diff options
Diffstat (limited to 'src/video_core')
45 files changed, 1473 insertions, 1742 deletions
diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index dd7ce8c99..b5dc68902 100644 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -524,11 +524,8 @@ private: void MarkRegionAsWritten(VAddr start, VAddr end) { const u64 page_end = end >> WRITE_PAGE_BIT; for (u64 page_start = start >> WRITE_PAGE_BIT; page_start <= page_end; ++page_start) { - auto it = written_pages.find(page_start); - if (it != written_pages.end()) { - it->second = it->second + 1; - } else { - written_pages.insert_or_assign(page_start, 1); + if (const auto [it, inserted] = written_pages.emplace(page_start, 1); !inserted) { + ++it->second; } } } @@ -539,7 +536,7 @@ private: auto it = written_pages.find(page_start); if (it != written_pages.end()) { if (it->second > 1) { - it->second = it->second - 1; + --it->second; } else { written_pages.erase(it); } diff --git a/src/video_core/compatible_formats.cpp b/src/video_core/compatible_formats.cpp index 6c426b035..b06c32c84 100644 --- a/src/video_core/compatible_formats.cpp +++ b/src/video_core/compatible_formats.cpp @@ -17,101 +17,94 @@ namespace { // https://www.khronos.org/registry/OpenGL/extensions/ARB/ARB_texture_view.txt constexpr std::array VIEW_CLASS_128_BITS = { - PixelFormat::RGBA32F, - PixelFormat::RGBA32UI, + PixelFormat::R32G32B32A32_FLOAT, + PixelFormat::R32G32B32A32_UINT, + PixelFormat::R32G32B32A32_SINT, }; -// Missing formats: -// PixelFormat::RGBA32I constexpr std::array VIEW_CLASS_96_BITS = { - PixelFormat::RGB32F, + PixelFormat::R32G32B32_FLOAT, }; // Missing formats: // PixelFormat::RGB32UI, // PixelFormat::RGB32I, constexpr std::array VIEW_CLASS_64_BITS = { - PixelFormat::RGBA16F, PixelFormat::RG32F, PixelFormat::RGBA16UI, PixelFormat::RG32UI, - PixelFormat::RGBA16U, PixelFormat::RGBA16F, PixelFormat::RGBA16S, + PixelFormat::R32G32_FLOAT, PixelFormat::R32G32_UINT, + PixelFormat::R32G32_SINT, PixelFormat::R16G16B16A16_FLOAT, + PixelFormat::R16G16B16A16_UNORM, PixelFormat::R16G16B16A16_SNORM, + PixelFormat::R16G16B16A16_UINT, PixelFormat::R16G16B16A16_SINT, }; -// Missing formats: -// PixelFormat::RGBA16I -// PixelFormat::RG32I // TODO: How should we handle 48 bits? constexpr std::array VIEW_CLASS_32_BITS = { - PixelFormat::RG16F, PixelFormat::R11FG11FB10F, PixelFormat::R32F, - PixelFormat::A2B10G10R10U, PixelFormat::RG16UI, PixelFormat::R32UI, - PixelFormat::RG16I, PixelFormat::R32I, PixelFormat::ABGR8U, - PixelFormat::RG16, PixelFormat::ABGR8S, PixelFormat::RG16S, - PixelFormat::RGBA8_SRGB, PixelFormat::E5B9G9R9F, PixelFormat::BGRA8, - PixelFormat::BGRA8_SRGB, + PixelFormat::R16G16_FLOAT, PixelFormat::B10G11R11_FLOAT, PixelFormat::R32_FLOAT, + PixelFormat::A2B10G10R10_UNORM, PixelFormat::R16G16_UINT, PixelFormat::R32_UINT, + PixelFormat::R16G16_SINT, PixelFormat::R32_SINT, PixelFormat::A8B8G8R8_UNORM, + PixelFormat::R16G16_UNORM, PixelFormat::A8B8G8R8_SNORM, PixelFormat::R16G16_SNORM, + PixelFormat::A8B8G8R8_SRGB, PixelFormat::E5B9G9R9_FLOAT, PixelFormat::B8G8R8A8_UNORM, + PixelFormat::B8G8R8A8_SRGB, PixelFormat::A8B8G8R8_UINT, PixelFormat::A8B8G8R8_SINT, + PixelFormat::A2B10G10R10_UINT, }; -// Missing formats: -// PixelFormat::RGBA8UI -// PixelFormat::RGBA8I -// PixelFormat::RGB10_A2_UI // TODO: How should we handle 24 bits? constexpr std::array VIEW_CLASS_16_BITS = { - PixelFormat::R16F, PixelFormat::RG8UI, PixelFormat::R16UI, PixelFormat::R16I, - PixelFormat::RG8U, PixelFormat::R16U, PixelFormat::RG8S, PixelFormat::R16S, + PixelFormat::R16_FLOAT, PixelFormat::R8G8_UINT, PixelFormat::R16_UINT, + PixelFormat::R16_SINT, PixelFormat::R8G8_UNORM, PixelFormat::R16_UNORM, + PixelFormat::R8G8_SNORM, PixelFormat::R16_SNORM, PixelFormat::R8G8_SINT, }; -// Missing formats: -// PixelFormat::RG8I constexpr std::array VIEW_CLASS_8_BITS = { - PixelFormat::R8UI, - PixelFormat::R8U, + PixelFormat::R8_UINT, + PixelFormat::R8_UNORM, + PixelFormat::R8_SINT, + PixelFormat::R8_SNORM, }; -// Missing formats: -// PixelFormat::R8I -// PixelFormat::R8S constexpr std::array VIEW_CLASS_RGTC1_RED = { - PixelFormat::DXN1, + PixelFormat::BC4_UNORM, + PixelFormat::BC4_SNORM, }; -// Missing formats: -// COMPRESSED_SIGNED_RED_RGTC1 constexpr std::array VIEW_CLASS_RGTC2_RG = { - PixelFormat::DXN2UNORM, - PixelFormat::DXN2SNORM, + PixelFormat::BC5_UNORM, + PixelFormat::BC5_SNORM, }; constexpr std::array VIEW_CLASS_BPTC_UNORM = { - PixelFormat::BC7U, - PixelFormat::BC7U_SRGB, + PixelFormat::BC7_UNORM, + PixelFormat::BC7_SRGB, }; constexpr std::array VIEW_CLASS_BPTC_FLOAT = { - PixelFormat::BC6H_SF16, - PixelFormat::BC6H_UF16, + PixelFormat::BC6H_SFLOAT, + PixelFormat::BC6H_UFLOAT, }; // Compatibility table taken from Table 4.X.1 in: // https://www.khronos.org/registry/OpenGL/extensions/ARB/ARB_copy_image.txt constexpr std::array COPY_CLASS_128_BITS = { - PixelFormat::RGBA32UI, PixelFormat::RGBA32F, PixelFormat::DXT23, - PixelFormat::DXT23_SRGB, PixelFormat::DXT45, PixelFormat::DXT45_SRGB, - PixelFormat::DXN2SNORM, PixelFormat::BC7U, PixelFormat::BC7U_SRGB, - PixelFormat::BC6H_SF16, PixelFormat::BC6H_UF16, + PixelFormat::R32G32B32A32_UINT, PixelFormat::R32G32B32A32_FLOAT, PixelFormat::R32G32B32A32_SINT, + PixelFormat::BC2_UNORM, PixelFormat::BC2_SRGB, PixelFormat::BC3_UNORM, + PixelFormat::BC3_SRGB, PixelFormat::BC5_UNORM, PixelFormat::BC5_SNORM, + PixelFormat::BC7_UNORM, PixelFormat::BC7_SRGB, PixelFormat::BC6H_SFLOAT, + PixelFormat::BC6H_UFLOAT, }; // Missing formats: // PixelFormat::RGBA32I // COMPRESSED_RG_RGTC2 constexpr std::array COPY_CLASS_64_BITS = { - PixelFormat::RGBA16F, PixelFormat::RG32F, PixelFormat::RGBA16UI, PixelFormat::RG32UI, - PixelFormat::RGBA16U, PixelFormat::RGBA16S, PixelFormat::DXT1_SRGB, PixelFormat::DXT1, - + PixelFormat::R16G16B16A16_FLOAT, PixelFormat::R16G16B16A16_UINT, + PixelFormat::R16G16B16A16_UNORM, PixelFormat::R16G16B16A16_SNORM, + PixelFormat::R16G16B16A16_SINT, PixelFormat::R32G32_UINT, + PixelFormat::R32G32_FLOAT, PixelFormat::R32G32_SINT, + PixelFormat::BC1_RGBA_UNORM, PixelFormat::BC1_RGBA_SRGB, }; // Missing formats: -// PixelFormat::RGBA16I -// PixelFormat::RG32I, // COMPRESSED_RGB_S3TC_DXT1_EXT // COMPRESSED_SRGB_S3TC_DXT1_EXT // COMPRESSED_RGBA_S3TC_DXT1_EXT diff --git a/src/video_core/compatible_formats.h b/src/video_core/compatible_formats.h index d1082566d..51766349b 100644 --- a/src/video_core/compatible_formats.h +++ b/src/video_core/compatible_formats.h @@ -2,6 +2,8 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#pragma once + #include <array> #include <bitset> #include <cstddef> diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 8e19c3373..512578c8b 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -81,7 +81,7 @@ void GPU::WaitFence(u32 syncpoint_id, u32 value) { } MICROPROFILE_SCOPE(GPU_wait); std::unique_lock lock{sync_mutex}; - sync_cv.wait(lock, [=]() { return syncpoints[syncpoint_id].load() >= value; }); + sync_cv.wait(lock, [=, this] { return syncpoints[syncpoint_id].load() >= value; }); } void GPU::IncrementSyncPoint(const u32 syncpoint_id) { diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h index 8d04d9fd9..ebfc7b0c7 100644 --- a/src/video_core/gpu.h +++ b/src/video_core/gpu.h @@ -40,53 +40,61 @@ namespace Tegra { enum class RenderTargetFormat : u32 { NONE = 0x0, - RGBA32_FLOAT = 0xC0, - RGBA32_UINT = 0xC2, - RGBA16_UNORM = 0xC6, - RGBA16_SNORM = 0xC7, - RGBA16_UINT = 0xC9, - RGBA16_FLOAT = 0xCA, - RG32_FLOAT = 0xCB, - RG32_UINT = 0xCD, - RGBX16_FLOAT = 0xCE, - BGRA8_UNORM = 0xCF, - BGRA8_SRGB = 0xD0, - RGB10_A2_UNORM = 0xD1, - RGBA8_UNORM = 0xD5, - RGBA8_SRGB = 0xD6, - RGBA8_SNORM = 0xD7, - RGBA8_UINT = 0xD9, - RG16_UNORM = 0xDA, - RG16_SNORM = 0xDB, - RG16_SINT = 0xDC, - RG16_UINT = 0xDD, - RG16_FLOAT = 0xDE, - R11G11B10_FLOAT = 0xE0, + R32B32G32A32_FLOAT = 0xC0, + R32G32B32A32_SINT = 0xC1, + R32G32B32A32_UINT = 0xC2, + R16G16B16A16_UNORM = 0xC6, + R16G16B16A16_SNORM = 0xC7, + R16G16B16A16_SINT = 0xC8, + R16G16B16A16_UINT = 0xC9, + R16G16B16A16_FLOAT = 0xCA, + R32G32_FLOAT = 0xCB, + R32G32_SINT = 0xCC, + R32G32_UINT = 0xCD, + R16G16B16X16_FLOAT = 0xCE, + B8G8R8A8_UNORM = 0xCF, + B8G8R8A8_SRGB = 0xD0, + A2B10G10R10_UNORM = 0xD1, + A2B10G10R10_UINT = 0xD2, + A8B8G8R8_UNORM = 0xD5, + A8B8G8R8_SRGB = 0xD6, + A8B8G8R8_SNORM = 0xD7, + A8B8G8R8_SINT = 0xD8, + A8B8G8R8_UINT = 0xD9, + R16G16_UNORM = 0xDA, + R16G16_SNORM = 0xDB, + R16G16_SINT = 0xDC, + R16G16_UINT = 0xDD, + R16G16_FLOAT = 0xDE, + B10G11R11_FLOAT = 0xE0, R32_SINT = 0xE3, R32_UINT = 0xE4, R32_FLOAT = 0xE5, - B5G6R5_UNORM = 0xE8, - BGR5A1_UNORM = 0xE9, - RG8_UNORM = 0xEA, - RG8_SNORM = 0xEB, - RG8_UINT = 0xED, + R5G6B5_UNORM = 0xE8, + A1R5G5B5_UNORM = 0xE9, + R8G8_UNORM = 0xEA, + R8G8_SNORM = 0xEB, + R8G8_SINT = 0xEC, + R8G8_UINT = 0xED, R16_UNORM = 0xEE, R16_SNORM = 0xEF, R16_SINT = 0xF0, R16_UINT = 0xF1, R16_FLOAT = 0xF2, R8_UNORM = 0xF3, + R8_SNORM = 0xF4, + R8_SINT = 0xF5, R8_UINT = 0xF6, }; enum class DepthFormat : u32 { - Z32_FLOAT = 0xA, - Z16_UNORM = 0x13, - S8_Z24_UNORM = 0x14, - Z24_X8_UNORM = 0x15, - Z24_S8_UNORM = 0x16, - Z24_C8_UNORM = 0x18, - Z32_S8_X24_FLOAT = 0x19, + D32_FLOAT = 0xA, + D16_UNORM = 0x13, + S8_UINT_Z24_UNORM = 0x14, + D24X8_UNORM = 0x15, + D24S8_UNORM = 0x16, + D24C8_UNORM = 0x18, + D32_FLOAT_S8X24_UINT = 0x19, }; struct CommandListHeader; @@ -97,9 +105,9 @@ class DebugContext; */ struct FramebufferConfig { enum class PixelFormat : u32 { - ABGR8 = 1, - RGB565 = 4, - BGRA8 = 5, + A8B8G8R8_UNORM = 1, + RGB565_UNORM = 4, + B8G8R8A8_UNORM = 5, }; VAddr address; @@ -244,7 +252,7 @@ public: const Tegra::DmaPusher& DmaPusher() const; struct Regs { - static constexpr size_t NUM_REGS = 0x100; + static constexpr size_t NUM_REGS = 0x40; union { struct { @@ -263,7 +271,7 @@ public: u32 semaphore_trigger; INSERT_UNION_PADDING_WORDS(0xC); - // The puser and the puller share the reference counter, the pusher only has read + // The pusher and the puller share the reference counter, the pusher only has read // access u32 reference_count; INSERT_UNION_PADDING_WORDS(0x5); diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index 738c6f0c1..bf761abf2 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp @@ -44,9 +44,9 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer, dma_pusher.DispatchCalls(); } else if (const auto data = std::get_if<SwapBuffersCommand>(&next.data)) { renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr); - } else if (const auto data = std::get_if<OnCommandListEndCommand>(&next.data)) { + } else if (std::holds_alternative<OnCommandListEndCommand>(next.data)) { renderer.Rasterizer().ReleaseFences(); - } else if (const auto data = std::get_if<GPUTickCommand>(&next.data)) { + } else if (std::holds_alternative<GPUTickCommand>(next.data)) { system.GPU().TickWork(); } else if (const auto data = std::get_if<FlushRegionCommand>(&next.data)) { renderer.Rasterizer().FlushRegion(data->addr, data->size); diff --git a/src/video_core/macro/macro_jit_x64.cpp b/src/video_core/macro/macro_jit_x64.cpp index 07292702f..c1b9e4ad9 100644 --- a/src/video_core/macro/macro_jit_x64.cpp +++ b/src/video_core/macro/macro_jit_x64.cpp @@ -419,7 +419,6 @@ void Tegra::MacroJITx64Impl::Optimizer_ScanFlags() { void MacroJITx64Impl::Compile() { MICROPROFILE_SCOPE(MacroJitCompile); - bool keep_executing = true; labels.fill(Xbyak::Label()); Common::X64::ABI_PushRegistersAndAdjustStack(*this, Common::X64::ABI_ALL_CALLEE_SAVED, 8); diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index ff5505d12..844164645 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -4,7 +4,6 @@ #include "common/alignment.h" #include "common/assert.h" -#include "common/logging/log.h" #include "core/core.h" #include "core/hle/kernel/memory/page_table.h" #include "core/hle/kernel/process.h" @@ -16,121 +15,137 @@ namespace Tegra { MemoryManager::MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer) - : rasterizer{rasterizer}, system{system} { - page_table.Resize(address_space_width, page_bits, false); - - // Initialize the map with a single free region covering the entire managed space. - VirtualMemoryArea initial_vma; - initial_vma.size = address_space_end; - vma_map.emplace(initial_vma.base, initial_vma); - - UpdatePageTableForVMA(initial_vma); -} + : system{system}, rasterizer{rasterizer}, page_table(page_table_size) {} MemoryManager::~MemoryManager() = default; -GPUVAddr