diff options
Diffstat (limited to 'src/video_core')
| -rw-r--r-- | src/video_core/engines/maxwell_dma.cpp | 300 | ||||
| -rw-r--r-- | src/video_core/engines/maxwell_dma.h | 348 | ||||
| -rw-r--r-- | src/video_core/renderer_vulkan/vk_blit_screen.cpp | 718 | ||||
| -rw-r--r-- | src/video_core/renderer_vulkan/vk_device.cpp | 276 | ||||
| -rw-r--r-- | src/video_core/texture_cache/surface_params.cpp | 5 | ||||
| -rw-r--r-- | src/video_core/texture_cache/surface_params.h | 2 | ||||
| -rw-r--r-- | src/video_core/textures/decoders.cpp | 134 | ||||
| -rw-r--r-- | src/video_core/textures/decoders.h | 38 |
8 files changed, 1004 insertions, 817 deletions
diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp index 01d7df405..a2d3d7823 100644 --- a/src/video_core/engines/maxwell_dma.cpp +++ b/src/video_core/engines/maxwell_dma.cpp @@ -14,50 +14,45 @@ namespace Tegra::Engines { +using namespace Texture; + MaxwellDMA::MaxwellDMA(Core::System& system, MemoryManager& memory_manager) : system{system}, memory_manager{memory_manager} {} void MaxwellDMA::CallMethod(u32 method, u32 method_argument, bool is_last_call) { - ASSERT_MSG(method < Regs::NUM_REGS, - "Invalid MaxwellDMA register, increase the size of the Regs structure"); + ASSERT_MSG(method < NUM_REGS, "Invalid MaxwellDMA register"); regs.reg_array[method] = method_argument; -#define MAXWELLDMA_REG_INDEX(field_name) \ - (offsetof(Tegra::Engines::MaxwellDMA::Regs, field_name) / sizeof(u32)) - - switch (method) { - case MAXWELLDMA_REG_INDEX(exec): { - HandleCopy(); - break; - } + if (method == offsetof(Regs, launch_dma) / sizeof(u32)) { + Launch(); } - -#undef MAXWELLDMA_REG_INDEX } void MaxwellDMA::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) { - for (std::size_t i = 0; i < amount; i++) { + for (size_t i = 0; i < amount; ++i) { CallMethod(method, base_start[i], methods_pending - static_cast<u32>(i) <= 1); } } -void MaxwellDMA::HandleCopy() { - LOG_TRACE(HW_GPU, "Requested a DMA copy"); - - const GPUVAddr source = regs.src_address.Address(); - const GPUVAddr dest = regs.dst_address.Address(); +void MaxwellDMA::Launch() { + LOG_TRACE(Render_OpenGL, "DMA copy 0x{:x} -> 0x{:x}", static_cast<GPUVAddr>(regs.offset_in), + static_cast<GPUVAddr>(regs.offset_out)); // TODO(Subv): Perform more research and implement all features of this engine. - ASSERT(regs.exec.enable_swizzle == 0); - ASSERT(regs.exec.query_mode == Regs::QueryMode::None); - ASSERT(regs.exec.query_intr == Regs::QueryIntr::None); - ASSERT(regs.exec.copy_mode == Regs::CopyMode::Unk2); - ASSERT(regs.dst_params.pos_x == 0); - ASSERT(regs.dst_params.pos_y == 0); - - if (!regs.exec.is_dst_linear && !regs.exec.is_src_linear) { + const LaunchDMA& launch = regs.launch_dma; + ASSERT(launch.remap_enable == 0); + ASSERT(launch.semaphore_type == LaunchDMA::SemaphoreType::NONE); + ASSERT(launch.interrupt_type == LaunchDMA::InterruptType::NONE); + ASSERT(launch.data_transfer_type == LaunchDMA::DataTransferType::NON_PIPELINED); + ASSERT(regs.dst_params.origin.x == 0); + ASSERT(regs.dst_params.origin.y == 0); + + const bool is_src_pitch = launch.src_memory_layout == LaunchDMA::MemoryLayout::PITCH; + const bool is_dst_pitch = launch.dst_memory_layout == LaunchDMA::MemoryLayout::PITCH; + + if (!is_src_pitch && !is_dst_pitch) { // If both the source and the destination are in block layout, assert. UNREACHABLE_MSG("Tiled->Tiled DMA transfers are not yet implemented"); return; @@ -66,144 +61,161 @@ void MaxwellDMA::HandleCopy() { // All copies here update the main memory, so mark all rasterizer states as invalid. system.GPU().Maxwell3D().OnMemoryWrite(); - if (regs.exec.is_dst_linear && regs.exec.is_src_linear) { - // When the enable_2d bit is disabled, the copy is performed as if we were copying a 1D - // buffer of length `x_count`, otherwise we copy a 2D image of dimensions (x_count, - // y_count). - if (!regs.exec.enable_2d) { - memory_manager.CopyBlock(dest, source, regs.x_count); - return; - } + if (is_src_pitch && is_dst_pitch) { + CopyPitchToPitch(); + } else { + ASSERT(launch.multi_line_enable == 1); - // If both the source and the destination are in linear layout, perform a line-by-line - // copy. We're going to take a subrect of size (x_count, y_count) from the source - // rectangle. There is no need to manually flush/invalidate the regions because - // CopyBlock does that for us. - for (u32 line = 0; line < regs.y_count; ++line) { - const GPUVAddr source_line = source + line * regs.src_pitch; - const GPUVAddr dest_line = dest + line * regs.dst_pitch; - memory_manager.CopyBlock(dest_line, source_line, regs.x_count); + if (!is_src_pitch && is_dst_pitch) { + CopyBlockLinearToPitch(); + } else { + CopyPitchToBlockLinear(); } - return; } +} - ASSERT(regs.exec.enable_2d == 1); - - if (regs.exec.is_dst_linear && !regs.exec.is_src_linear) { - - ASSERT(regs.src_params.BlockDepth() == 0); - // Optimized path for micro copies. - if (regs.dst_pitch * regs.y_count < Texture::GetGOBSize() && regs.dst_pitch <= 64) { - const u32 bytes_per_pixel = regs.dst_pitch / regs.x_count; - const std::size_t src_size = Texture::GetGOBSize(); - const std::size_t dst_size = regs.dst_pitch * regs.y_count; - u32 pos_x = regs.src_params.pos_x; - u32 pos_y = regs.src_params.pos_y; - const u64 offset = - Texture::GetGOBOffset(regs.src_params.size_x, regs.src_params.size_y, pos_x, pos_y, - regs.src_params.BlockDepth(), bytes_per_pixel); - const u32 x_in_gob = 64 / bytes_per_pixel; - pos_x = pos_x % x_in_gob; - pos_y = pos_y % 8; - - if (read_buffer.size() < src_size) { - read_buffer.resize(src_size); - } - - if (write_buffer.size() < dst_size) { - write_buffer.resize(dst_size); - } - - if (Settings::IsGPULevelExtreme()) { - memory_manager.ReadBlock(source + offset, read_buffer.data(), src_size); - memory_manager.ReadBlock(dest, write_buffer.data(), dst_size); - } else { - memory_manager.ReadBlockUnsafe(source + offset, read_buffer.data(), src_size); - memory_manager.ReadBlockUnsafe(dest, write_buffer.data(), dst_size); - } - - Texture::UnswizzleSubrect(regs.x_count, regs.y_count, regs.dst_pitch, - regs.src_params.size_x, bytes_per_pixel, read_buffer.data(), - write_buffer.data(), regs.src_params.BlockHeight(), pos_x, - pos_y); - - memory_manager.WriteBlock(dest, write_buffer.data(), dst_size); - - return; - } - // If the input is tiled and the output is linear, deswizzle the input and copy it over. - const u32 bytes_per_pixel = regs.dst_pitch / regs.x_count; - const std::size_t src_size = Texture::CalculateSize( - true, bytes_per_pixel, regs.src_params.size_x, regs.src_params.size_y, - regs.src_params.size_z, regs.src_params.BlockHeight(), regs.src_params.BlockDepth()); - - const std::size_t src_layer_size = Texture::CalculateSize( - true, bytes_per_pixel, regs.src_params.size_x, regs.src_params.size_y, 1, - regs.src_params.BlockHeight(), regs.src_params.BlockDepth()); - - const std::size_t dst_size = regs.dst_pitch * regs.y_count; +void MaxwellDMA::CopyPitchToPitch() { + // When `multi_line_enable` bit is disabled the copy is performed as if we were copying a 1D + // buffer of length `line_length_in`. + // Otherwise we copy a 2D image of dimensions (line_length_in, line_count). + if (!regs.launch_dma.multi_line_enable) { + memory_manager.CopyBlock(regs.offset_out, regs.offset_in, regs.line_length_in); + return; + } - if (read_buffer.size() < src_size) { - read_buffer.resize(src_size); - } + // Perform a line-by-line copy. + // We're going to take a subrect of size (line_length_in, line_count) from the source rectangle. + // There is no need to manually flush/invalidate the regions because CopyBlock does that for us. + for (u32 line = 0; line < regs.line_count; ++line) { + const GPUVAddr source_line = regs.offset_in + static_cast<size_t>(line) * regs.pitch_in; + const GPUVAddr dest_line = regs.offset_out + static_cast<size_t>(line) * regs.pitch_out; + memory_manager.CopyBlock(dest_line, source_line, regs.line_length_in); + } +} - if (write_buffer.size() < dst_size) { - write_buffer.resize(dst_size); - } +void MaxwellDMA::CopyBlockLinearToPitch() { + ASSERT(regs.src_params.block_size.depth == 0); - if (Settings::IsGPULevelExtreme()) { - memory_manager.ReadBlock(source, read_buffer.data(), src_size); - memory_manager.ReadBlock(dest, write_buffer.data(), dst_size); - } else { - memory_manager.ReadBlockUnsafe(source, read_buffer.data(), src_size); - memory_manager.ReadBlockUnsafe(dest, write_buffer.data(), dst_size); - } + // Optimized path for micro copies. + const size_t dst_size = static_cast<size_t>(regs.pitch_out) * regs.line_count; + if (dst_size < GOB_SIZE && regs.pitch_out <= GOB_SIZE_X) { + FastCopyBlockLinearToPitch(); + return; + } - Texture::UnswizzleSubrect( - regs.x_count, regs.y_count, regs.dst_pitch, regs.src_params.size_x, bytes_per_pixel, - read_buffer.data() + src_layer_size * regs.src_params.pos_z, write_buffer.data(), - regs.src_params.BlockHeight(), regs.src_params.pos_x, regs.src_params.pos_y); + // Deswizzle the input and copy it over. + const u32 bytes_per_pixel = regs.pitch_out / regs.line_length_in; + const Parameters& src_params = regs.src_params; + const u32 width = src_params.width; + const u32 height = src_params.height; + const u32 depth = src_params.depth; + const u32 block_height = src_params.block_size.height; + const u32 block_depth = src_params.block_size.depth; + const size_t src_size = + CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth); + const size_t src_layer_size = + CalculateSize(true, bytes_per_pixel, width, height, 1, block_height, block_depth); + + if (read_buffer.size() < src_size) { + read_buffer.resize(src_size); + } + if (write_buffer.size() < dst_size) { + write_buffer.resize(dst_size); + } - memory_manager.WriteBlock(dest, write_buffer.data(), dst_size); + if (Settings::IsGPULevelExtreme()) { + memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size); + memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size); } else { - ASSERT(regs.dst_params.BlockDepth() == 0); + memory_manager.ReadBlockUnsafe(regs.offset_in, read_buffer.data(), src_size); + memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size); + } - const u32 bytes_per_pixel = regs.src_pitch / regs.x_count; + UnswizzleSubrect(regs.line_length_in, regs.line_count, regs.pitch_out, width, bytes_per_pixel, + read_buffer.data() + src_layer_size * src_params.layer, write_buffer.data(), + block_height, src_params.origin.x, src_params.origin.y); - const std::size_t dst_size = Texture::CalculateSize( - true, bytes_per_pixel, regs.dst_params.size_x, regs.dst_params.size_y, - regs.dst_params.size_z, regs.dst_params.BlockHeight(), regs.dst_params.BlockDepth()); + memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size); +} - const std::size_t dst_layer_size = Texture::CalculateSize( - true, bytes_per_pixel, regs.dst_params.size_x, regs.dst_params.size_y, 1, - regs.dst_params.BlockHeight(), regs.dst_params.BlockDepth()); +void MaxwellDMA::CopyPitchToBlockLinear() { + const auto& dst_params = regs.dst_params; + const u32 bytes_per_pixel = regs.pitch_in / regs.line_length_in; + const u32 width = dst_params.width; + const u32 height = dst_params.height; + const u32 depth = dst_params.depth; + const u32 block_height = dst_params.block_size.height; + const u32 block_depth = dst_params.block_size.depth; + const size_t dst_size = + CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth); + const size_t dst_layer_size = + CalculateSize(true, bytes_per_pixel, width, height, 1, block_height, block_depth); + + const size_t src_size = static_cast<size_t>(regs.pitch_in) * regs.line_count; + + if (read_buffer.size() < src_size) { + read_buffer.resize(src_size); + } + if (write_buffer.size() < dst_size) { + write_buffer.resize(dst_size); + } - const std::size_t src_size = regs.src_pitch * regs.y_count; + if (Settings::IsGPULevelExtreme()) { + memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size); + memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size); + } else { + memory_manager.ReadBlockUnsafe(regs.offset_in, read_buffer.data(), src_size); + memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size); + } - if (read_buffer.size() < src_size) { - read_buffer.resize(src_size); - } + // If the input is linear and the output is tiled, swizzle the input and copy it over. + if (regs.dst_params.block_size.depth > 0) { + ASSERT(dst_params.layer == 0); + SwizzleSliceToVoxel(regs.line_length_in, regs.line_count, regs.pitch_in, width, height, + bytes_per_pixel, block_height, block_depth, dst_params.origin.x, + dst_params.origin.y, write_buffer.data(), read_buffer.data()); + } else { + SwizzleSubrect(regs.line_length_in, regs.line_count, regs.pitch_in, width, bytes_per_pixel, + write_buffer.data() + dst_layer_size * dst_params.layer, read_buffer.data(), + block_height, dst_params.origin.x, dst_params.origin.y); + } - if (write_buffer.size() < dst_size) { - write_buffer.resize(dst_size); - } + memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size); +} - if (Settings::IsGPULevelExtreme()) { - memory_manager.ReadBlock(source, read_buffer.data(), src_size); - memory_manager.ReadBlock(dest, write_buffer.data(), dst_size); - } else { - memory_manager.ReadBlockUnsafe(source, read_buffer.data(), src_size); - memory_manager.ReadBlockUnsafe(dest, write_buffer.data(), dst_size); - } +void MaxwellDMA::FastCopyBlockLinearToPitch() { + const u32 bytes_per_pixel = regs.pitch_out / regs.line_length_in; + const size_t src_size = GOB_SIZE; + const size_t dst_size = static_cast<size_t>(regs.pitch_out) * regs.line_count; + u32 pos_x = regs.src_params.origin.x; + u32 pos_y = regs.src_params.origin.y; + const u64 offset = GetGOBOffset(regs.src_params.width, regs.src_params.height, pos_x, pos_y, + regs.src_params.block_size.height, bytes_per_pixel); + const u32 x_in_gob = 64 / bytes_per_pixel; + pos_x = pos_x % x_in_gob; + pos_y = pos_y % 8; + + if (read_buffer.size() < src_size) { + read_buffer.resize(src_size); + } - // If the input is linear and the output is tiled, swizzle the input and copy it over. - Texture::SwizzleSubrect( - regs.x_count, regs.y_count, regs.src_pitch, regs.dst_params.size_x, bytes_per_pixel, - write_buffer.data() + dst_layer_size * regs.dst_params.pos_z, read_buffer.data(), - regs.dst_params.BlockHeight(), regs.dst_params.pos_x, regs.dst_params.pos_y); + if (write_buffer.size() < dst_size) { + write_buffer.resize(dst_size); + } - memory_manager.WriteBlock(dest, write_buffer.data(), dst_size); + if (Settings::IsGPULevelExtreme()) { + memory_manager.ReadBlock(regs.offset_in + offset, read_buffer.data(), src_size); + memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size); + } else { + memory_manager.ReadBlockUnsafe(regs.offset_in + offset, read_buffer.data(), src_size); + memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size); } + + UnswizzleSubrect(regs.line_length_in, regs.line_count, regs.pitch_out, regs.src_params.width, + bytes_per_pixel, read_buffer.data(), write_buffer.data(), + regs.src_params.block_size.height, pos_x, pos_y); + + memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size); } } // namespace Tegra::Engines diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h index 502dd8509..50f445efc 100644 --- a/src/video_core/engines/maxwell_dma.h +++ b/src/video_core/engines/maxwell_dma.h @@ -24,160 +24,190 @@ class MemoryManager; namespace Tegra::Engines { /** - * This Engine is known as GK104_Copy. Documentation can be found in: + * This engine is known as gk104_copy. Documentation can be found in: + * https://github.com/NVIDIA/open-gpu-doc/blob/master/classes/dma-copy/clb0b5.h * https://github.com/envytools/envytools/blob/master/rnndb/fifo/gk104_copy.xml */ class MaxwellDMA final : public EngineInterface { public: - explicit MaxwellDMA(Core::System& system, MemoryManager& memory_manager); - ~MaxwellDMA() = default; - - /// Write the value to the register identified by method. - void CallMethod(u32 method, u32 method_argument, bool is_last_call) override; - - /// Write multiple values to the register identified by method. - void CallMultiMethod(u32 method, const u32* base_start, u32 amount, - u32 methods_pending) override; + struct PackedGPUVAddr { + u32 upper; + u32 lower; + + constexpr operator GPUVAddr() const noexcept { + return (static_cast<GPUVAddr>(upper & 0xff) << 32) | lower; + } + }; + + union BlockSize { + BitField<0, 4, u32> width; + BitField<4, 4, u32> height; + BitField<8, 4, u32> depth; + BitField<12, 4, u32> gob_height; + }; + static_assert(sizeof(BlockSize) == 4); + + union Origin { + BitField<0, 16, u32> x; + BitField<16, 16, u32> y; + }; + static_assert(sizeof(Origin) == 4); + + struct Parameters { + BlockSize block_size; + u32 width; + u32 height; + u32 