format; } void VKSwapchain::CreateSemaphores() { - const auto dev{device.GetLogical()}; - const auto& dld{device.GetDispatchLoader()}; - present_semaphores.resize(image_count); - for (std::size_t i = 0; i < image_count; i++) { - present_semaphores[i] = dev.createSemaphoreUnique({}, nullptr, dld); - } + std::generate(present_semaphores.begin(), present_semaphores.end(), + [this] { return device.GetLogical().CreateSemaphore(); }); } void VKSwapchain::CreateImageViews() { - const auto dev{device.GetLogical()}; - const auto& dld{device.GetDispatchLoader()}; + VkImageViewCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + // ci.image + ci.viewType = VK_IMAGE_VIEW_TYPE_2D; + ci.format = image_format; + ci.components = {VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY, + VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY}; + ci.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; + ci.subresourceRange.baseMipLevel = 0; + ci.subresourceRange.levelCount = 1; + ci.subresourceRange.baseArrayLayer = 0; + ci.subresourceRange.layerCount = 1; image_views.resize(image_count); for (std::size_t i = 0; i < image_count; i++) { - const vk::ImageViewCreateInfo image_view_ci({}, images[i], vk::ImageViewType::e2D, - image_format, {}, - {vk::ImageAspectFlagBits::eColor, 0, 1, 0, 1}); - image_views[i] = dev.createImageViewUnique(image_view_ci, nullptr, dld); + ci.image = images[i]; + image_views[i] = device.GetLogical().CreateImageView(ci); } } diff --git a/src/video_core/renderer_vulkan/vk_swapchain.h b/src/video_core/renderer_vulkan/vk_swapchain.h index 2f3b2ccd5..a35d61345 100644 --- a/src/video_core/renderer_vulkan/vk_swapchain.h +++ b/src/video_core/renderer_vulkan/vk_swapchain.h @@ -7,7 +7,7 @@ #include <vector> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Layout { struct FramebufferLayout; @@ -20,7 +20,7 @@ class VKFence; class VKSwapchain { public: - explicit VKSwapchain(vk::SurfaceKHR surface, const VKDevice& device); + explicit VKSwapchain(VkSurfaceKHR surface, const VKDevice& device); ~VKSwapchain(); /// Creates (or recreates) the swapchain with a given size. @@ -31,12 +31,12 @@ public: /// Presents the rendered image to the swapchain. Returns true when the swapchains had to be /// recreated. Takes responsability for the ownership of fence. - bool Present(vk::Semaphore render_semaphore, VKFence& fence); + bool Present(VkSemaphore render_semaphore, VKFence& fence); /// Returns true when the framebuffer layout has changed. bool HasFramebufferChanged(const Layout::FramebufferLayout& framebuffer) const; - const vk::Extent2D& GetSize() const { + VkExtent2D GetSize() const { return extent; } @@ -48,15 +48,15 @@ public: return image_index; } - vk::Image GetImageIndex(std::size_t index) const { + VkImage GetImageIndex(std::size_t index) const { return images[index]; } - vk::ImageView GetImageViewIndex(std::size_t index) const { + VkImageView GetImageViewIndex(std::size_t index) const { return *image_views[index]; } - vk::Format GetImageFormat() const { + VkFormat GetImageFormat() const { return image_format; } @@ -65,30 +65,30 @@ public: } private: - void CreateSwapchain(const vk::SurfaceCapabilitiesKHR& capabilities, u32 width, u32 height, + void CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities, u32 width, u32 height, bool srgb); void CreateSemaphores(); void CreateImageViews(); void Destroy(); - const vk::SurfaceKHR surface; + const VkSurfaceKHR surface; const VKDevice& device; - UniqueSwapchainKHR swapchain; + vk::SwapchainKHR swapchain; std::size_t image_count{}; - std::vector<vk::Image> images; - std::vector<UniqueImageView> image_views; - std::vector<UniqueFramebuffer> framebuffers; + std::vector<VkImage> images; + std::vector<vk::ImageView> image_views; + std::vector<vk::Framebuffer> framebuffers; std::vector<VKFence*> fences; - std::vector<UniqueSemaphore> present_semaphores; + std::vector<vk::Semaphore> present_semaphores; u32 image_index{}; u32 frame_index{}; - vk::Format image_format{}; - vk::Extent2D extent{}; + VkFormat image_format{}; + VkExtent2D extent{}; u32 current_width{}; u32 current_height{}; diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index 26175921b..de4c23120 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -17,7 +17,6 @@ #include "core/memory.h" #include "video_core/engines/maxwell_3d.h" #include "video_core/morton.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/maxwell_to_vk.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_memory_manager.h" @@ -25,6 +24,7 @@ #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" #include "video_core/renderer_vulkan/vk_texture_cache.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/surface.h" #include "video_core/textures/convert.h" @@ -35,23 +35,22 @@ using VideoCore::MortonSwizzleMode; using Tegra::Texture::SwizzleSource; using VideoCore::Surface::PixelFormat; -using VideoCore::Surface::SurfaceCompression; using VideoCore::Surface::SurfaceTarget; namespace { -vk::ImageType SurfaceTargetToImage(SurfaceTarget target) { +VkImageType SurfaceTargetToImage(SurfaceTarget target) { switch (target) { case SurfaceTarget::Texture1D: case SurfaceTarget::Texture1DArray: - return vk::ImageType::e1D; + return VK_IMAGE_TYPE_1D; case SurfaceTarget::Texture2D: case SurfaceTarget::Texture2DArray: case SurfaceTarget::TextureCubemap: case SurfaceTarget::TextureCubeArray: - return vk::ImageType::e2D; + return VK_IMAGE_TYPE_2D; case SurfaceTarget::Texture3D: - return vk::ImageType::e3D; + return VK_IMAGE_TYPE_3D; case SurfaceTarget::TextureBuffer: UNREACHABLE(); return {}; @@ -60,35 +59,35 @@ vk::ImageType SurfaceTargetToImage(SurfaceTarget target) { return {}; } -vk::ImageAspectFlags PixelFormatToImageAspect(PixelFormat pixel_format) { +VkImageAspectFlags PixelFormatToImageAspect(PixelFormat pixel_format) { if (pixel_format < PixelFormat::MaxColorFormat) { - return vk::ImageAspectFlagBits::eColor; + return VK_IMAGE_ASPECT_COLOR_BIT; } else if (pixel_format < PixelFormat::MaxDepthFormat) { - return vk::ImageAspectFlagBits::eDepth; + return VK_IMAGE_ASPECT_DEPTH_BIT; } else if (pixel_format < PixelFormat::MaxDepthStencilFormat) { - return vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil; + return VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; } else { - UNREACHABLE_MSG("Invalid pixel format={}", static_cast<u32>(pixel_format)); - return vk::ImageAspectFlagBits::eColor; + UNREACHABLE_MSG("Invalid pixel format={}", static_cast<int>(pixel_format)); + return VK_IMAGE_ASPECT_COLOR_BIT; } } -vk::ImageViewType GetImageViewType(SurfaceTarget target) { +VkImageViewType GetImageViewType(SurfaceTarget target) { switch (target) { case SurfaceTarget::Texture1D: - return vk::ImageViewType::e1D; + return VK_IMAGE_VIEW_TYPE_1D; case SurfaceTarget::Texture2D: - return vk::ImageViewType::e2D; + return VK_IMAGE_VIEW_TYPE_2D; case SurfaceTarget::Texture3D: - return vk::ImageViewType::e3D; + return VK_IMAGE_VIEW_TYPE_3D; case SurfaceTarget::Texture1DArray: - return vk::ImageViewType::e1DArray; + return VK_IMAGE_VIEW_TYPE_1D_ARRAY; case SurfaceTarget::Texture2DArray: - return vk::ImageViewType::e2DArray; + return VK_IMAGE_VIEW_TYPE_2D_ARRAY; case SurfaceTarget::TextureCubemap: - return vk::ImageViewType::eCube; + return VK_IMAGE_VIEW_TYPE_CUBE; case SurfaceTarget::TextureCubeArray: - return vk::ImageViewType::eCubeArray; + return VK_IMAGE_VIEW_TYPE_CUBE_ARRAY; case SurfaceTarget::TextureBuffer: break; } @@ -96,71 +95,88 @@ vk::ImageViewType GetImageViewType(SurfaceTarget target) { return {}; } -UniqueBuffer CreateBuffer(const VKDevice& device, const SurfaceParams& params) { +vk::Buffer CreateBuffer(const VKDevice& device, const SurfaceParams& params, + std::size_t host_memory_size) { // TODO(Rodrigo): Move texture buffer creation to the buffer cache - const vk::BufferCreateInfo buffer_ci({}, params.GetHostSizeInBytes(), - vk::BufferUsageFlagBits::eUniformTexelBuffer | - vk::BufferUsageFlagBits::eTransferSrc | - vk::BufferUsageFlagBits::eTransferDst, - vk::SharingMode::eExclusive, 0, nullptr); - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - return dev.createBufferUnique(buffer_ci, nullptr, dld); + VkBufferCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.size = static_cast<VkDeviceSize>(host_memory_size); + ci.usage = VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_SRC_BIT | + VK_BUFFER_USAGE_TRANSFER_DST_BIT; + ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; + ci.queueFamilyIndexCount = 0; + ci.pQueueFamilyIndices = nullptr; + return device.GetLogical().CreateBuffer(ci); } -vk::BufferViewCreateInfo GenerateBufferViewCreateInfo(const VKDevice& device, - const SurfaceParams& params, - vk::Buffer buffer) { +VkBufferViewCreateInfo GenerateBufferViewCreateInfo(const VKDevice& device, + const SurfaceParams& params, VkBuffer buffer, + std::size_t host_memory_size) { ASSERT(params.IsBuffer()); - const auto format = - MaxwellToVK::SurfaceFormat(device, FormatType::Buffer, params.pixel_format).format; - return vk::BufferViewCreateInfo({}, buffer, format, 0, params.GetHostSizeInBytes()); + VkBufferViewCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_BUFFER_VIEW_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.buffer = buffer; + ci.format = MaxwellToVK::SurfaceFormat(device, FormatType::Buffer, params.pixel_format).format; + ci.offset = 0; + ci.range = static_cast<VkDeviceSize>(host_memory_size); + return ci; } -vk::ImageCreateInfo GenerateImageCreateInfo(const VKDevice& device, const SurfaceParams& params) { - constexpr auto sample_count = vk::SampleCountFlagBits::e1; - constexpr auto tiling = vk::ImageTiling::eOptimal; - +VkImageCreateInfo GenerateImageCreateInfo(const VKDevice& device, const SurfaceParams& params) { ASSERT(!params.IsBuffer()); const auto [format, attachable, storage] = MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, params.pixel_format); - auto image_usage = vk::ImageUsageFlagBits::eSampled | vk::ImageUsageFlagBits::eTransferDst | - vk::ImageUsageFlagBits::eTransferSrc; + VkImageCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.imageType = SurfaceTargetToImage(params.target); + ci.format = format; + ci.mipLevels = params.num_levels; + ci.arrayLayers = static_cast<u32>(params.GetNumLayers()); + ci.samples = VK_SAMPLE_COUNT_1_BIT; + ci.tiling = VK_IMAGE_TILING_OPTIMAL; + ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; + ci.queueFamilyIndexCount = 0; + ci.pQueueFamilyIndices = nullptr; + ci.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; + + ci.usage = VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT | + VK_IMAGE_USAGE_TRANSFER_SRC_BIT; if (attachable) { - image_usage |= params.IsPixelFormatZeta() ? vk::ImageUsageFlagBits::eDepthStencilAttachment - : vk::ImageUsageFlagBits::eColorAttachment; + ci.usage |= params.IsPixelFormatZeta() ? VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT + : VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; } if (storage) { - image_usage |= vk::ImageUsageFlagBits::eStorage; + ci.usage |= VK_IMAGE_USAGE_STORAGE_BIT; } - vk::ImageCreateFlags flags; - vk::Extent3D extent; switch (params.target) { case SurfaceTarget::TextureCubemap: case SurfaceTarget::TextureCubeArray: - flags |= vk::ImageCreateFlagBits::eCubeCompatible; + ci.flags |= VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT; [[fallthrough]]; case SurfaceTarget::Texture1D: case SurfaceTarget::Texture1DArray: case SurfaceTarget::Texture2D: case SurfaceTarget::Texture2DArray: - extent = vk::Extent3D(params.width, params.height, 1); + ci.extent = {params.width, params.height, 1}; break; case SurfaceTarget::Texture3D: - extent = vk::Extent3D(params.width, params.height, params.depth); + ci.extent = {params.width, params.height, params.depth}; break; case SurfaceTarget::TextureBuffer: UNREACHABLE(); } - return vk::ImageCreateInfo(flags, SurfaceTargetToImage(params.target), format, extent, - params.num_levels, static_cast<u32>(params.GetNumLayers()), - sample_count, tiling, image_usage, vk::SharingMode::eExclusive, 0, - nullptr, vk::ImageLayout::eUndefined); + return ci; } } // Anonymous namespace @@ -169,19 +185,18 @@ CachedSurface::CachedSurface(Core::System& system, const VKDevice& device, VKResourceManager& resource_manager, VKMemoryManager& memory_manager, VKScheduler& scheduler, VKStagingBufferPool& staging_pool, GPUVAddr gpu_addr, const SurfaceParams& params) - : SurfaceBase<View>{gpu_addr, params}, system{system}, device{device}, - resource_manager{resource_manager}, memory_manager{memory_manager}, scheduler{scheduler}, - staging_pool{staging_pool} { + : SurfaceBase<View>{gpu_addr, params, device.IsOptimalAstcSupported()}, system{system}, + device{device}, resource_manager{resource_manager}, + memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{staging_pool} { if (params.IsBuffer()) { - buffer = CreateBuffer(device, params); - commit = memory_manager.Commit(*buffer, false); + buffer = CreateBuffer(device, params, host_memory_size); + commit = memory_manager.Commit(buffer, false); - const auto buffer_view_ci = GenerateBufferViewCreateInfo(device, params, *buffer); + const auto buffer_view_ci = + GenerateBufferViewCreateInfo(device, params, *buffer, host_memory_size); format = buffer_view_ci.format; - const auto dev = device.GetLogical(); - const auto& dld = device.GetDispatchLoader(); - buffer_view = dev.createBufferViewUnique(buffer_view_ci, nullptr, dld); + buffer_view = device.GetLogical().CreateBufferView(buffer_view_ci); } else { const auto image_ci = GenerateImageCreateInfo(device, params); format = image_ci.format; @@ -219,16 +234,15 @@ void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) { // We can't copy images to buffers inside a renderpass scheduler.RequestOutsideRenderPassOperationContext(); - FullTransition(vk::PipelineStageFlagBits::eTransfer, vk::AccessFlagBits::eTransferRead, - vk::ImageLayout::eTransferSrcOptimal); + FullTransition(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_READ_BIT, + VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL); const auto& buffer = staging_pool.GetUnusedBuffer(host_memory_size, true); // TODO(Rodrigo): Do this in a single copy for (u32 level = 0; level < params.num_levels; ++level) { - scheduler.Record([image = image->GetHandle(), buffer = *buffer.handle, - copy = GetBufferImageCopy(level)](auto cmdbuf, auto& dld) { - cmdbuf.copyImageToBuffer(image, vk::ImageLayout::eTransferSrcOptimal, buffer, {copy}, - dld); + scheduler.Record([image = *image->GetHandle(), buffer = *buffer.handle, + copy = GetBufferImageCopy(level)](vk::CommandBuffer cmdbuf) { + cmdbuf.CopyImageToBuffer(image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, buffer, copy); }); } scheduler.Finish(); @@ -255,15 +269,27 @@ void CachedSurface::UploadBuffer(const std::vector<u8>& staging_buffer) { std::memcpy(src_buffer.commit->Map(host_memory_size), staging_buffer.data(), host_memory_size); scheduler.Record([src_buffer = *src_buffer.handle, dst_buffer = *buffer, - size = params.GetHostSizeInBytes()](auto cmdbuf, auto& dld) { - const vk::BufferCopy copy(0, 0, size); - cmdbuf.copyBuffer(src_buffer, dst_buffer, {copy}, dld); - - cmdbuf.pipelineBarrier( - vk::PipelineStageFlagBits::eTransfer, vk::PipelineStageFlagBits::eVertexShader, {}, {}, - {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, - vk::AccessFlagBits::eShaderRead, 0, 0, dst_buffer, 0, size)}, - {}, dld); + size = host_memory_size](vk::CommandBuffer cmdbuf) { + VkBufferCopy copy; + copy.srcOffset = 0; + copy.dstOffset = 0; + copy.size = size; + cmdbuf.CopyBuffer(src_buffer, dst_buffer, copy); + + VkBufferMemoryBarrier barrier; + barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; + barrier.pNext = nullptr; + barrier.srcAccessMask = VK_PIPELINE_STAGE_TRANSFER_BIT; + barrier.dstAccessMask = VK_PIPELINE_STAGE_VERTEX_SHADER_BIT; + barrier.srcQueueFamilyIndex = VK_ACCESS_TRANSFER_WRITE_BIT; + barrier.dstQueueFamilyIndex = VK_ACCESS_SHADER_READ_BIT; + barrier.srcQueueFamilyIndex = 0; + barrier.dstQueueFamilyIndex = 0; + barrier.buffer = dst_buffer; + barrier.offset = 0; + barrier.size = size; + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT, + 0, {}, barrier, {}); }); } @@ -271,46 +297,49 @@ void CachedSurface::UploadImage(const std::vector<u8>& staging_buffer) { const auto& src_buffer = staging_pool.GetUnusedBuffer(host_memory_size, true); std::memcpy(src_buffer.commit->Map(host_memory_size), staging_buffer.data(), host_memory_size); - FullTransition(vk::PipelineStageFlagBits::eTransfer, vk::AccessFlagBits::eTransferWrite, - vk::ImageLayout::eTransferDstOptimal); + FullTransition(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, + VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL); for (u32 level = 0; level < params.num_levels; ++level) { - vk::BufferImageCopy copy = GetBufferImageCopy(level); - if (image->GetAspectMask() == - (vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil)) { - vk::BufferImageCopy depth = copy; - vk::BufferImageCopy stencil = copy; - depth.imageSubresource.aspectMask = vk::ImageAspectFlagBits::eDepth; - stencil.imageSubresource.aspectMask = vk::ImageAspectFlagBits::eStencil; - scheduler.Record([buffer = *src_buffer.handle, image = image->GetHandle(), depth, - stencil](auto cmdbuf, auto& dld) { - cmdbuf.copyBufferToImage(buffer, image, vk::ImageLayout::eTransferDstOptimal, - {depth, stencil}, dld); + const VkBufferImageCopy copy = GetBufferImageCopy(level); + if (image->GetAspectMask() == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) { + scheduler.Record([buffer = *src_buffer.handle, image = *image->GetHandle(), + copy](vk::CommandBuffer cmdbuf) { + std::array<VkBufferImageCopy, 2> copies = {copy, copy}; + copies[0].imageSubresource.