MemoryManager::AllocateSpace(u64 size, u64 align) { - const u64 aligned_size{Common::AlignUp(size, page_size)}; - const GPUVAddr gpu_addr{FindFreeRegion(address_space_base, aligned_size)}; - - AllocateMemory(gpu_addr, 0, aligned_size); - +GPUVAddr MemoryManager::UpdateRange(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size) { + u64 remaining_size{size}; + for (u64 offset{}; offset < size; offset += page_size) { + if (remaining_size < page_size) { + SetPageEntry(gpu_addr + offset, page_entry + offset, remaining_size); + } else { + SetPageEntry(gpu_addr + offset, page_entry + offset); + } + remaining_size -= page_size; + } return gpu_addr; } -GPUVAddr MemoryManager::AllocateSpace(GPUVAddr gpu_addr, u64 size, u64 align) { - const u64 aligned_size{Common::AlignUp(size, page_size)}; - - AllocateMemory(gpu_addr, 0, aligned_size); +GPUVAddr MemoryManager::Map(VAddr cpu_addr, GPUVAddr gpu_addr, std::size_t size) { + return UpdateRange(gpu_addr, cpu_addr, size); +} - return gpu_addr; +GPUVAddr MemoryManager::MapAllocate(VAddr cpu_addr, std::size_t size, std::size_t align) { + return Map(cpu_addr, *FindFreeRange(size, align), size); } -GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, u64 size) { - const u64 aligned_size{Common::AlignUp(size, page_size)}; - const GPUVAddr gpu_addr{FindFreeRegion(address_space_base, aligned_size)}; +void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) { + if (!size) { + return; + } - MapBackingMemory(gpu_addr, system.Memory().GetPointer(cpu_addr), aligned_size, cpu_addr); - ASSERT( - system.CurrentProcess()->PageTable().LockForDeviceAddressSpace(cpu_addr, size).IsSuccess()); + // Flush and invalidate through the GPU interface, to be asynchronous if possible. + system.GPU().FlushAndInvalidateRegion(*GpuToCpuAddress(gpu_addr), size); - return gpu_addr; + UpdateRange(gpu_addr, PageEntry::State::Unmapped, size); } -GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size) { - ASSERT((gpu_addr & page_mask) == 0); +std::optional<GPUVAddr> MemoryManager::AllocateFixed(GPUVAddr gpu_addr, std::size_t size) { + for (u64 offset{}; offset < size; offset += page_size) { + if (!GetPageEntry(gpu_addr + offset).IsUnmapped()) { + return {}; + } + } - const u64 aligned_size{Common::AlignUp(size, page_size)}; + return UpdateRange(gpu_addr, PageEntry::State::Allocated, size); +} - MapBackingMemory(gpu_addr, system.Memory().GetPointer(cpu_addr), aligned_size, cpu_addr); - ASSERT( - system.CurrentProcess()->PageTable().LockForDeviceAddressSpace(cpu_addr, size).IsSuccess()); - return gpu_addr; +GPUVAddr MemoryManager::Allocate(std::size_t size, std::size_t align) { + return *AllocateFixed(*FindFreeRange(size, align), size); } -GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) { - ASSERT((gpu_addr & page_mask) == 0); +void MemoryManager::TryLockPage(PageEntry page_entry, std::size_t size) { + if (!page_entry.IsValid()) { + return; + } - const u64 aligned_size{Common::AlignUp(size, page_size)}; - const auto cpu_addr = GpuToCpuAddress(gpu_addr); - ASSERT(cpu_addr); + ASSERT(system.CurrentProcess() + ->PageTable() + .LockForDeviceAddressSpace(page_entry.ToAddress(), size) + .IsSuccess()); +} - // Flush and invalidate through the GPU interface, to be asynchronous if possible. - system.GPU().FlushAndInvalidateRegion(*cpu_addr, aligned_size); +void MemoryManager::TryUnlockPage(PageEntry page_entry, std::size_t size) { + if (!page_entry.IsValid()) { + return; + } - UnmapRange(gpu_addr, aligned_size); ASSERT(system.CurrentProcess() ->PageTable() - .UnlockForDeviceAddressSpace(cpu_addr.value(), size) + .UnlockForDeviceAddressSpace(page_entry.ToAddress(), size) .IsSuccess()); - - return gpu_addr; } -GPUVAddr MemoryManager::FindFreeRegion(GPUVAddr region_start, u64 size) const { - // Find the first Free VMA. - const VMAHandle vma_handle{ - std::find_if(vma_map.begin(), vma_map.end(), [region_start, size](const auto& vma) { - if (vma.second.type != VirtualMemoryArea::Type::Unmapped) { - return false; - } +PageEntry MemoryManager::GetPageEntry(GPUVAddr gpu_addr) const { + return page_table[PageEntryIndex(gpu_addr)]; +} - const VAddr vma_end{vma.second.base + vma.second.size}; - return vma_end > region_start && vma_end >= region_start + size; - })}; +void MemoryManager::SetPageEntry(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size) { + // TODO(bunnei): We should lock/unlock device regions. This currently causes issues due to + // improper tracking, but should be fixed in the future. - if (vma_handle == vma_map.end()) { - return {}; - } + //// Unlock the old page + // TryUnlockPage(page_table[PageEntryIndex(gpu_addr)], size); - return std::max(region_start, vma_handle->second.base); -} + //// Lock the new page + // TryLockPage(page_entry, size); -bool MemoryManager::IsAddressValid(GPUVAddr addr) const { - return (addr >> page_bits) < page_table.pointers.size(); + page_table[PageEntryIndex(gpu_addr)] = page_entry; } -std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr addr) const { - if (!IsAddressValid(addr)) { - return {}; +std::optional<GPUVAddr> MemoryManager::FindFreeRange(std::size_t size, std::size_t align) const { + if (!align) { + align = page_size; + } else { + align = Common::AlignUp(align, page_size); } - const VAddr cpu_addr{page_table.backing_addr[addr >> page_bits]}; - if (cpu_addr) { - return cpu_addr + (addr & page_mask); + u64 available_size{}; + GPUVAddr gpu_addr{address_space_start}; + while (gpu_addr + available_size < address_space_size) { + if (GetPageEntry(gpu_addr + available_size).IsUnmapped()) { + available_size += page_size; + + if (available_size >= size) { + return gpu_addr; + } + } else { + gpu_addr += available_size + page_size; + available_size = 0; + + const auto remainder{gpu_addr % align}; + if (remainder) { + gpu_addr = (gpu_addr - remainder) + align; + } + } } return {}; } -template <typename T> -T MemoryManager::Read(GPUVAddr addr) const { - if (!IsAddressValid(addr)) { +std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) const { + const auto page_entry{GetPageEntry(gpu_addr)}; + if (!page_entry.IsValid()) { return {}; } - const u8* page_pointer{GetPointer(addr)}; - if (page_pointer) { + return page_entry.ToAddress() + (gpu_addr & page_mask); +} + +template <typename T> +T MemoryManager::Read(GPUVAddr addr) const { + if (auto page_pointer{GetPointer(addr)}; page_pointer) { // NOTE: Avoid adding any extra logic to this fast-path block T value; std::memcpy(&value, page_pointer, sizeof(T)); @@ -144,12 +159,7 @@ T MemoryManager::Read(GPUVAddr addr) const { template <typename T> void MemoryManager::Write(GPUVAddr addr, T data) { - if (!IsAddressValid(addr)) { - return; - } - - u8* page_pointer{GetPointer(addr)}; - if (page_pointer) { + if (auto page_pointer{GetPointer(addr)}; page_pointer) { // NOTE: Avoid adding any extra logic to this fast-path block std::memcpy(page_pointer, &data, sizeof(T)); return; @@ -167,66 +177,49 @@ template void MemoryManager::Write<u16>(GPUVAddr addr, u16 data); template void MemoryManager::Write<u32>(GPUVAddr addr, u32 data); template void MemoryManager::Write<u64>(GPUVAddr addr, u64 data); -u8* MemoryManager::GetPointer(GPUVAddr addr) { - if (!IsAddressValid(addr)) { +u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) { + if (!GetPageEntry(gpu_addr).IsValid()) { return {}; } - 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)); + const auto address{GpuToCpuAddress(gpu_addr)}; + if (!address) { + return {}; } - LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr); - return {}; + return system.Memory().GetPointer(*address); } -const u8* MemoryManager::GetPointer(GPUVAddr addr) const { - if (!IsAddressValid(addr)) { +const u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) const { + if (!GetPageEntry(gpu_addr).IsValid()) { return {}; } - 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)); + const auto address{GpuToCpuAddress(gpu_addr)}; + if (!address) { + return {}; } - LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr); - return {}; -} - -bool MemoryManager::IsBlockContinuous(const GPUVAddr start, const std::size_t size) const { - const std::size_t inner_size = size - 1; - const GPUVAddr end = start + inner_size; - const auto host_ptr_start = reinterpret_cast<std::uintptr_t>(GetPointer(start)); - const auto host_ptr_end = reinterpret_cast<std::uintptr_t>(GetPointer(end)); - const auto range = static_cast<std::size_t>(host_ptr_end - host_ptr_start); - return range == inner_size; + return system.Memory().GetPointer(*address); } -void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, - const std::size_t size) const { +void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size) const { std::size_t remaining_size{size}; std::size_t page_index{gpu_src_addr >> page_bits}; std::size_t page_offset{gpu_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 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(src_addr, copy_amount); - memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); + if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) { + const auto src_addr{*page_addr + 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(src_addr, copy_amount); + system.Memory().ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); + } page_index++; page_offset = 0; @@ -241,18 +234,17 @@ void MemoryManager::ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t page_index{gpu_src_addr >> page_bits}; std::size_t page_offset{gpu_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 VAddr src_addr{page_table.backing_addr[page_index] + page_offset}; - memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); + + if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) { + const auto src_addr{*page_addr + page_offset}; + system.Memory().ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); } else { std::memset(dest_buffer, 0, copy_amount); } + page_index++; page_offset = 0; dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount; @@ -260,23 +252,23 @@ void MemoryManager::ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer, } } -void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, - const std::size_t size) { +void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size) { std::size_t remaining_size{size}; std::size_t page_index{gpu_dest_addr >> page_bits}; std::size_t page_offset{gpu_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)}; - 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(dest_addr, copy_amount); - memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); + if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) { + const auto dest_addr{*page_addr + 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(dest_addr, copy_amount); + system.Memory().WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); + } page_index++; page_offset = 0; @@ -286,21 +278,20 @@ void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, } void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buffer, - const std::size_t size) { + std::size_t size) { std::size_t remaining_size{size}; std::size_t page_index{gpu_dest_addr >> page_bits}; std::size_t page_offset{gpu_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) { - const VAddr dest_addr{page_table.backing_addr[page_index] + page_offset}; - memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); + + if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) { + const auto dest_addr{*page_addr + page_offset}; + system.Memory().WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); } + page_index++; page_offset = 0; src_buffer = static_cast<const u8*>(src_buffer) + copy_amount; @@ -308,273 +299,26 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buf } } -void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, - const std::size_t size) { +void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size) { std::vector<u8> tmp_buffer(size); ReadBlock(gpu_src_addr, tmp_buffer.data(), size); WriteBlock(gpu_dest_addr, tmp_buffer.data(), size); } void MemoryManager::CopyBlockUnsafe(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, - const std::size_t size) { + std::size_t size) { std::vector<u8> tmp_buffer(size); ReadBlockUnsafe(gpu_src_addr, tmp_buffer.data(), size); WriteBlockUnsafe(gpu_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 & Core::Memory::PAGE_MASK) + size; - return page <= Core::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, - (base + size) * page_size); - - const VAddr end{base + size}; - ASSERT_MSG(end <= page_table.pointers.size(), "out of range mapping at {:016X}", - base + page_table.pointers.size()); - - if (memory == nullptr) { - while (base != end) { - page_table.pointers[base] = nullptr; - page_table.backing_addr[base] = 0; - - base += 1; - } - } else { - while (base != end) { - page_table.pointers[base] = memory; - page_table.backing_addr[base] = backing_addr; - - base += 1; - memory += page_size; - backing_addr += page_size; - } - } -} - -void MemoryManager::MapMemoryRegion(GPUVAddr base, u64 size, u8* target, VAddr backing_addr) { - ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: {:016X}", size); - ASSERT_MSG((base & page_mask) == 0, "non-page aligned base: {:016X}", base); - MapPages(base / page_size, size / page_size, target, Common::PageType::Memory, backing_addr); -} - -void MemoryManager::UnmapRegion(GPUVAddr base, u64 size) { - ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: {:016X}", size); - ASSERT_MSG((base & page_mask) == 0, "non-page aligned base: {:016X}", base); - MapPages(base / page_size, size / page_size, nullptr, Common::PageType::Unmapped); -} - -bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const { - ASSERT(base + size == next.base); - if (type != next.type) { - return {}; - } - if (type == VirtualMemoryArea::Type::Allocated && (offset + size != next.offset)) { - return {}; - } - if (type == VirtualMemoryArea::Type::Mapped && backing_memory + size != next.backing_memory) { - return {}; - } - return true; -} - -MemoryManager::VMAHandle MemoryManager::FindVMA(GPUVAddr target) const { - if (target >= address_space_end) { - return vma_map.end(); - } else { - return std::prev(vma_map.upper_bound(target)); - } -} - -MemoryManager::VMAIter MemoryManager::Allocate(VMAIter vma_handle) { - VirtualMemoryArea& vma{vma_handle->second}; - - vma.type = VirtualMemoryArea::Type::Allocated; - vma.backing_addr = 0; - vma.backing_memory = {}; - UpdatePageTableForVMA(vma); - - return MergeAdjacent(vma_handle); -} - -MemoryManager::VMAHandle MemoryManager::AllocateMemory(GPUVAddr target, std::size_t offset, - u64 size) { - - // This is the appropriately sized VMA that will turn into our allocation. - VMAIter vma_handle{CarveVMA(target, size)}; - VirtualMemoryArea& vma{vma_handle->second}; - - ASSERT(vma.size == size); - - vma.offset = offset; - - return Allocate(vma_handle); -} - -MemoryManager::VMAHandle MemoryManager::MapBackingMemory(GPUVAddr target, u8* memory, u64 size, - VAddr backing_addr) { - // This is the appropriately sized VMA that will turn into our allocation. - VMAIter vma_handle{CarveVMA(target, size)}; - VirtualMemoryArea& vma{vma_handle->second}; - - ASSERT(vma.size == size); - - vma.type = VirtualMemoryArea::Type::Mapped; - vma.backing_memory = memory; - vma.backing_addr = backing_addr; - UpdatePageTableForVMA(vma); - - return MergeAdjacent(vma_handle); -} - -void MemoryManager::UnmapRange(GPUVAddr target, u64 size) { - VMAIter vma{CarveVMARange(target, size)}; - const VAddr target_end{target + size}; - const VMAIter end{vma_map.end()}; - - // The comparison against the end of the range must be done using addresses since VMAs can be - // merged during this process, causing invalidation of the iterators. - while (vma != end && vma->second.base < target_end) { - // Unmapped ranges return to allocated state and can be reused - // This behavior is used by Super Mario Odyssey, Sonic Forces, and likely other games - vma = std::next(Allocate(vma)); - } - - ASSERT(FindVMA(target)->second.size >= size); -} - -MemoryManager::VMAIter MemoryManager::StripIterConstness(const VMAHandle& iter) { - // This uses a neat C++ trick to convert a const_iterator to a regular iterator, given - // non-const access to its container. - return vma_map.erase(iter, iter); // Erases an empty range of elements -} - -MemoryManager::VMAIter MemoryManager::CarveVMA(GPUVAddr base, u64 size) { - ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: 0x{:016X}", size); - ASSERT_MSG((base & page_mask) == 0, "non-page aligned base: 0x{:016X}", base); - - VMAIter vma_handle{StripIterConstness(FindVMA(base))}; - if (vma_handle == vma_map.end()) { - // Target address is outside the managed range - return {}; - } - - const VirtualMemoryArea& vma{vma_handle->second}; - if (vma.type == VirtualMemoryArea::Type::Mapped) { - // Region is already allocated - return vma_handle; - } - - const VAddr start_in_vma{base - vma.base}; - const VAddr end_in_vma{start_in_vma + size}; - - ASSERT_MSG(end_in_vma <= vma.size, "region size 0x{:016X} is less than required size 0x{:016X}", - vma.size, end_in_vma); - - if (end_in_vma < vma.size) { - // Split VMA at the end of the allocated region - SplitVMA(vma_handle, end_in_vma); - } - if (start_in_vma != 0) { - // Split VMA at the start of the allocated region - vma_handle = SplitVMA(vma_handle, start_in_vma); - } - - return vma_handle; -} - -MemoryManager::VMAIter MemoryManager::CarveVMARange(GPUVAddr target, u64 size) { - ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: 0x{:016X}", size); - ASSERT_MSG((target & page_mask) == 0, "non-page aligned base: 0x{:016X}", target); - - const VAddr target_end{target + size}; - ASSERT(target_end >= target); - ASSERT(size > 0); - - VMAIter begin_vma{StripIterConstness(FindVMA(target))}; - const VMAIter i_end{vma_map.lower_bound(target_end)}; - if (std::any_of(begin_vma, i_end, [](const auto& entry) { - return entry.second.type == VirtualMemoryArea::Type::Unmapped; - })) { + const auto cpu_addr{GpuToCpuAddress(gpu_addr)}; + if (!cpu_addr) { return {}; } - - if (target != begin_vma->second.base) { - begin_vma = SplitVMA(begin_vma, target - begin_vma->second.base); - } - - VMAIter end_vma{StripIterConstness(FindVMA(target_end))}; - if (end_vma != vma_map.end() && target_end != end_vma->second.base) { - end_vma = SplitVMA(end_vma, target_end - end_vma->second.base); - } - - return begin_vma; -} - -MemoryManager::VMAIter MemoryManager::SplitVMA(VMAIter vma_handle, u64 offset_in_vma) { - VirtualMemoryArea& old_vma{vma_handle->second}; - VirtualMemoryArea new_vma{old_vma}; // Make a copy of the VMA - - // For now, don't allow no-op VMA splits (trying to split at a boundary) because it's probably - // a bug. This restriction might be removed later. - ASSERT(offset_in_vma < old_vma.size); - ASSERT(offset_in_vma > 0); - - old_vma.size = offset_in_vma; - new_vma.base += offset_in_vma; - new_vma.size -= offset_in_vma; - - switch (new_vma.type) { - case VirtualMemoryArea::Type::Unmapped: - break; - case VirtualMemoryArea::Type::Allocated: - new_vma.offset += offset_in_vma; - break; - case VirtualMemoryArea::Type::Mapped: - new_vma.backing_memory += offset_in_vma; - break; - } - - ASSERT(old_vma.CanBeMergedWith(new_vma)); - - return vma_map.emplace_hint(std::next(vma_handle), new_vma.base, new_vma); -} - -MemoryManager::VMAIter MemoryManager::MergeAdjacent(VMAIter iter) { - const VMAIter next_vma{std::next(iter)}; - if (next_vma != vma_map.end() && iter->second.CanBeMergedWith(next_vma->second)) { - iter->second.size += next_vma->second.size; - vma_map.erase(next_vma); - } - - if (iter != vma_map.begin()) { - VMAIter prev_vma{std::prev(iter)}; - if (prev_vma->second.CanBeMergedWith(iter->second)) { - prev_vma->second.size += iter->second.size; - vma_map.erase(iter); - iter = prev_vma; - } - } - - return iter; -} - -void MemoryManager::UpdatePageTableForVMA(const VirtualMemoryArea& vma) { - switch (vma.type) { - case VirtualMemoryArea::Type::Unmapped: - UnmapRegion(vma.base, vma.size); - break; - case VirtualMemoryArea::Type::Allocated: - MapMemoryRegion(vma.base, vma.size, nullptr, vma.backing_addr); - break; - case VirtualMemoryArea::Type::Mapped: - MapMemoryRegion(vma.base, vma.size, vma.backing_memory, vma.backing_addr); - break; - } + const std::size_t page{(*cpu_addr & Core::Memory::PAGE_MASK) + size}; + return page <= Core::Memory::PAGE_SIZE; } } // namespace Tegra diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index 87658e87a..681bd9588 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -6,9 +6,9 @@ #include <map> #include <optional> +#include <vector> #include "common/common_types.h" -#include "common/page_table.h" namespace VideoCore { class RasterizerInterface; @@ -20,45 +20,57 @@ class System; namespace Tegra { -/** - * Represents a VMA in an address space. A VMA is a contiguous region of virtual addressing space - * with homogeneous attributes across its extents. In this particular implementation each VMA is - * also backed by a single host memory allocation. - */ -struct VirtualMemoryArea { - enum class Type : u8 { - Unmapped, - Allocated, - Mapped, +class PageEntry final { +public: + enum class State : u32 { + Unmapped = static_cast<u32>(-1), + Allocated = static_cast<u32>(-2), }; - /// Virtual base address of the region. - GPUVAddr base{}; - /// Size of the region. - u64 size{}; - /// Memory area mapping type. - Type type{Type::Unmapped}; - /// CPU memory mapped address corresponding to this memory area. - VAddr backing_addr{}; - /// Offset into the backing_memory the mapping starts from. - std::size_t offset{}; - /// Pointer backing this VMA. - u8* backing_memory{}; - - /// Tests if this area can be merged to the right with `next`. - bool CanBeMergedWith(const VirtualMemoryArea& next) const; + constexpr PageEntry() = default; + constexpr PageEntry(State state) : state{state} {} + constexpr PageEntry(VAddr addr) : state{static_cast<State>(addr >> ShiftBits)} {} + + constexpr bool IsUnmapped() const { + return state == State::Unmapped; + } + + constexpr bool IsAllocated() const { + return state == State::Allocated; + } + + constexpr bool IsValid() const { + return !IsUnmapped() && !IsAllocated(); + } + + constexpr VAddr ToAddress() const { + if (!IsValid()) { + return {}; + } + + return static_cast<VAddr>(state) << ShiftBits; + } + + constexpr PageEntry operator+(u64 offset) { + // If this is a reserved value, offsets do not apply + if (!IsValid()) { + return *this; + } + return PageEntry{(static_cast<VAddr>(state) << ShiftBits) + offset}; + } + +private: + static constexpr std::size_t ShiftBits{12}; + + State state{State::Unmapped}; }; +static_assert(sizeof(PageEntry) == 4, "PageEntry is too large"); class MemoryManager final { public: explicit MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer); ~MemoryManager(); - GPUVAddr AllocateSpace(u64 size, u64 align); - GPUVAddr AllocateSpace(GPUVAddr addr, u64 size, u64 align); - GPUVAddr MapBufferEx(VAddr cpu_addr, u64 size); - GPUVAddr MapBufferEx(VAddr cpu_addr, GPUVAddr addr, u64 size); - GPUVAddr UnmapBuffer(GPUVAddr addr, u64 size); std::optional<VAddr> GpuToCpuAddress(GPUVAddr addr) const; template <typename T> @@ -70,9 +82,6 @@ public: u8* GetPointer(GPUVAddr addr); const u8* GetPointer(GPUVAddr addr) const; - /// Returns true if the block is continuous in host memory, false otherwise - bool IsBlockContinuous(GPUVAddr start, std::size_t size) const; - /** * ReadBlock and WriteBlock are full read and write operations over virtual * GPU Memory. It's important to use these when GPU memory may not be continuous @@ -98,92 +107,43 @@ public: void CopyBlockUnsafe(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size); /** - * IsGranularRange checks if a gpu region can be simply read with a pointer + * 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; - using VMAIter = VMAMap::iterator; - - bool IsAddressValid(GPUVAddr addr) const; - void MapPages(GPUVAddr base, u64 size, u8* memory, Common::PageType type, - VAddr backing_addr = 0); - void MapMemoryRegion(GPUVAddr base, u64 size, u8* target, VAddr backing_addr); - void UnmapRegion(GPUVAddr base, u64 size); - - /// Finds the VMA in which the given address is included in, or `vma_map.end()`. - VMAHandle FindVMA(GPUVAddr target) const; - - VMAHandle AllocateMemory(GPUVAddr target, std::size_t offset, u64 size); - - /** - * Maps an unmanaged host memory pointer at a given address. - * - * @param target The guest address to start the mapping at. - * @param memory The memory to be mapped. - * @param size Size of the mapping in bytes. - * @param backing_addr The base address of the range