depth; + u32 layer; + Origin origin; + }; + static_assert(sizeof(Parameters) == 24); + + struct Semaphore { + PackedGPUVAddr address; + u32 payload; + }; + static_assert(sizeof(Semaphore) == 12); + + struct RenderEnable { + enum class Mode : u32 { + FALSE = 0, + TRUE = 1, + CONDITIONAL = 2, + RENDER_IF_EQUAL = 3, + RENDER_IF_NOT_EQUAL = 4, + }; - struct Regs { - static constexpr std::size_t NUM_REGS = 0x1D6; + PackedGPUVAddr address; + BitField<0, 3, Mode> mode; + }; + static_assert(sizeof(RenderEnable) == 12); + + enum class PhysModeTarget : u32 { + LOCAL_FB = 0, + COHERENT_SYSMEM = 1, + NONCOHERENT_SYSMEM = 2, + }; + using PhysMode = BitField<0, 2, PhysModeTarget>; + + union LaunchDMA { + enum class DataTransferType : u32 { + NONE = 0, + PIPELINED = 1, + NON_PIPELINED = 2, + }; - struct Parameters { - union { - BitField<0, 4, u32> block_depth; - BitField<4, 4, u32> block_height; - BitField<8, 4, u32> block_width; - }; - u32 size_x; - u32 size_y; - u32 size_z; - u32 pos_z; - union { - BitField<0, 16, u32> pos_x; - BitField<16, 16, u32> pos_y; - }; + enum class SemaphoreType : u32 { + NONE = 0, + RELEASE_ONE_WORD_SEMAPHORE = 1, + RELEASE_FOUR_WORD_SEMAPHORE = 2, + }; - u32 BlockHeight() const { - return block_height.Value(); - } + enum class InterruptType : u32 { + NONE = 0, + BLOCKING = 1, + NON_BLOCKING = 2, + }; - u32 BlockDepth() const { - return block_depth.Value(); - } + enum class MemoryLayout : u32 { + BLOCKLINEAR = 0, + PITCH = 1, }; - static_assert(sizeof(Parameters) == 24, "Parameters has wrong size"); + enum class Type : u32 { + VIRTUAL = 0, + PHYSICAL = 1, + }; - enum class ComponentMode : u32 { - Src0 = 0, - Src1 = 1, - Src2 = 2, - Src3 = 3, - Const0 = 4, - Const1 = 5, - Zero = 6, + enum class SemaphoreReduction : u32 { + IMIN = 0, + IMAX = 1, + IXOR = 2, + IAND = 3, + IOR = 4, + IADD = 5, + INC = 6, + DEC = 7, + FADD = 0xA, }; - enum class CopyMode : u32 { - None = 0, - Unk1 = 1, - Unk2 = 2, + enum class SemaphoreReductionSign : u32 { + SIGNED = 0, + UNSIGNED = 1, }; - enum class QueryMode : u32 { - None = 0, - Short = 1, - Long = 2, + enum class BypassL2 : u32 { + USE_PTE_SETTING = 0, + FORCE_VOLATILE = 1, }; - enum class QueryIntr : u32 { - None = 0, - Block = 1, - NonBlock = 2, + BitField<0, 2, DataTransferType> data_transfer_type; + BitField<2, 1, u32> flush_enable; + BitField<3, 2, SemaphoreType> semaphore_type; + BitField<5, 2, InterruptType> interrupt_type; + BitField<7, 1, MemoryLayout> src_memory_layout; + BitField<8, 1, MemoryLayout> dst_memory_layout; + BitField<9, 1, u32> multi_line_enable; + BitField<10, 1, u32> remap_enable; + BitField<11, 1, u32> rmwdisable; + BitField<12, 1, Type> src_type; + BitField<13, 1, Type> dst_type; + BitField<14, 4, SemaphoreReduction> semaphore_reduction; + BitField<18, 1, SemaphoreReductionSign> semaphore_reduction_sign; + BitField<19, 1, u32> reduction_enable; + BitField<20, 1, BypassL2> bypass_l2; + }; + static_assert(sizeof(LaunchDMA) == 4); + + struct RemapConst { + enum Swizzle : u32 { + SRC_X = 0, + SRC_Y = 1, + SRC_Z = 2, + SRC_W = 3, + CONST_A = 4, + CONST_B = 5, + NO_WRITE = 6, }; - union { - struct { - INSERT_UNION_PADDING_WORDS(0xC0); - - struct { - union { - BitField<0, 2, CopyMode> copy_mode; - BitField<2, 1, u32> flush; - - BitField<3, 2, QueryMode> query_mode; - BitField<5, 2, QueryIntr> query_intr; - - BitField<7, 1, u32> is_src_linear; - BitField<8, 1, u32> is_dst_linear; - - BitField<9, 1, u32> enable_2d; - BitField<10, 1, u32> enable_swizzle; - }; - } exec; - - INSERT_UNION_PADDING_WORDS(0x3F); - - struct { - u32 address_high; - u32 address_low; - - GPUVAddr Address() const { - return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) | - address_low); - } - } src_address; - - struct { - u32 address_high; - u32 address_low; - - GPUVAddr Address() const { - return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) | - address_low); - } - } dst_address; - - u32 src_pitch; - u32 dst_pitch; - u32 x_count; - u32 y_count; - - INSERT_UNION_PADDING_WORDS(0xB8); - - u32 const0; - u32 const1; - union { - BitField<0, 4, ComponentMode> component0; - BitField<4, 4, ComponentMode> component1; - BitField<8, 4, ComponentMode> component2; - BitField<12, 4, ComponentMode> component3; - BitField<16, 2, u32> component_size; - BitField<20, 3, u32> src_num_components; - BitField<24, 3, u32> dst_num_components; - - u32 SrcBytePerPixel() const { - return src_num_components.Value() * component_size.Value(); - } - u32 DstBytePerPixel() const { - return dst_num_components.Value() * component_size.Value(); - } - } swizzle_config; + PackedGPUVAddr address; - Parameters dst_params; + union { + BitField<0, 3, Swizzle> dst_x; + BitField<4, 3, Swizzle> dst_y; + BitField<8, 3, Swizzle> dst_z; + BitField<12, 3, Swizzle> dst_w; + BitField<16, 2, u32> component_size_minus_one; + BitField<20, 2, u32> num_src_components_minus_one; + BitField<24, 2, u32> num_dst_components_minus_one; + }; + }; + static_assert(sizeof(RemapConst) == 12); - INSERT_UNION_PADDING_WORDS(1); + explicit MaxwellDMA(Core::System& system, MemoryManager& memory_manager); + ~MaxwellDMA() = default; - Parameters src_params; + /// Write the value to the register identified by method. + void CallMethod(u32 method, u32 method_argument, bool is_last_call) override; - INSERT_UNION_PADDING_WORDS(0x13); - }; - std::array<u32, NUM_REGS> reg_array; - }; - } regs{}; + /// Write multiple values to the register identified by method. + void CallMultiMethod(u32 method, const u32* base_start, u32 amount, + u32 methods_pending) override; private: + /// Performs the copy from the source buffer to the destination buffer as configured in the + /// registers. + void Launch(); + + void CopyPitchToPitch(); + + void CopyBlockLinearToPitch(); + + void CopyPitchToBlockLinear(); + + void FastCopyBlockLinearToPitch(); + Core::System& system; MemoryManager& memory_manager; @@ -185,28 +215,58 @@ private: std::vector<u8> read_buffer; std::vector<u8> write_buffer; - /// Performs the copy from the source buffer to the destination buffer as configured in the - /// registers. - void HandleCopy(); -}; + static constexpr std::size_t NUM_REGS = 0x800; + struct Regs { + union { + struct { + u32 reserved[0x40]; + u32 nop; + u32 reserved01[0xf]; + u32 pm_trigger; + u32 reserved02[0x3f]; + Semaphore semaphore; + u32 reserved03[0x2]; + RenderEnable render_enable; + PhysMode src_phys_mode; + PhysMode dst_phys_mode; + u32 reserved04[0x26]; + LaunchDMA launch_dma; + u32 reserved05[0x3f]; + PackedGPUVAddr offset_in; + PackedGPUVAddr offset_out; + u32 pitch_in; + u32 pitch_out; + u32 line_length_in; + u32 line_count; + u32 reserved06[0xb8]; + RemapConst remap_const; + Parameters dst_params; + u32 reserved07[0x1]; + Parameters src_params; + u32 reserved08[0x275]; + u32 pm_trigger_end; + u32 reserved09[0x3ba]; + }; + std::array<u32, NUM_REGS> reg_array; + }; + } regs{}; #define ASSERT_REG_POSITION(field_name, position) \ static_assert(offsetof(MaxwellDMA::Regs, field_name) == position * 4, \ "Field " #field_name " has invalid position") -ASSERT_REG_POSITION(exec, 0xC0); -ASSERT_REG_POSITION(src_address, 0x100); -ASSERT_REG_POSITION(dst_address, 0x102); -ASSERT_REG_POSITION(src_pitch, 0x104); -ASSERT_REG_POSITION(dst_pitch, 0x105); -ASSERT_REG_POSITION(x_count, 0x106); -ASSERT_REG_POSITION(y_count, 0x107); -ASSERT_REG_POSITION(const0, 0x1C0); -ASSERT_REG_POSITION(const1, 0x1C1); -ASSERT_REG_POSITION(swizzle_config, 0x1C2); -ASSERT_REG_POSITION(dst_params, 0x1C3); -ASSERT_REG_POSITION(src_params, 0x1CA); + ASSERT_REG_POSITION(launch_dma, 0xC0); + ASSERT_REG_POSITION(offset_in, 0x100); + ASSERT_REG_POSITION(offset_out, 0x102); + ASSERT_REG_POSITION(pitch_in, 0x104); + ASSERT_REG_POSITION(pitch_out, 0x105); + ASSERT_REG_POSITION(line_length_in, 0x106); + ASSERT_REG_POSITION(line_count, 0x107); + ASSERT_REG_POSITION(remap_const, 0x1C0); + ASSERT_REG_POSITION(dst_params, 0x1C3); + ASSERT_REG_POSITION(src_params, 0x1CA); #undef ASSERT_REG_POSITION +}; } // namespace Tegra::Engines diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp index fbd406f2b..866813465 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp +++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp @@ -141,24 +141,28 @@ struct ScreenRectVertex { std::array<f32, 2> tex_coord; static VkVertexInputBindingDescription GetDescription() { - VkVertexInputBindingDescription description; - description.binding = 0; - description.stride = sizeof(ScreenRectVertex); - description.inputRate = VK_VERTEX_INPUT_RATE_VERTEX; - return description; + return { + .binding = 0, + .stride = sizeof(ScreenRectVertex), + .inputRate = VK_VERTEX_INPUT_RATE_VERTEX, + }; } 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; + return {{ + { + .location = 0, + .binding = 0, + .format = VK_FORMAT_R32G32_SFLOAT, + .offset = offsetof(ScreenRectVertex, position), + }, + { + .location = 1, + .binding = 0, + .format = VK_FORMAT_R32G32_SFLOAT, + .offset = offsetof(ScreenRectVertex, tex_coord), + }, + }}; } }; @@ -267,20 +271,25 @@ std::tuple<VKFence&, VkSemaphore> VKBlitScreen::Draw(const Tegra::FramebufferCon 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; + const VkBufferImageCopy copy{ + .bufferOffset = image_offset, + .bufferRowLength = 0, + .bufferImageHeight = 0, + .imageSubresource = + { + .