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT; + copies[1].imageSubresource.aspectMask = VK_IMAGE_ASPECT_STENCIL_BIT; + cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, + copies); }); } else { - scheduler.Record([buffer = *src_buffer.handle, image = image->GetHandle(), - copy](auto cmdbuf, auto& dld) { - cmdbuf.copyBufferToImage(buffer, image, vk::ImageLayout::eTransferDstOptimal, - {copy}, dld); + scheduler.Record([buffer = *src_buffer.handle, image = *image->GetHandle(), + copy](vk::CommandBuffer cmdbuf) { + cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy); }); } } } -vk::BufferImageCopy CachedSurface::GetBufferImageCopy(u32 level) const { - const u32 vk_depth = params.target == SurfaceTarget::Texture3D ? params.GetMipDepth(level) : 1; - const auto compression_type = params.GetCompressionType(); - const std::size_t mip_offset = compression_type == SurfaceCompression::Converted - ? params.GetConvertedMipmapOffset(level) - : params.GetHostMipmapLevelOffset(level); - - return vk::BufferImageCopy( - mip_offset, 0, 0, - {image->GetAspectMask(), level, 0, static_cast<u32>(params.GetNumLayers())}, {0, 0, 0}, - {params.GetMipWidth(level), params.GetMipHeight(level), vk_depth}); +VkBufferImageCopy CachedSurface::GetBufferImageCopy(u32 level) const { + VkBufferImageCopy copy; + copy.bufferOffset = params.GetHostMipmapLevelOffset(level, is_converted); + copy.bufferRowLength = 0; + copy.bufferImageHeight = 0; + copy.imageSubresource.aspectMask = image->GetAspectMask(); + copy.imageSubresource.mipLevel = level; + copy.imageSubresource.baseArrayLayer = 0; + copy.imageSubresource.layerCount = static_cast<u32>(params.GetNumLayers()); + copy.imageOffset.x = 0; + copy.imageOffset.y = 0; + copy.imageOffset.z = 0; + copy.imageExtent.width = params.GetMipWidth(level); + copy.imageExtent.height = params.GetMipHeight(level); + copy.imageExtent.depth = + params.target == SurfaceTarget::Texture3D ? params.GetMipDepth(level) : 1; + return copy; } -vk::ImageSubresourceRange CachedSurface::GetImageSubresourceRange() const { +VkImageSubresourceRange CachedSurface::GetImageSubresourceRange() const { return {image->GetAspectMask(), 0, params.num_levels, 0, static_cast<u32>(params.GetNumLayers())}; } @@ -322,12 +351,12 @@ CachedSurfaceView::CachedSurfaceView(const VKDevice& device, CachedSurface& surf aspect_mask{surface.GetAspectMask()}, device{device}, surface{surface}, base_layer{params.base_layer}, num_layers{params.num_layers}, base_level{params.base_level}, num_levels{params.num_levels}, image_view_type{image ? GetImageViewType(params.target) - : vk::ImageViewType{}} {} + : VK_IMAGE_VIEW_TYPE_1D} {} CachedSurfaceView::~CachedSurfaceView() = default; -vk::ImageView CachedSurfaceView::GetHandle(SwizzleSource x_source, SwizzleSource y_source, - SwizzleSource z_source, SwizzleSource w_source) { +VkImageView CachedSurfaceView::GetHandle(SwizzleSource x_source, SwizzleSource y_source, + SwizzleSource z_source, SwizzleSource w_source) { const u32 swizzle = EncodeSwizzle(x_source, y_source, z_source, w_source); if (last_image_view && last_swizzle == swizzle) { return last_image_view; @@ -352,37 +381,45 @@ vk::ImageView CachedSurfaceView::GetHandle(SwizzleSource x_source, SwizzleSource // Games can sample depth or stencil values on textures. This is decided by the swizzle value on // hardware. To emulate this on Vulkan we specify it in the aspect. - vk::ImageAspectFlags aspect = aspect_mask; - if (aspect == (vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil)) { + VkImageAspectFlags aspect = aspect_mask; + if (aspect == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) { UNIMPLEMENTED_IF(x_source != SwizzleSource::R && x_source != SwizzleSource::G); const bool is_first = x_source == SwizzleSource::R; switch (params.pixel_format) { case VideoCore::Surface::PixelFormat::Z24S8: case VideoCore::Surface::PixelFormat::Z32FS8: - aspect = is_first ? vk::ImageAspectFlagBits::eDepth : vk::ImageAspectFlagBits::eStencil; + aspect = is_first ? VK_IMAGE_ASPECT_DEPTH_BIT : VK_IMAGE_ASPECT_STENCIL_BIT; break; case VideoCore::Surface::PixelFormat::S8Z24: - aspect = is_first ? vk::ImageAspectFlagBits::eStencil : vk::ImageAspectFlagBits::eDepth; + aspect = is_first ? VK_IMAGE_ASPECT_STENCIL_BIT : VK_IMAGE_ASPECT_DEPTH_BIT; break; default: - aspect = vk::ImageAspectFlagBits::eDepth; + aspect = VK_IMAGE_ASPECT_DEPTH_BIT; UNIMPLEMENTED(); } // Vulkan doesn't seem to understand swizzling of a depth stencil image, use identity - swizzle_x = vk::ComponentSwizzle::eR; - swizzle_y = vk::ComponentSwizzle::eG; - swizzle_z = vk::ComponentSwizzle::eB; - swizzle_w = vk::ComponentSwizzle::eA; + swizzle_x = VK_COMPONENT_SWIZZLE_R; + swizzle_y = VK_COMPONENT_SWIZZLE_G; + swizzle_z = VK_COMPONENT_SWIZZLE_B; + swizzle_w = VK_COMPONENT_SWIZZLE_A; } - const vk::ImageViewCreateInfo image_view_ci( - {}, surface.GetImageHandle(), image_view_type, surface.GetImage().GetFormat(), - {swizzle_x, swizzle_y, swizzle_z, swizzle_w}, - {aspect, base_level, num_levels, base_layer, num_layers}); + VkImageViewCreateInfo ci; + ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; + ci.pNext = nullptr; + ci.flags = 0; + ci.image = surface.GetImageHandle(); + ci.viewType = image_view_type; + ci.format = surface.GetImage().GetFormat(); + ci.components = {swizzle_x, swizzle_y, swizzle_z, swizzle_w}; + ci.subresourceRange.aspectMask = aspect; + ci.subresourceRange.baseMipLevel = base_level; + ci.subresourceRange.levelCount = num_levels; + ci.subresourceRange.baseArrayLayer = base_layer; + ci.subresourceRange.layerCount = num_layers; + image_view = device.GetLogical().CreateImageView(ci); - const auto dev = device.GetLogical(); - image_view = dev.createImageViewUnique(image_view_ci, nullptr, device.GetDispatchLoader()); return last_image_view = *image_view; } @@ -390,8 +427,9 @@ VKTextureCache::VKTextureCache(Core::System& system, VideoCore::RasterizerInterf const VKDevice& device, VKResourceManager& resource_manager, VKMemoryManager& memory_manager, VKScheduler& scheduler, VKStagingBufferPool& staging_pool) - : TextureCache(system, rasterizer), device{device}, resource_manager{resource_manager}, - memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{staging_pool} {} + : TextureCache(system, rasterizer, device.IsOptimalAstcSupported()), device{device}, + resource_manager{resource_manager}, memory_manager{memory_manager}, scheduler{scheduler}, + staging_pool{staging_pool} {} VKTextureCache::~VKTextureCache() = default; @@ -418,25 +456,36 @@ void VKTextureCache::ImageCopy(Surface& src_surface, Surface& dst_surface, scheduler.RequestOutsideRenderPassOperationContext(); src_surface->Transition(copy_params.source_z, copy_params.depth, copy_params.source_level, 1, - vk::PipelineStageFlagBits::eTransfer, vk::AccessFlagBits::eTransferRead, - vk::ImageLayout::eTransferSrcOptimal); - dst_surface->Transition( - dst_base_layer, num_layers, copy_params.dest_level, 1, vk::PipelineStageFlagBits::eTransfer, - vk::AccessFlagBits::eTransferWrite, vk::ImageLayout::eTransferDstOptimal); - - const vk::ImageSubresourceLayers src_subresource( - src_surface->GetAspectMask(), copy_params.source_level, copy_params.source_z, num_layers); - const vk::ImageSubresourceLayers dst_subresource( - dst_surface->GetAspectMask(), copy_params.dest_level, dst_base_layer, num_layers); - const vk::Offset3D src_offset(copy_params.source_x, copy_params.source_y, 0); - const vk::Offset3D dst_offset(copy_params.dest_x, copy_params.dest_y, dst_offset_z); - const vk::Extent3D extent(copy_params.width, copy_params.height, extent_z); - const vk::ImageCopy copy(src_subresource, src_offset, dst_subresource, dst_offset, extent); - const vk::Image src_image = src_surface->GetImageHandle(); - const vk::Image dst_image = dst_surface->GetImageHandle(); - scheduler.Record([src_image, dst_image, copy](auto cmdbuf, auto& dld) { - cmdbuf.copyImage(src_image, vk::ImageLayout::eTransferSrcOptimal, dst_image, - vk::ImageLayout::eTransferDstOptimal, {copy}, dld); + VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_READ_BIT, + VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL); + dst_surface->Transition(dst_base_layer, num_layers, copy_params.dest_level, 1, + VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, + VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL); + + VkImageCopy copy; + copy.srcSubresource.aspectMask = src_surface->GetAspectMask(); + copy.srcSubresource.mipLevel = copy_params.source_level; + copy.srcSubresource.baseArrayLayer = copy_params.source_z; + copy.srcSubresource.layerCount = num_layers; + copy.srcOffset.x = copy_params.source_x; + copy.srcOffset.y = copy_params.source_y; + copy.srcOffset.z = 0; + copy.dstSubresource.aspectMask = dst_surface->GetAspectMask(); + copy.dstSubresource.mipLevel = copy_params.dest_level; + copy.dstSubresource.baseArrayLayer = dst_base_layer; + copy.dstSubresource.layerCount = num_layers; + copy.dstOffset.x = copy_params.dest_x; + copy.dstOffset.y = copy_params.dest_y; + copy.dstOffset.z = dst_offset_z; + copy.extent.width = copy_params.width; + copy.extent.height = copy_params.height; + copy.extent.depth = extent_z; + + const VkImage src_image = src_surface->GetImageHandle(); + const VkImage dst_image = dst_surface->GetImageHandle(); + scheduler.Record([src_image, dst_image, copy](vk::CommandBuffer cmdbuf) { + cmdbuf.CopyImage(src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_image, + VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy); }); } @@ -445,25 +494,34 @@ void VKTextureCache::ImageBlit(View& src_view, View& dst_view, // We can't blit inside a renderpass scheduler.RequestOutsideRenderPassOperationContext(); - src_view->Transition(vk::ImageLayout::eTransferSrcOptimal, vk::PipelineStageFlagBits::eTransfer, - vk::AccessFlagBits::eTransferRead); - dst_view->Transition(vk::ImageLayout::eTransferDstOptimal, vk::PipelineStageFlagBits::eTransfer, - vk::AccessFlagBits::eTransferWrite); - - const auto& cfg = copy_config; - const auto src_top_left = vk::Offset3D(cfg.src_rect.left, cfg.src_rect.top, 0); - const auto src_bot_right = vk::Offset3D(cfg.src_rect.right, cfg.src_rect.bottom, 1); - const auto dst_top_left = vk::Offset3D(cfg.dst_rect.left, cfg.dst_rect.top, 0); - const auto dst_bot_right = vk::Offset3D(cfg.dst_rect.right, cfg.dst_rect.bottom, 1); - const vk::ImageBlit blit(src_view->GetImageSubresourceLayers(), {src_top_left, src_bot_right}, - dst_view->GetImageSubresourceLayers(), {dst_top_left, dst_bot_right}); + src_view->Transition(VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, VK_PIPELINE_STAGE_TRANSFER_BIT, + VK_ACCESS_TRANSFER_READ_BIT); + dst_view->Transition(VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, VK_PIPELINE_STAGE_TRANSFER_BIT, + VK_ACCESS_TRANSFER_WRITE_BIT); + + VkImageBlit blit; + blit.srcSubresource = src_view->GetImageSubresourceLayers(); + blit.srcOffsets[0].x = copy_config.src_rect.left; + blit.srcOffsets[0].y = copy_config.src_rect.top; + blit.srcOffsets[0].z = 0; + blit.srcOffsets[1].x = copy_config.src_rect.right; + blit.srcOffsets[1].y = copy_config.src_rect.bottom; + blit.srcOffsets[1].z = 1; + blit.dstSubresource = dst_view->GetImageSubresourceLayers(); + blit.dstOffsets[0].x = copy_config.dst_rect.left; + blit.dstOffsets[0].y = copy_config.dst_rect.top; + blit.dstOffsets[0].z = 0; + blit.dstOffsets[1].x = copy_config.dst_rect.right; + blit.dstOffsets[1].y = copy_config.dst_rect.bottom; + blit.dstOffsets[1].z = 1; + const bool is_linear = copy_config.filter == Tegra::Engines::Fermi2D::Filter::Linear; scheduler.Record([src_image = src_view->GetImage(), dst_image = dst_view->GetImage(), blit, - is_linear](auto cmdbuf, auto& dld) { - cmdbuf.blitImage(src_image, vk::ImageLayout::eTransferSrcOptimal, dst_image, - vk::ImageLayout::eTransferDstOptimal, {blit}, - is_linear ? vk::Filter::eLinear : vk::Filter::eNearest, dld); + is_linear](vk::CommandBuffer cmdbuf) { + cmdbuf.BlitImage(src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_image, + VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, blit, + is_linear ? VK_FILTER_LINEAR : VK_FILTER_NEAREST); }); } diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h index 22e3d34de..115595f28 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.h +++ b/src/video_core/renderer_vulkan/vk_texture_cache.h @@ -13,10 +13,10 @@ #include "common/math_util.h" #include "video_core/gpu.h" #include "video_core/rasterizer_cache.