to back this mapping. - */ - VMAHandle MapBackingMemory(GPUVAddr target, u8* memory, u64 size, VAddr backing_addr); - - /// Unmaps a range of addresses, splitting VMAs as necessary. - void UnmapRange(GPUVAddr target, u64 size); - - /// Converts a VMAHandle to a mutable VMAIter. - VMAIter StripIterConstness(const VMAHandle& iter); - - /// Marks as the specified VMA as allocated. - VMAIter Allocate(VMAIter vma); - - /** - * Carves a VMA of a specific size at the specified address by splitting Free VMAs while doing - * the appropriate error checking. - */ - VMAIter CarveVMA(GPUVAddr base, u64 size); - - /** - * Splits the edges of the given range of non-Free VMAs so that there is a VMA split at each - * end of the range. - */ - VMAIter CarveVMARange(GPUVAddr base, u64 size); - - /** - * Splits a VMA in two, at the specified offset. - * @returns the right side of the split, with the original iterator becoming the left side. - */ - VMAIter SplitVMA(VMAIter vma, u64 offset_in_vma); + GPUVAddr Map(VAddr cpu_addr, GPUVAddr gpu_addr, std::size_t size); + GPUVAddr MapAllocate(VAddr cpu_addr, std::size_t size, std::size_t align); + std::optional<GPUVAddr> AllocateFixed(GPUVAddr gpu_addr, std::size_t size); + GPUVAddr Allocate(std::size_t size, std::size_t align); + void Unmap(GPUVAddr gpu_addr, std::size_t size); - /** - * Checks for and merges the specified VMA with adjacent ones if possible. - * @returns the merged VMA or the original if no merging was possible. - */ - VMAIter MergeAdjacent(VMAIter vma); +private: + PageEntry GetPageEntry(GPUVAddr gpu_addr) const; + void SetPageEntry(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size = page_size); + GPUVAddr UpdateRange(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size); + std::optional<GPUVAddr> FindFreeRange(std::size_t size, std::size_t align) const; - /// Updates the pages corresponding to this VMA so they match the VMA's attributes. - void UpdatePageTableForVMA(const VirtualMemoryArea& vma); + void TryLockPage(PageEntry page_entry, std::size_t size); + void TryUnlockPage(PageEntry page_entry, std::size_t size); - /// Finds a free (unmapped region) of the specified size starting at the specified address. - GPUVAddr FindFreeRegion(GPUVAddr region_start, u64 size) const; + static constexpr std::size_t PageEntryIndex(GPUVAddr gpu_addr) { + return (gpu_addr >> page_bits) & page_table_mask; + } -private: + static constexpr u64 address_space_size = 1ULL << 40; + static constexpr u64 address_space_start = 1ULL << 32; static constexpr u64 page_bits{16}; static constexpr u64 page_size{1 << page_bits}; static constexpr u64 page_mask{page_size - 1}; + static constexpr u64 page_table_bits{24}; + static constexpr u64 page_table_size{1 << page_table_bits}; + static constexpr u64 page_table_mask{page_table_size - 1}; - /// Address space in bits, according to Tegra X1 TRM - static constexpr u32 address_space_width{40}; - /// Start address for mapping, this is fairly arbitrary but must be non-zero. - static constexpr GPUVAddr address_space_base{0x100000}; - /// End of address space, based on address space in bits. - static constexpr GPUVAddr address_space_end{1ULL << address_space_width}; + Core::System& system; - Common::PageTable page_table; - VMAMap vma_map; VideoCore::RasterizerInterface& rasterizer; - Core::System& system; + std::vector<PageEntry> page_table; }; } // namespace Tegra diff --git a/src/video_core/morton.cpp b/src/video_core/morton.cpp index 836b25c1d..9da9fb4ff 100644 --- a/src/video_core/morton.cpp +++ b/src/video_core/morton.cpp @@ -41,146 +41,168 @@ static void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth } static constexpr ConversionArray morton_to_linear_fns = { - MortonCopy<true, PixelFormat::ABGR8U>, - MortonCopy<true, PixelFormat::ABGR8S>, - MortonCopy<true, PixelFormat::ABGR8UI>, - MortonCopy<true, PixelFormat::B5G6R5U>, - MortonCopy<true, PixelFormat::A2B10G10R10U>, - MortonCopy<true, PixelFormat::A1B5G5R5U>, - MortonCopy<true, PixelFormat::R8U>, - MortonCopy<true, PixelFormat::R8UI>, - MortonCopy<true, PixelFormat::RGBA16F>, - MortonCopy<true, PixelFormat::RGBA16U>, - MortonCopy<true, PixelFormat::RGBA16S>, - MortonCopy<true, PixelFormat::RGBA16UI>, - MortonCopy<true, PixelFormat::R11FG11FB10F>, - MortonCopy<true, PixelFormat::RGBA32UI>, - MortonCopy<true, PixelFormat::DXT1>, - MortonCopy<true, PixelFormat::DXT23>, - MortonCopy<true, PixelFormat::DXT45>, - MortonCopy<true, PixelFormat::DXN1>, - MortonCopy<true, PixelFormat::DXN2UNORM>, - MortonCopy<true, PixelFormat::DXN2SNORM>, - MortonCopy<true, PixelFormat::BC7U>, - MortonCopy<true, PixelFormat::BC6H_UF16>, - MortonCopy<true, PixelFormat::BC6H_SF16>, - MortonCopy<true, PixelFormat::ASTC_2D_4X4>, - MortonCopy<true, PixelFormat::BGRA8>, - MortonCopy<true, PixelFormat::RGBA32F>, - MortonCopy<true, PixelFormat::RG32F>, - MortonCopy<true, PixelFormat::R32F>, - MortonCopy<true, PixelFormat::R16F>, - MortonCopy<true, PixelFormat::R16U>, - MortonCopy<true, PixelFormat::R16S>, - MortonCopy<true, PixelFormat::R16UI>, - MortonCopy<true, PixelFormat::R16I>, - MortonCopy<true, PixelFormat::RG16>, - MortonCopy<true, PixelFormat::RG16F>, - MortonCopy<true, PixelFormat::RG16UI>, - MortonCopy<true, PixelFormat::RG16I>, - MortonCopy<true, PixelFormat::RG16S>, - MortonCopy<true, PixelFormat::RGB32F>, - MortonCopy<true, PixelFormat::RGBA8_SRGB>, - MortonCopy<true, PixelFormat::RG8U>, - MortonCopy<true, PixelFormat::RG8S>, - MortonCopy<true, PixelFormat::RG8UI>, - MortonCopy<true, PixelFormat::RG32UI>, - MortonCopy<true, PixelFormat::RGBX16F>, - MortonCopy<true, PixelFormat::R32UI>, - MortonCopy<true, PixelFormat::R32I>, - MortonCopy<true, PixelFormat::ASTC_2D_8X8>, - MortonCopy<true, PixelFormat::ASTC_2D_8X5>, - MortonCopy<true, PixelFormat::ASTC_2D_5X4>, - MortonCopy<true, PixelFormat::BGRA8_SRGB>, - MortonCopy<true, PixelFormat::DXT1_SRGB>, - MortonCopy<true, PixelFormat::DXT23_SRGB>, - MortonCopy<true, PixelFormat::DXT45_SRGB>, - MortonCopy<true, PixelFormat::BC7U_SRGB>, - MortonCopy<true, PixelFormat::R4G4B4A4U>, + MortonCopy<true, PixelFormat::A8B8G8R8_UNORM>, + MortonCopy<true, PixelFormat::A8B8G8R8_SNORM>, + MortonCopy<true, PixelFormat::A8B8G8R8_SINT>, + MortonCopy<true, PixelFormat::A8B8G8R8_UINT>, + MortonCopy<true, PixelFormat::R5G6B5_UNORM>, + MortonCopy<true, PixelFormat::B5G6R5_UNORM>, + MortonCopy<true, PixelFormat::A1R5G5B5_UNORM>, + MortonCopy<true, PixelFormat::A2B10G10R10_UNORM>, + MortonCopy<true, PixelFormat::A2B10G10R10_UINT>, + MortonCopy<true, PixelFormat::A1B5G5R5_UNORM>, + MortonCopy<true, PixelFormat::R8_UNORM>, + MortonCopy<true, PixelFormat::R8_SNORM>, + MortonCopy<true, PixelFormat::R8_SINT>, + MortonCopy<true, PixelFormat::R8_UINT>, + MortonCopy<true, PixelFormat::R16G16B16A16_FLOAT>, + MortonCopy<true, PixelFormat::R16G16B16A16_UNORM>, + MortonCopy<true, PixelFormat::R16G16B16A16_SNORM>, + MortonCopy<true, PixelFormat::R16G16B16A16_SINT>, + MortonCopy<true, PixelFormat::R16G16B16A16_UINT>, + MortonCopy<true, PixelFormat::B10G11R11_FLOAT>, + MortonCopy<true, PixelFormat::R32G32B32A32_UINT>, + MortonCopy<true, PixelFormat::BC1_RGBA_UNORM>, + MortonCopy<true, PixelFormat::BC2_UNORM>, + MortonCopy<true, PixelFormat::BC3_UNORM>, + MortonCopy<true, PixelFormat::BC4_UNORM>, + MortonCopy<true, PixelFormat::BC4_SNORM>, + MortonCopy<true, PixelFormat::BC5_UNORM>, + MortonCopy<true, PixelFormat::BC5_SNORM>, + MortonCopy<true, PixelFormat::BC7_UNORM>, + MortonCopy<true, PixelFormat::BC6H_UFLOAT>, + MortonCopy<true, PixelFormat::BC6H_SFLOAT>, + MortonCopy<true, PixelFormat::ASTC_2D_4X4_UNORM>, + MortonCopy<true, PixelFormat::B8G8R8A8_UNORM>, + MortonCopy<true, PixelFormat::R32G32B32A32_FLOAT>, + MortonCopy<true, PixelFormat::R32G32B32A32_SINT>, + MortonCopy<true, PixelFormat::R32G32_FLOAT>, + MortonCopy<true, PixelFormat::R32G32_SINT>, + MortonCopy<true, PixelFormat::R32_FLOAT>, + MortonCopy<true, PixelFormat::R16_FLOAT>, + MortonCopy<true, PixelFormat::R16_UNORM>, + MortonCopy<true, PixelFormat::R16_SNORM>, + MortonCopy<true, PixelFormat::R16_UINT>, + MortonCopy<true, PixelFormat::R16_SINT>, + MortonCopy<true, PixelFormat::R16G16_UNORM>, + MortonCopy<true, PixelFormat::R16G16_FLOAT>, + MortonCopy<true, PixelFormat::R16G16_UINT>, + MortonCopy<true, PixelFormat::R16G16_SINT>, + MortonCopy<true, PixelFormat::R16G16_SNORM>, + MortonCopy<true, PixelFormat::R32G32B32_FLOAT>, + MortonCopy<true, PixelFormat::A8B8G8R8_SRGB>, + MortonCopy<true, PixelFormat::R8G8_UNORM>, + MortonCopy<true, PixelFormat::R8G8_SNORM>, + MortonCopy<true, PixelFormat::R8G8_SINT>, + MortonCopy<true, PixelFormat::R8G8_UINT>, + MortonCopy<true, PixelFormat::R32G32_UINT>, + MortonCopy<true, PixelFormat::R16G16B16X16_FLOAT>, + MortonCopy<true, PixelFormat::R32_UINT>, + MortonCopy<true, PixelFormat::R32_SINT>, + MortonCopy<true, PixelFormat::ASTC_2D_8X8_UNORM>, + MortonCopy<true, PixelFormat::ASTC_2D_8X5_UNORM>, + MortonCopy<true, PixelFormat::ASTC_2D_5X4_UNORM>, + MortonCopy<true, PixelFormat::B8G8R8A8_SRGB>, + MortonCopy<true, PixelFormat::BC1_RGBA_SRGB>, + MortonCopy<true, PixelFormat::BC2_SRGB>, + MortonCopy<true, PixelFormat::BC3_SRGB>, + MortonCopy<true, PixelFormat::BC7_SRGB>, + MortonCopy<true, PixelFormat::A4B4G4R4_UNORM>, MortonCopy<true, PixelFormat::ASTC_2D_4X4_SRGB>, MortonCopy<true, PixelFormat::ASTC_2D_8X8_SRGB>, MortonCopy<true, PixelFormat::ASTC_2D_8X5_SRGB>, MortonCopy<true, PixelFormat::ASTC_2D_5X4_SRGB>, - MortonCopy<true, PixelFormat::ASTC_2D_5X5>, + MortonCopy<true, PixelFormat::ASTC_2D_5X5_UNORM>, MortonCopy<true, PixelFormat::ASTC_2D_5X5_SRGB>, - MortonCopy<true, PixelFormat::ASTC_2D_10X8>, + MortonCopy<true, PixelFormat::ASTC_2D_10X8_UNORM>, MortonCopy<true, PixelFormat::ASTC_2D_10X8_SRGB>, - MortonCopy<true, PixelFormat::ASTC_2D_6X6>, + MortonCopy<true, PixelFormat::ASTC_2D_6X6_UNORM>, MortonCopy<true, PixelFormat::ASTC_2D_6X6_SRGB>, - MortonCopy<true, PixelFormat::ASTC_2D_10X10>, + MortonCopy<true, PixelFormat::ASTC_2D_10X10_UNORM>, MortonCopy<true, PixelFormat::ASTC_2D_10X10_SRGB>, - MortonCopy<true, PixelFormat::ASTC_2D_12X12>, + MortonCopy<true, PixelFormat::ASTC_2D_12X12_UNORM>, MortonCopy<true, PixelFormat::ASTC_2D_12X12_SRGB>, - MortonCopy<true, PixelFormat::ASTC_2D_8X6>, + MortonCopy<true, PixelFormat::ASTC_2D_8X6_UNORM>, MortonCopy<true, PixelFormat::ASTC_2D_8X6_SRGB>, - MortonCopy<true, PixelFormat::ASTC_2D_6X5>, + MortonCopy<true, PixelFormat::ASTC_2D_6X5_UNORM>, MortonCopy<true, PixelFormat::ASTC_2D_6X5_SRGB>, - MortonCopy<true, PixelFormat::E5B9G9R9F>, - MortonCopy<true, PixelFormat::Z32F>, - MortonCopy<true, PixelFormat::Z16>, - MortonCopy<true, PixelFormat::Z24S8>, - MortonCopy<true, PixelFormat::S8Z24>, - MortonCopy<true, PixelFormat::Z32FS8>, + MortonCopy<true, PixelFormat::E5B9G9R9_FLOAT>, + MortonCopy<true, PixelFormat::D32_FLOAT>, + MortonCopy<true, PixelFormat::D16_UNORM>, + MortonCopy<true, PixelFormat::D24_UNORM_S8_UINT>, + MortonCopy<true, PixelFormat::S8_UINT_D24_UNORM>, + MortonCopy<true, PixelFormat::D32_FLOAT_S8_UINT>, }; static constexpr ConversionArray linear_to_morton_fns = { - MortonCopy<false, PixelFormat::ABGR8U>, - MortonCopy<false, PixelFormat::ABGR8S>, - MortonCopy<false, PixelFormat::ABGR8UI>, - MortonCopy<false, PixelFormat::B5G6R5U>, - MortonCopy<false, PixelFormat::A2B10G10R10U>, - MortonCopy<false, PixelFormat::A1B5G5R5U>, - MortonCopy<false, PixelFormat::R8U>, - MortonCopy<false, PixelFormat::R8UI>, - MortonCopy<false, PixelFormat::RGBA16F>, - MortonCopy<false, PixelFormat::RGBA16S>, - MortonCopy<false, PixelFormat::RGBA16U>, - MortonCopy<false, PixelFormat::RGBA16UI>, - MortonCopy<false, PixelFormat::R11FG11FB10F>, - MortonCopy<false, PixelFormat::RGBA32UI>, - MortonCopy<false, PixelFormat::DXT1>, - MortonCopy<false, PixelFormat::DXT23>, - MortonCopy<false, PixelFormat::DXT45>, - MortonCopy<false, PixelFormat::DXN1>, - MortonCopy<false, PixelFormat::DXN2UNORM>, - MortonCopy<false, PixelFormat::DXN2SNORM>, - MortonCopy<false, PixelFormat::BC7U>, - MortonCopy<false, PixelFormat::BC6H_UF16>, - MortonCopy<false, PixelFormat::BC6H_SF16>, + MortonCopy<false, PixelFormat::A8B8G8R8_UNORM>, + MortonCopy<false, PixelFormat::A8B8G8R8_SNORM>, + MortonCopy<false, PixelFormat::A8B8G8R8_SINT>, + MortonCopy<false, PixelFormat::A8B8G8R8_UINT>, + MortonCopy<false, PixelFormat::R5G6B5_UNORM>, + MortonCopy<false, PixelFormat::B5G6R5_UNORM>, + MortonCopy<false, PixelFormat::A1R5G5B5_UNORM>, + MortonCopy<false, PixelFormat::A2B10G10R10_UNORM>, + MortonCopy<false, PixelFormat::A2B10G10R10_UINT>, + MortonCopy<false, PixelFormat::A1B5G5R5_UNORM>, + MortonCopy<false, PixelFormat::R8_UNORM>, + MortonCopy<false, PixelFormat::R8_SNORM>, + MortonCopy<false, PixelFormat::R8_SINT>, + MortonCopy<false, PixelFormat::R8_UINT>, + MortonCopy<false, PixelFormat::R16G16B16A16_FLOAT>, + MortonCopy<false, PixelFormat::R16G16B16A16_SNORM>, + MortonCopy<false, PixelFormat::R16G16B16A16_SINT>, + MortonCopy<false, PixelFormat::R16G16B16A16_UNORM>, + MortonCopy<false, PixelFormat::R16G16B16A16_UINT>, + MortonCopy<false, PixelFormat::B10G11R11_FLOAT>, + MortonCopy<false, PixelFormat::R32G32B32A32_UINT>, + MortonCopy<false, PixelFormat::BC1_RGBA_UNORM>, + MortonCopy<false, PixelFormat::BC2_UNORM>, + MortonCopy<false, PixelFormat::BC3_UNORM>, + MortonCopy<false, PixelFormat::BC4_UNORM>, + MortonCopy<false, PixelFormat::BC4_SNORM>, + MortonCopy<false, PixelFormat::BC5_UNORM>, + MortonCopy<false, PixelFormat::BC5_SNORM>, + MortonCopy<false, PixelFormat::BC7_UNORM>, + MortonCopy<false, PixelFormat::BC6H_UFLOAT>, + MortonCopy<false, PixelFormat::BC6H_SFLOAT>, // TODO(Subv): Swizzling