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, + .mipLevel = 0, + .baseArrayLayer = 0, + .layerCount = 1, + }, + .imageOffset = {.x = 0, .y = 0, .z = 0}, + .imageExtent = + { + .width = framebuffer.width, + .height = framebuffer.height, + .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); @@ -295,11 +304,9 @@ std::tuple<VKFence&, VkSemaphore> VKBlitScreen::Draw(const Tegra::FramebufferCon descriptor_set = descriptor_sets[image_index], buffer = *buffer, size = swapchain.GetSize(), pipeline = *pipeline, 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; + const VkClearValue clear_color{ + .color = {.float32 = {0.0f, 0.0f, 0.0f, 0.0f}}, + }; VkRenderPassBeginInfo renderpass_bi; renderpass_bi.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO; @@ -379,93 +386,109 @@ void VKBlitScreen::CreateSemaphores() { } void VKBlitScreen::CreateDescriptorPool() { - 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(); + const std::array<VkDescriptorPoolSize, 2> pool_sizes{{ + { + .type = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, + .descriptorCount = static_cast<u32>(image_count), + }, + { + .type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, + .descriptorCount = static_cast<u32>(image_count), + }, + }}; + + const VkDescriptorPoolCreateInfo ci{ + .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO, + .pNext = nullptr, + .flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT, + .maxSets = static_cast<u32>(image_count), + .poolSizeCount = static_cast<u32>(pool_sizes.size()), + .pPoolSizes = pool_sizes.data(), + }; descriptor_pool = device.GetLogical().CreateDescriptorPool(ci); } void VKBlitScreen::CreateRenderPass() { - 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; + const VkAttachmentDescription color_attachment{ + .flags = 0, + .format = swapchain.GetImageFormat(), + .samples = VK_SAMPLE_COUNT_1_BIT, + .loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR, + .storeOp = VK_ATTACHMENT_STORE_OP_STORE, + .stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE, + .stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE, + .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED, + .finalLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR, + }; + + const VkAttachmentReference color_attachment_ref{ + .attachment = 0, + .layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL, + }; + + const VkSubpassDescription subpass_description{ + .flags = 0, + .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS, + .inputAttachmentCount = 0, + .pInputAttachments = nullptr, + .colorAttachmentCount = 1, + .pColorAttachments = &color_attachment_ref, + .pResolveAttachments = nullptr, + .pDepthStencilAttachment = nullptr, + .preserveAttachmentCount = 0, + .pPreserveAttachments = nullptr, + }; + + const VkSubpassDependency dependency{ + .srcSubpass = VK_SUBPASS_EXTERNAL, + .dstSubpass = 0, + .srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, + .dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, + .srcAccessMask = 0, + .dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT, + .dependencyFlags = 0, + }; + + const VkRenderPassCreateInfo renderpass_ci{ + .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .attachmentCount = 1, + .pAttachments = &color_attachment, + .subpassCount = 1, + .pSubpasses = &subpass_description, + .dependencyCount = 1, + .pDependencies = &dependency, + }; renderpass = device.GetLogical().CreateRenderPass(renderpass_ci); } void VKBlitScreen::CreateDescriptorSetLayout() { - 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(); + const std::array<VkDescriptorSetLayoutBinding, 2> layout_bindings{{ + { + .binding = 0, + .descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, + .descriptorCount = 1, + .stageFlags = VK_SHADER_STAGE_VERTEX_BIT, + .pImmutableSamplers = nullptr, + }, + { + .binding = 1, + .descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, + .descriptorCount = 1, + .stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT, + .pImmutableSamplers = nullptr, + }, + }}; + + const VkDescriptorSetLayoutCreateInfo ci{ + .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .bindingCount = static_cast<u32>(layout_bindings.size()), + .pBindings = layout_bindings.data(), + }; descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout(ci); } @@ -473,175 +496,192 @@ void VKBlitScreen::CreateDescriptorSetLayout() { void VKBlitScreen::CreateDescriptorSets() { 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(); + const VkDescriptorSetAllocateInfo ai{ + .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO, + .pNext = nullptr, + .descriptorPool = *descriptor_pool, + .descriptorSetCount = static_cast<u32>(image_count), + .pSetLayouts = layouts.data(), + }; + descriptor_sets = descriptor_pool.Allocate(ai); } void VKBlitScreen::CreatePipelineLayout() { - 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; + const VkPipelineLayoutCreateInfo ci{ + .sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .setLayoutCount = 1, + .pSetLayouts = descriptor_set_layout.address(), + .pushConstantRangeCount = 0, + .pPushConstantRanges = nullptr, + }; pipeline_layout = device.GetLogical().CreatePipelineLayout(ci); } void VKBlitScreen::CreateGraphicsPipeline() { - 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 std::array<VkPipelineShaderStageCreateInfo, 2> shader_stages{{ + { + .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .stage = VK_SHADER_STAGE_VERTEX_BIT, + .module = *vertex_shader, + .pName = "main", + .pSpecializationInfo = nullptr, + }, + { + .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .stage = VK_SHADER_STAGE_FRAGMENT_BIT, + .module = *fragment_shader, + .pName = "main", + .pSpecializationInfo = nullptr, + }, + }}; const auto vertex_binding_description = ScreenRectVertex::GetDescription(); const auto vertex_attrs_description = ScreenRectVertex::GetAttributes(); - 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; + const VkPipelineVertexInputStateCreateInfo vertex_input_ci{ + .sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .vertexBindingDescriptionCount = 1, + .pVertexBindingDescriptions = &vertex_binding_description, + .vertexAttributeDescriptionCount = u32{vertex_attrs_description.size()}, + .pVertexAttributeDescriptions = vertex_attrs_description.data(), + }; + + const VkPipelineInputAssemblyStateCreateInfo input_assembly_ci{ + .sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP, + .primitiveRestartEnable = VK_FALSE, + }; + + const VkPipelineViewportStateCreateInfo viewport_state_ci{ + .sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .viewportCount = 1, + .pViewports = nullptr, + .scissorCount = 1, + .pScissors = nullptr, + }; + + const VkPipelineRasterizationStateCreateInfo rasterization_ci{ + .sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .depthClampEnable = VK_FALSE, + .rasterizerDiscardEnable = VK_FALSE, + .polygonMode = VK_POLYGON_MODE_FILL, + .cullMode = VK_CULL_MODE_NONE, + .frontFace = VK_FRONT_FACE_CLOCKWISE, + .depthBiasEnable = VK_FALSE, + .depthBiasConstantFactor = 0.0f, + .depthBiasClamp = 0.0f, + .depthBiasSlopeFactor = 0.0f, + .lineWidth = 1.0f, + }; + + const VkPipelineMultisampleStateCreateInfo multisampling_ci{ + .