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_image.h" #include "video_core/renderer_vulkan/vk_memory_manager.h" #include "video_core/renderer_vulkan/vk_scheduler.h" +#include "video_core/renderer_vulkan/wrapper.h" #include "video_core/texture_cache/surface_base.h" #include "video_core/texture_cache/texture_cache.h" #include "video_core/textures/decoders.h" @@ -60,15 +60,15 @@ public: void UploadTexture(const std::vector<u8>& staging_buffer) override; void DownloadTexture(std::vector<u8>& staging_buffer) override; - void FullTransition(vk::PipelineStageFlags new_stage_mask, vk::AccessFlags new_access, - vk::ImageLayout new_layout) { + void FullTransition(VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access, + VkImageLayout new_layout) { image->Transition(0, static_cast<u32>(params.GetNumLayers()), 0, params.num_levels, new_stage_mask, new_access, new_layout); } void Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels, - vk::PipelineStageFlags new_stage_mask, vk::AccessFlags new_access, - vk::ImageLayout new_layout) { + VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access, + VkImageLayout new_layout) { image->Transition(base_layer, num_layers, base_level, num_levels, new_stage_mask, new_access, new_layout); } @@ -81,15 +81,15 @@ public: return *image; } - vk::Image GetImageHandle() const { - return image->GetHandle(); + VkImage GetImageHandle() const { + return *image->GetHandle(); } - vk::ImageAspectFlags GetAspectMask() const { + VkImageAspectFlags GetAspectMask() const { return image->GetAspectMask(); } - vk::BufferView GetBufferViewHandle() const { + VkBufferView GetBufferViewHandle() const { return *buffer_view; } @@ -104,9 +104,9 @@ private: void UploadImage(const std::vector<u8>& staging_buffer); - vk::BufferImageCopy GetBufferImageCopy(u32 level) const; + VkBufferImageCopy GetBufferImageCopy(u32 level) const; - vk::ImageSubresourceRange GetImageSubresourceRange() const; + VkImageSubresourceRange GetImageSubresourceRange() const; Core::System& system; const VKDevice& device; @@ -116,11 +116,11 @@ private: VKStagingBufferPool& staging_pool; std::optional<VKImage> image; - UniqueBuffer buffer; - UniqueBufferView buffer_view; + vk::Buffer buffer; + vk::BufferView buffer_view; VKMemoryCommit commit; - vk::Format format; + VkFormat format = VK_FORMAT_UNDEFINED; }; class CachedSurfaceView final : public VideoCommon::ViewBase { @@ -129,16 +129,16 @@ public: const ViewParams& params, bool is_proxy); ~CachedSurfaceView(); - vk::ImageView GetHandle(Tegra::Texture::SwizzleSource x_source, - Tegra::Texture::SwizzleSource y_source, - Tegra::Texture::SwizzleSource z_source, - Tegra::Texture::SwizzleSource w_source); + VkImageView GetHandle(Tegra::Texture::SwizzleSource x_source, + Tegra::Texture::SwizzleSource y_source, + Tegra::Texture::SwizzleSource z_source, + Tegra::Texture::SwizzleSource w_source); bool IsSameSurface(const CachedSurfaceView& rhs) const { return &surface == &rhs.surface; } - vk::ImageView GetHandle() { + VkImageView GetHandle() { return GetHandle(Tegra::Texture::SwizzleSource::R, Tegra::Texture::SwizzleSource::G, Tegra::Texture::SwizzleSource::B, Tegra::Texture::SwizzleSource::A); } @@ -159,24 +159,24 @@ public: return buffer_view; } - vk::Image GetImage() const { + VkImage GetImage() const { return image; } - vk::BufferView GetBufferView() const { + VkBufferView GetBufferView() const { return buffer_view; } - vk::ImageSubresourceRange GetImageSubresourceRange() const { + VkImageSubresourceRange GetImageSubresourceRange() const { return {aspect_mask, base_level, num_levels, base_layer, num_layers}; } - vk::ImageSubresourceLayers GetImageSubresourceLayers() const { + VkImageSubresourceLayers GetImageSubresourceLayers() const { return {surface.GetAspectMask(), base_level, base_layer, num_layers}; } - void Transition(vk::ImageLayout new_layout, vk::PipelineStageFlags new_stage_mask, - vk::AccessFlags new_access) const { + void Transition(VkImageLayout new_layout, VkPipelineStageFlags new_stage_mask, + VkAccessFlags new_access) const { surface.Transition(base_layer, num_layers, base_level, num_levels, new_stage_mask, new_access, new_layout); } @@ -196,9 +196,9 @@ private: // Store a copy of these values to avoid double dereference when reading them const SurfaceParams params; - const vk::Image image; - const vk::BufferView buffer_view; - const vk::ImageAspectFlags aspect_mask; + const VkImage image; + const VkBufferView buffer_view; + const VkImageAspectFlags aspect_mask; const VKDevice& device; CachedSurface& surface; @@ -206,12 +206,12 @@ private: const u32 num_layers; const u32 base_level; const u32 num_levels; - const vk::ImageViewType image_view_type; + const VkImageViewType image_view_type; - vk::ImageView last_image_view; - u32 last_swizzle{}; + VkImageView last_image_view = nullptr; + u32 last_swizzle = 0; - std::unordered_map<u32, UniqueImageView> view_cache; + std::unordered_map<u32, vk::ImageView> view_cache; }; class VKTextureCache final : public TextureCacheBase { diff --git a/src/video_core/renderer_vulkan/vk_update_descriptor.cpp b/src/video_core/renderer_vulkan/vk_update_descriptor.cpp index 0e577b9ff..4bfec0077 100644 --- a/src/video_core/renderer_vulkan/vk_update_descriptor.cpp +++ b/src/video_core/renderer_vulkan/vk_update_descriptor.cpp @@ -7,10 +7,10 @@ #include "common/assert.h" #include "common/logging/log.h" -#include "video_core/renderer_vulkan/declarations.h" #include "video_core/renderer_vulkan/vk_device.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -27,8 +27,8 @@ void VKUpdateDescriptorQueue::Acquire() { entries.clear(); } -void VKUpdateDescriptorQueue::Send(vk::DescriptorUpdateTemplate update_template, - vk::DescriptorSet set) { +void VKUpdateDescriptorQueue::Send(VkDescriptorUpdateTemplateKHR update_template, + VkDescriptorSet set) { if (payload.size() + entries.size() >= payload.max_size()) { LOG_WARNING(Render_Vulkan, "Payload overflow, waiting for worker thread"); scheduler.WaitWorker(); @@ -37,21 +37,21 @@ void VKUpdateDescriptorQueue::Send(vk::DescriptorUpdateTemplate update_template, const auto payload_start = payload.data() + payload.size(); for (const auto& entry : entries) { - if (const auto image = std::get_if<vk::DescriptorImageInfo>(&entry)) { + if (const auto image = std::get_if<VkDescriptorImageInfo>(&entry)) { payload.push_back(*image); } else if (const auto buffer = std::get_if<Buffer>(&entry)) { payload.emplace_back(*buffer->buffer, buffer->offset, buffer->size); - } else if (const auto texel = std::get_if<vk::BufferView>(&entry)) { + } else if (const auto texel = std::get_if<VkBufferView>(&entry)) { payload.push_back(*texel); } else { UNREACHABLE(); } } - scheduler.Record([dev = device.GetLogical(), payload_start, set, - update_template]([[maybe_unused]] auto cmdbuf, auto& dld) { - dev.updateDescriptorSetWithTemplate(set, update_template, payload_start, dld); - }); + scheduler.Record( + [payload_start, set, update_template, logical = &device.GetLogical()](vk::CommandBuffer) { + logical->UpdateDescriptorSet(set, update_template, payload_start); + }); } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_update_descriptor.h b/src/video_core/renderer_vulkan/vk_update_descriptor.h index 8c825aa29..a9e3d5dba 100644 --- a/src/video_core/renderer_vulkan/vk_update_descriptor.h +++ b/src/video_core/renderer_vulkan/vk_update_descriptor.h @@ -9,7 +9,7 @@ #include <boost/container/static_vector.hpp> #include "common/common_types.h" -#include "video_core/renderer_vulkan/declarations.h" +#include "video_core/renderer_vulkan/wrapper.h" namespace Vulkan { @@ -20,18 +20,18 @@ class DescriptorUpdateEntry { public: explicit DescriptorUpdateEntry() : image{} {} - DescriptorUpdateEntry(vk::DescriptorImageInfo image) : image{image} {} + DescriptorUpdateEntry(VkDescriptorImageInfo image) : image{image} {} - DescriptorUpdateEntry(vk::Buffer buffer, vk::DeviceSize offset, vk::DeviceSize size) + DescriptorUpdateEntry(VkBuffer buffer, VkDeviceSize offset, VkDeviceSize size) : buffer{buffer, offset, size} {} - DescriptorUpdateEntry(vk::BufferView texel_buffer) : texel_buffer{texel_buffer} {} + DescriptorUpdateEntry(VkBufferView texel_buffer) : texel_buffer{texel_buffer} {} private: union { - vk::DescriptorImageInfo image; - vk::DescriptorBufferInfo buffer; - vk::BufferView texel_buffer; + VkDescriptorImageInfo image; + VkDescriptorBufferInfo buffer; + VkBufferView texel_buffer; }; }; @@ -44,37 +44,35 @@ public: void Acquire(); - void Send(vk::DescriptorUpdateTemplate update_template, vk::DescriptorSet set); + void Send(VkDescriptorUpdateTemplateKHR update_template, VkDescriptorSet set); - void AddSampledImage(vk::Sampler sampler, vk::ImageView image_view) { - entries.emplace_back(vk::DescriptorImageInfo{sampler, image_view, {}}); + void AddSampledImage(VkSampler sampler, VkImageView image_view) { + entries.emplace_back(VkDescriptorImageInfo{sampler, image_view, {}}); } - void AddImage(vk::ImageView image_view) { - entries.emplace_back(vk::DescriptorImageInfo{{}, image_view, {}}); + void AddImage(VkImageView image_view) { + entries.emplace_back(VkDescriptorImageInfo{{}, image_view, {}}); } - void AddBuffer(const vk::Buffer* buffer, u64 offset, std::size_t size) { + void AddBuffer(const VkBuffer* buffer, u64 offset, std::size_t size) { entries.push_back(Buffer{buffer, offset, size}); } - void AddTexelBuffer(vk::BufferView texel_buffer) { + void AddTexelBuffer(VkBufferView texel_buffer) { entries.emplace_back(texel_buffer); } - vk::ImageLayout* GetLastImageLayout() { - return &std::get<vk::DescriptorImageInfo>(entries.back()).imageLayout; + VkImageLayout* GetLastImageLayout() { + return &std::get<VkDescriptorImageInfo>(entries.back()).imageLayout; } private: struct Buffer { - const vk::Buffer* buffer{}; - u64 offset{}; - std::size_t size{}; + const VkBuffer* buffer = nullptr; + u64 offset = 0; + std::size_t size = 0; }; - using Variant = std::variant<vk::DescriptorImageInfo, Buffer, vk::BufferView>; - // Old gcc versions don't consider this trivially copyable. - // static_assert(std::is_trivially_copyable_v<Variant>); + using Variant = std::variant<VkDescriptorImageInfo, Buffer, VkBufferView>; const VKDevice& device; VKScheduler& scheduler; diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp index 478394682..4db329fa5 100644 --- a/src/video_core/shader/decode/arithmetic.cpp +++ b/src/video_core/shader/decode/arithmetic.cpp @@ -136,7 +136,8 @@ u32 ShaderIR::DecodeArithmetic(NodeBlock& bb, u32 pc) { SetRegister(bb, instr.gpr0, value); break; } - case OpCode::Id::FCMP_R: { + case OpCode::Id::FCMP_RR: + case OpCode::Id::FCMP_RC: { UNIMPLEMENTED_IF(instr.fcmp.ftz == 0); Node op_c = GetRegister(instr.gpr39); Node comp = GetPredicateComparisonFloat(instr.fcmp.cond, std::move(op_c), Immediate(0.0f)); diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp index c72690b2b..b9989c88c 100644 --- a/src/video_core/shader/decode/conversion.cpp +++ b/src/video_core/shader/decode/conversion.cpp @@ -2,6 +2,10 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <limits> +#include <optional> +#include <utility> + #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" @@ -15,9 +19,49 @@ using Tegra::Shader::OpCode; using Tegra::Shader::Register; namespace { + constexpr OperationCode GetFloatSelector(u64 selector) { return selector == 0 ? OperationCode::FCastHalf0 : OperationCode::FCastHalf1; } + +constexpr u32 SizeInBits(Register::Size size) { + switch (size) { + case Register::Size::Byte: + return 8; + case Register::Size::Short: + return 16; + case Register::Size::Word: + return 32; + case Register::Size::Long: + return 64; + } + return 0; +} + +constexpr std::optional<std::pair<s32, s32>> IntegerSaturateBounds(Register::Size src_size, + Register::Size dst_size, + bool src_signed, + bool dst_signed) { + const u32 dst_bits = SizeInBits(dst_size); + if (src_size == Register::Size::Word && dst_size == Register::Size::Word) { + if (src_signed == dst_signed) { + return std::nullopt; + } + return std::make_pair(0, std::numeric_limits<s32>::max()); + } + if (dst_signed) { + // Signed destination, clamp to [-128, 127] for instance + return std::make_pair(-(1 << (dst_bits - 1)), (1 << (dst_bits - 1)) - 1); + } else { + // Unsigned destination + if (dst_bits == 32) { + // Avoid shifting by 32, that is undefined behavior + return std::make_pair(0, s32(std::numeric_limits<u32>::max())); + } + return std::make_pair(0, (1 << dst_bits) - 1); + } +} + } // Anonymous namespace u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) { @@ -28,14 +72,13 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) { case OpCode::Id::I2I_R: case OpCode::Id::I2I_C: case OpCode::Id::I2I_IMM: { - UNIMPLEMENTED_IF(instr.conversion.int_src.selector != 0); - UNIMPLEMENTED_IF(instr.conversion.dst_size != Register::Size::Word); - UNIMPLEMENTED_IF(instr.alu.saturate_d); + const bool src_signed = instr.conversion.is_input_signed; + const bool dst_signed = instr.conversion.is_output_signed; + const Register::Size src_size = instr.conversion.src_size; + const Register::Size dst_size = instr.conversion.dst_size; + const u32 selector = static_cast<u32>(instr.conversion.int_src.selector); - const bool input_signed = instr.conversion.is_input_signed; - const bool output_signed = instr.conversion.is_output_signed; - - Node value = [&]() { + Node value = [this, instr, opcode] { switch (opcode->get().GetId()) { case OpCode::Id::I2I_R: return GetRegister(instr.gpr20); @@ -48,16 +91,60 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) { return Immediate(0); } }(); - value = ConvertIntegerSize(value, instr.conversion.src_size, input_signed); - value = GetOperandAbsNegInteger(value, instr.conversion.abs_a, instr.conversion.negate_a, - input_signed); - if (input_signed != output_signed) { - value = SignedOperation(OperationCode::ICastUnsigned, output_signed, NO_PRECISE, value); + // Ensure the source selector is valid + switch (instr.conversion.src_size) { + case Register::Size::Byte: + break; + case Register::Size::Short: + ASSERT(selector == 0 || selector == 2); + break; + default: + ASSERT(selector == 0); + break; + } + + if (src_size != Register::Size::Word || selector != 0) { + value = SignedOperation(OperationCode::IBitfieldExtract, src_signed, std::move(value), + Immediate(selector * 8), Immediate(SizeInBits(src_size))); + } + + value = GetOperandAbsNegInteger(std::move(value), instr.conversion.abs_a, + instr.conversion.negate_a, src_signed); + + if (instr.alu.saturate_d) { + if (src_signed && !dst_signed) { + Node is_negative = Operation(OperationCode::LogicalUGreaterEqual, value, + Immediate(1 << (SizeInBits(src_size) - 1))); + value = Operation(OperationCode::Select, std::move(is_negative), Immediate(0), + std::move(value)); + + // Simplify generated expressions, this can be removed without semantic impact + SetTemporary(bb, 0, std::move(value)); + value = GetTemporary(0); + + if (dst_size != Register::Size::Word) { + const Node limit = Immediate((1 << SizeInBits(dst_size)) - 1); + Node is_large = + Operation(OperationCode::LogicalUGreaterThan, std::move(value), limit); + value = Operation(OperationCode::Select, std::move(is_large), limit, + std::move(value)); + } + } else if (const std::optional bounds = + IntegerSaturateBounds(src_size, dst_size, src_signed, dst_signed)) { + value = SignedOperation(OperationCode::IMax, src_signed, std::move(value), + Immediate(bounds->first)); + value = SignedOperation(OperationCode::IMin, src_signed, std::move(value), + Immediate(bounds->second)); + } + } else if (dst_size != Register::Size::Word) { + // No saturation, we only have to mask the result + Node mask = Immediate((1 << SizeInBits(dst_size)) - 1); + value = Operation(OperationCode::UBitwiseAnd, std::move(value), std::move(mask)); } SetInternalFlagsFromInteger(bb, value, instr.generates_cc); - SetRegister(bb, instr.gpr0, value); + SetRegister(bb, instr.gpr0, std::move(value)); break; } case OpCode::Id::I2F_R: diff --git a/src/video_core/shader/decode/image.cpp b/src/video_core/shader/decode/image.cpp index d2fe4ec5d..0dd7a1196 100644 --- a/src/video_core/shader/decode/image.cpp +++ b/src/video_core/shader/decode/image.cpp @@ -13,13 +13,247 @@ #include "video_core/engines/shader_bytecode.h" #include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" +#include "video_core/textures/texture.h" namespace VideoCommon::Shader { using Tegra::Shader::Instruction; using Tegra::Shader::OpCode; +using Tegra::Shader::PredCondition; +using Tegra::Shader::StoreType; +using Tegra::Texture::ComponentType; +using Tegra::Texture::TextureFormat; +using Tegra::Texture::TICEntry; namespace { + +ComponentType GetComponentType(Tegra::Engines::SamplerDescriptor descriptor, + std::size_t component) { + const TextureFormat format{descriptor.format}; + switch (format) { + case