ASTC formats are not supported nullptr, - MortonCopy<false, PixelFormat::BGRA8>, - MortonCopy<false, PixelFormat::RGBA32F>, - MortonCopy<false, PixelFormat::RG32F>, - MortonCopy<false, PixelFormat::R32F>, - MortonCopy<false, PixelFormat::R16F>, - MortonCopy<false, PixelFormat::R16U>, - MortonCopy<false, PixelFormat::R16S>, - MortonCopy<false, PixelFormat::R16UI>, - MortonCopy<false, PixelFormat::R16I>, - MortonCopy<false, PixelFormat::RG16>, - MortonCopy<false, PixelFormat::RG16F>, - MortonCopy<false, PixelFormat::RG16UI>, - MortonCopy<false, PixelFormat::RG16I>, - MortonCopy<false, PixelFormat::RG16S>, - MortonCopy<false, PixelFormat::RGB32F>, - MortonCopy<false, PixelFormat::RGBA8_SRGB>, - MortonCopy<false, PixelFormat::RG8U>, - MortonCopy<false, PixelFormat::RG8S>, - MortonCopy<false, PixelFormat::RG8UI>, - MortonCopy<false, PixelFormat::RG32UI>, - MortonCopy<false, PixelFormat::RGBX16F>, - MortonCopy<false, PixelFormat::R32UI>, - MortonCopy<false, PixelFormat::R32I>, + MortonCopy<false, PixelFormat::B8G8R8A8_UNORM>, + MortonCopy<false, PixelFormat::R32G32B32A32_FLOAT>, + MortonCopy<false, PixelFormat::R32G32B32A32_SINT>, + MortonCopy<false, PixelFormat::R32G32_FLOAT>, + MortonCopy<false, PixelFormat::R32G32_SINT>, + MortonCopy<false, PixelFormat::R32_FLOAT>, + MortonCopy<false, PixelFormat::R16_FLOAT>, + MortonCopy<false, PixelFormat::R16_UNORM>, + MortonCopy<false, PixelFormat::R16_SNORM>, + MortonCopy<false, PixelFormat::R16_UINT>, + MortonCopy<false, PixelFormat::R16_SINT>, + MortonCopy<false, PixelFormat::R16G16_UNORM>, + MortonCopy<false, PixelFormat::R16G16_FLOAT>, + MortonCopy<false, PixelFormat::R16G16_UINT>, + MortonCopy<false, PixelFormat::R16G16_SINT>, + MortonCopy<false, PixelFormat::R16G16_SNORM>, + MortonCopy<false, PixelFormat::R32G32B32_FLOAT>, + MortonCopy<false, PixelFormat::A8B8G8R8_SRGB>, + MortonCopy<false, PixelFormat::R8G8_UNORM>, + MortonCopy<false, PixelFormat::R8G8_SNORM>, + MortonCopy<false, PixelFormat::R8G8_SINT>, + MortonCopy<false, PixelFormat::R8G8_UINT>, + MortonCopy<false, PixelFormat::R32G32_UINT>, + MortonCopy<false, PixelFormat::R16G16B16X16_FLOAT>, + MortonCopy<false, PixelFormat::R32_UINT>, + MortonCopy<false, PixelFormat::R32_SINT>, nullptr, nullptr, nullptr, - MortonCopy<false, PixelFormat::BGRA8_SRGB>, - MortonCopy<false, PixelFormat::DXT1_SRGB>, - MortonCopy<false, PixelFormat::DXT23_SRGB>, - MortonCopy<false, PixelFormat::DXT45_SRGB>, - MortonCopy<false, PixelFormat::BC7U_SRGB>, - MortonCopy<false, PixelFormat::R4G4B4A4U>, + MortonCopy<false, PixelFormat::B8G8R8A8_SRGB>, + MortonCopy<false, PixelFormat::BC1_RGBA_SRGB>, + MortonCopy<false, PixelFormat::BC2_SRGB>, + MortonCopy<false, PixelFormat::BC3_SRGB>, + MortonCopy<false, PixelFormat::BC7_SRGB>, + MortonCopy<false, PixelFormat::A4B4G4R4_UNORM>, nullptr, nullptr, nullptr, @@ -199,12 +221,12 @@ static constexpr ConversionArray linear_to_morton_fns = { nullptr, nullptr, nullptr, - MortonCopy<false, PixelFormat::E5B9G9R9F>, - MortonCopy<false, PixelFormat::Z32F>, - MortonCopy<false, PixelFormat::Z16>, - MortonCopy<false, PixelFormat::Z24S8>, - MortonCopy<false, PixelFormat::S8Z24>, - MortonCopy<false, PixelFormat::Z32FS8>, + MortonCopy<false, PixelFormat::E5B9G9R9_FLOAT>, + MortonCopy<false, PixelFormat::D32_FLOAT>, + MortonCopy<false, PixelFormat::D16_UNORM>, + MortonCopy<false, PixelFormat::D24_UNORM_S8_UINT>, + MortonCopy<false, PixelFormat::S8_UINT_D24_UNORM>, + MortonCopy<false, PixelFormat::D32_FLOAT_S8_UINT>, }; static MortonCopyFn GetSwizzleFunction(MortonSwizzleMode mode, Surface::PixelFormat format) { diff --git a/src/video_core/renderer_opengl/gl_arb_decompiler.cpp b/src/video_core/renderer_opengl/gl_arb_decompiler.cpp index 4489abf61..b7e9ed2e9 100644 --- a/src/video_core/renderer_opengl/gl_arb_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_arb_decompiler.cpp @@ -913,11 +913,19 @@ void ARBDecompiler::DeclareCompute() { const ComputeInfo& info = registry.GetComputeInfo(); AddLine("GROUP_SIZE {} {} {};", info.workgroup_size[0], info.workgroup_size[1], info.workgroup_size[2]); - if (info.shared_memory_size_in_words > 0) { - const u32 size_in_bytes = info.shared_memory_size_in_words * 4; - AddLine("SHARED_MEMORY {};", size_in_bytes); - AddLine("SHARED shared_mem[] = {{program.sharedmem}};"); + if (info.shared_memory_size_in_words == 0) { + return; + } + const u32 limit = device.GetMaxComputeSharedMemorySize(); + u32 size_in_bytes = info.shared_memory_size_in_words * 4; + if (size_in_bytes > limit) { + LOG_ERROR(Render_OpenGL, "Shared memory size {} is clamped to host's limit {}", + size_in_bytes, limit); + size_in_bytes = limit; } + + AddLine("SHARED_MEMORY {};", size_in_bytes); + AddLine("SHARED shared_mem[] = {{program.sharedmem}};"); } void ARBDecompiler::DeclareInputAttributes() { @@ -1283,13 +1291,6 @@ std::string ARBDecompiler::Visit(const Node& node) { return "{0, 0, 0, 0}.x"; } - const auto buffer_index = [this, &abuf]() -> std::string { - if (stage != ShaderType::Geometry) { - return ""; - } - return fmt::format("[{}]", Visit(abuf->GetBuffer())); - }; - const Attribute::Index index = abuf->GetIndex(); const u32 element = abuf->GetElement(); const char swizzle = Swizzle(element); @@ -1395,7 +1396,7 @@ std::string ARBDecompiler::Visit(const Node& node) { return {}; } - if (const auto cmt = std::get_if<CommentNode>(&*node)) { + if ([[maybe_unused]] const auto cmt = std::get_if<CommentNode>(&*node)) { // Uncommenting this will generate invalid code. GLASM lacks comments. // AddLine("// {}", cmt->GetText()); return {}; @@ -1703,7 +1704,7 @@ std::string ARBDecompiler::HCastFloat(Operation operation) { } std::string ARBDecompiler::HUnpack(Operation operation) { - const std::string operand = Visit(operation[0]); + std::string operand = Visit(operation[0]); switch (std::get<Tegra::Shader::HalfType>(operation.GetMeta())) { case Tegra::Shader::HalfType::H0_H1: return operand; @@ -2053,7 +2054,7 @@ std::string ARBDecompiler::InvocationId(Operation) { std::string ARBDecompiler::YNegate(Operation) { LOG_WARNING(Render_OpenGL, "(STUBBED)"); - const std::string temporary = AllocTemporary(); + std::string temporary = AllocTemporary(); AddLine("MOV.F {}, 1;", temporary); return temporary; } @@ -2076,10 +2077,6 @@ std::string ARBDecompiler::ShuffleIndexed(Operation operation) { } std::string ARBDecompiler::Barrier(Operation) { - if (!ir.IsDecompiled()) { - LOG_ERROR(Render_OpenGL, "BAR used but shader is not decompiled"); - return {}; - } AddLine("BAR;"); return {}; } diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp index 630acb73b..e7d95149f 100644 --- a/src/video_core/renderer_opengl/gl_device.cpp +++ b/src/video_core/renderer_opengl/gl_device.cpp @@ -212,6 +212,7 @@ Device::Device() shader_storage_alignment = GetInteger<std::size_t>(GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT); max_vertex_attributes = GetInteger<u32>(GL_MAX_VERTEX_ATTRIBS); max_varyings = GetInteger<u32>(GL_MAX_VARYING_VECTORS); + max_compute_shared_memory_size = GetInteger<u32>(GL_MAX_COMPUTE_SHARED_MEMORY_SIZE); has_warp_intrinsics = GLAD_GL_NV_gpu_shader5 && GLAD_GL_NV_shader_thread_group && GLAD_GL_NV_shader_thread_shuffle; has_shader_ballot = GLAD_GL_ARB_shader_ballot; @@ -250,6 +251,7 @@ Device::Device(std::nullptr_t) { shader_storage_alignment = 4; max_vertex_attributes = 16; max_varyings = 15; + max_compute_shared_memory_size = 0x10000; has_warp_intrinsics = true; has_shader_ballot = true; has_vertex_viewport_layer = true; diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h index 94d38d7d1..8a4b6b9fc 100644 --- a/src/video_core/renderer_opengl/gl_device.h +++ b/src/video_core/renderer_opengl/gl_device.h @@ -52,6 +52,10 @@ public: return max_varyings; } + u32 GetMaxComputeSharedMemorySize() const { + return max_compute_shared_memory_size; + } + bool HasWarpIntrinsics() const { return has_warp_intrinsics; } @@ -118,6 +122,7 @@ private: std::size_t shader_storage_alignment{}; u32 max_vertex_attributes{}; u32 max_varyings{}; + u32 max_compute_shared_memory_size{}; bool has_warp_intrinsics{}; bool has_shader_ballot{}; bool has_vertex_viewport_layer{}; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 03e82c599..cb284db77 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -178,16 +178,11 @@ RasterizerOpenGL::RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWind if (device.UseAsynchronousShaders()) { // Max worker threads we should allow - constexpr auto MAX_THREADS = 2u; - // Amount of threads we should reserve for other parts of yuzu - constexpr auto RESERVED_THREADS = 6u; - // Get the amount of threads we can use(this can return zero) - const auto cpu_thread_count = - std::max(RESERVED_THREADS, std::thread::hardware_concurrency()); - // Deduce how many "extra" threads we have to use. - const auto max_threads_unused = cpu_thread_count - RESERVED_THREADS; + constexpr u32 MAX_THREADS = 4; + // Deduce how many threads we can use + const u32 threads_used = std::thread::hardware_concurrency() / 4; // Always allow at least 1 thread regardless of our settings - const auto max_worker_count = std::max(1u, max_threads_unused); + const auto max_worker_count = std::max(1U, threads_used); // Don't use more than MAX_THREADS const auto worker_count = std::min(max_worker_count, MAX_THREADS); async_shaders.AllocateWorkers(worker_count); diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index f469ed656..be71e1733 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -126,7 +126,7 @@ std::shared_ptr<Registry> MakeRegistry(const ShaderDiskCacheEntry& entry) { const VideoCore::GuestDriverProfile guest_profile{entry.texture_handler_size}; const VideoCommon::Shader::SerializedRegistryInfo info{guest_profile, entry.bound_buffer, entry.graphics_info, entry.compute_info}; - const auto registry = std::make_shared<Registry>(entry.type, info); + auto registry = std::make_shared<Registry>(entry.type, info); for (const auto& [address, value] : entry.keys) { const auto [buffer, offset] = address; registry->InsertKey(buffer, offset, value); @@ -237,7 +237,6 @@ std::unique_ptr<Shader> Shader::CreateStageFromMemory( const ShaderParameters& params, Maxwell::ShaderProgram program_type, ProgramCode code, ProgramCode code_b, VideoCommon::Shader::AsyncShaders& async_shaders, VAddr cpu_addr) { const auto shader_type = GetShaderType(program_type); - const std::size_t size_in_bytes = code.size() * sizeof(u64); auto& gpu = params.system.GPU(); gpu.ShaderNotify().MarkSharderBuilding(); @@ -287,8 +286,6 @@ std::unique_ptr<Shader> Shader::CreateStageFromMemory( std::unique_ptr<Shader> Shader::CreateKernelFromMemory(const ShaderParameters& params, ProgramCode code) { - const std::size_t size_in_bytes = code.size() * sizeof(u64); - auto& gpu = params.system.GPU(); gpu.ShaderNotify().MarkSharderBuilding(); diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index 2c49aeaac..3f75fcd2b 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -602,8 +602,15 @@ private: return; } const auto& info = registry.GetComputeInfo(); - if (const u32 size = info.shared_memory_size_in_words; size > 0) { - code.AddLine("shared uint smem[{}];", size); + if (u32 size = info.shared_memory_size_in_words * 4; size > 0) { + const u32 limit = device.GetMaxComputeSharedMemorySize(); + if (size > limit) { + LOG_ERROR(Render_OpenGL, "Shared memory size {} is clamped to host's limit {}", + size, limit); + size = limit; + } + + code.AddLine("shared uint smem[{}];", size / 4); code.AddNewLine(); } code.AddLine("layout (local_size_x = {}, local_size_y = {}, local_size_z = {}) in;", @@ -1912,7 +1919,7 @@ private: Expression Comparison(Operation operation) { static_assert(!unordered || type == Type::Float); - const Expression expr = GenerateBinaryInfix(operation, op, Type::Bool, type, type); + Expression expr = GenerateBinaryInfix(operation, op, Type::Bool, type, type); if constexpr (op.compare("!