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .rasterizationSamples = VK_SAMPLE_COUNT_1_BIT, + .sampleShadingEnable = VK_FALSE, + .minSampleShading = 0.0f, + .pSampleMask = nullptr, + .alphaToCoverageEnable = VK_FALSE, + .alphaToOneEnable = VK_FALSE, + }; + + const VkPipelineColorBlendAttachmentState color_blend_attachment{ + .blendEnable = VK_FALSE, + .srcColorBlendFactor = VK_BLEND_FACTOR_ZERO, + .dstColorBlendFactor = VK_BLEND_FACTOR_ZERO, + .colorBlendOp = VK_BLEND_OP_ADD, + .srcAlphaBlendFactor = VK_BLEND_FACTOR_ZERO, + .dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO, + .alphaBlendOp = VK_BLEND_OP_ADD, + .colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT | + VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT, + }; + + const VkPipelineColorBlendStateCreateInfo color_blend_ci{ + .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .logicOpEnable = VK_FALSE, + .logicOp = VK_LOGIC_OP_COPY, + .attachmentCount = 1, + .pAttachments = &color_blend_attachment, + .blendConstants = {0.0f, 0.0f, 0.0f, 0.0f}, + }; + + static constexpr std::array dynamic_states{ + VK_DYNAMIC_STATE_VIEWPORT, + VK_DYNAMIC_STATE_SCISSOR, + }; + const VkPipelineDynamicStateCreateInfo dynamic_state_ci{ + .sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .dynamicStateCount = static_cast<u32>(dynamic_states.size()), + .pDynamicStates = dynamic_states.data(), + }; + + const VkGraphicsPipelineCreateInfo pipeline_ci{ + .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .stageCount = static_cast<u32>(shader_stages.size()), + .pStages = shader_stages.data(), + .pVertexInputState = &vertex_input_ci, + .pInputAssemblyState = &input_assembly_ci, + .pTessellationState = nullptr, + .pViewportState = &viewport_state_ci, + .pRasterizationState = &rasterization_ci, + .pMultisampleState = &multisampling_ci, + .pDepthStencilState = nullptr, + .pColorBlendState = &color_blend_ci, + .pDynamicState = &dynamic_state_ci, + .layout = *pipeline_layout, + .renderPass = *renderpass, + .subpass = 0, + .basePipelineHandle = 0, + .basePipelineIndex = 0, + }; pipeline = device.GetLogical().CreateGraphicsPipeline(pipeline_ci); } void VKBlitScreen::CreateSampler() { - 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; + const VkSamplerCreateInfo ci{ + .sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .magFilter = VK_FILTER_LINEAR, + .minFilter = VK_FILTER_NEAREST, + .mipmapMode = VK_SAMPLER_MIPMAP_MODE_LINEAR, + .addressModeU = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER, + .addressModeV = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER, + .addressModeW = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER, + .mipLodBias = 0.0f, + .anisotropyEnable = VK_FALSE, + .maxAnisotropy = 0.0f, + .compareEnable = VK_FALSE, + .compareOp = VK_COMPARE_OP_NEVER, + .minLod = 0.0f, + .maxLod = 0.0f, + .borderColor = VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK, + .unnormalizedCoordinates = VK_FALSE, + }; sampler = device.GetLogical().CreateSampler(ci); } @@ -650,15 +690,16 @@ void VKBlitScreen::CreateFramebuffers() { const VkExtent2D size{swapchain.GetSize()}; framebuffers.resize(image_count); - 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; + VkFramebufferCreateInfo ci{ + .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .renderPass = *renderpass, + .attachmentCount = 1, + .width = size.width, + .height = size.height, + .layers = 1, + }; for (std::size_t i = 0; i < image_count; ++i) { const VkImageView image_view{swapchain.GetImageViewIndex(i)}; @@ -678,16 +719,17 @@ void VKBlitScreen::ReleaseRawImages() { } void VKBlitScreen::CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer) { - 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; + const VkBufferCreateInfo ci{ + .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .size = CalculateBufferSize(framebuffer), + .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, + .sharingMode = VK_SHARING_MODE_EXCLUSIVE, + .queueFamilyIndexCount = 0, + .pQueueFamilyIndices = nullptr, + }; buffer = device.GetLogical().CreateBuffer(ci); buffer_commit = memory_manager.Commit(buffer, true); @@ -697,24 +739,28 @@ void VKBlitScreen::CreateRawImages(const Tegra::FramebufferConfig& framebuffer) raw_images.resize(image_count); raw_buffer_commits.resize(image_count); - 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; + const VkImageCreateInfo ci{ + .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .imageType = VK_IMAGE_TYPE_2D, + .format = GetFormat(framebuffer), + .extent = + { + .width = framebuffer.width, + .height = framebuffer.height, + .depth = 1, + }, + .mipLevels = 1, + .arrayLayers = 1, + .samples = VK_SAMPLE_COUNT_1_BIT, + .tiling = VK_IMAGE_TILING_LINEAR, + .usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT, + .sharingMode = VK_SHARING_MODE_EXCLUSIVE, + .queueFamilyIndexCount = 0, + .pQueueFamilyIndices = nullptr, + .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED, + }; for (std::size_t i = 0; i < image_count; ++i) { raw_images[i] = std::make_unique<VKImage>(device, scheduler, ci, VK_IMAGE_ASPECT_COLOR_BIT); @@ -723,39 +769,43 @@ void VKBlitScreen::CreateRawImages(const Tegra::FramebufferConfig& framebuffer) } 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; + const VkDescriptorBufferInfo buffer_info{ + .buffer = *buffer, + .offset = offsetof(BufferData, uniform), + .range = sizeof(BufferData::uniform), + }; + + const VkWriteDescriptorSet ubo_write{ + .sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, + .pNext = nullptr, + .dstSet = descriptor_sets[image_index], + .dstBinding = 0, + .dstArrayElement = 0, + .descriptorCount = 1, + .descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, + .pImageInfo = nullptr, + .pBufferInfo = &buffer_info, + .pTexelBufferView = nullptr, + }; + + const VkDescriptorImageInfo image_info{ + .sampler = *sampler, + .imageView = image_view, + .imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL, + }; + + const VkWriteDescriptorSet sampler_write{ + .sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, + .pNext = nullptr, + .dstSet = descriptor_sets[image_index], + .dstBinding = 1, + .dstArrayElement = 0, + .descriptorCount = 1, + .descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, + .pImageInfo = &image_info, + .pBufferInfo = nullptr, + .pTexelBufferView = nullptr, + }; device.GetLogical().UpdateDescriptorSets(std::array{ubo_write, sampler_write}, {}); } diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp index fdaea4210..9226e591c 100644 --- a/src/video_core/renderer_vulkan/vk_device.cpp +++ b/src/video_core/renderer_vulkan/vk_device.cpp @@ -22,14 +22,21 @@ namespace { namespace Alternatives { -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{}}; +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 = { +constexpr std::array REQUIRED_EXTENSIONS{ VK_KHR_SWAPCHAIN_EXTENSION_NAME, VK_KHR_16BIT_STORAGE_EXTENSION_NAME, VK_KHR_8BIT_STORAGE_EXTENSION_NAME, @@ -169,97 +176,104 @@ bool VKDevice::Create() { const auto queue_cis = GetDeviceQueueCreateInfos(); const std::vector extensions = LoadExtensions(); - VkPhysicalDeviceFeatures2 features2; - features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; - features2.pNext = nullptr; + VkPhysicalDeviceFeatures2 features2{ + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2, + .pNext = nullptr, + }; const void* first_next = &features2; void** next = &features2.pNext; - auto& features = features2.features; - 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.depthBiasClamp = true; - features.fillModeNonSolid = false; - features.depthBounds = false; - features.wideLines = false; - features.largePoints = true; - features.alphaToOne = false; - features.multiViewport = 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.