TextureFormat::R16_G16_B16_A16: + case TextureFormat::R32_G32_B32_A32: + case TextureFormat::R32_G32_B32: + case TextureFormat::R32_G32: + case TextureFormat::R16_G16: + case TextureFormat::R32: + case TextureFormat::R16: + case TextureFormat::R8: + case TextureFormat::R1: + if (component == 0) { + return descriptor.r_type; + } + if (component == 1) { + return descriptor.g_type; + } + if (component == 2) { + return descriptor.b_type; + } + if (component == 3) { + return descriptor.a_type; + } + break; + case TextureFormat::A8R8G8B8: + if (component == 0) { + return descriptor.a_type; + } + if (component == 1) { + return descriptor.r_type; + } + if (component == 2) { + return descriptor.g_type; + } + if (component == 3) { + return descriptor.b_type; + } + break; + case TextureFormat::A2B10G10R10: + case TextureFormat::A4B4G4R4: + case TextureFormat::A5B5G5R1: + case TextureFormat::A1B5G5R5: + if (component == 0) { + return descriptor.a_type; + } + if (component == 1) { + return descriptor.b_type; + } + if (component == 2) { + return descriptor.g_type; + } + if (component == 3) { + return descriptor.r_type; + } + break; + case TextureFormat::R32_B24G8: + if (component == 0) { + return descriptor.r_type; + } + if (component == 1) { + return descriptor.b_type; + } + if (component == 2) { + return descriptor.g_type; + } + break; + case TextureFormat::B5G6R5: + case TextureFormat::B6G5R5: + if (component == 0) { + return descriptor.b_type; + } + if (component == 1) { + return descriptor.g_type; + } + if (component == 2) { + return descriptor.r_type; + } + break; + case TextureFormat::G8R24: + case TextureFormat::G24R8: + case TextureFormat::G8R8: + case TextureFormat::G4R4: + if (component == 0) { + return descriptor.g_type; + } + if (component == 1) { + return descriptor.r_type; + } + break; + } + UNIMPLEMENTED_MSG("texture format not implement={}", format); + return ComponentType::FLOAT; +} + +bool IsComponentEnabled(std::size_t component_mask, std::size_t component) { + constexpr u8 R = 0b0001; + constexpr u8 G = 0b0010; + constexpr u8 B = 0b0100; + constexpr u8 A = 0b1000; + constexpr std::array<u8, 16> mask = { + 0, (R), (G), (R | G), (B), (R | B), (G | B), (R | G | B), + (A), (R | A), (G | A), (R | G | A), (B | A), (R | B | A), (G | B | A), (R | G | B | A)}; + return std::bitset<4>{mask.at(component_mask)}.test(component); +} + +u32 GetComponentSize(TextureFormat format, std::size_t component) { + switch (format) { + case TextureFormat::R32_G32_B32_A32: + return 32; + case TextureFormat::R16_G16_B16_A16: + return 16; + case TextureFormat::R32_G32_B32: + return component <= 2 ? 32 : 0; + case TextureFormat::R32_G32: + return component <= 1 ? 32 : 0; + case TextureFormat::R16_G16: + return component <= 1 ? 16 : 0; + case TextureFormat::R32: + return component == 0 ? 32 : 0; + case TextureFormat::R16: + return component == 0 ? 16 : 0; + case TextureFormat::R8: + return component == 0 ? 8 : 0; + case TextureFormat::R1: + return component == 0 ? 1 : 0; + case TextureFormat::A8R8G8B8: + return 8; + case TextureFormat::A2B10G10R10: + return (component == 3 || component == 2 || component == 1) ? 10 : 2; + case TextureFormat::A4B4G4R4: + return 4; + case TextureFormat::A5B5G5R1: + return (component == 0 || component == 1 || component == 2) ? 5 : 1; + case TextureFormat::A1B5G5R5: + return (component == 1 || component == 2 || component == 3) ? 5 : 1; + case TextureFormat::R32_B24G8: + if (component == 0) { + return 32; + } + if (component == 1) { + return 24; + } + if (component == 2) { + return 8; + } + return 0; + case TextureFormat::B5G6R5: + if (component == 0 || component == 2) { + return 5; + } + if (component == 1) { + return 6; + } + return 0; + case TextureFormat::B6G5R5: + if (component == 1 || component == 2) { + return 5; + } + if (component == 0) { + return 6; + } + return 0; + case TextureFormat::G8R24: + if (component == 0) { + return 8; + } + if (component == 1) { + return 24; + } + return 0; + case TextureFormat::G24R8: + if (component == 0) { + return 8; + } + if (component == 1) { + return 24; + } + return 0; + case TextureFormat::G8R8: + return (component == 0 || component == 1) ? 8 : 0; + case TextureFormat::G4R4: + return (component == 0 || component == 1) ? 4 : 0; + default: + UNIMPLEMENTED_MSG("texture format not implement={}", format); + return 0; + } +} + +std::size_t GetImageComponentMask(TextureFormat format) { + constexpr u8 R = 0b0001; + constexpr u8 G = 0b0010; + constexpr u8 B = 0b0100; + constexpr u8 A = 0b1000; + switch (format) { + case TextureFormat::R32_G32_B32_A32: + case TextureFormat::R16_G16_B16_A16: + case TextureFormat::A8R8G8B8: + case TextureFormat::A2B10G10R10: + case TextureFormat::A4B4G4R4: + case TextureFormat::A5B5G5R1: + case TextureFormat::A1B5G5R5: + return std::size_t{R | G | B | A}; + case TextureFormat::R32_G32_B32: + case TextureFormat::R32_B24G8: + case TextureFormat::B5G6R5: + case TextureFormat::B6G5R5: + return std::size_t{R | G | B}; + case TextureFormat::R32_G32: + case TextureFormat::R16_G16: + case TextureFormat::G8R24: + case TextureFormat::G24R8: + case TextureFormat::G8R8: + case TextureFormat::G4R4: + return std::size_t{R | G}; + case TextureFormat::R32: + case TextureFormat::R16: + case TextureFormat::R8: + case TextureFormat::R1: + return std::size_t{R}; + default: + UNIMPLEMENTED_MSG("texture format not implement={}", format); + return std::size_t{R | G | B | A}; + } +} + std::size_t GetImageTypeNumCoordinates(Tegra::Shader::ImageType image_type) { switch (image_type) { case Tegra::Shader::ImageType::Texture1D: @@ -37,6 +271,39 @@ std::size_t GetImageTypeNumCoordinates(Tegra::Shader::ImageType image_type) { } } // Anonymous namespace +std::pair<Node, bool> ShaderIR::GetComponentValue(ComponentType component_type, u32 component_size, + Node original_value) { + switch (component_type) { + case ComponentType::SNORM: { + // range [-1.0, 1.0] + auto cnv_value = Operation(OperationCode::FMul, original_value, + Immediate(static_cast<float>(1 << component_size) / 2.f - 1.f)); + cnv_value = Operation(OperationCode::ICastFloat, std::move(cnv_value)); + return {BitfieldExtract(std::move(cnv_value), 0, component_size), true}; + } + case ComponentType::SINT: + case ComponentType::UNORM: { + bool is_signed = component_type == ComponentType::SINT; + // range [0.0, 1.0] + auto cnv_value = Operation(OperationCode::FMul, original_value, + Immediate(static_cast<float>(1 << component_size) - 1.f)); + return {SignedOperation(OperationCode::ICastFloat, is_signed, std::move(cnv_value)), + is_signed}; + } + case ComponentType::UINT: // range [0, (1 << component_size) - 1] + return {std::move(original_value), false}; + case ComponentType::FLOAT: + if (component_size == 16) { + return {Operation(OperationCode::HCastFloat, original_value), true}; + } else { + return {std::move(original_value), true}; + } + default: + UNIMPLEMENTED_MSG("Unimplement component type={}", component_type); + return {std::move(original_value), true}; + } +} + u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) { const Instruction instr = {program_code[pc]}; const auto opcode = OpCode::Decode(instr); @@ -53,7 +320,6 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) { switch (opcode->get().GetId()) { case OpCode::Id::SULD: { - UNIMPLEMENTED_IF(instr.suldst.mode != Tegra::Shader::SurfaceDataMode::P); UNIMPLEMENTED_IF(instr.suldst.out_of_bounds_store != Tegra::Shader::OutOfBoundsStore::Ignore); @@ -62,17 +328,89 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) { : GetBindlessImage(instr.gpr39, type)}; image.MarkRead(); - u32 indexer = 0; - for (u32 element = 0; element < 4; ++element) { - if (!instr.suldst.IsComponentEnabled(element)) { - continue; + if (instr.suldst.mode == Tegra::Shader::SurfaceDataMode::P) { + u32 indexer = 0; + for (u32 element = 0; element < 4; ++element) { + if (!instr.suldst.IsComponentEnabled(element)) { + continue; + } + MetaImage meta{image, {}, element}; + Node value = Operation(OperationCode::ImageLoad, meta, GetCoordinates(type)); + SetTemporary(bb, indexer++, std::move(value)); + } + for (u32 i = 0; i < indexer; ++i) { + SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i)); + } + } else if (instr.suldst.mode == Tegra::Shader::SurfaceDataMode::D_BA) { + UNIMPLEMENTED_IF(instr.suldst.GetStoreDataLayout() != StoreType::Bits32 && + instr.suldst.GetStoreDataLayout() != StoreType::Bits64); + + auto descriptor = [this, instr] { + std::optional<Tegra::Engines::SamplerDescriptor> descriptor; + if (instr.suldst.is_immediate) { + descriptor = + registry.ObtainBoundSampler(static_cast<u32>(instr.image.index.Value())); + } else { + const Node image_register = GetRegister(instr.gpr39); + const auto [base_image, buffer, offset] = TrackCbuf( + image_register, global_code, static_cast<s64>(global_code.size())); + descriptor = registry.ObtainBindlessSampler(buffer, offset); + } + if (!descriptor) { + UNREACHABLE_MSG("Failed to obtain image descriptor"); + } + return *descriptor; + }(); + + const auto comp_mask = GetImageComponentMask(descriptor.format); + + switch (instr.suldst.GetStoreDataLayout()) { + case StoreType::Bits32: + case StoreType::Bits64: { + u32 indexer = 0; + u32 shifted_counter = 0; + Node value = Immediate(0); + for (u32 element = 0; element < 4; ++element) { + if (!IsComponentEnabled(comp_mask, element)) { + continue; + } + const auto component_type = GetComponentType(descriptor, element); + const auto component_size = GetComponentSize(descriptor.format, element); + MetaImage meta{image, {}, element}; + + auto [converted_value, is_signed] = GetComponentValue( + component_type, component_size, + Operation(OperationCode::ImageLoad, meta, GetCoordinates(type))); + + // shift element to correct position + const auto shifted = shifted_counter; + if (shifted > 0) { + converted_value = + SignedOperation(OperationCode::ILogicalShiftLeft, is_signed, + std::move(converted_value), Immediate(shifted)); + } + shifted_counter += component_size; + + // add value into result + value = Operation(OperationCode::UBitwiseOr, value, std::move(converted_value)); + + // if we shifted enough for 1 byte -> we save it into temp + if (shifted_counter >= 32) { + SetTemporary(bb, indexer++, std::move(value)); + // reset counter and value to prepare pack next byte + value = Immediate(0); + shifted_counter = 0; + } + } + for (u32 i = 0; i < indexer; ++i) { + SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i)); + } + break; + } + default: + UNREACHABLE(); + break; } - MetaImage meta{image, {}, element}; - Node value = Operation(OperationCode::ImageLoad, meta, GetCoordinates(type)); - SetTemporary(bb, indexer++, std::move(value)); - } - for (u32 i = 0; i < indexer; ++i) { - SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i)); } break; } diff --git a/src/video_core/shader/decode/other.cpp b/src/video_core/shader/decode/other.cpp index 4944e9d69..d4f95b18c 100644 --- a/src/video_core/shader/decode/other.cpp +++ b/src/video_core/shader/decode/other.cpp @@ -11,12 +11,17 @@ namespace VideoCommon::Shader { +using std::move; using Tegra::Shader::ConditionCode; using Tegra::Shader::Instruction; +using Tegra::Shader::IpaInterpMode; using Tegra::Shader::OpCode; +using Tegra::Shader::PixelImap; using Tegra::Shader::Register; using Tegra::Shader::SystemVariable; +using Index = Tegra::Shader::Attribute::Index; + u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { const Instruction instr = {program_code[pc]}; const auto opcode = OpCode::Decode(instr); @@ -66,18 +71,24 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { bb.push_back(Operation(OperationCode::Discard)); break; } - case OpCode::Id::MOV_SYS: { + case OpCode::Id::S2R: { const Node value = [this, instr] { switch (instr.sys20) { case SystemVariable::LaneId: - LOG_WARNING(HW_GPU, "MOV_SYS instruction with LaneId is incomplete"); + LOG_WARNING(HW_GPU, "S2R instruction with LaneId is incomplete"); return Immediate(0U); case SystemVariable::InvocationId: return Operation(OperationCode::InvocationId); case SystemVariable::Ydirection: return Operation(OperationCode::YNegate); case SystemVariable::InvocationInfo: - LOG_WARNING(HW_GPU, "MOV_SYS instruction with InvocationInfo is incomplete"); + LOG_WARNING(HW_GPU, "S2R instruction with InvocationInfo is incomplete"); + return Immediate(0U); + case SystemVariable::WscaleFactorXY: + UNIMPLEMENTED_MSG("S2R WscaleFactorXY is not implemented"); + return Immediate(0U); + case SystemVariable::WscaleFactorZ: + UNIMPLEMENTED_MSG("S2R WscaleFactorZ is not implemented"); return Immediate(0U); case SystemVariable::Tid: { Node value = Immediate(0); @@ -213,27 +224,28 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { } case OpCode::Id::IPA: { const bool is_physical = instr.ipa.idx && instr.gpr8.Value() != 0xff; - const auto attribute = instr.attribute.fmt28; - const Tegra::Shader::IpaMode input_mode{instr.ipa.interp_mode.Value(), - instr.ipa.sample_mode.Value()}; + const Index index = attribute.index; Node value = is_physical ? GetPhysicalInputAttribute(instr.gpr8) - : GetInputAttribute(attribute.index, attribute.element); - const Tegra::Shader::Attribute::Index index = attribute.index.Value(); - const bool is_generic = index >= Tegra::Shader::Attribute::Index::Attribute_0 && - index <= Tegra::Shader::Attribute::Index::Attribute_31; - if (is_generic || is_physical) { - // TODO(Blinkhawk): There are cases where a perspective attribute use PASS. - // In theory by setting them as perspective, OpenGL does the perspective correction. - // A way must figured to reverse the last step of it. - if (input_mode.interpolation_mode == Tegra::Shader::IpaInterpMode::Multiply) { - value = Operation(OperationCode::FMul, PRECISE, value, GetRegister(instr.gpr20)); + : GetInputAttribute(index, attribute.element); + + // Code taken from Ryujinx. + if (index >= Index::Attribute_0 && index <= Index::Attribute_31) { + const u32 location = static_cast<u32>(index) - static_cast<u32>(Index::Attribute_0); + if (header.ps.GetPixelImap(location) == PixelImap::Perspective) { + Node position_w = GetInputAttribute(Index::Position, 3); + value = Operation(OperationCode::FMul, move(value), move(position_w)); } } - value = GetSaturatedFloat(value, instr.ipa.saturate); - SetRegister(bb, instr.gpr0, value); + if (instr.ipa.interp_mode == IpaInterpMode::Multiply) { + value = Operation(OperationCode::FMul, move(value), GetRegister(instr.gpr20)); + } + + value = GetSaturatedFloat(move(value), instr.ipa.saturate); + + SetRegister(bb, instr.gpr0, move(value)); break; } case OpCode::Id::OUT_R: { diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp index 48350e042..6c4a1358b 100644 --- a/src/video_core/shader/decode/texture.cpp +++ b/src/video_core/shader/decode/texture.cpp @@ -780,20 +780,6 @@ Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is // When lod is used always is in gpr20 const Node lod = lod_enabled ? GetRegister(instr.gpr20) : Immediate(0); - // Fill empty entries from the guest sampler - const std::size_t entry_coord_count = GetCoordCount(sampler.GetType()); - if (type_coord_count != entry_coord_count) { - LOG_WARNING(HW_GPU, "Bound and built texture types mismatch"); - - // When the size is higher we insert zeroes - for (std::size_t i = type_coord_count; i < entry_coord_count; ++i) { - coords.push_back(GetRegister(Register::ZeroIndex)); - } - - // Then we ensure the size matches the number of entries (dropping unused values) - coords.resize(entry_coord_count); - } - Node4 values; for (u32 element = 0; element < values.size(); ++element) { auto coords_copy = coords; diff --git a/src/video_core/shader/decode/video.cpp