=") == 0 && type == Type::Float && !unordered) { // GLSL's operator!=(float, float) doesn't seem be ordered. This happens on both AMD's @@ -1952,10 +1959,6 @@ private: return {fmt::format("({} != 0)", carry), Type::Bool}; } - Expression LogicalFIsNan(Operation operation) { - return GenerateUnary(operation, "isnan", Type::Bool, Type::Float); - } - Expression LogicalAssign(Operation operation) { const Node& dest = operation[0]; const Node& src = operation[1]; @@ -2771,15 +2774,6 @@ private: return std::min<u32>(device.GetMaxVaryings(), Maxwell::NumVaryings); } - bool IsRenderTargetEnabled(u32 render_target) const { - for (u32 component = 0; component < 4; ++component) { - if (header.ps.IsColorComponentOutputEnabled(render_target, component)) { - return true; - } - } - return false; - } - const Device& device; const ShaderIR& ir; const Registry& registry; diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index 61505879b..0a7bc9e2b 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -41,91 +41,103 @@ struct FormatTuple { }; constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format_tuples = {{ - {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_RG8UI, GL_RG_INTEGER, GL_UNSIGNED_INT}, // RG8UI - {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT}, // RG32UI - {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT}, // RGBX16F - {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT}, // R32UI - {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 + {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // A8B8G8R8_UNORM + {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE}, // A8B8G8R8_SNORM + {GL_RGBA8I, GL_RGBA_INTEGER, GL_BYTE}, // A8B8G8R8_SINT + {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE}, // A8B8G8R8_UINT + {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5}, // R5G6B5_UNORM + {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV}, // B5G6R5_UNORM + {GL_RGB5_A1, GL_BGRA, GL_UNSIGNED_SHORT_1_5_5_5_REV}, // A1R5G5B5_UNORM + {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV}, // A2B10G10R10_UNORM + {GL_RGB10_A2UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT_2_10_10_10_REV}, // A2B10G10R10_UINT + {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV}, // A1B5G5R5_UNORM + {GL_R8, GL_RED, GL_UNSIGNED_BYTE}, // R8_UNORM + {GL_R8_SNORM, GL_RED, GL_BYTE}, // R8_SNORM + {GL_R8I, GL_RED_INTEGER, GL_BYTE}, // R8_SINT + {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE}, // R8_UINT + {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT}, // R16G16B16A16_FLOAT + {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT}, // R16G16B16A16_UNORM + {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT}, // R16G16B16A16_SNORM + {GL_RGBA16I, GL_RGBA_INTEGER, GL_SHORT}, // R16G16B16A16_SINT + {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT}, // R16G16B16A16_UINT + {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV}, // B10G11R11_FLOAT + {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT}, // R32G32B32A32_UINT + {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT}, // BC1_RGBA_UNORM + {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT}, // BC2_UNORM + {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT}, // BC3_UNORM + {GL_COMPRESSED_RED_RGTC1}, // BC4_UNORM + {GL_COMPRESSED_SIGNED_RED_RGTC1}, // BC4_SNORM + {GL_COMPRESSED_RG_RGTC2}, // BC5_UNORM + {GL_COMPRESSED_SIGNED_RG_RGTC2}, // BC5_SNORM + {GL_COMPRESSED_RGBA_BPTC_UNORM}, // BC7_UNORM + {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT}, // BC6H_UFLOAT + {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT}, // BC6H_SFLOAT + {GL_COMPRESSED_RGBA_ASTC_4x4_KHR}, // ASTC_2D_4X4_UNORM + {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE}, // B8G8R8A8_UNORM + {GL_RGBA32F, GL_RGBA, GL_FLOAT}, // R32G32B32A32_FLOAT + {GL_RGBA32I, GL_RGBA_INTEGER, GL_INT}, // R32G32B32A32_SINT + {GL_RG32F, GL_RG, GL_FLOAT}, // R32G32_FLOAT + {GL_RG32I, GL_RG_INTEGER, GL_INT}, // R32G32_SINT + {GL_R32F, GL_RED, GL_FLOAT}, // R32_FLOAT + {GL_R16F, GL_RED, GL_HALF_FLOAT}, // R16_FLOAT + {GL_R16, GL_RED, GL_UNSIGNED_SHORT}, // R16_UNORM + {GL_R16_SNORM, GL_RED, GL_SHORT}, // R16_SNORM + {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT}, // R16_UINT + {GL_R16I, GL_RED_INTEGER, GL_SHORT}, // R16_SINT + {GL_RG16, GL_RG, GL_UNSIGNED_SHORT}, // R16G16_UNORM + {GL_RG16F, GL_RG, GL_HALF_FLOAT}, // R16G16_FLOAT + {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT}, // R16G16_UINT + {GL_RG16I, GL_RG_INTEGER, GL_SHORT}, // R16G16_SINT + {GL_RG16_SNORM, GL_RG, GL_SHORT}, // R16G16_SNORM + {GL_RGB32F, GL_RGB, GL_FLOAT}, // R32G32B32_FLOAT + {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // A8B8G8R8_SRGB + {GL_RG8, GL_RG, GL_UNSIGNED_BYTE}, // R8G8_UNORM + {GL_RG8_SNORM, GL_RG, GL_BYTE}, // R8G8_SNORM + {GL_RG8I, GL_RG_INTEGER, GL_BYTE}, // R8G8_SINT + {GL_RG8UI, GL_RG_INTEGER, GL_UNSIGNED_BYTE}, // R8G8_UINT + {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT}, // R32G32_UINT + {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT}, // R16G16B16X16_FLOAT + {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT}, // R32_UINT + {GL_R32I, GL_RED_INTEGER, GL_INT}, // R32_SINT + {GL_COMPRESSED_RGBA_ASTC_8x8_KHR}, // ASTC_2D_8X8_UNORM + {GL_COMPRESSED_RGBA_ASTC_8x5_KHR}, // ASTC_2D_8X5_UNORM + {GL_COMPRESSED_RGBA_ASTC_5x4_KHR}, // ASTC_2D_5X4_UNORM + {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE}, // B8G8R8A8_UNORM // Compressed sRGB formats - {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_SRGB_ALPHA_S3TC_DXT1_EXT}, // BC1_RGBA_SRGB + {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT}, // BC2_SRGB + {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT}, // BC3_SRGB + {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM}, // BC7_SRGB + {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4_REV}, // A4B4G4R4_UNORM {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_RGBA_ASTC_5x5_KHR}, // ASTC_2D_5X5_UNORM {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x5_KHR}, // ASTC_2D_5X5_SRGB - {GL_COMPRESSED_RGBA_ASTC_10x8_KHR}, // ASTC_2D_10X8 + {GL_COMPRESSED_RGBA_ASTC_10x8_KHR}, // ASTC_2D_10X8_UNORM {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x8_KHR}, // ASTC_2D_10X8_SRGB - {GL_COMPRESSED_RGBA_ASTC_6x6_KHR}, // ASTC_2D_6X6 + {GL_COMPRESSED_RGBA_ASTC_6x6_KHR}, // ASTC_2D_6X6_UNORM {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x6_KHR}, // ASTC_2D_6X6_SRGB - {GL_COMPRESSED_RGBA_ASTC_10x10_KHR}, // ASTC_2D_10X10 + {GL_COMPRESSED_RGBA_ASTC_10x10_KHR}, // ASTC_2D_10X10_UNORM {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x10_KHR}, // ASTC_2D_10X10_SRGB - {GL_COMPRESSED_RGBA_ASTC_12x12_KHR}, // ASTC_2D_12X12 + {GL_COMPRESSED_RGBA_ASTC_12x12_KHR}, // ASTC_2D_12X12_UNORM {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x12_KHR}, // ASTC_2D_12X12_SRGB - {GL_COMPRESSED_RGBA_ASTC_8x6_KHR}, // ASTC_2D_8X6 + {GL_COMPRESSED_RGBA_ASTC_8x6_KHR}, // ASTC_2D_8X6_UNORM {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x6_KHR}, // ASTC_2D_8X6_SRGB - {GL_COMPRESSED_RGBA_ASTC_6x5_KHR}, // ASTC_2D_6X5 + {GL_COMPRESSED_RGBA_ASTC_6x5_KHR}, // ASTC_2D_6X5_UNORM {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 + {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV}, // E5B9G9R9_FLOAT // Depth formats - {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT}, // Z32F - {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // Z16 + {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT}, // D32_FLOAT + {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // D16_UNORM // DepthStencil formats - {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 + {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // D24_UNORM_S8_UINT + {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // S8_UINT_D24_UNORM + {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, + GL_FLOAT_32_UNSIGNED_INT_24_8_REV}, // D32_FLOAT_S8_UINT }}; const FormatTuple& GetFormatTuple(PixelFormat pixel_format) { @@ -178,10 +190,10 @@ GLint GetSwizzleSource(SwizzleSource source) { GLenum GetComponent(PixelFormat format, bool is_first) { switch (format) { - case PixelFormat::Z24S8: - case PixelFormat::Z32FS8: + case PixelFormat::D24_UNORM_S8_UINT: + case PixelFormat::D32_FLOAT_S8_UINT: return is_first ? GL_DEPTH_COMPONENT : GL_STENCIL_INDEX; - case PixelFormat::S8Z24: + case PixelFormat::S8_UINT_D24_UNORM: return is_first ? GL_STENCIL_INDEX : GL_DEPTH_COMPONENT; default: UNREACHABLE(); @@ -482,9 +494,9 @@ GLuint CachedSurfaceView::GetTexture(SwizzleSource x_source, SwizzleSource y_sou std::array swizzle{x_source, y_source, z_source, w_source}; switch (const PixelFormat format = GetSurfaceParams().pixel_format) { - case PixelFormat::Z24S8: - case PixelFormat::Z32FS8: - case PixelFormat::S8Z24: + case PixelFormat::D24_UNORM_S8_UINT: + case PixelFormat::D32_FLOAT_S8_UINT: + case PixelFormat::S8_UINT_D24_UNORM: UNIMPLEMENTED_IF(x_source != SwizzleSource::R && x_source != SwizzleSource::G); glTextureParameteri(view.handle, GL_DEPTH_STENCIL_TEXTURE_MODE, GetComponent(format, x_source == SwizzleSource::R)); diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index e66cdc083..52e9e8250 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -535,12 +535,12 @@ void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture, GLint internal_format; switch (framebuffer.pixel_format) { - case Tegra::FramebufferConfig::PixelFormat::ABGR8: + case Tegra::FramebufferConfig::PixelFormat::A8B8G8R8_UNORM: internal_format = GL_RGBA8; texture.gl_format = GL_RGBA; texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; break; - case Tegra::FramebufferConfig::PixelFormat::RGB565: + case Tegra::FramebufferConfig::PixelFormat::RGB565_UNORM: internal_format = GL_RGB565; texture.gl_format = GL_RGB; texture.gl_type = GL_UNSIGNED_SHORT_5_6_5; diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp index d1f0ea932..81a39a3b8 100644 --- a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp +++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp @@ -40,7 +40,6 @@ constexpr std::array POLYGON_OFFSET_ENABLE_LUT = { } // Anonymous namespace void FixedPipelineState::Fill(const Maxwell& regs, bool has_extended_dynamic_state) { - const auto& clip = regs.view_volume_clip_control; const std::array enabled_lut = {regs.polygon_offset_point_enable, regs.polygon_offset_line_enable, regs.polygon_offset_fill_enable}; diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp index d7f1ae89f..f8c77f4fa 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp @@ -117,90 +117,101 @@ struct FormatTuple { VkFormat format; ///< Vulkan format int usage = 0; ///< Describes image format usage } constexpr tex_format_tuples[] = { - {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, Attachable}, // 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_R16_UINT, Attachable | Storage}, // 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_R8G8_UINT, Attachable | Storage}, // RG8UI - {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_B8G8R8A8_SRGB, Attachable}, // 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 + {VK_FORMAT_A8B8G8R8_UNORM_PACK32, Attachable | Storage}, // A8B8G8R8_UNORM + {VK_FORMAT_A8B8G8R8_SNORM_PACK32, Attachable | Storage}, // A8B8G8R8_SNORM + {VK_FORMAT_A8B8G8R8_SINT_PACK32, Attachable | Storage}, // A8B8G8R8_SINT + {VK_FORMAT_A8B8G8R8_UINT_PACK32, Attachable | Storage}, // A8B8G8R8_UINT + {VK_FORMAT_R5G6B5_UNORM_PACK16, Attachable}, // R5G6B5_UNORM + {VK_FORMAT_B5G6R5_UNORM_PACK16, Attachable}, // B5G6R5_UNORM + {VK_FORMAT_A1R5G5B5_UNORM_PACK16, Attachable}, // A1R5G5B5_UNORM + {VK_FORMAT_A2B10G10R10_UNORM_PACK32, Attachable | Storage}, // A2B10G10R10_UNORM + {VK_FORMAT_A2B10G10R10_UINT_PACK32, Attachable | Storage}, // A2B10G10R10_UINT + {VK_FORMAT_A1R5G5B5_UNORM_PACK16, Attachable}, // A1B5G5R5_UNORM (flipped with swizzle) + {VK_FORMAT_R8_UNORM, Attachable | Storage}, // R8_UNORM + {VK_FORMAT_R8_SNORM, Attachable | Storage}, // R8_SNORM + {VK_FORMAT_R8_SINT, Attachable | Storage}, // R8_SINT + {VK_FORMAT_R8_UINT, Attachable | Storage}, // R8_UINT + {VK_FORMAT_R16G16B16A16_SFLOAT, Attachable | Storage}, // R16G16B16A16_FLOAT + {VK_FORMAT_R16G16B16A16_UNORM, Attachable | Storage}, // R16G16B16A16_UNORM + {VK_FORMAT_R16G16B16A16_SNORM, Attachable | Storage}, // R16G16B16A16_SNORM + {VK_FORMAT_R16G16B16A16_SINT, Attachable | Storage}, // R16G16B16A16_SINT + {VK_FORMAT_R16G16B16A16_UINT, Attachable | Storage}, // R16G16B16A16_UINT + {VK_FORMAT_B10G11R11_UFLOAT_PACK32, Attachable | Storage}, // B10G11R11_FLOAT + {VK_FORMAT_R32G32B32A32_UINT, Attachable | Storage}, // R32G32B32A32_UINT + {VK_FORMAT_BC1_RGBA_UNORM_BLOCK}, // BC1_RGBA_UNORM + {VK_FORMAT_BC2_UNORM_BLOCK}, // BC2_UNORM + {VK_FORMAT_BC3_UNORM_BLOCK}, // BC3_UNORM + {VK_FORMAT_BC4_UNORM_BLOCK}, // BC4_UNORM + {VK_FORMAT_BC4_SNORM_BLOCK}, // BC4_SNORM + {VK_FORMAT_BC5_UNORM_BLOCK}, // BC5_UNORM + {VK_FORMAT_BC5_SNORM_BLOCK}, // BC5_SNORM + {VK_FORMAT_BC7_UNORM_BLOCK}, // BC7_UNORM + {VK_FORMAT_BC6H_UFLOAT_BLOCK}, // BC6H_UFLOAT + {VK_FORMAT_BC6H_SFLOAT_BLOCK}, // BC6H_SFLOAT + {VK_FORMAT_ASTC_4x4_UNORM_BLOCK}, // ASTC_2D_4X4_UNORM + {VK_FORMAT_B8G8R8A8_UNORM, Attachable}, // B8G8R8A8_UNORM + {VK_FORMAT_R32G32B32A32_SFLOAT, Attachable | Storage}, // R32G32B32A32_FLOAT + {VK_FORMAT_R32G32B32A32_SINT, Attachable | Storage}, // R32G32B32A32_SINT + {VK_FORMAT_R32G32_SFLOAT, Attachable | Storage}, // R32G32_FLOAT + {VK_FORMAT_R32G32_SINT, Attachable | Storage}, // R32G32_SINT + {VK_FORMAT_R32_SFLOAT, Attachable | Storage}, // R32_FLOAT + {VK_FORMAT_R16_SFLOAT, Attachable | Storage}, // R16_FLOAT + {VK_FORMAT_R16_UNORM, Attachable | Storage}, // R16_UNORM + {VK_FORMAT_UNDEFINED}, // R16_SNORM + {VK_FORMAT_R16_UINT, Attachable | Storage}, // R16_UINT + {VK_FORMAT_UNDEFINED}, // R16_SINT + {VK_FORMAT_R16G16_UNORM, Attachable | Storage}, // R16G16_UNORM + {VK_FORMAT_R16G16_SFLOAT, Attachable | Storage}, // R16G16_FLOAT + {VK_FORMAT_UNDEFINED}, // R16G16_UINT + {VK_FORMAT_UNDEFINED}, // R16G16_SINT + {VK_FORMAT_R16G16_SNORM, Attachable | Storage}, // R16G16_SNORM + {VK_FORMAT_UNDEFINED}, // R32G32B32_FLOAT + {VK_FORMAT_R8G8B8A8_SRGB, Attachable}, // A8B8G8R8_SRGB + {VK_FORMAT_R8G8_UNORM, Attachable | Storage}, // R8G8_UNORM + {VK_FORMAT_R8G8_SNORM, Attachable | Storage}, // R8G8_SNORM + {VK_FORMAT_R8G8_SINT, Attachable | Storage}, // R8G8_SINT + {VK_FORMAT_R8G8_UINT, Attachable | Storage}, // R8G8_UINT + {VK_FORMAT_R32G32_UINT, Attachable | Storage}, // R32G32_UINT + {VK_FORMAT_UNDEFINED}, // R16G16B16X16_FLOAT + {VK_FORMAT_R32_UINT, Attachable | Storage}, // R32_UINT + {VK_FORMAT_R32_SINT, Attachable | Storage}, // R32_SINT + {VK_FORMAT_ASTC_8x8_UNORM_BLOCK}, // ASTC_2D_8X8_UNORM + {VK_FORMAT_UNDEFINED}, // ASTC_2D_8X5_UNORM + {VK_FORMAT_UNDEFINED}, // ASTC_2D_5X4_UNORM + {VK_FORMAT_B8G8R8A8_SRGB, Attachable}, // B8G8R8A8_SRGB + {VK_FORMAT_BC1_RGBA_SRGB_BLOCK}, // BC1_RGBA_SRGB + {VK_FORMAT_BC2_SRGB_BLOCK}, // BC2_SRGB + {VK_FORMAT_BC3_SRGB_BLOCK}, // BC3_SRGB + {VK_FORMAT_BC7_SRGB_BLOCK}, // BC7_SRGB + {VK_FORMAT_R4G4B4A4_UNORM_PACK16, Attachable}, // A4B4G4R4_UNORM + {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_UNORM + {VK_FORMAT_ASTC_5x5_SRGB_BLOCK}, // ASTC_2D_5X5_SRGB + {VK_FORMAT_ASTC_10x8_UNORM_BLOCK}, // ASTC_2D_10X8_UNORM + {VK_FORMAT_ASTC_10x8_SRGB_BLOCK}, // ASTC_2D_10X8_SRGB + {VK_FORMAT_ASTC_6x6_UNORM_BLOCK}, // ASTC_2D_6X6_UNORM + {VK_FORMAT_ASTC_6x6_SRGB_BLOCK}, // ASTC_2D_6X6_SRGB + {VK_FORMAT_ASTC_10x10_UNORM_BLOCK}, // ASTC_2D_10X10_UNORM + {VK_FORMAT_ASTC_10x10_SRGB_BLOCK}, // ASTC_2D_10X10_SRGB + {VK_FORMAT_ASTC_12x12_UNORM_BLOCK}, // ASTC_2D_12X12_UNORM + {VK_FORMAT_ASTC_12x12_SRGB_BLOCK}, // ASTC_2D_12X12_SRGB + {VK_FORMAT_ASTC_8x6_UNORM_BLOCK}, // ASTC_2D_8X6_UNORM + {VK_FORMAT_ASTC_8x6_SRGB_BLOCK}, // ASTC_2D_8X6_SRGB + {VK_FORMAT_ASTC_6x5_UNORM_BLOCK}, // ASTC_2D_6X5_UNORM + {VK_FORMAT_ASTC_6x5_SRGB_BLOCK}, // ASTC_2D_6X5_SRGB + {VK_FORMAT_E5B9G9R9_UFLOAT_PACK32}, // E5B9G9R9_FLOAT // Depth formats - {VK_FORMAT_D32_SFLOAT, Attachable}, // Z32F - {VK_FORMAT_D16_UNORM, Attachable}, // Z16 + {VK_FORMAT_D32_SFLOAT, Attachable}, // D32_FLOAT + {VK_FORMAT_D16_UNORM, Attachable}, // D16_UNORM // DepthStencil formats - {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 + {VK_FORMAT_D24_UNORM_S8_UINT, Attachable}, // D24_UNORM_S8_UINT + {VK_FORMAT_D24_UNORM_S8_UINT, Attachable}, // S8_UINT_D24_UNORM (emulated) + {VK_FORMAT_D32_SFLOAT_S8_UINT, Attachable}, // D32_FLOAT_S8_UINT }; static_assert(std::size(tex_format_tuples) == VideoCore::Surface::MaxPixelFormat); @@ -221,7 +232,7 @@ FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFo return {VK_FORMAT_A8B8G8R8_UNORM_PACK32, true, true}; } - // Use ABGR8 on hardware that doesn't support ASTC natively + // Use A8B8G8R8_UNORM 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_A8B8G8R8_SRGB_PACK32 diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp index 866813465..a551e3de8 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp +++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp @@ -187,9 +187,9 @@ std::size_t GetSizeInBytes(const Tegra::FramebufferConfig& framebuffer) { VkFormat GetFormat(const Tegra::FramebufferConfig& framebuffer) { switch (framebuffer.pixel_format) { - case Tegra::FramebufferConfig::PixelFormat::ABGR8: + case Tegra::FramebufferConfig::PixelFormat::A8B8G8R8_UNORM: return VK_FORMAT_A8B8G8R8_UNORM_PACK32; - case Tegra::FramebufferConfig::PixelFormat::RGB565: + case Tegra::FramebufferConfig::PixelFormat::RGB565_UNORM: return VK_FORMAT_R5G6B5_UNORM_PACK16; default: UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", @@ -696,6 +696,7 @@ void VKBlitScreen::CreateFramebuffers() { .flags = 0, .renderPass = *renderpass, .attachmentCount = 1, + .pAttachments = nullptr, .width = size.width, .height = size.height, .layers = 1, diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp index 26379ee01..0c03e4d83 100644 --- a/src/video_core/renderer_vulkan/vk_device.cpp +++ b/src/video_core/renderer_vulkan/vk_device.cpp @@ -84,14 +84,19 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties( VK_FORMAT_A8B8G8R8_UNORM_PACK32, VK_FORMAT_A8B8G8R8_UINT_PACK32, VK_FORMAT_A8B8G8R8_SNORM_PACK32, + VK_FORMAT_A8B8G8R8_SINT_PACK32, VK_FORMAT_A8B8G8R8_SRGB_PACK32, VK_FORMAT_B5G6R5_UNORM_PACK16, VK_FORMAT_A2B10G10R10_UNORM_PACK32, + VK_FORMAT_A2B10G10R10_UINT_PACK32, VK_FORMAT_A1R5G5B5_UNORM_PACK16, VK_FORMAT_R32G32B32A32_SFLOAT, + VK_FORMAT_R32G32B32A32_SINT, VK_FORMAT_R32G32B32A32_UINT, VK_FORMAT_R32G32_SFLOAT, + VK_FORMAT_R32G32_SINT, VK_FORMAT_R32G32_UINT, + VK_FORMAT_R16G16B16A16_SINT, VK_FORMAT_R16G16B16A16_UINT, VK_FORMAT_R16G16B16A16_SNORM, VK_FORMAT_R16G16B16A16_UNORM, @@ -103,8 +108,11 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties( VK_FORMAT_R8G8B8A8_SRGB, VK_FORMAT_R8G8_UNORM, VK_FORMAT_R8G8_SNORM, + VK_FORMAT_R8G8_SINT, VK_FORMAT_R8G8_UINT, VK_FORMAT_R8_UNORM, + VK_FORMAT_R8_SNORM, + VK_FORMAT_R8_SINT, VK_FORMAT_R8_UINT, VK_FORMAT_B10G11R11_UFLOAT_PACK32, VK_FORMAT_R32_SFLOAT, @@ -124,6 +132,7 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties( VK_FORMAT_BC2_UNORM_BLOCK, VK_FORMAT_BC3_UNORM_BLOCK, VK_FORMAT_BC4_UNORM_BLOCK, + VK_FORMAT_BC4_SNORM_BLOCK, VK_FORMAT_BC5_UNORM_BLOCK, VK_FORMAT_BC5_SNORM_BLOCK, VK_FORMAT_BC7_UNORM_BLOCK, @@ -762,8 +771,9 @@ std::vector<VkDeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const .pNext = nullptr, .flags = 0, .queueFamilyIndex = queue_family, + .queueCount = 1, + .pQueuePriorities = nullptr, }); - ci.queueCount = 1; ci.pQueuePriorities = &QUEUE_PRIORITY; } diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h index ae5c21baa..529744f2d 100644 --- a/src/video_core/renderer_vulkan/vk_device.h +++ b/src/video_core/renderer_vulkan/vk_device.h @@ -122,6 +122,11 @@ public: return properties.limits.maxPushConstantsSize; } + /// Returns the maximum size for shared memory. + u32 GetMaxComputeSharedMemorySize() const { + return properties.limits.maxComputeSharedMemorySize; + } + /// Returns true if ASTC is natively supported. bool IsOptimalAstcSupported() const { return is_optimal_astc_supported; diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp index 42b3a744c..418c62bc4 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp @@ -261,8 +261,13 @@ VKComputePipeline& VKPipelineCache::GetComputePipeline(const ComputePipelineCach } const Specialization specialization{ + .base_binding = 0, .workgroup_size = key.workgroup_size, .shared_memory_size = key.shared_memory_size, + .point_size = std::nullopt, + .enabled_attributes = {}, + .attribute_types = {}, + .ndc_minus_one_to_one = false, }; const SPIRVShader spirv_shader{Decompile(device, shader->GetIR(), ShaderType::Compute, shader->GetRegistry(), specialization), diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index 31e44aa2b..7500e8244 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -532,10 +532,6 @@ void RasterizerVulkan::Clear() { scheduler.Record([clear_depth = regs.clear_depth, clear_stencil = regs.clear_stencil, 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; @@ -819,8 +815,13 @@ bool RasterizerVulkan::WalkAttachmentOverlaps(const CachedSurfaceView& attachmen 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()}; + FramebufferCacheKey key{ + .renderpass = renderpass, + .width = std::numeric_limits<u32>::max(), + .height = std::numeric_limits<u32>::max(), + .layers = std::numeric_limits<u32>::max(), + .views = {}, + }; const auto try_push = [&key](const View& view) { if (!view) { diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp index 2d5460776..b068888f9 100644 --- a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp @@ -47,6 +47,7 @@ vk::Sampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) c VkSamplerCustomBorderColorCreateInfoEXT border{ .sType = VK_STRUCTURE_TYPE_SAMPLER_CUSTOM_BORDER_COLOR_CREATE_INFO_EXT, .pNext = nullptr, + .customBorderColor = {}, .format = VK_FORMAT_UNDEFINED, }; std::memcpy(&border.customBorderColor, color.data(), sizeof(color)); diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp index 97429cc59..cd7d7a4e4 100644 --- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp +++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp @@ -685,13 +685,19 @@ private: } t_smem_uint = TypePointer(spv::StorageClass::Workgroup, t_uint); - const u32 smem_size = specialization.shared_memory_size; + u32 smem_size = specialization.shared_memory_size * 4; if (smem_size == 0) { // Avoid declaring an empty array. return; } - const auto element_count = static_cast<u32>(Common::AlignUp(smem_size, 4) / 4); - const Id type_array = TypeArray(t_uint, Constant(t_uint, element_count)); + const u32 limit = device.GetMaxComputeSharedMemorySize(); + if (smem_size > limit) { + LOG_ERROR(Render_Vulkan, "Shared memory size {} is clamped to host's limit {}", + smem_size, limit); + smem_size = limit; + } + + const Id type_array = TypeArray(t_uint, Constant(t_uint, smem_size / 4)); const Id type_pointer = TypePointer(spv::StorageClass::Workgroup, type_array); Name(type_pointer, "SharedMemory"); @@ -700,9 +706,9 @@ private: } void DeclareInternalFlags() { - constexpr std::array names = {"zero", "sign", "carry", "overflow"}; + static constexpr std::array names{"zero", "sign", "carry", "overflow"}; + for (std::size_t flag = 0; flag < INTERNAL_FLAGS_COUNT; ++flag) { - const auto flag_code = static_cast<InternalFlag>(flag); const Id id = OpVariable(t_prv_bool, spv::StorageClass::Private, v_false); internal_flags[flag] = AddGlobalVariable(Name(id, names[flag])); } @@ -2798,7 +2804,6 @@ private: std::map<GlobalMemoryBase, Id> global_buffers; std::map<u32, TexelBuffer> uniform_texels; std::map<u32, SampledImage> sampled_images; - std::map<u32, TexelBuffer> storage_texels; std::map<u32, StorageImage> images; std::array<Id, Maxwell::NumRenderTargets> frag_colors{}; diff --git a/src/video_core/renderer_vulkan/vk_swapchain.cpp b/src/video_core/renderer_vulkan/vk_swapchain.cpp index c25e312b6..6bfd2abae 100644 --- a/src/video_core/renderer_vulkan/vk_swapchain.cpp +++ b/src/video_core/renderer_vulkan/vk_swapchain.cpp @@ -156,6 +156,7 @@ void VKSwapchain::CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities, .minImageCount = requested_image_count, .imageFormat = surface_format.format, .imageColorSpace = surface_format.colorSpace, + .imageExtent = {}, .imageArrayLayers = 1, .imageUsage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT, .imageSharingMode = VK_SHARING_MODE_EXCLUSIVE, @@ -204,6 +205,7 @@ void VKSwapchain::CreateImageViews() { .