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; + features2.features = { + .robustBufferAccess = false, + .fullDrawIndexUint32 = false, + .imageCubeArray = false, + .independentBlend = true, + .geometryShader = true, + .tessellationShader = true, + .sampleRateShading = false, + .dualSrcBlend = false, + .logicOp = false, + .multiDrawIndirect = false, + .drawIndirectFirstInstance = false, + .depthClamp = true, + .depthBiasClamp = true, + .fillModeNonSolid = false, + .depthBounds = false, + .wideLines = false, + .largePoints = true, + .alphaToOne = false, + .multiViewport = true, + .samplerAnisotropy = true, + .textureCompressionETC2 = false, + .textureCompressionASTC_LDR = is_optimal_astc_supported, + .textureCompressionBC = false, + .occlusionQueryPrecise = true, + .pipelineStatisticsQuery = false, + .vertexPipelineStoresAndAtomics = true, + .fragmentStoresAndAtomics = true, + .shaderTessellationAndGeometryPointSize = false, + .shaderImageGatherExtended = true, + .shaderStorageImageExtendedFormats = false, + .shaderStorageImageMultisample = false, + .shaderStorageImageReadWithoutFormat = is_formatless_image_load_supported, + .shaderStorageImageWriteWithoutFormat = true, + .shaderUniformBufferArrayDynamicIndexing = false, + .shaderSampledImageArrayDynamicIndexing = false, + .shaderStorageBufferArrayDynamicIndexing = false, + .shaderStorageImageArrayDynamicIndexing = false, + .shaderClipDistance = false, + .shaderCullDistance = false, + .shaderFloat64 = false, + .shaderInt64 = false, + .shaderInt16 = false, + .shaderResourceResidency = false, + .shaderResourceMinLod = false, + .sparseBinding = false, + .sparseResidencyBuffer = false, + .sparseResidencyImage2D = false, + .sparseResidencyImage3D = false, + .sparseResidency2Samples = false, + .sparseResidency4Samples = false, + .sparseResidency8Samples = false, + .sparseResidency16Samples = false, + .sparseResidencyAliased = false, + .variableMultisampleRate = false, + .inheritedQueries = false, + }; + + VkPhysicalDevice16BitStorageFeaturesKHR bit16_storage{ + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES_KHR, + .pNext = nullptr, + .storageBuffer16BitAccess = false, + .uniformAndStorageBuffer16BitAccess = true, + .storagePushConstant16 = false, + .storageInputOutput16 = false, + }; SetNext(next, bit16_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; + VkPhysicalDevice8BitStorageFeaturesKHR bit8_storage{ + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_8BIT_STORAGE_FEATURES_KHR, + .pNext = nullptr, + .storageBuffer8BitAccess = false, + .uniformAndStorageBuffer8BitAccess = true, + .storagePushConstant8 = false, + }; SetNext(next, bit8_storage); - VkPhysicalDeviceHostQueryResetFeaturesEXT host_query_reset; - host_query_reset.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_HOST_QUERY_RESET_FEATURES_EXT; - host_query_reset.hostQueryReset = true; + VkPhysicalDeviceHostQueryResetFeaturesEXT host_query_reset{ + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_HOST_QUERY_RESET_FEATURES_EXT, + .hostQueryReset = true, + }; SetNext(next, host_query_reset); 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; + float16_int8 = { + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT16_INT8_FEATURES_KHR, + .pNext = nullptr, + .shaderFloat16 = true, + .shaderInt8 = false, + }; SetNext(next, float16_int8); } else { LOG_INFO(Render_Vulkan, "Device doesn't support float16 natively"); @@ -271,10 +285,11 @@ bool VKDevice::Create() { 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; + std430_layout = { + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_UNIFORM_BUFFER_STANDARD_LAYOUT_FEATURES_KHR, + .pNext = nullptr, + .uniformBufferStandardLayout = true, + }; SetNext(next, std430_layout); } else { LOG_INFO(Render_Vulkan, "Device doesn't support packed UBOs"); @@ -282,9 +297,11 @@ bool VKDevice::Create() { 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; + index_type_uint8 = { + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_INDEX_TYPE_UINT8_FEATURES_EXT, + .pNext = nullptr, + .indexTypeUint8 = true, + }; SetNext(next, index_type_uint8); } else { LOG_INFO(Render_Vulkan, "Device doesn't support uint8 indexes"); @@ -292,11 +309,12 @@ bool VKDevice::Create() { 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; + transform_feedback = { + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT, + .pNext = nullptr, + .transformFeedback = true, + .geometryStreams = true, + }; SetNext(next, transform_feedback); } else { LOG_INFO(Render_Vulkan, "Device doesn't support transform feedbacks"); @@ -304,10 +322,12 @@ bool VKDevice::Create() { VkPhysicalDeviceCustomBorderColorFeaturesEXT custom_border; if (ext_custom_border_color) { - custom_border.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CUSTOM_BORDER_COLOR_FEATURES_EXT; - custom_border.pNext = nullptr; - custom_border.customBorderColors = VK_TRUE; - custom_border.customBorderColorWithoutFormat = VK_TRUE; + custom_border = { + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CUSTOM_BORDER_COLOR_FEATURES_EXT, + .pNext = nullptr, + .customBorderColors = VK_TRUE, + .customBorderColorWithoutFormat = VK_TRUE, + }; SetNext(next, custom_border); } else { LOG_INFO(Render_Vulkan, "Device doesn't support custom border colors"); @@ -315,9 +335,11 @@ bool VKDevice::Create() { VkPhysicalDeviceExtendedDynamicStateFeaturesEXT dynamic_state; if (ext_extended_dynamic_state) { - dynamic_state.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_FEATURES_EXT; - dynamic_state.pNext = nullptr; - dynamic_state.extendedDynamicState = VK_TRUE; + dynamic_state = { + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_FEATURES_EXT, + .pNext = nullptr, + .extendedDynamicState = VK_TRUE, + }; SetNext(next, dynamic_state); } else { LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state"); @@ -331,11 +353,13 @@ bool VKDevice::Create() { if (nv_device_diagnostics_config) { nsight_aftermath_tracker.Initialize(); - diagnostics_nv.sType = VK_STRUCTURE_TYPE_DEVICE_DIAGNOSTICS_CONFIG_CREATE_INFO_NV; - diagnostics_nv.pNext = &features2; - diagnostics_nv.flags = VK_DEVICE_DIAGNOSTICS_CONFIG_ENABLE_SHADER_DEBUG_INFO_BIT_NV | - VK_DEVICE_DIAGNOSTICS_CONFIG_ENABLE_RESOURCE_TRACKING_BIT_NV | - VK_DEVICE_DIAGNOSTICS_CONFIG_ENABLE_AUTOMATIC_CHECKPOINTS_BIT_NV; + diagnostics_nv = { + .sType = VK_STRUCTURE_TYPE_DEVICE_DIAGNOSTICS_CONFIG_CREATE_INFO_NV, + .pNext = &features2, + .flags = VK_DEVICE_DIAGNOSTICS_CONFIG_ENABLE_SHADER_DEBUG_INFO_BIT_NV | + VK_DEVICE_DIAGNOSTICS_CONFIG_ENABLE_RESOURCE_TRACKING_BIT_NV | + VK_DEVICE_DIAGNOSTICS_CONFIG_ENABLE_AUTOMATIC_CHECKPOINTS_BIT_NV, + }; first_next = &diagnostics_nv; } @@ -704,13 +728,15 @@ void VKDevice::SetupFeatures() { } void VKDevice::CollectTelemetryParameters() { - VkPhysicalDeviceDriverPropertiesKHR driver; - driver.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES_KHR; - driver.pNext = nullptr; + VkPhysicalDeviceDriverPropertiesKHR driver{ + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES_KHR, + .pNext = nullptr, + }; - VkPhysicalDeviceProperties2KHR properties; - properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2_KHR; - properties.pNext = &driver; + VkPhysicalDeviceProperties2KHR properties{ + .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2_KHR, + .pNext = &driver, + }; physical.GetProperties2KHR(properties); driver_id = driver.driverID; @@ -719,24 +745,26 @@ void VKDevice::CollectTelemetryParameters() { const std::vector extensions = physical.EnumerateDeviceExtensionProperties(); reported_extensions.reserve(std::size(extensions)); for (const auto& extension : extensions) { - reported_extensions.push_back(extension.extensionName); + reported_extensions.emplace_back(extension.extensionName); } } std::vector<VkDeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const { static constexpr float QUEUE_PRIORITY = 1.0f; - std::unordered_set<u32> unique_queue_families = {graphics_family, present_family}; + std::unordered_set<u32> unique_queue_families{graphics_family, present_family}; std::vector<VkDeviceQueueCreateInfo> queue_cis; + queue_cis.reserve(unique_queue_families.size()); 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; + queue_cis.push_back({ + .sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .queueFamilyIndex = queue_family, + .queueCount = 1, + .pQueuePriorities = &QUEUE_PRIORITY, + }); } return queue_cis; diff --git a/src/video_core/texture_cache/surface_params.cpp b/src/video_core/texture_cache/surface_params.cpp index 0b2b2b8c4..921562c1f 100644 --- a/src/video_core/texture_cache/surface_params.cpp +++ b/src/video_core/texture_cache/surface_params.cpp @@ -343,8 +343,7 @@ std::size_t SurfaceParams::GetLayerSize(bool as_host_size, bool uncompressed) co size += GetInnerMipmapMemorySize(level, as_host_size, uncompressed); } if (is_tiled && is_layered) { - return Common::AlignBits(size, - Tegra::Texture::GetGOBSizeShift() + block_height + block_depth); + return Common::AlignBits(size, Tegra::Texture::GOB_SIZE_SHIFT + block_height + block_depth); } return size; } @@ -418,7 +417,7 @@ std::tuple<u32, u32, u32> SurfaceParams::GetBlockOffsetXYZ(u32 offset) const { const u32 block_size = GetBlockSize(); const