b/src/video_core/shader/decode/video.cpp index b047cf870..64ba60ea2 100644 --- a/src/video_core/shader/decode/video.cpp +++ b/src/video_core/shader/decode/video.cpp @@ -10,16 +10,24 @@ namespace VideoCommon::Shader { +using std::move; using Tegra::Shader::Instruction; using Tegra::Shader::OpCode; using Tegra::Shader::Pred; using Tegra::Shader::VideoType; using Tegra::Shader::VmadShr; +using Tegra::Shader::VmnmxOperation; +using Tegra::Shader::VmnmxType; u32 ShaderIR::DecodeVideo(NodeBlock& bb, u32 pc) { const Instruction instr = {program_code[pc]}; const auto opcode = OpCode::Decode(instr); + if (opcode->get().GetId() == OpCode::Id::VMNMX) { + DecodeVMNMX(bb, instr); + return pc; + } + const Node op_a = GetVideoOperand(GetRegister(instr.gpr8), instr.video.is_byte_chunk_a, instr.video.signed_a, instr.video.type_a, instr.video.byte_height_a); @@ -109,4 +117,54 @@ Node ShaderIR::GetVideoOperand(Node op, bool is_chunk, bool is_signed, } } +void ShaderIR::DecodeVMNMX(NodeBlock& bb, Tegra::Shader::Instruction instr) { + UNIMPLEMENTED_IF(!instr.vmnmx.is_op_b_register); + UNIMPLEMENTED_IF(instr.vmnmx.SourceFormatA() != VmnmxType::Bits32); + UNIMPLEMENTED_IF(instr.vmnmx.SourceFormatB() != VmnmxType::Bits32); + UNIMPLEMENTED_IF(instr.vmnmx.is_src_a_signed != instr.vmnmx.is_src_b_signed); + UNIMPLEMENTED_IF(instr.vmnmx.sat); + UNIMPLEMENTED_IF(instr.generates_cc); + + Node op_a = GetRegister(instr.gpr8); + Node op_b = GetRegister(instr.gpr20); + Node op_c = GetRegister(instr.gpr39); + + const bool is_oper1_signed = instr.vmnmx.is_src_a_signed; // Stubbed + const bool is_oper2_signed = instr.vmnmx.is_dest_signed; + + const auto operation_a = instr.vmnmx.mx ? OperationCode::IMax : OperationCode::IMin; + Node value = SignedOperation(operation_a, is_oper1_signed, move(op_a), move(op_b)); + + switch (instr.vmnmx.operation) { + case VmnmxOperation::Mrg_16H: + value = BitfieldInsert(move(op_c), move(value), 16, 16); + break; + case VmnmxOperation::Mrg_16L: + value = BitfieldInsert(move(op_c), move(value), 0, 16); + break; + case VmnmxOperation::Mrg_8B0: + value = BitfieldInsert(move(op_c), move(value), 0, 8); + break; + case VmnmxOperation::Mrg_8B2: + value = BitfieldInsert(move(op_c), move(value), 16, 8); + break; + case VmnmxOperation::Acc: + value = Operation(OperationCode::IAdd, move(value), move(op_c)); + break; + case VmnmxOperation::Min: + value = SignedOperation(OperationCode::IMin, is_oper2_signed, move(value), move(op_c)); + break; + case VmnmxOperation::Max: + value = SignedOperation(OperationCode::IMax, is_oper2_signed, move(value), move(op_c)); + break; + case VmnmxOperation::Nop: + break; + default: + UNREACHABLE(); + break; + } + + SetRegister(bb, instr.gpr0, move(value)); +} + } // namespace VideoCommon::Shader diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp index baf7188d2..8852c8a1b 100644 --- a/src/video_core/shader/shader_ir.cpp +++ b/src/video_core/shader/shader_ir.cpp @@ -359,6 +359,9 @@ Node ShaderIR::GetConditionCode(Tegra::Shader::ConditionCode cc) const { switch (cc) { case Tegra::Shader::ConditionCode::NEU: return GetInternalFlag(InternalFlag::Zero, true); + case Tegra::Shader::ConditionCode::FCSM_TR: + UNIMPLEMENTED_MSG("EXIT.FCSM_TR is not implemented"); + return MakeNode<PredicateNode>(Pred::NeverExecute, false); default: UNIMPLEMENTED_MSG("Unimplemented condition code: {}", static_cast<u32>(cc)); return MakeNode<PredicateNode>(Pred::NeverExecute, false); diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h index 80fc9b82c..c6e7bdf50 100644 --- a/src/video_core/shader/shader_ir.h +++ b/src/video_core/shader/shader_ir.h @@ -312,6 +312,10 @@ private: /// Conditionally saturates a half float pair Node GetSaturatedHalfFloat(Node value, bool saturate = true); + /// Get image component value by type and size + std::pair<Node, bool> GetComponentValue(Tegra::Texture::ComponentType component_type, + u32 component_size, Node original_value); + /// Returns a predicate comparing two floats Node GetPredicateComparisonFloat(Tegra::Shader::PredCondition condition, Node op_a, Node op_b); /// Returns a predicate comparing two integers @@ -350,6 +354,9 @@ private: /// Marks the usage of a input or output attribute. void MarkAttributeUsage(Tegra::Shader::Attribute::Index index, u64 element); + /// Decodes VMNMX instruction and inserts its code into the passed basic block. + void DecodeVMNMX(NodeBlock& bb, Tegra::Shader::Instruction instr); + void WriteTexInstructionFloat(NodeBlock& bb, Tegra::Shader::Instruction instr, const Node4& components); diff --git a/src/video_core/surface.h b/src/video_core/surface.h index ae8817465..e0acd44d3 100644 --- a/src/video_core/surface.h +++ b/src/video_core/surface.h @@ -504,103 +504,6 @@ static constexpr u32 GetBytesPerPixel(PixelFormat pixel_format) { return GetFormatBpp(pixel_format) / CHAR_BIT; } -enum class SurfaceCompression { - None, // Not compressed - Compressed, // Texture is compressed - Converted, // Texture is converted before upload or after download - Rearranged, // Texture is swizzled before upload or after download -}; - -constexpr std::array<SurfaceCompression, MaxPixelFormat> compression_type_table = {{ - SurfaceCompression::None, // ABGR8U - SurfaceCompression::None, // ABGR8S - SurfaceCompression::None, // ABGR8UI - SurfaceCompression::None, // B5G6R5U - SurfaceCompression::None, // A2B10G10R10U - SurfaceCompression::None, // A1B5G5R5U - SurfaceCompression::None, // R8U - SurfaceCompression::None, // R8UI - SurfaceCompression::None, // RGBA16F - SurfaceCompression::None, // RGBA16U - SurfaceCompression::None, // RGBA16S - SurfaceCompression::None, // RGBA16UI - SurfaceCompression::None, // R11FG11FB10F - SurfaceCompression::None, // RGBA32UI - SurfaceCompression::Compressed, // DXT1 - SurfaceCompression::Compressed, // DXT23 - SurfaceCompression::Compressed, // DXT45 - SurfaceCompression::Compressed, // DXN1 - SurfaceCompression::Compressed, // DXN2UNORM - SurfaceCompression::Compressed, // DXN2SNORM - SurfaceCompression::Compressed, // BC7U - SurfaceCompression::Compressed, // BC6H_UF16 - SurfaceCompression::Compressed, // BC6H_SF16 - SurfaceCompression::Converted, // ASTC_2D_4X4 - SurfaceCompression::None, // BGRA8 - SurfaceCompression::None, // RGBA32F - SurfaceCompression::None, // RG32F - SurfaceCompression::None, // R32F - SurfaceCompression::None, // R16F - SurfaceCompression::None, // R16U - SurfaceCompression::None, // R16S - SurfaceCompression::None, // R16UI - SurfaceCompression::None, // R16I - SurfaceCompression::None, // RG16 - SurfaceCompression::None, // RG16F - SurfaceCompression::None, // RG16UI - SurfaceCompression::None, // RG16I - SurfaceCompression::None, // RG16S - SurfaceCompression::None, // RGB32F - SurfaceCompression::None, // RGBA8_SRGB - SurfaceCompression::None, // RG8U - SurfaceCompression::None, // RG8S - SurfaceCompression::None, // RG32UI - SurfaceCompression::None, // RGBX16F - SurfaceCompression::None, // R32UI - SurfaceCompression::None, // R32I - SurfaceCompression::Converted, // ASTC_2D_8X8 - SurfaceCompression::Converted, // ASTC_2D_8X5 - SurfaceCompression::Converted, // ASTC_2D_5X4 - SurfaceCompression::None, // BGRA8_SRGB - SurfaceCompression::Compressed, // DXT1_SRGB - SurfaceCompression::Compressed, // DXT23_SRGB - SurfaceCompression::Compressed, // DXT45_SRGB - SurfaceCompression::Compressed, // BC7U_SRGB - SurfaceCompression::None, // R4G4B4A4U - SurfaceCompression::Converted, // ASTC_2D_4X4_SRGB - SurfaceCompression::Converted, // ASTC_2D_8X8_SRGB - SurfaceCompression::Converted, // ASTC_2D_8X5_SRGB - SurfaceCompression::Converted, // ASTC_2D_5X4_SRGB - SurfaceCompression::Converted, // ASTC_2D_5X5 - SurfaceCompression::Converted, // ASTC_2D_5X5_SRGB - SurfaceCompression::Converted, // ASTC_2D_10X8 - SurfaceCompression::Converted, // ASTC_2D_10X8_SRGB - SurfaceCompression::Converted, // ASTC_2D_6X6 - SurfaceCompression::Converted, // ASTC_2D_6X6_SRGB - SurfaceCompression::Converted, // ASTC_2D_10X10 - SurfaceCompression::Converted, // ASTC_2D_10X10_SRGB - SurfaceCompression::Converted, // ASTC_2D_12X12 - SurfaceCompression::Converted, // ASTC_2D_12X12_SRGB - SurfaceCompression::Converted, // ASTC_2D_8X6 - SurfaceCompression::Converted, // ASTC_2D_8X6_SRGB - SurfaceCompression::Converted, // ASTC_2D_6X5 - SurfaceCompression::Converted, // ASTC_2D_6X5_SRGB - SurfaceCompression::None, // E5B9G9R9F - SurfaceCompression::None, // Z32F - SurfaceCompression::None, // Z16 - SurfaceCompression::None, // Z24S8 - SurfaceCompression::Rearranged, // S8Z24 - SurfaceCompression::None, // Z32FS8 -}}; - -constexpr SurfaceCompression GetFormatCompressionType(PixelFormat format) { - if (format == PixelFormat::Invalid) { - return SurfaceCompression::None; - } - DEBUG_ASSERT(static_cast<std::size_t>(format) < compression_type_table.size()); - return compression_type_table[static_cast<std::size_t>(format)]; -} - SurfaceTarget SurfaceTargetFromTextureType(Tegra::Texture::TextureType texture_type); bool SurfaceTargetIsLayered(SurfaceTarget target); diff --git a/src/video_core/texture_cache/surface_base.cpp b/src/video_core/texture_cache/surface_base.cpp index 002df414f..7af0e792c 100644 --- a/src/video_core/texture_cache/surface_base.cpp +++ b/src/video_core/texture_cache/surface_base.cpp @@ -18,15 +18,20 @@ MICROPROFILE_DEFINE(GPU_Flush_Texture, "GPU", "Texture Flush", MP_RGB(128, 192, using Tegra::Texture::ConvertFromGuestToHost; using VideoCore::MortonSwizzleMode; -using VideoCore::Surface::SurfaceCompression; +using VideoCore::Surface::IsPixelFormatASTC; +using VideoCore::Surface::PixelFormat; StagingCache::StagingCache() = default; StagingCache::~StagingCache() = default; -SurfaceBaseImpl::SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params) - : params{params}, host_memory_size{params.GetHostSizeInBytes()}, gpu_addr{gpu_addr}, - mipmap_sizes(params.num_levels), mipmap_offsets(params.num_levels) { +SurfaceBaseImpl::SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params, + bool is_astc_supported) + : params{params}, gpu_addr{gpu_addr}, mipmap_sizes(params.num_levels), + mipmap_offsets(params.num_levels) { + is_converted = IsPixelFormatASTC(params.pixel_format) && !is_astc_supported; + host_memory_size = params.GetHostSizeInBytes(is_converted); + std::size_t offset = 0; for (u32 level = 0; level < params.num_levels; ++level) { const std::size_t mipmap_size{params.GetGuestMipmapSize(level)}; @@ -164,7 +169,7 @@ void SurfaceBaseImpl::SwizzleFunc(MortonSwizzleMode mode, u8* memory, const Surf std::size_t guest_offset{mipmap_offsets[level]}; if (params.is_layered) { - std::size_t host_offset{0}; + std::size_t host_offset = 0; const std::size_t guest_stride = layer_size; const std::size_t host_stride = params.GetHostLayerSize(level); for (u32 layer = 0; layer < params.depth; ++layer) { @@ -185,28 +190,17 @@ void SurfaceBaseImpl::LoadBuffer(Tegra::MemoryManager& memory_manager, MICROPROFILE_SCOPE(GPU_Load_Texture); auto& staging_buffer = staging_cache.GetBuffer(0); u8* host_ptr; - is_continuous = memory_manager.IsBlockContinuous(gpu_addr, guest_memory_size); - - // Handle continuouty - if (is_continuous) { - // Use physical memory directly - host_ptr = memory_manager.GetPointer(gpu_addr); - if (!host_ptr) { - return; - } - } else { - // Use an extra temporal buffer - auto& tmp_buffer = staging_cache.GetBuffer(1); - tmp_buffer.resize(guest_memory_size); - host_ptr = tmp_buffer.data(); - memory_manager.ReadBlockUnsafe(gpu_addr, host_ptr, guest_memory_size); - } + // Use an extra temporal buffer + auto& tmp_buffer = staging_cache.GetBuffer(1); + tmp_buffer.resize(guest_memory_size); + host_ptr = tmp_buffer.data(); + memory_manager.ReadBlockUnsafe(gpu_addr, host_ptr, guest_memory_size); if (params.is_tiled) { ASSERT_MSG(params.block_width == 0, "Block width is defined as {} on texture target {}", params.block_width, static_cast<u32>(params.target)); for (u32 level = 0; level < params.num_levels; ++level) { - const std::size_t host_offset{params.GetHostMipmapLevelOffset(level)}; + const std::size_t host_offset{params.GetHostMipmapLevelOffset(level, false)}; SwizzleFunc(MortonSwizzleMode::MortonToLinear, host_ptr, params, staging_buffer.data() + host_offset, level); } @@ -219,7 +213,7 @@ void SurfaceBaseImpl::LoadBuffer(Tegra::MemoryManager& memory_manager, const u32 height{(params.height + block_height - 1) / block_height}; const u32 copy_size{width * bpp}; if (params.pitch == copy_size) { - std::memcpy(staging_buffer.data(), host_ptr, params.GetHostSizeInBytes()); + std::memcpy(staging_buffer.data(), host_ptr, params.GetHostSizeInBytes(false)); } else { const u8* start{host_ptr}; u8* write_to{staging_buffer.data()}; @@ -231,19 +225,15 @@ void SurfaceBaseImpl::LoadBuffer(Tegra::MemoryManager& memory_manager, } } - auto compression_type = params.GetCompressionType(); - if (compression_type == SurfaceCompression::None || - compression_type == SurfaceCompression::Compressed) + if (!is_converted && params.pixel_format != PixelFormat::S8Z24) { return; + } - for (u32 level_up = params.num_levels; level_up > 0; --level_up) { - const u32 level = level_up - 1; - const std::size_t in_host_offset{params.GetHostMipmapLevelOffset(level)}; - const std::size_t out_host_offset = compression_type == SurfaceCompression::Rearranged - ? in_host_offset - : params.GetConvertedMipmapOffset(level); - u8* in_buffer = staging_buffer.data() + in_host_offset; - u8* out_buffer = staging_buffer.data() + out_host_offset; + for (u32 level = params.num_levels; level--;) { + const std::size_t in_host_offset{params.GetHostMipmapLevelOffset(level, false)}; + const std::size_t out_host_offset{params.GetHostMipmapLevelOffset(level, is_converted)}; + u8* const in_buffer = staging_buffer.data() + in_host_offset; + u8* const out_buffer = staging_buffer.data() + out_host_offset; ConvertFromGuestToHost(in_buffer, out_buffer, params.pixel_format, params.GetMipWidth(level), params.GetMipHeight(level), params.GetMipDepth(level), true, true); @@ -256,24 +246,15 @@ void SurfaceBaseImpl::FlushBuffer(Tegra::MemoryManager& memory_manager, auto& staging_buffer = staging_cache.GetBuffer(0); u8* host_ptr; - // Handle continuouty - if (is_continuous) { - // Use physical memory directly - host_ptr = memory_manager.GetPointer(gpu_addr); - if (!host_ptr) { - return; - } - } else { - // Use an extra temporal buffer - auto& tmp_buffer = staging_cache.GetBuffer(1); - tmp_buffer.resize(guest_memory_size); - host_ptr = tmp_buffer.data(); - } + // Use an extra temporal buffer + auto& tmp_buffer = staging_cache.GetBuffer(1); + tmp_buffer.resize(guest_memory_size); + host_ptr = tmp_buffer.data(); if (params.is_tiled) { ASSERT_MSG(params.block_width == 0, "Block width is defined as {}", params.block_width); for (u32 level = 0; level < params.num_levels; ++level) { - const std::size_t host_offset{params.GetHostMipmapLevelOffset(level)}; + const std::size_t host_offset{params.GetHostMipmapLevelOffset(level, false)}; SwizzleFunc(MortonSwizzleMode::LinearToMorton, host_ptr, params, staging_buffer.data() + host_offset, level); } @@ -299,9 +280,7 @@ void SurfaceBaseImpl::FlushBuffer(Tegra::MemoryManager& memory_manager, } } } - if (!is_continuous) { - memory_manager.WriteBlockUnsafe(gpu_addr, host_ptr, guest_memory_size); - } + memory_manager.WriteBlockUnsafe(gpu_addr, host_ptr, guest_memory_size); } } // namespace VideoCommon diff --git a/src/video_core/texture_cache/surface_base.h b/src/video_core/texture_cache/surface_base.h index 5f79bb0aa..a39a8661b 100644 --- a/src/video_core/texture_cache/surface_base.h +++ b/src/video_core/texture_cache/surface_base.h @@ -68,8 +68,8 @@ public: return gpu_addr; } - bool Overlaps(const CacheAddr start, const CacheAddr end) const { - return (cache_addr < end) && (cache_addr_end > start); + bool Overlaps(const VAddr start, const VAddr end) const { + return (cpu_addr < end) && (cpu_addr_end > start); } bool IsInside(const GPUVAddr other_start, const GPUVAddr other_end) { @@ -86,21 +86,13 @@ public: return cpu_addr; } - void SetCpuAddr(const VAddr new_addr) { - cpu_addr = new_addr; - } - - CacheAddr GetCacheAddr() const { - return cache_addr; + VAddr GetCpuAddrEnd() const { + return cpu_addr_end; } - CacheAddr GetCacheAddrEnd() const { - return cache_addr_end; - } - - void SetCacheAddr(const CacheAddr new_addr) { - cache_addr = new_addr; - cache_addr_end = new_addr + guest_memory_size; + void SetCpuAddr(const VAddr new_addr) { + cpu_addr = new_addr; + cpu_addr_end = new_addr + guest_memory_size; } const SurfaceParams& GetSurfaceParams() const { @@ -119,18 +111,14 @@ public: return mipmap_sizes[level]; } - void MarkAsContinuous(const bool is_continuous) { - this->is_continuous = is_continuous; - } - - bool IsContinuous() const { - return is_continuous; - } - bool IsLinear() const { return !params.is_tiled; } + bool IsConverted() const { + return is_converted; + } + bool MatchFormat(VideoCore::Surface::PixelFormat pixel_format) const { return params.pixel_format == pixel_format; } @@ -160,7 +148,8 @@ public: } protected: - explicit SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params); + explicit SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params, + bool is_astc_supported); ~SurfaceBaseImpl() = default; virtual void DecorateSurfaceName() = 0; @@ -168,12 +157,11 @@ protected: const SurfaceParams params; std::size_t layer_size; std::size_t guest_memory_size; - const std::size_t host_memory_size; + std::size_t host_memory_size; GPUVAddr gpu_addr{}; - CacheAddr cache_addr{}; - CacheAddr cache_addr_end{}; VAddr cpu_addr{}; - bool is_continuous{}; + VAddr cpu_addr_end{}; + bool is_converted{}; std::vector<std::size_t> mipmap_sizes; std::vector<std::size_t> mipmap_offsets; @@ -288,8 +276,9 @@ public: } protected: - explicit SurfaceBase(const GPUVAddr gpu_addr, const SurfaceParams& params) - : SurfaceBaseImpl(gpu_addr, params) {} + explicit SurfaceBase(const GPUVAddr gpu_addr, const SurfaceParams& params, + bool is_astc_supported) + : SurfaceBaseImpl(gpu_addr, params, is_astc_supported) {} ~SurfaceBase() = default; diff --git a/src/video_core/texture_cache/surface_params.cpp b/src/video_core/texture_cache/surface_params.cpp index 9931c5ef7..6f3ef45be 100644 --- a/src/video_core/texture_cache/surface_params.cpp +++ b/src/video_core/texture_cache/surface_params.cpp @@ -113,10 +113,8 @@ SurfaceParams SurfaceParams::CreateForTexture(const FormatLookupTable& lookup_ta params.height = tic.Height(); params.depth = tic.Depth(); params.pitch = params.is_tiled ? 0 : tic.Pitch(); - if (params.target == SurfaceTarget::Texture2D && params.depth > 1) { - params.depth = 1; - } else if (params.target == SurfaceTarget::TextureCubemap || - params.target == SurfaceTarget::TextureCubeArray) { + if (params.target == SurfaceTarget::TextureCubemap || + params.target == SurfaceTarget::TextureCubeArray) { params.depth *= 6; } params.num_levels = tic.max_mip_level + 1; @@ -309,28 +307,26 @@ std::size_t SurfaceParams::GetGuestMipmapLevelOffset(u32 level) const { return offset; } -std::size_t SurfaceParams::GetHostMipmapLevelOffset(u32 level) const { - std::size_t offset = 0; - for (u32 i = 0; i < level; i++) { - offset += GetInnerMipmapMemorySize(i, true, false) * GetNumLayers(); - } - return offset; -} - -std::size_t SurfaceParams::GetConvertedMipmapOffset(u32 level) const { +std::size_t SurfaceParams::GetHostMipmapLevelOffset(u32 level, bool is_converted) const { std::size_t offset = 0; - for (u32 i = 0; i < level; i++) { - offset += GetConvertedMipmapSize(i); + if (is_converted) { + for (u32 i = 0; i < level; ++i) { + offset += GetConvertedMipmapSize(i) * GetNumLayers(); + } + } else { + for (u32 i = 0; i < level; ++i) { + offset += GetInnerMipmapMemorySize(i, true, false) * GetNumLayers(); + } } return offset; } std::size_t SurfaceParams::GetConvertedMipmapSize(u32 level) const { constexpr std::size_t rgba8_bpp = 4ULL; - const std::size_t width_t = GetMipWidth(level); - const std::size_t height_t = GetMipHeight(level); - const std::size_t depth_t = is_layered ? depth : GetMipDepth(level); - return width_t * height_t * depth_t * rgba8_bpp; + const std::size_t mip_width = GetMipWidth(level); + const std::size_t mip_height = GetMipHeight(level); + const std::size_t mip_depth = is_layered ? 1 : GetMipDepth(level); + return mip_width * mip_height * mip_depth * rgba8_bpp; } std::size_t SurfaceParams::GetLayerSize(bool as_host_size, bool uncompressed) const { diff --git a/src/video_core/texture_cache/surface_params.h b/src/video_core/texture_cache/surface_params.h index 995cc3818..24957df8d 100644 --- a/src/video_core/texture_cache/surface_params.h +++ b/src/video_core/texture_cache/surface_params.h @@ -20,8 +20,6 @@ namespace VideoCommon { class FormatLookupTable; -using VideoCore::Surface::SurfaceCompression; - class SurfaceParams { public: /// Creates SurfaceCachedParams from a texture configuration. @@ -67,16 +65,14 @@ public: return GetInnerMemorySize(false, false, false); } - std::size_t GetHostSizeInBytes() const { - std::size_t host_size_in_bytes; - if (GetCompressionType() == SurfaceCompression::Converted) { - // ASTC is uncompressed in software, in emulated as RGBA8 - host_size_in_bytes = 0; - for (u32 level = 0; level < num_levels; ++level) { - host_size_in_bytes += GetConvertedMipmapSize(level); - } - } else { - host_size_in_bytes = GetInnerMemorySize(true, false, false); + std::size_t GetHostSizeInBytes(bool is_converted) const { + if (!is_converted) { + return GetInnerMemorySize(true, false, false); + } + // ASTC is uncompressed in software, in emulated as RGBA8 + std::size_t host_size_in_bytes = 0; + for (u32 level = 0; level < num_levels; ++level) { + host_size_in_bytes += GetConvertedMipmapSize(level) * GetNumLayers(); } return host_size_in_bytes; } @@ -107,9 +103,8 @@ public: u32 GetMipBlockDepth(u32 level) const; /// Returns the best possible row/pitch alignment for the surface. - u32 GetRowAlignment(u32 level) const { - const u32 bpp = - GetCompressionType() == SurfaceCompression::Converted ? 4 : GetBytesPerPixel(); + u32 GetRowAlignment(u32 level, bool is_converted) const { + const u32 bpp = is_converted ? 4 : GetBytesPerPixel(); return 1U << Common::CountTrailingZeroes32(GetMipWidth(level) * bpp); } @@ -117,11 +112,7 @@ public: std::size_t GetGuestMipmapLevelOffset(u32 level) const; /// Returns the offset in bytes in host memory (linear) of a given mipmap level. - std::size_t GetHostMipmapLevelOffset(u32 level) const; - - /// Returns the offset in bytes in host memory (linear) of a given mipmap level - /// for a texture that is converted in host gpu. - std::size_t GetConvertedMipmapOffset(u32 level) const; + std::size_t GetHostMipmapLevelOffset(u32 level, bool is_converted) const; /// Returns the size in bytes in guest memory of a given mipmap level. std::size_t GetGuestMipmapSize(u32 level) const { @@ -196,11 +187,6 @@ public: pixel_format < VideoCore::Surface::PixelFormat::MaxDepthStencilFormat; } - /// Returns how the compression should be handled for this texture. - SurfaceCompression GetCompressionType() const { - return VideoCore::Surface::GetFormatCompressionType(pixel_format); - } - /// Returns is the surface is a TextureBuffer type of surface. bool IsBuffer() const { return target == VideoCore::Surface::SurfaceTarget::TextureBuffer; diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 6cdbe63d0..4edd4313b 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -52,11 +52,9 @@ using RenderTargetConfig = Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig; template <typename TSurface, typename TView> class TextureCache { - using IntervalMap = boost::icl::interval_map<CacheAddr, std::set<TSurface>>; - using IntervalType = typename IntervalMap::interval_type; public: - void InvalidateRegion(CacheAddr addr, std::size_t size) { + void InvalidateRegion(VAddr addr, std::size_t size) { std::lock_guard lock{mutex}; for (const auto& surface : GetSurfacesInRegion(addr, size)) { @@ -76,7 +74,7 @@ public: guard_samplers = new_guard; } - void FlushRegion(CacheAddr addr, std::size_t size) { + void FlushRegion(VAddr addr, std::size_t size) { std::lock_guard lock{mutex}; auto surfaces = GetSurfacesInRegion(addr, size); @@ -99,9 +97,9 @@ public: return GetNullSurface(SurfaceParams::ExpectedTarget(entry)); } - const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)}; - const auto cache_addr{ToCacheAddr(host_ptr)}; - if (!cache_addr) { + const std::optional<VAddr> cpu_addr = + system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr); + if (!cpu_addr) { return GetNullSurface(SurfaceParams::ExpectedTarget(entry)); } @@ -110,7 +108,7 @@ public: } const auto params{SurfaceParams::CreateForTexture(format_lookup_table, tic, entry)}; - const auto [surface, view] = GetSurface(gpu_addr, cache_addr, params, true, false); + const auto [surface, view] = GetSurface(gpu_addr, *cpu_addr, params, false); if (guard_samplers) { sampled_textures.push_back(surface); } @@ -124,13 +122,13 @@ public: if (!gpu_addr) { return GetNullSurface(SurfaceParams::ExpectedTarget(entry)); } - const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)}; - const auto cache_addr{ToCacheAddr(host_ptr)}; - if (!cache_addr) { + const std::optional<VAddr> cpu_addr = + system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr); + if (!cpu_addr) { return GetNullSurface(SurfaceParams::ExpectedTarget(entry)); } const auto params{SurfaceParams::CreateForImage(format_lookup_table, tic, entry)}; - const auto [surface, view] = GetSurface(gpu_addr, cache_addr, params, true, false); + const auto [surface, view] = GetSurface(gpu_addr, *cpu_addr, params, false); if (guard_samplers) { sampled_textures.push_back(surface); } @@ -145,7 +143,7 @@ public: return any_rt; } - TView GetDepthBufferSurface(bool preserve_contents) { + TView GetDepthBufferSurface() { std::lock_guard lock{mutex}; auto& maxwell3d = system.GPU().Maxwell3D(); if (!maxwell3d.dirty.flags[VideoCommon::Dirty::ZetaBuffer]) { @@ -159,14 +157,14 @@ public: SetEmptyDepthBuffer(); return {}; } - const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)}; - const auto cache_addr{ToCacheAddr(host_ptr)}; - if (!cache_addr) { + const std::optional<VAddr> cpu_addr = + system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr); + if (!cpu_addr) { SetEmptyDepthBuffer(); return {}; } const auto depth_params{SurfaceParams::CreateForDepthBuffer(system)}; - auto surface_view = GetSurface(gpu_addr, cache_addr, depth_params, preserve_contents, true); + auto surface_view = GetSurface(gpu_addr, *cpu_addr, depth_params, true); if (depth_buffer.target) depth_buffer.target->MarkAsRenderTarget(false, NO_RT); depth_buffer.target = surface_view.first; @@ -176,7 +174,7 @@ public: return surface_view.second; } - TView GetColorBufferSurface(std::size_t index, bool preserve_contents) { + TView GetColorBufferSurface(std::size_t index) { std::lock_guard lock{mutex}; ASSERT(index < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets); auto& maxwell3d = system.GPU().Maxwell3D(); @@ -199,16 +197,15 @@ public: return {}; } - const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)}; - const auto cache_addr{ToCacheAddr(host_ptr)}; - if (!cache_addr) { + const std::optional<VAddr> cpu_addr = + system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr); + if (!cpu_addr) { SetEmptyColorBuffer(index); return {}; } - auto surface_view = - GetSurface(gpu_addr, cache_addr, SurfaceParams::CreateForFramebuffer(system, index), - preserve_contents, true); + auto surface_view = GetSurface(gpu_addr, *cpu_addr, + SurfaceParams::CreateForFramebuffer(system, index), true); if (render_targets[index].target) render_targets[index].target->MarkAsRenderTarget(false, NO_RT); render_targets[index].target = surface_view.first; @@ -257,27 +254,26 @@ public: const GPUVAddr src_gpu_addr = src_config.Address(); const GPUVAddr dst_gpu_addr = dst_config.Address(); DeduceBestBlit(src_params, dst_params, src_gpu_addr, dst_gpu_addr); - const auto dst_host_ptr{system.GPU().MemoryManager().GetPointer(dst_gpu_addr)}; - const auto dst_cache_addr{ToCacheAddr(dst_host_ptr)}; - const auto src_host_ptr{system.GPU().MemoryManager().GetPointer(src_gpu_addr)}; - const auto src_cache_addr{ToCacheAddr(src_host_ptr)}; + const std::optional<VAddr> dst_cpu_addr = + system.GPU().MemoryManager().GpuToCpuAddress(dst_gpu_addr); + const std::optional<VAddr> src_cpu_addr = + system.GPU().MemoryManager().GpuToCpuAddress(src_gpu_addr); std::pair<TSurface, TView> dst_surface = - GetSurface(dst_gpu_addr, dst_cache_addr, dst_params, true, false); + GetSurface(dst_gpu_addr, *dst_cpu_addr, dst_params, false); std::pair<TSurface, TView> src_surface = - GetSurface(src_gpu_addr, src_cache_addr, src_params, true, false); + GetSurface(src_gpu_addr, *src_cpu_addr, src_params, false); ImageBlit(src_surface.second, dst_surface.second, copy_config); dst_surface.first->MarkAsModified(true, Tick()); } - TSurface TryFindFramebufferSurface(const u8* host_ptr) { - const CacheAddr cache_addr = ToCacheAddr(host_ptr); - if (!cache_addr) { + TSurface TryFindFramebufferSurface(VAddr addr) { + if (!addr) { return nullptr; } - const CacheAddr page = cache_addr >> registry_page_bits; + const VAddr page = addr >> registry_page_bits; std::vector<TSurface>& list = registry[page]; for (auto& surface : list) { - if (surface->GetCacheAddr() == cache_addr) { + if (surface->GetCpuAddr() == addr) { return surface; } } @@ -289,8 +285,9 @@ public: } protected: - TextureCache(Core::System& system, VideoCore::RasterizerInterface& rasterizer) - : system{system}, rasterizer{rasterizer} { + explicit TextureCache(Core::System& system, VideoCore::RasterizerInterface& rasterizer, + bool is_astc_supported) + : system{system}, is_astc_supported{is_astc_supported}, rasterizer{rasterizer} { for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) { SetEmptyColorBuffer(i); } @@ -337,18 +334,14 @@ protected: void Register(TSurface surface) { const GPUVAddr gpu_addr = surface->GetGpuAddr(); - const CacheAddr cache_ptr = ToCacheAddr(system.GPU().MemoryManager().GetPointer(gpu_addr)); const std::size_t size = surface->GetSizeInBytes(); const std::optional<VAddr> cpu_addr = system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr); - if (!cache_ptr || !cpu_addr) { + if (!cpu_addr) { LOG_CRITICAL(HW_GPU, "Failed to register surface with unmapped gpu_address 0x{:016x}", gpu_addr); return; } - const bool continuous = system.GPU().MemoryManager().IsBlockContinuous(gpu_addr, size); - surface->MarkAsContinuous(continuous); - surface->SetCacheAddr(cache_ptr); surface->SetCpuAddr(*cpu_addr); RegisterInnerCache(surface); surface->MarkAsRegistered(true); @@ -381,6 +374,7 @@ protected: } Core::System& system; + const bool is_astc_supported; private: enum class RecycleStrategy : u32 { @@ -456,22 +450,18 @@ private: * @param overlaps The overlapping surfaces registered in the cache. * @param params The parameters for the new surface. * @param gpu_addr The starting address of the new surface. - * @param preserve_contents Indicates that the new surface should be loaded from memory or left - * blank. * @param untopological Indicates to the recycler that the texture has no way to match the * overlaps due to topological reasons. **/ std::pair<TSurface, TView> RecycleSurface(std::vector<TSurface>& overlaps, const