u32 block_index = offset / block_size; const u32 gob_offset = offset % block_size; - const u32 gob_index = gob_offset / static_cast<u32>(Tegra::Texture::GetGOBSize()); + const u32 gob_index = gob_offset / static_cast<u32>(Tegra::Texture::GOB_SIZE); const u32 x_gob_pixels = 64U / GetBytesPerPixel(); const u32 x_block_pixels = x_gob_pixels << block_width; const u32 y_block_pixels = 8U << block_height; diff --git a/src/video_core/texture_cache/surface_params.h b/src/video_core/texture_cache/surface_params.h index 24957df8d..118aa689e 100644 --- a/src/video_core/texture_cache/surface_params.h +++ b/src/video_core/texture_cache/surface_params.h @@ -204,7 +204,7 @@ public: static std::size_t AlignLayered(const std::size_t out_size, const u32 block_height, const u32 block_depth) { return Common::AlignBits(out_size, - Tegra::Texture::GetGOBSizeShift() + block_height + block_depth); + Tegra::Texture::GOB_SIZE_SHIFT + block_height + block_depth); } /// Converts a width from a type of surface into another. This helps represent the diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 548e4c3fe..98beabef1 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -6,6 +6,7 @@ #include <cstring> #include "common/alignment.h" #include "common/assert.h" +#include "common/bit_util.h" #include "video_core/gpu.h" #include "video_core/textures/decoders.h" #include "video_core/textures/texture.h" @@ -37,20 +38,10 @@ struct alignas(64) SwizzleTable { std::array<std::array<u16, M>, N> values{}; }; -constexpr u32 gob_size_x_shift = 6; -constexpr u32 gob_size_y_shift = 3; -constexpr u32 gob_size_z_shift = 0; -constexpr u32 gob_size_shift = gob_size_x_shift + gob_size_y_shift + gob_size_z_shift; +constexpr u32 FAST_SWIZZLE_ALIGN = 16; -constexpr u32 gob_size_x = 1U << gob_size_x_shift; -constexpr u32 gob_size_y = 1U << gob_size_y_shift; -constexpr u32 gob_size_z = 1U << gob_size_z_shift; -constexpr u32 gob_size = 1U << gob_size_shift; - -constexpr u32 fast_swizzle_align = 16; - -constexpr auto legacy_swizzle_table = SwizzleTable<gob_size_y, gob_size_x, gob_size_z>(); -constexpr auto fast_swizzle_table = SwizzleTable<gob_size_y, 4, fast_swizzle_align>(); +constexpr auto LEGACY_SWIZZLE_TABLE = SwizzleTable<GOB_SIZE_X, GOB_SIZE_X, GOB_SIZE_Z>(); +constexpr auto FAST_SWIZZLE_TABLE = SwizzleTable<GOB_SIZE_Y, 4, FAST_SWIZZLE_ALIGN>(); /** * This function manages ALL the GOBs(Group of Bytes) Inside a single block. @@ -69,17 +60,17 @@ void PreciseProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, con u32 y_address = z_address; u32 pixel_base = layer_z * z + y_start * stride_x; for (u32 y = y_start; y < y_end; y++) { - const auto& table = legacy_swizzle_table[y % gob_size_y]; + const auto& table = LEGACY_SWIZZLE_TABLE[y % GOB_SIZE_Y]; for (u32 x = x_start; x < x_end; x++) { - const u32 swizzle_offset{y_address + table[x * bytes_per_pixel % gob_size_x]}; + const u32 swizzle_offset{y_address + table[x * bytes_per_pixel % GOB_SIZE_X]}; const u32 pixel_index{x * out_bytes_per_pixel + pixel_base}; data_ptrs[unswizzle] = swizzled_data + swizzle_offset; data_ptrs[!unswizzle] = unswizzled_data + pixel_index; std::memcpy(data_ptrs[0], data_ptrs[1], bytes_per_pixel); } pixel_base += stride_x; - if ((y + 1) % gob_size_y == 0) - y_address += gob_size; + if ((y + 1) % GOB_SIZE_Y == 0) + y_address += GOB_SIZE; } z_address += xy_block_size; } @@ -104,18 +95,18 @@ void FastProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, const u32 y_address = z_address; u32 pixel_base = layer_z * z + y_start * stride_x; for (u32 y = y_start; y < y_end; y++) { - const auto& table = fast_swizzle_table[y % gob_size_y]; - for (u32 xb = x_startb; xb < x_endb; xb += fast_swizzle_align) { - const u32 swizzle_offset{y_address + table[(xb / fast_swizzle_align) % 4]}; + const auto& table = FAST_SWIZZLE_TABLE[y % GOB_SIZE_Y]; + for (u32 xb = x_startb; xb < x_endb; xb += FAST_SWIZZLE_ALIGN) { + const u32 swizzle_offset{y_address + table[(xb / FAST_SWIZZLE_ALIGN) % 4]}; const u32 out_x = xb * out_bytes_per_pixel / bytes_per_pixel; const u32 pixel_index{out_x + pixel_base}; data_ptrs[unswizzle ? 1 : 0] = swizzled_data + swizzle_offset; data_ptrs[unswizzle ? 0 : 1] = unswizzled_data + pixel_index; - std::memcpy(data_ptrs[0], data_ptrs[1], fast_swizzle_align); + std::memcpy(data_ptrs[0], data_ptrs[1], FAST_SWIZZLE_ALIGN); } pixel_base += stride_x; - if ((y + 1) % gob_size_y == 0) - y_address += gob_size; + if ((y + 1) % GOB_SIZE_Y == 0) + y_address += GOB_SIZE; } z_address += xy_block_size; } @@ -138,9 +129,9 @@ void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool auto div_ceil = [](const u32 x, const u32 y) { return ((x + y - 1) / y); }; const u32 stride_x = width * out_bytes_per_pixel; const u32 layer_z = height * stride_x; - const u32 gob_elements_x = gob_size_x / bytes_per_pixel; - constexpr u32 gob_elements_y = gob_size_y; - constexpr u32 gob_elements_z = gob_size_z; + const u32 gob_elements_x = GOB_SIZE_X / bytes_per_pixel; + constexpr u32 gob_elements_y = GOB_SIZE_Y; + constexpr u32 gob_elements_z = GOB_SIZE_Z; const u32 block_x_elements = gob_elements_x; const u32 block_y_elements = gob_elements_y * block_height; const u32 block_z_elements = gob_elements_z * block_depth; @@ -148,7 +139,7 @@ void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool const u32 blocks_on_x = div_ceil(aligned_width, block_x_elements); const u32 blocks_on_y = div_ceil(height, block_y_elements); const u32 blocks_on_z = div_ceil(depth, block_z_elements); - const u32 xy_block_size = gob_size * block_height; + const u32 xy_block_size = GOB_SIZE * block_height; const u32 block_size = xy_block_size * block_depth; u32 tile_offset = 0; for (u32 zb = 0; zb < blocks_on_z; zb++) { @@ -182,7 +173,7 @@ void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel, bool unswizzle, u32 block_height, u32 block_depth, u32 width_spacing) { const u32 block_height_size{1U << block_height}; const u32 block_depth_size{1U << block_depth}; - if (bytes_per_pixel % 3 != 0 && (width * bytes_per_pixel) % fast_swizzle_align == 0) { + if (bytes_per_pixel % 3 != 0 && (width * bytes_per_pixel) % FAST_SWIZZLE_ALIGN == 0) { SwizzledData<true>(swizzled_data, unswizzled_data, unswizzle, width, height, depth, bytes_per_pixel, out_bytes_per_pixel, block_height_size, block_depth_size, width_spacing); @@ -259,25 +250,26 @@ std::vector<u8> UnswizzleTexture(u8* address, u32 tile_size_x, u32 tile_size_y, } void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width, - u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, + u32 bytes_per_pixel, u8* swizzled_data, const u8* unswizzled_data, u32 block_height_bit, u32 offset_x, u32 offset_y) { const u32 block_height = 1U << block_height_bit; - const u32 image_width_in_gobs{(swizzled_width * bytes_per_pixel + (gob_size_x - 1)) / - gob_size_x}; + const u32 image_width_in_gobs = + (swizzled_width * bytes_per_pixel + (GOB_SIZE_X - 1)) / GOB_SIZE_X; for (u32 line = 0; line < subrect_height; ++line) { const u32 dst_y = line + offset_y; const u32 gob_address_y = - (dst_y / (gob_size_y * block_height)) * gob_size * block_height * image_width_in_gobs + - ((dst_y % (gob_size_y * block_height)) / gob_size_y) * gob_size; - const auto& table = legacy_swizzle_table[dst_y % gob_size_y]; + (dst_y / (GOB_SIZE_Y * block_height)) * GOB_SIZE * block_height * image_width_in_gobs + + ((dst_y % (GOB_SIZE_Y * block_height)) / GOB_SIZE_Y) * GOB_SIZE; + const auto& table = LEGACY_SWIZZLE_TABLE[dst_y % GOB_SIZE_Y]; for (u32 x = 0; x < subrect_width; ++x) { const u32 dst_x = x + offset_x; const u32 gob_address = - gob_address_y + (dst_x * bytes_per_pixel / gob_size_x) * gob_size * block_height; - const u32 swizzled_offset = gob_address + table[(dst_x * bytes_per_pixel) % gob_size_x]; - u8* source_line = unswizzled_data + line * source_pitch + x * bytes_per_pixel; - u8* dest_addr = swizzled_data + swizzled_offset; + gob_address_y + (dst_x * bytes_per_pixel / GOB_SIZE_X) * GOB_SIZE * block_height; + const u32 swizzled_offset = gob_address + table[(dst_x * bytes_per_pixel) % GOB_SIZE_X]; + const u32 unswizzled_offset = line * source_pitch + x * bytes_per_pixel; + const u8* const source_line = unswizzled_data + unswizzled_offset; + u8* const dest_addr = swizzled_data + swizzled_offset; std::memcpy(dest_addr, source_line, bytes_per_pixel); } } @@ -289,14 +281,15 @@ void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 const u32 block_height = 1U << block_height_bit; for (u32 line = 0; line < subrect_height; ++line) { const u32 y2 = line + offset_y; - const u32 gob_address_y = (y2 / (gob_size_y * block_height)) * gob_size * block_height + - ((y2 % (gob_size_y * block_height)) / gob_size_y) * gob_size; - const auto& table = legacy_swizzle_table[y2 % gob_size_y]; + const u32 gob_address_y = (y2 / (GOB_SIZE_Y * block_height)) * GOB_SIZE * block_height + + ((y2 % (GOB_SIZE_Y * block_height)) / GOB_SIZE_Y) * GOB_SIZE; + const auto& table = LEGACY_SWIZZLE_TABLE[y2 % GOB_SIZE_Y]; for (u32 x = 0; x < subrect_width; ++x) { const u32 x2 = (x + offset_x) * bytes_per_pixel; - const u32 gob_address = gob_address_y + (x2 / gob_size_x) * gob_size * block_height; - const u32 swizzled_offset = gob_address + table[x2 % gob_size_x]; - u8* dest_line = unswizzled_data + line * dest_pitch + x * bytes_per_pixel; + const u32 gob_address = gob_address_y + (x2 / GOB_SIZE_X) * GOB_SIZE * block_height; + const u32 swizzled_offset = gob_address + table[x2 % GOB_SIZE_X]; + const u32 unswizzled_offset = line * dest_pitch + x * bytes_per_pixel; + u8* dest_line = unswizzled_data + unswizzled_offset; u8* source_addr = swizzled_data + swizzled_offset; std::memcpy(dest_line, source_addr, bytes_per_pixel); @@ -304,21 +297,48 @@ void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 } } +void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height, + u32 bytes_per_pixel, u32 block_height, u32 block_depth, u32 origin_x, + u32 origin_y, u8* output, const u8* input) { + UNIMPLEMENTED_IF(origin_x > 0); + UNIMPLEMENTED_IF(origin_y > 0); + + const u32 stride = width * bytes_per_pixel; + const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) / GOB_SIZE_X; + const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth); + + const u32 block_height_mask = (1U << block_height) - 1; + const u32 x_shift = Common::CountTrailingZeroes32(GOB_SIZE << (block_height + block_depth)); + + for (u32 line = 0; line < line_count; ++line) { + const auto& table = LEGACY_SWIZZLE_TABLE[line % GOB_SIZE_Y]; + const u32 block_y = line / GOB_SIZE_Y; + const u32 dst_offset_y = + (block_y >> block_height) * block_size + (block_y & block_height_mask) * GOB_SIZE; + for (u32 x = 0; x < line_length_in; ++x) { + const u32 dst_offset = + ((x / GOB_SIZE_X) << x_shift) + dst_offset_y + table[x % GOB_SIZE_X]; + const u32 src_offset = x * bytes_per_pixel + line * pitch; + std::memcpy(output + dst_offset, input + src_offset, bytes_per_pixel); + } + } +} + void SwizzleKepler(const u32 width, const u32 height, const u32 dst_x, const u32 dst_y, const u32 block_height_bit, const std::size_t copy_size, const u8* source_data, u8* swizzle_data) { const u32 block_height = 1U << block_height_bit; - const u32 image_width_in_gobs{(width + gob_size_x - 1) / gob_size_x}; + const u32 image_width_in_gobs{(width + GOB_SIZE_X - 1) / GOB_SIZE_X}; std::size_t count = 0; for (std::size_t y = dst_y; y < height && count < copy_size; ++y) { const std::size_t gob_address_y = - (y / (gob_size_y * block_height)) * gob_size * block_height * image_width_in_gobs + - ((y % (gob_size_y * block_height)) / gob_size_y) * gob_size; - const auto& table = legacy_swizzle_table[y % gob_size_y]; + (y / (GOB_SIZE_Y * block_height)) * GOB_SIZE * block_height * image_width_in_gobs + + ((y % (GOB_SIZE_Y * block_height)) / GOB_SIZE_Y) * GOB_SIZE; + const auto& table = LEGACY_SWIZZLE_TABLE[y % GOB_SIZE_Y]; for (std::size_t x = dst_x; x < width && count < copy_size; ++x) { const std::size_t gob_address = - gob_address_y + (x / gob_size_x) * gob_size * block_height; - const std::size_t swizzled_offset = gob_address + table[x % gob_size_x]; + gob_address_y + (x / GOB_SIZE_X) * GOB_SIZE * block_height; + const std::size_t swizzled_offset = gob_address + table[x % GOB_SIZE_X]; const u8* source_line = source_data + count; u8* dest_addr = swizzle_data + swizzled_offset; count++; @@ -373,9 +393,9 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth) { if (tiled) { - const u32 aligned_width = Common::AlignBits(width * bytes_per_pixel, gob_size_x_shift); - const u32 aligned_height = Common::AlignBits(height, gob_size_y_shift + block_height); - const u32 aligned_depth = Common::AlignBits(depth, gob_size_z_shift + block_depth); + const u32 aligned_width = Common::AlignBits(width * bytes_per_pixel, GOB_SIZE_X_SHIFT); + const u32 aligned_height = Common::AlignBits(height, GOB_SIZE_Y_SHIFT + block_height); + const u32 aligned_depth = Common::AlignBits(depth, GOB_SIZE_Z_SHIFT + block_depth); return aligned_width * aligned_height * aligned_depth; } else { return width * height * depth * bytes_per_pixel; @@ -386,14 +406,14 @@ u64 GetGOBOffset(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height, u32 bytes_per_pixel) { auto div_ceil = [](const u32 x, const u32 y) { return ((x + y - 1) / y); }; const u32 gobs_in_block = 1 << block_height; - const u32 y_blocks = gob_size_y << block_height; - const u32 x_per_gob = gob_size_x / bytes_per_pixel; + const u32 y_blocks = GOB_SIZE_Y << block_height; + const u32 x_per_gob = GOB_SIZE_X / bytes_per_pixel; const u32 x_blocks = div_ceil(width, x_per_gob); - const u32 block_size = gob_size * gobs_in_block; + const u32 block_size = GOB_SIZE * gobs_in_block; const u32 stride = block_size * x_blocks; const u32 base = (dst_y / y_blocks) * stride + (dst_x / x_per_gob) * block_size; const u32 relative_y = dst_y % y_blocks; - return base + (relative_y / gob_size_y) * gob_size; + return base + (relative_y / GOB_SIZE_Y) * GOB_SIZE; } } // namespace Tegra::Texture diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h index 06f3ebf87..232b696b3 100644 --- a/src/video_core/textures/decoders.h +++ b/src/video_core/textures/decoders.h @@ -10,15 +10,15 @@ namespace Tegra::Texture { -// GOBSize constant. Calculated by 64 bytes in x multiplied by 8 y coords, represents -// an small rect of (64/bytes_per_pixel)X8. -inline std::size_t GetGOBSize() { - return 512; -} +constexpr u32 GOB_SIZE_X = 64; +constexpr u32 GOB_SIZE_Y = 8; +constexpr u32 GOB_SIZE_Z = 1; +constexpr u32 GOB_SIZE = GOB_SIZE_X * GOB_SIZE_Y * GOB_SIZE_Z; -inline std::size_t GetGOBSizeShift() { - return 9; -} +constexpr std::size_t GOB_SIZE_X_SHIFT = 6; +constexpr std::size_t GOB_SIZE_Y_SHIFT = 3; +constexpr std::size_t GOB_SIZE_Z_SHIFT = 0; +constexpr std::size_t GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT + GOB_SIZE_Z_SHIFT; /// Unswizzles a swizzled texture without changing its format. void UnswizzleTexture(u8* unswizzled_data, u8* address, u32 tile_size_x, u32 tile_size_y, @@ -48,14 +48,32 @@ std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height /// Copies an untiled subrectangle into a tiled surface. void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width, - u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height, - u32 offset_x, u32 offset_y); + u32 bytes_per_pixel, u8* swizzled_data, const u8* unswizzled_data, + u32 block_height_bit, u32 offset_x, u32 offset_y); /// Copies a tiled subrectangle into a linear surface. void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 swizzled_width, u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height, u32 offset_x, u32 offset_y); +/// @brief Swizzles a 2D array of pixels into a 3D texture +/// @param line_length_in Number of pixels per line +/// @param line_count Number of lines +/// @param pitch Number of bytes per line +/// @param width Width of the swizzled texture +/// @param height Height of the swizzled texture +/// @param bytes_per_pixel Number of bytes used per pixel +/// @param block_height Block height shift +/// @param block_depth Block depth shift +/// @param origin_x Column offset in pixels of the swizzled texture +/// @param origin_y Row offset in pixels of the swizzled texture +/// @param output Pointer to the pixels of the swizzled texture +/// @param input Pointer to the 2D array of pixels used as input +/// @pre input and output points to an array large enough to hold the number of bytes used +void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height, + u32 bytes_per_pixel, u32 block_height, u32 block_depth, u32 origin_x, + u32 origin_y, u8* output, const u8* input); + void SwizzleKepler(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height, std::size_t copy_size, const u8* source_data, u8* swizzle_data); |