SurfaceParams& params, const GPUVAddr gpu_addr, - const bool preserve_contents, const MatchTopologyResult untopological) { - const bool do_load = preserve_contents && Settings::values.use_accurate_gpu_emulation; for (auto& surface : overlaps) { Unregister(surface); } switch (PickStrategy(overlaps, params, gpu_addr, untopological)) { case RecycleStrategy::Ignore: { - return InitializeSurface(gpu_addr, params, do_load); + return InitializeSurface(gpu_addr, params, Settings::values.use_accurate_gpu_emulation); } case RecycleStrategy::Flush: { std::sort(overlaps.begin(), overlaps.end(), @@ -481,7 +471,7 @@ private: for (auto& surface : overlaps) { FlushSurface(surface); } - return InitializeSurface(gpu_addr, params, preserve_contents); + return InitializeSurface(gpu_addr, params); } case RecycleStrategy::BufferCopy: { auto new_surface = GetUncachedSurface(gpu_addr, params); @@ -490,7 +480,7 @@ private: } default: { UNIMPLEMENTED_MSG("Unimplemented Texture Cache Recycling Strategy!"); - return InitializeSurface(gpu_addr, params, do_load); + return InitializeSurface(gpu_addr, params); } } } @@ -519,7 +509,9 @@ private: } const auto& final_params = new_surface->GetSurfaceParams(); if (cr_params.type != final_params.type) { - BufferCopy(current_surface, new_surface); + if (Settings::values.use_accurate_gpu_emulation) { + BufferCopy(current_surface, new_surface); + } } else { std::vector<CopyParams> bricks = current_surface->BreakDown(final_params); for (auto& brick : bricks) { @@ -626,14 +618,11 @@ private: * @param params The parameters on the new surface. * @param gpu_addr The starting address of the new surface. * @param cache_addr The starting address of the new surface on physical memory. - * @param preserve_contents Indicates that the new surface should be loaded from memory or - * left blank. */ std::optional<std::pair<TSurface, TView>> Manage3DSurfaces(std::vector<TSurface>& overlaps, const SurfaceParams& params, const GPUVAddr gpu_addr, - const CacheAddr cache_addr, - bool preserve_contents) { + const VAddr cpu_addr) { if (params.target == SurfaceTarget::Texture3D) { bool failed = false; if (params.num_levels > 1) { @@ -657,7 +646,7 @@ private: failed = true; break; } - const u32 offset = static_cast<u32>(surface->GetCacheAddr() - cache_addr); + const u32 offset = static_cast<u32>(surface->GetCpuAddr() - cpu_addr); const auto [x, y, z] = params.GetBlockOffsetXYZ(offset); modified |= surface->IsModified(); const CopyParams copy_params(0, 0, 0, 0, 0, z, 0, 0, params.width, params.height, @@ -677,23 +666,23 @@ private: } else { for (const auto& surface : overlaps) { if (!surface->MatchTarget(params.target)) { - if (overlaps.size() == 1 && surface->GetCacheAddr() == cache_addr) { + if (overlaps.size() == 1 && surface->GetCpuAddr() == cpu_addr) { if (Settings::values.use_accurate_gpu_emulation) { return std::nullopt; } Unregister(surface); - return InitializeSurface(gpu_addr, params, preserve_contents); + return InitializeSurface(gpu_addr, params); } return std::nullopt; } - if (surface->GetCacheAddr() != cache_addr) { + if (surface->GetCpuAddr() != cpu_addr) { continue; } if (surface->MatchesStructure(params) == MatchStructureResult::FullMatch) { return {{surface, surface->GetMainView()}}; } } - return InitializeSurface(gpu_addr, params, preserve_contents); + return InitializeSurface(gpu_addr, params); } } @@ -716,23 +705,19 @@ private: * * @param gpu_addr The starting address of the candidate surface. * @param params The parameters on the candidate surface. - * @param preserve_contents Indicates that the new surface should be loaded from memory or - * left blank. * @param is_render Whether or not the surface is a render target. **/ - std::pair<TSurface, TView> GetSurface(const GPUVAddr gpu_addr, const CacheAddr cache_addr, - const SurfaceParams& params, bool preserve_contents, - bool is_render) { + std::pair<TSurface, TView> GetSurface(const GPUVAddr gpu_addr, const VAddr cpu_addr, + const SurfaceParams& params, bool is_render) { // Step 1 // Check Level 1 Cache for a fast structural match. If candidate surface // matches at certain level we are pretty much done. - if (const auto iter = l1_cache.find(cache_addr); iter != l1_cache.end()) { + if (const auto iter = l1_cache.find(cpu_addr); iter != l1_cache.end()) { TSurface& current_surface = iter->second; const auto topological_result = current_surface->MatchesTopology(params); if (topological_result != MatchTopologyResult::FullMatch) { std::vector<TSurface> overlaps{current_surface}; - return RecycleSurface(overlaps, params, gpu_addr, preserve_contents, - topological_result); + return RecycleSurface(overlaps, params, gpu_addr, topological_result); } const auto struct_result = current_surface->MatchesStructure(params); @@ -753,11 +738,11 @@ private: // Step 2 // Obtain all possible overlaps in the memory region const std::size_t candidate_size = params.GetGuestSizeInBytes(); - auto overlaps{GetSurfacesInRegion(cache_addr, candidate_size)}; + auto overlaps{GetSurfacesInRegion(cpu_addr, candidate_size)}; // If none are found, we are done. we just load the surface and create it. if (overlaps.empty()) { - return InitializeSurface(gpu_addr, params, preserve_contents); + return InitializeSurface(gpu_addr, params); } // Step 3 @@ -767,15 +752,13 @@ private: for (const auto& surface : overlaps) { const auto topological_result = surface->MatchesTopology(params); if (topological_result != MatchTopologyResult::FullMatch) { - return RecycleSurface(overlaps, params, gpu_addr, preserve_contents, - topological_result); + return RecycleSurface(overlaps, params, gpu_addr, topological_result); } } // Check if it's a 3D texture if (params.block_depth > 0) { - auto surface = - Manage3DSurfaces(overlaps, params, gpu_addr, cache_addr, preserve_contents); + auto surface = Manage3DSurfaces(overlaps, params, gpu_addr, cpu_addr); if (surface) { return *surface; } @@ -795,8 +778,7 @@ private: return *view; } } - return RecycleSurface(overlaps, params, gpu_addr, preserve_contents, - MatchTopologyResult::FullMatch); + return RecycleSurface(overlaps, params, gpu_addr, MatchTopologyResult::FullMatch); } // Now we check if the candidate is a mipmap/layer of the overlap std::optional<TView> view = @@ -820,7 +802,7 @@ private: pair.first->EmplaceView(params, gpu_addr, candidate_size); if (mirage_view) return {pair.first, *mirage_view}; - return RecycleSurface(overlaps, params, gpu_addr, preserve_contents, + return RecycleSurface(overlaps, params, gpu_addr, MatchTopologyResult::FullMatch); } return {current_surface, *view}; @@ -836,8 +818,7 @@ private: } } // We failed all the tests, recycle the overlaps into a new texture. - return RecycleSurface(overlaps, params, gpu_addr, preserve_contents, - MatchTopologyResult::FullMatch); + return RecycleSurface(overlaps, params, gpu_addr, MatchTopologyResult::FullMatch); } /** @@ -850,16 +831,16 @@ private: * @param params The parameters on the candidate surface. **/ Deduction DeduceSurface(const GPUVAddr gpu_addr, const SurfaceParams& params) { - const auto host_ptr{system.GPU().MemoryManager().GetPointer(gpu_addr)}; - const auto cache_addr{ToCacheAddr(host_ptr)}; + const std::optional<VAddr> cpu_addr = + system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr); - if (!cache_addr) { + if (!cpu_addr) { Deduction result{}; result.type = DeductionType::DeductionFailed; return result; } - if (const auto iter = l1_cache.find(cache_addr); iter != l1_cache.end()) { + if (const auto iter = l1_cache.find(*cpu_addr); iter != l1_cache.end()) { TSurface& current_surface = iter->second; const auto topological_result = current_surface->MatchesTopology(params); if (topological_result != MatchTopologyResult::FullMatch) { @@ -878,7 +859,7 @@ private: } const std::size_t candidate_size = params.GetGuestSizeInBytes(); - auto overlaps{GetSurfacesInRegion(cache_addr, candidate_size)}; + auto overlaps{GetSurfacesInRegion(*cpu_addr, candidate_size)}; if (overlaps.empty()) { Deduction result{}; @@ -995,10 +976,10 @@ private: } std::pair<TSurface, TView> InitializeSurface(GPUVAddr gpu_addr, const SurfaceParams& params, - bool preserve_contents) { + bool do_load = true) { auto new_surface{GetUncachedSurface(gpu_addr, params)}; Register(new_surface); - if (preserve_contents) { + if (do_load) { LoadSurface(new_surface); } return {new_surface, new_surface->GetMainView()}; @@ -1022,10 +1003,10 @@ private: } void RegisterInnerCache(TSurface& surface) { - const CacheAddr cache_addr = surface->GetCacheAddr(); - CacheAddr start = cache_addr >> registry_page_bits; - const CacheAddr end = (surface->GetCacheAddrEnd() - 1) >> registry_page_bits; - l1_cache[cache_addr] = surface; + const VAddr cpu_addr = surface->GetCpuAddr(); + VAddr start = cpu_addr >> registry_page_bits; + const VAddr end = (surface->GetCpuAddrEnd() - 1) >> registry_page_bits; + l1_cache[cpu_addr] = surface; while (start <= end) { registry[start].push_back(surface); start++; @@ -1033,10 +1014,10 @@ private: } void UnregisterInnerCache(TSurface& surface) { - const CacheAddr cache_addr = surface->GetCacheAddr(); - CacheAddr start = cache_addr >> registry_page_bits; - const CacheAddr end = (surface->GetCacheAddrEnd() - 1) >> registry_page_bits; - l1_cache.erase(cache_addr); + const VAddr cpu_addr = surface->GetCpuAddr(); + VAddr start = cpu_addr >> registry_page_bits; + const VAddr end = (surface->GetCpuAddrEnd() - 1) >> registry_page_bits; + l1_cache.erase(cpu_addr); while (start <= end) { auto& reg{registry[start]}; reg.erase(std::find(reg.begin(), reg.end(), surface)); @@ -1044,18 +1025,18 @@ private: } } - std::vector<TSurface> GetSurfacesInRegion(const CacheAddr cache_addr, const std::size_t size) { + std::vector<TSurface> GetSurfacesInRegion(const VAddr cpu_addr, const std::size_t size) { if (size == 0) { return {}; } - const CacheAddr cache_addr_end = cache_addr + size; - CacheAddr start = cache_addr >> registry_page_bits; - const CacheAddr end = (cache_addr_end - 1) >> registry_page_bits; + const VAddr cpu_addr_end = cpu_addr + size; + VAddr start = cpu_addr >> registry_page_bits; + const VAddr end = (cpu_addr_end - 1) >> registry_page_bits; std::vector<TSurface> surfaces; while (start <= end) { std::vector<TSurface>& list = registry[start]; for (auto& surface : list) { - if (!surface->IsPicked() && surface->Overlaps(cache_addr, cache_addr_end)) { + if (!surface->IsPicked() && surface->Overlaps(cpu_addr, cpu_addr_end)) { surface->MarkAsPicked(true); surfaces.push_back(surface); } @@ -1144,14 +1125,14 @@ private: // large in size. static constexpr u64 registry_page_bits{20}; static constexpr u64 registry_page_size{1 << registry_page_bits}; - std::unordered_map<CacheAddr, std::vector<TSurface>> registry; + std::unordered_map<VAddr, std::vector<TSurface>> registry; static constexpr u32 DEPTH_RT = 8; static constexpr u32 NO_RT = 0xFFFFFFFF; // The L1 Cache is used for fast texture lookup before checking the overlaps // This avoids calculating size and other stuffs. - std::unordered_map<CacheAddr, TSurface> l1_cache; + std::unordered_map<VAddr, TSurface> l1_cache; /// The surface reserve is a "backup" cache, this is where we put unique surfaces that have /// previously been used. This is to prevent surfaces from being constantly created and diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp index 062b4f252..365bde2f1 100644 --- a/src/video_core/textures/astc.cpp +++ b/src/video_core/textures/astc.cpp @@ -20,6 +20,8 @@ #include <cstring> #include <vector> +#include <boost/container/static_vector.hpp> + #include "common/common_types.h" #include "video_core/textures/astc.h" @@ -39,25 +41,25 @@ constexpr u32 Popcnt(u32 n) { class InputBitStream { public: - explicit InputBitStream(const u8* ptr, std::size_t start_offset = 0) - : m_CurByte(ptr), m_NextBit(start_offset % 8) {} + constexpr explicit InputBitStream(const u8* ptr, std::size_t start_offset = 0) + : cur_byte{ptr}, next_bit{start_offset % 8} {} - std::size_t GetBitsRead() const { - return m_BitsRead; + constexpr std::size_t GetBitsRead() const { + return bits_read; } - u32 ReadBit() { - u32 bit = *m_CurByte >> m_NextBit++; - while (m_NextBit >= 8) { - m_NextBit -= 8; - m_CurByte++; + constexpr bool ReadBit() { + const bool bit = (*cur_byte >> next_bit++) & 1; + while (next_bit >= 8) { + next_bit -= 8; + cur_byte++; } - m_BitsRead++; - return bit & 1; + bits_read++; + return bit; } - u32 ReadBits(std::size_t nBits) { + constexpr u32 ReadBits(std::size_t nBits) { u32 ret = 0; for (std::size_t i = 0; i < nBits; ++i) { ret |= (ReadBit() & 1) << i; @@ -66,7 +68,7 @@ public: } template <std::size_t nBits> - u32 ReadBits() { + constexpr u32 ReadBits() { u32 ret = 0; for (std::size_t i = 0; i < nBits; ++i) { ret |= (ReadBit() & 1) << i; @@ -75,64 +77,58 @@ public: } private: - const u8* m_CurByte; - std::size_t m_NextBit = 0; - std::size_t m_BitsRead = 0; + const u8* cur_byte; + std::size_t next_bit = 0; + std::size_t bits_read = 0; }; class OutputBitStream { public: - explicit OutputBitStream(u8* ptr, s32 nBits = 0, s32 start_offset = 0) - : m_NumBits(nBits), m_CurByte(ptr), m_NextBit(start_offset % 8) {} - - ~OutputBitStream() = default; + constexpr explicit OutputBitStream(u8* ptr, std::size_t bits = 0, std::size_t start_offset = 0) + : cur_byte{ptr}, num_bits{bits}, next_bit{start_offset % 8} {} - s32 GetBitsWritten() const { - return m_BitsWritten; + constexpr std::size_t GetBitsWritten() const { + return bits_written; } - void WriteBitsR(u32 val, u32 nBits) { + constexpr void WriteBitsR(u32 val, u32 nBits) { for (u32 i = 0; i < nBits; i++) { WriteBit((val >> (nBits - i - 1)) & 1); } } - void WriteBits(u32 val, u32 nBits) { + constexpr void WriteBits(u32 val, u32 nBits) { for (u32 i = 0; i < nBits; i++) { WriteBit((val >> i) & 1); } } private: - void WriteBit(s32 b) { - - if (done) + constexpr void WriteBit(bool b) { + if (bits_written >= num_bits) { return; + } - const u32 mask = 1 << m_NextBit++; + const u32 mask = 1 << next_bit++; // clear the bit - *m_CurByte &= static_cast<u8>(~mask); + *cur_byte &= static_cast<u8>(~mask); // Write the bit, if necessary if (b) - *m_CurByte |= static_cast<u8>(mask); + *cur_byte |= static_cast<u8>(mask); // Next byte? - if (m_NextBit >= 8) { - m_CurByte += 1; - m_NextBit = 0; + if (next_bit >= 8) { + cur_byte += 1; + next_bit = 0; } - - done = done || ++m_BitsWritten >= m_NumBits; } - s32 m_BitsWritten = 0; - const s32 m_NumBits; - u8* m_CurByte; - s32 m_NextBit = 0; - - bool done = false; + u8* cur_byte; + std::size_t num_bits; + std::size_t bits_written = 0; + std::size_t next_bit = 0; }; template <typename IntType> @@ -195,9 +191,13 @@ struct IntegerEncodedValue { u32 trit_value; }; }; +using IntegerEncodedVector = boost::container::static_vector< + IntegerEncodedValue, 64, + boost::container::static_vector_options< + boost::container::inplace_alignment<alignof(IntegerEncodedValue)>, + boost::container::throw_on_overflow<false>>::type>; -static void DecodeTritBlock(InputBitStream& bits, std::vector<IntegerEncodedValue>& result, - u32 nBitsPerValue) { +static void DecodeTritBlock(InputBitStream& bits, IntegerEncodedVector& result, u32 nBitsPerValue) { // Implement the algorithm in section C.2.12 u32 m[5]; u32 t[5]; @@ -255,7 +255,7 @@ static void DecodeTritBlock(InputBitStream& bits, std::vector<IntegerEncodedValu } } -static void DecodeQus32Block(InputBitStream& bits, std::vector<IntegerEncodedValue>& result, +static void DecodeQus32Block(InputBitStream& bits, IntegerEncodedVector& result, u32 nBitsPerValue) { // Implement the algorithm in section C.2.12 u32 m[3]; @@ -343,8 +343,8 @@ static constexpr std::array EncodingsValues = MakeEncodedValues(); // Fills result with the values that are encoded in the given // bitstream. We must know beforehand what the maximum possible // value is, and how many values we're decoding. -static void DecodeIntegerSequence(std::vector<IntegerEncodedValue>& result, InputBitStream& bits, - u32 maxRange, u32 nValues) { +static void DecodeIntegerSequence(IntegerEncodedVector& result, InputBitStream& bits, u32 maxRange, + u32 nValues) { // Determine encoding parameters IntegerEncodedValue val = EncodingsValues[maxRange]; @@ -634,12 +634,14 @@ static void FillError(u32* outBuf, u32 blockWidth, u32 blockHeight) { // Replicates low numBits such that [(toBit - 1):(toBit - 1 - fromBit)] // is the same as [(numBits - 1):0] and repeats all the way down. template <typename IntType> -static IntType Replicate(IntType val, u32 numBits, u32 toBit) { - if (numBits == 0) +static constexpr IntType Replicate(IntType val, u32 numBits, u32 toBit) { + if (numBits == 0) { return 0; - if (toBit == 0) + } + if (toBit == 0) { return 0; - IntType v = val & static_cast<IntType>((1 << numBits) - 1); + } + const IntType v = val & static_cast<IntType>((1 << numBits) - 1); IntType res = v; u32 reslen = numBits; while (reslen < toBit) { @@ -656,6 +658,89 @@ static IntType Replicate(IntType val, u32 numBits, u32 toBit) { return res; } +static constexpr std::size_t NumReplicateEntries(u32 num_bits) { + return std::size_t(1) << num_bits; +} + +template <typename IntType, u32 num_bits, u32 to_bit> +static constexpr auto MakeReplicateTable() { + std::array<IntType, NumReplicateEntries(num_bits)> table{}; + for (IntType value = 0; value < static_cast<IntType>(std::size(table)); ++value) { + table[value] = Replicate(value, num_bits, to_bit); + } + return table; +} + +static constexpr auto REPLICATE_BYTE_TO_16_TABLE = MakeReplicateTable<u32, 8, 16>(); +static constexpr u32 ReplicateByteTo16(std::size_t value) { + return REPLICATE_BYTE_TO_16_TABLE[value]; +} + +static constexpr auto REPLICATE_BIT_TO_7_TABLE = MakeReplicateTable<u32, 1, 7>(); +static constexpr u32 ReplicateBitTo7(std::size_t value) { + return REPLICATE_BIT_TO_7_TABLE[value]; +} + +static constexpr auto REPLICATE_BIT_TO_9_TABLE = MakeReplicateTable<u32, 1, 9>(); +static constexpr u32 ReplicateBitTo9(std::size_t value) { + return REPLICATE_BIT_TO_9_TABLE[value]; +} + +static constexpr auto REPLICATE_1_BIT_TO_8_TABLE = MakeReplicateTable<u32, 1, 8>(); +static constexpr auto REPLICATE_2_BIT_TO_8_TABLE = MakeReplicateTable<u32, 2, 8>(); +static constexpr auto REPLICATE_3_BIT_TO_8_TABLE = MakeReplicateTable<u32, 3, 8>(); +static constexpr auto REPLICATE_4_BIT_TO_8_TABLE = MakeReplicateTable<u32, 4, 8>(); +static constexpr auto REPLICATE_5_BIT_TO_8_TABLE = MakeReplicateTable<u32, 5, 8>(); +static constexpr auto REPLICATE_6_BIT_TO_8_TABLE = MakeReplicateTable<u32, 6, 8>(); +static constexpr auto REPLICATE_7_BIT_TO_8_TABLE = MakeReplicateTable<u32, 7, 8>(); +static constexpr auto REPLICATE_8_BIT_TO_8_TABLE = MakeReplicateTable<u32, 8, 8>(); +/// Use a precompiled table with the most common usages, if it's not in the expected range, fallback +/// to the runtime implementation +static constexpr u32 FastReplicateTo8(u32 value, u32 num_bits) { + switch (num_bits) { + case 1: + return REPLICATE_1_BIT_TO_8_TABLE[value]; + case 2: + return REPLICATE_2_BIT_TO_8_TABLE[value]; + case 3: + return REPLICATE_3_BIT_TO_8_TABLE[value]; + case 4: + return REPLICATE_4_BIT_TO_8_TABLE[value]; + case 5: + return REPLICATE_5_BIT_TO_8_TABLE[value]; + case 6: + return REPLICATE_6_BIT_TO_8_TABLE[value]; + case 7: + return REPLICATE_7_BIT_TO_8_TABLE[value]; + case 8: + return REPLICATE_8_BIT_TO_8_TABLE[value]; + default: + return Replicate(value, num_bits, 8); + } +} + +static constexpr auto REPLICATE_1_BIT_TO_6_TABLE = MakeReplicateTable<u32, 1, 6>(); +static constexpr auto REPLICATE_2_BIT_TO_6_TABLE = MakeReplicateTable<u32, 2, 6>(); +static constexpr auto REPLICATE_3_BIT_TO_6_TABLE = MakeReplicateTable<u32, 3, 6>(); +static constexpr auto REPLICATE_4_BIT_TO_6_TABLE = MakeReplicateTable<u32, 4, 6>(); +static constexpr auto REPLICATE_5_BIT_TO_6_TABLE = MakeReplicateTable<u32, 5, 6>(); +static constexpr u32 FastReplicateTo6(u32 value, u32 num_bits) { + switch (num_bits) { + case 1: + return REPLICATE_1_BIT_TO_6_TABLE[value]; + case 2: + return REPLICATE_2_BIT_TO_6_TABLE[value]; + case 3: + return REPLICATE_3_BIT_TO_6_TABLE[value]; + case 4: + return REPLICATE_4_BIT_TO_6_TABLE[value]; + case 5: + return REPLICATE_5_BIT_TO_6_TABLE[value]; + default: + return Replicate(value, num_bits, 6); + } +} + class Pixel { protected: using ChannelType = s16; @@ -674,10 +759,10 @@ public: // significant bits when going from larger to smaller bit depth // or by repeating the most significant bits when going from // smaller to larger bit depths. - void ChangeBitDepth(const u8 (&depth)[4]) { + void ChangeBitDepth() { for (u32 i = 0; i < 4; i++) { - Component(i) = ChangeBitDepth(Component(i), m_BitDepth[i], depth[i]); - m_BitDepth[i] = depth[i]; + Component(i) = ChangeBitDepth(Component(i), m_BitDepth[i]); + m_BitDepth[i] = 8; } } @@ -689,28 +774,23 @@ public: // Changes the bit depth of a single component. See the comment // above for how we do this. - static ChannelType ChangeBitDepth(Pixel::ChannelType val, u8 oldDepth, u8 newDepth) { - assert(newDepth <= 8); + static ChannelType ChangeBitDepth(Pixel::ChannelType val, u8 oldDepth) { assert(oldDepth <= 8); - if (oldDepth == newDepth) { + if (oldDepth == 8) { // Do nothing return val; - } else if (oldDepth == 0 && newDepth != 0) { - return static_cast<ChannelType>((1 << newDepth) - 1); - } else if (newDepth > oldDepth) { - return Replicate(val, oldDepth, newDepth); + } else if (oldDepth == 0) { + return static_cast<ChannelType>((1 << 8) - 1); + } else if (8 > oldDepth) { + return static_cast<ChannelType>(FastReplicateTo8(static_cast<u32>(val), oldDepth)); } else { // oldDepth > newDepth - if (newDepth == 0) { - return 0xFF; - } else { - u8 bitsWasted = static_cast<u8>(oldDepth - newDepth); - u16 v = static_cast<u16>(val); - v = static_cast<u16>((v + (1 << (bitsWasted - 1))) >> bitsWasted); - v = ::std::min<u16>(::std::max<u16>(0, v), static_cast<u16>((1 << newDepth) - 1)); - return static_cast<u8>(v); - } + const u8 bitsWasted = static_cast<u8>(oldDepth - 8); + u16 v = static_cast<u16>(val); + v = static_cast<u16>((v + (1 << (bitsWasted - 1))) >> bitsWasted); + v = ::std::min<u16>(::std::max<u16>(0, v), static_cast<u16>((1 << 8) - 1)); + return static_cast<u8>(v); } assert(false && "We shouldn't get here."); @@ -760,8 +840,7 @@ public: // up in the most-significant byte. u32 Pack() const { Pixel eightBit(*this); - const u8 eightBitDepth[4] = {8, 8, 8, 8}; - eightBit.ChangeBitDepth(eightBitDepth); + eightBit.ChangeBitDepth(); u32 r = 0; r |= eightBit.A(); @@ -816,8 +895,7 @@ static void DecodeColorValues(u32* out, u8* data, const u32* modes, const u32 nP } // We now have enough to decode our integer sequence. - std::vector<IntegerEncodedValue> decodedColorValues; - decodedColorValues.reserve(32); + IntegerEncodedVector decodedColorValues; InputBitStream colorStream(data); DecodeIntegerSequence(decodedColorValues, colorStream, range, nValues); @@ -839,12 +917,12 @@ static void DecodeColorValues(u32* out, u8* data, const u32* modes, const u32 nP u32 A = 0, B = 0, C = 0, D = 0; // A is just the lsb replicated 9 times. - A = Replicate(bitval & 1, 1, 9); + A = ReplicateBitTo9(bitval & 1); switch (val.encoding) { // Replicate bits case IntegerEncoding::JustBits: - out[outIdx++] = Replicate(bitval, bitlen, 8); + out[outIdx++] = FastReplicateTo8(bitval, bitlen); break; // Use algorithm in C.2.13 @@ -962,13 +1040,13 @@ static u32 UnquantizeTexelWeight(const IntegerEncodedValue& val) { u32 bitval = val.bit_value; u32 bitlen = val.num_bits; - u32 A = Replicate(bitval & 1, 1, 7); + u32 A = ReplicateBitTo7(bitval & 1); u32 B = 0, C = 0, D = 0; u32 result = 0; switch (val.encoding) { case IntegerEncoding::JustBits: - result = Replicate(bitval, bitlen, 6); + result = FastReplicateTo6(bitval, bitlen); break; case IntegerEncoding::Trit: { @@ -1047,7 +1125,7 @@ static u32 UnquantizeTexelWeight(const IntegerEncodedValue& val) { return result; } -static void UnquantizeTexelWeights(u32 out[2][144], const std::vector<IntegerEncodedValue>& weights, +static void UnquantizeTexelWeights(u32 out[2][144], const IntegerEncodedVector& weights, const TexelWeightParams& params, const u32 blockWidth, const u32 blockHeight) { u32 weightIdx = 0; @@ -1545,8 +1623,7 @@ static void DecompressBlock(const u8 inBuf[16], const u32 blockWidth, const u32 static_cast<u8>((1 << (weightParams.GetPackedBitSize() % 8)) - 1); memset(texelWeightData + clearByteStart, 0, 16 - clearByteStart); - std::vector<IntegerEncodedValue> texelWeightValues; - texelWeightValues.reserve(64); + IntegerEncodedVector texelWeightValues; InputBitStream weightStream(texelWeightData); @@ -1568,9 +1645,9 @@ static void DecompressBlock(const u8 inBuf[16], const u32 blockWidth, const u32 Pixel p; for (u32 c = 0; c < 4; c++) { u32 C0 = endpos32s[partition][0].Component(c); - C0 = Replicate(C0, 8, 16); + C0 = ReplicateByteTo16(C0); u32 C1 = endpos32s[partition][1].Component(c); - C1 = Replicate(C1, 8, 16); + C1 = ReplicateByteTo16(C1); u32 plane = 0; if (weightParams.m_bDualPlane && (((planeIdx + 1) & 3) == c)) { diff --git a/src/video_core/textures/texture.cpp b/src/video_core/textures/texture.cpp new file mode 100644 index 000000000..d1939d744 --- /dev/null +++ b/src/video_core/textures/texture.cpp @@ -0,0 +1,80 @@ +// Copyright 2020 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <algorithm> +#include <array> + +#include "core/settings.h" +#include "video_core/textures/texture.h" + +namespace Tegra::Texture { + +namespace { + +constexpr std::array<float, 256> SRGB_CONVERSION_LUT = { + 0.000000f, 0.000000f, 0.000000f, 0.000012f, 0.000021f, 0.000033f, 0.000046f, 0.000062f, + 0.000081f, 0.000102f, 0.000125f, 0.000151f, 0.000181f, 0.000214f, 0.000251f, 0.000293f, + 0.000338f, 0.000388f, 0.000443f, 0.000503f, 0.000568f, 0.000639f, 0.000715f, 0.000798f, + 0.000887f, 0.000983f, 0.001085f, 0.001195f, 0.001312f, 0.001437f, 0.001569f, 0.001710f, + 0.001860f, 0.002019f, 0.002186f, 0.002364f, 0.002551f, 0.002748f, 0.002955f, 0.003174f, + 0.003403f, 0.003643f, 0.003896f, 0.004160f, 0.004436f, 0.004725f, 0.005028f, 0.005343f, + 0.005672f, 0.006015f, 0.006372f, 0.006744f, 0.007130f, 0.007533f, 0.007950f, 0.008384f, + 0.008834f, 0.009301f, 0.009785f, 0.010286f, 0.010805f, 0.011342f, 0.011898f, 0.012472f, + 0.013066f, 0.013680f, 0.014313f, 0.014967f, 0.015641f, 0.016337f, 0.017054f, 0.017793f, + 0.018554f, 0.019337f, 0.020144f, 0.020974f, 0.021828f, 0.022706f, 0.023609f, 0.024536f, + 0.025489f, 0.026468f, 0.027473f, 0.028504f, 0.029563f, 0.030649f, 0.031762f, 0.032904f, + 0.034074f, 0.035274f, 0.036503f, 0.037762f, 0.039050f, 0.040370f, 0.041721f, 0.043103f, + 0.044518f, 0.045964f, 0.047444f, 0.048956f, 0.050503f, 0.052083f, 0.053699f, 0.055349f, + 0.057034f, 0.058755f, 0.060513f, 0.062307f, 0.064139f, 0.066008f, 0.067915f, 0.069861f, + 0.071845f, 0.073869f, 0.075933f, 0.078037f, 0.080182f, 0.082369f, 0.084597f, 0.086867f, + 0.089180f, 0.091535f, 0.093935f, 0.096378f, 0.098866f, 0.101398f, 0.103977f, 0.106601f, + 0.109271f, 0.111988f, 0.114753f, 0.117565f, 0.120426f, 0.123335f, 0.126293f, 0.129301f, + 0.132360f, 0.135469f, 0.138629f, 0.141841f, 0.145105f, 0.148421f, 0.151791f, 0.155214f, + 0.158691f, 0.162224f, 0.165810f, 0.169453f, 0.173152f, 0.176907f, 0.180720f, 0.184589f, + 0.188517f, 0.192504f, 0.196549f, 0.200655f, 0.204820f, 0.209046f, 0.213334f, 0.217682f, + 0.222093f, 0.226567f, 0.231104f, 0.235704f, 0.240369f, 0.245099f, 0.249894f, 0.254754f, + 0.259681f, 0.264674f, 0.269736f, 0.274864f, 0.280062f, 0.285328f, 0.290664f, 0.296070f, + 0.301546f, 0.307094f, 0.312713f, 0.318404f, 0.324168f, 0.330006f, 0.335916f, 0.341902f, + 0.347962f, 0.354097f, 0.360309f, 0.366597f, 0.372961f, 0.379403f, 0.385924f, 0.392524f, + 0.399202f, 0.405960f, 0.412798f, 0.419718f, 0.426719f, 0.433802f, 0.440967f, 0.448216f, + 0.455548f, 0.462965f, 0.470465f, 0.478052f, 0.485725f, 0.493484f, 0.501329f, 0.509263f, + 0.517285f, 0.525396f, 0.533595f, 0.541885f, 0.550265f, 0.558736f, 0.567299f, 0.575954f, + 0.584702f, 0.593542f, 0.602477f, 0.611507f, 0.620632f, 0.629852f, 0.639168f, 0.648581f, + 0.658092f, 0.667700f, 0.677408f, 0.687214f, 0.697120f, 0.707127f, 0.717234f, 0.727443f, + 0.737753f, 0.748167f, 0.758685f, 0.769305f, 0.780031f, 0.790861f, 0.801798f, 0.812839f, + 0.823989f, 0.835246f, 0.846611f, 0.858085f, 0.869668f, 0.881360f, 0.893164f, 0.905078f, + 0.917104f, 0.929242f, 0.941493f, 0.953859f, 0.966338f, 1.000000f, 1.000000f, 1.000000f, +}; + +unsigned SettingsMinimumAnisotropy() noexcept { + switch (static_cast<Anisotropy>(Settings::values.max_anisotropy)) { + default: + case Anisotropy::Default: + return 1U; + case Anisotropy::Filter2x: + return 2U; + case Anisotropy::Filter4x: + return 4U; + case Anisotropy::Filter8x: + return 8U; + case Anisotropy::Filter16x: + return 16U; + } +} + +} // Anonymous namespace + +std::array<float, 4> TSCEntry::GetBorderColor() const noexcept { + if (!srgb_conversion) { + return border_color; + } + return {SRGB_CONVERSION_LUT[srgb_border_color_r], SRGB_CONVERSION_LUT[srgb_border_color_g], + SRGB_CONVERSION_LUT[srgb_border_color_b], border_color[3]}; +} + +float TSCEntry::GetMaxAnisotropy() const noexcept { + return static_cast<float>(std::max(1U << max_anisotropy, SettingsMinimumAnisotropy())); +} + +} // namespace Tegra::Texture diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h index 7edc4abe1..eba05aced 100644 --- a/src/video_core/textures/texture.h +++ b/src/video_core/textures/texture.h @@ -8,7 +8,6 @@ #include "common/assert.h" #include "common/bit_field.h" #include "common/common_types.h" -#include "core/settings.h" namespace Tegra::Texture { @@ -132,6 +131,20 @@ enum class SwizzleSource : u32 { OneFloat = 7, }; +enum class MsaaMode : u32 { + Msaa1x1 = 0, + Msaa2x1 = 1, + Msaa2x2 = 2, + Msaa4x2 = 3, + Msaa4x2_D3D = 4, + Msaa2x1_D3D = 5, + Msaa4x4 = 6, + Msaa2x2_VC4 = 8, + Msaa2x2_VC12 = 9, + Msaa4x2_VC8 = 10, + Msaa4x2_VC24 = 11, +}; + union TextureHandle { TextureHandle(u32 raw) : raw{raw} {} @@ -198,6 +211,7 @@ struct TICEntry { union { BitField<0, 4, u32> res_min_mip_level; BitField<4, 4, u32> res_max_mip_level; + BitField<8, 4, MsaaMode> msaa_mode; BitField<12, 12, u32> min_lod_clamp; }; @@ -336,24 +350,9 @@ struct TSCEntry { std::array<u8, 0x20> raw; }; - float GetMaxAnisotropy() const { - const u32 min_value = [] { - switch (static_cast<Anisotropy>(Settings::values.max_anisotropy)) { - default: - case Anisotropy::Default: - return 1U; - case Anisotropy::Filter2x: - return 2U; - case Anisotropy::Filter4x: - return 4U; - case Anisotropy::Filter8x: - return 8U; - case Anisotropy::Filter16x: - return 16U; - } - }(); - return static_cast<float>(std::max(1U << max_anisotropy, min_value)); - } + std::array<float, 4> GetBorderColor() const noexcept; + + float GetMaxAnisotropy() const noexcept; float GetMinLod() const { return static_cast<float>(min_lod_clamp) / 256.0f; @@ -368,15 +367,6 @@ struct TSCEntry { constexpr u32 mask = 1U << (13 - 1); return static_cast<float>(static_cast<s32>((mip_lod_bias ^ mask) - mask)) / 256.0f; } - - std::array<float, 4> GetBorderColor() const { - if (srgb_conversion) { - return {static_cast<float>(srgb_border_color_r) / 255.0f, - static_cast<float>(srgb_border_color_g) / 255.0f, - static_cast<float>(srgb_border_color_b) / 255.0f, border_color[3]}; - } - return border_color; - } }; static_assert(sizeof(TSCEntry) == 0x20, "TSCEntry has wrong size"); |