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO, .pNext = nullptr, .flags = 0, + .image = {}, .viewType = VK_IMAGE_VIEW_TYPE_2D, .format = image_format, .components = diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index 9bc18c21a..2c6f54101 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -138,6 +138,7 @@ VkImageCreateInfo GenerateImageCreateInfo(const VKDevice& device, const SurfaceP .flags = 0, .imageType = SurfaceTargetToImage(params.target), .format = format, + .extent = {}, .mipLevels = params.num_levels, .arrayLayers = static_cast<u32>(params.GetNumLayers()), .samples = VK_SAMPLE_COUNT_1_BIT, @@ -235,7 +236,7 @@ void CachedSurface::UploadTexture(const std::vector<u8>& staging_buffer) { void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) { UNIMPLEMENTED_IF(params.IsBuffer()); - if (params.pixel_format == VideoCore::Surface::PixelFormat::A1B5G5R5U) { + if (params.pixel_format == VideoCore::Surface::PixelFormat::A1B5G5R5_UNORM) { LOG_WARNING(Render_Vulkan, "A1B5G5R5 flushing is stubbed"); } @@ -385,7 +386,7 @@ VkImageView CachedSurfaceView::GetImageView(SwizzleSource x_source, SwizzleSourc std::array swizzle{MaxwellToVK::SwizzleSource(x_source), MaxwellToVK::SwizzleSource(y_source), MaxwellToVK::SwizzleSource(z_source), MaxwellToVK::SwizzleSource(w_source)}; - if (params.pixel_format == VideoCore::Surface::PixelFormat::A1B5G5R5U) { + if (params.pixel_format == VideoCore::Surface::PixelFormat::A1B5G5R5_UNORM) { // A1B5G5R5 is implemented as A1R5G5B5, we have to change the swizzle here. std::swap(swizzle[0], swizzle[2]); } @@ -397,11 +398,11 @@ VkImageView CachedSurfaceView::GetImageView(SwizzleSource x_source, SwizzleSourc 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: + case VideoCore::Surface::PixelFormat::D24_UNORM_S8_UINT: + case VideoCore::Surface::PixelFormat::D32_FLOAT_S8_UINT: aspect = is_first ? VK_IMAGE_ASPECT_DEPTH_BIT : VK_IMAGE_ASPECT_STENCIL_BIT; break; - case VideoCore::Surface::PixelFormat::S8Z24: + case VideoCore::Surface::PixelFormat::S8_UINT_D24_UNORM: aspect = is_first ? VK_IMAGE_ASPECT_STENCIL_BIT : VK_IMAGE_ASPECT_DEPTH_BIT; break; default: @@ -458,6 +459,7 @@ VkImageView CachedSurfaceView::GetAttachment() { .pNext = nullptr, .flags = 0, .image = surface.GetImageHandle(), + .viewType = VK_IMAGE_VIEW_TYPE_1D, .format = surface.GetImage().GetFormat(), .components = { @@ -471,6 +473,8 @@ VkImageView CachedSurfaceView::GetAttachment() { .aspectMask = aspect_mask, .baseMipLevel = base_level, .levelCount = num_levels, + .baseArrayLayer = 0, + .layerCount = 0, }, }; if (image_view_type == VK_IMAGE_VIEW_TYPE_3D) { diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp index a041519b7..73155966f 100644 --- a/src/video_core/shader/decode/arithmetic_integer.cpp +++ b/src/video_core/shader/decode/arithmetic_integer.cpp @@ -98,12 +98,12 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) { op_b = GetOperandAbsNegInteger(op_b, false, instr.iadd3.neg_b, true); op_c = GetOperandAbsNegInteger(op_c, false, instr.iadd3.neg_c, true); - const Node value = [&]() { - const Node add_ab = Operation(OperationCode::IAdd, NO_PRECISE, op_a, op_b); + const Node value = [&] { + Node add_ab = Operation(OperationCode::IAdd, NO_PRECISE, op_a, op_b); if (opcode->get().GetId() != OpCode::Id::IADD3_R) { return Operation(OperationCode::IAdd, NO_PRECISE, add_ab, op_c); } - const Node shifted = [&]() { + const Node shifted = [&] { switch (instr.iadd3.mode) { case Tegra::Shader::IAdd3Mode::RightShift: // TODO(tech4me): According to diff --git a/src/video_core/shader/decode/image.cpp b/src/video_core/shader/decode/image.cpp index 07778dc3e..e75ca4fdb 100644 --- a/src/video_core/shader/decode/image.cpp +++ b/src/video_core/shader/decode/image.cpp @@ -31,11 +31,11 @@ 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::R16G16B16A16: + case TextureFormat::R32G32B32A32: + case TextureFormat::R32G32B32: + case TextureFormat::R32G32: + case TextureFormat::R16G16: case TextureFormat::R32: case TextureFormat::R16: case TextureFormat::R8: @@ -97,7 +97,7 @@ ComponentType GetComponentType(Tegra::Engines::SamplerDescriptor descriptor, break; case TextureFormat::B5G6R5: case TextureFormat::B6G5R5: - case TextureFormat::BF10GF11RF11: + case TextureFormat::B10G11R11: if (component == 0) { return descriptor.b_type; } @@ -108,9 +108,9 @@ ComponentType GetComponentType(Tegra::Engines::SamplerDescriptor descriptor, return descriptor.r_type; } break; - case TextureFormat::G8R24: - case TextureFormat::G24R8: - case TextureFormat::G8R8: + case TextureFormat::R24G8: + case TextureFormat::R8G24: + case TextureFormat::R8G8: case TextureFormat::G4R4: if (component == 0) { return descriptor.g_type; @@ -137,15 +137,15 @@ bool IsComponentEnabled(std::size_t component_mask, std::size_t component) { u32 GetComponentSize(TextureFormat format, std::size_t component) { switch (format) { - case TextureFormat::R32_G32_B32_A32: + case TextureFormat::R32G32B32A32: return 32; - case TextureFormat::R16_G16_B16_A16: + case TextureFormat::R16G16B16A16: return 16; - case TextureFormat::R32_G32_B32: + case TextureFormat::R32G32B32: return component <= 2 ? 32 : 0; - case TextureFormat::R32_G32: + case TextureFormat::R32G32: return component <= 1 ? 32 : 0; - case TextureFormat::R16_G16: + case TextureFormat::R16G16: return component <= 1 ? 16 : 0; case TextureFormat::R32: return component == 0 ? 32 : 0; @@ -192,7 +192,7 @@ u32 GetComponentSize(TextureFormat format, std::size_t component) { return 6; } return 0; - case TextureFormat::BF10GF11RF11: + case TextureFormat::B10G11R11: if (component == 1 || component == 2) { return 11; } @@ -200,7 +200,7 @@ u32 GetComponentSize(TextureFormat format, std::size_t component) { return 10; } return 0; - case TextureFormat::G8R24: + case TextureFormat::R24G8: if (component == 0) { return 8; } @@ -208,7 +208,7 @@ u32 GetComponentSize(TextureFormat format, std::size_t component) { return 24; } return 0; - case TextureFormat::G24R8: + case TextureFormat::R8G24: if (component == 0) { return 8; } @@ -216,7 +216,7 @@ u32 GetComponentSize(TextureFormat format, std::size_t component) { return 24; } return 0; - case TextureFormat::G8R8: + case TextureFormat::R8G8: return (component == 0 || component == 1) ? 8 : 0; case TextureFormat::G4R4: return (component == 0 || component == 1) ? 4 : 0; @@ -231,25 +231,25 @@ std::size_t GetImageComponentMask(TextureFormat format) { constexpr u8 B = 0b0100; constexpr u8 A = 0b1000; switch (format) { - case TextureFormat::R32_G32_B32_A32: - case TextureFormat::R16_G16_B16_A16: + case TextureFormat::R32G32B32A32: + case TextureFormat::R16G16B16A16: 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::R32G32B32: case TextureFormat::R32_B24G8: case TextureFormat::B5G6R5: case TextureFormat::B6G5R5: - case TextureFormat::BF10GF11RF11: + case TextureFormat::B10G11R11: 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::R32G32: + case TextureFormat::R16G16: + case TextureFormat::R24G8: + case TextureFormat::R8G24: + case TextureFormat::R8G8: case TextureFormat::G4R4: return std::size_t{R | G}; case TextureFormat::R32: diff --git a/src/video_core/shader/decode/other.cpp b/src/video_core/shader/decode/other.cpp index c0a8f233f..29a7cfbfe 100644 --- a/src/video_core/shader/decode/other.cpp +++ b/src/video_core/shader/decode/other.cpp @@ -75,8 +75,7 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { const Node value = [this, instr] { switch (instr.sys20) { case SystemVariable::LaneId: - LOG_WARNING(HW_GPU, "S2R instruction with LaneId is incomplete"); - return Immediate(0U); + return Operation(OperationCode::ThreadId); case SystemVariable::InvocationId: return Operation(OperationCode::InvocationId); case SystemVariable::Ydirection: diff --git a/src/video_core/shader/decode/video.cpp b/src/video_core/shader/decode/video.cpp index 64ba60ea2..1c0957277 100644 --- a/src/video_core/shader/decode/video.cpp +++ b/src/video_core/shader/decode/video.cpp @@ -91,29 +91,28 @@ u32 ShaderIR::DecodeVideo(NodeBlock& bb, u32 pc) { return pc; } -Node ShaderIR::GetVideoOperand(Node op, bool is_chunk, bool is_signed, - Tegra::Shader::VideoType type, u64 byte_height) { +Node ShaderIR::GetVideoOperand(Node op, bool is_chunk, bool is_signed, VideoType type, + u64 byte_height) { if (!is_chunk) { return BitfieldExtract(op, static_cast<u32>(byte_height * 8), 8); } - const Node zero = Immediate(0); switch (type) { - case Tegra::Shader::VideoType::Size16_Low: + case VideoType::Size16_Low: return BitfieldExtract(op, 0, 16); - case Tegra::Shader::VideoType::Size16_High: + case VideoType::Size16_High: return BitfieldExtract(op, 16, 16); - case Tegra::Shader::VideoType::Size32: + case VideoType::Size32: // TODO(Rodrigo): From my hardware tests it becomes a bit "mad" when this type is used // (1 * 1 + 0 == 0x5b800000). Until a better explanation is found: abort. UNIMPLEMENTED(); - return zero; - case Tegra::Shader::VideoType::Invalid: + return Immediate(0); + case VideoType::Invalid: UNREACHABLE_MSG("Invalid instruction encoding"); - return zero; + return Immediate(0); default: UNREACHABLE(); - return zero; + return Immediate(0); } } diff --git a/src/video_core/shader/decode/xmad.cpp b/src/video_core/shader/decode/xmad.cpp index c83dc6615..233b8fa42 100644 --- a/src/video_core/shader/decode/xmad.cpp +++ b/src/video_core/shader/decode/xmad.cpp @@ -81,20 +81,21 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) { SetTemporary(bb, 0, product); product = GetTemporary(0); - const Node original_c = op_c; + Node original_c = op_c; const Tegra::Shader::XmadMode set_mode = mode; // Workaround to clang compile error - op_c = [&]() { + op_c = [&] { switch (set_mode) { case Tegra::Shader::XmadMode::None: return original_c; case Tegra::Shader::XmadMode::CLo: - return BitfieldExtract(original_c, 0, 16); + return BitfieldExtract(std::move(original_c), 0, 16); case Tegra::Shader::XmadMode::CHi: - return BitfieldExtract(original_c, 16, 16); + return BitfieldExtract(std::move(original_c), 16, 16); case Tegra::Shader::XmadMode::CBcc: { - const Node shifted_b = SignedOperation(OperationCode::ILogicalShiftLeft, is_signed_b, - original_b, Immediate(16)); - return SignedOperation(OperationCode::IAdd, is_signed_c, original_c, shifted_b); + Node shifted_b = SignedOperation(OperationCode::ILogicalShiftLeft, is_signed_b, + original_b, Immediate(16)); + return SignedOperation(OperationCode::IAdd, is_signed_c, std::move(original_c), + std::move(shifted_b)); } case Tegra::Shader::XmadMode::CSfu: { const Node comp_a = diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp index e322c3402..29d794b34 100644 --- a/src/video_core/shader/shader_ir.cpp +++ b/src/video_core/shader/shader_ir.cpp @@ -112,9 +112,9 @@ Node ShaderIR::GetOutputAttribute(Attribute::Index index, u64 element, Node buff } Node ShaderIR::GetInternalFlag(InternalFlag flag, bool negated) const { - const Node node = MakeNode<InternalFlagNode>(flag); + Node node = MakeNode<InternalFlagNode>(flag); if (negated) { - return Operation(OperationCode::LogicalNegate, node); + return Operation(OperationCode::LogicalNegate, std::move(node)); } return node; } diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp index bbe93903c..1688267bb 100644 --- a/src/video_core/surface.cpp +++ b/src/video_core/surface.cpp @@ -74,117 +74,131 @@ bool SurfaceTargetIsArray(SurfaceTarget target) { PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) { switch (format) { - case Tegra::DepthFormat::S8_Z24_UNORM: - return PixelFormat::S8Z24; - case Tegra::DepthFormat::Z24_S8_UNORM: - return PixelFormat::Z24S8; - case Tegra::DepthFormat::Z32_FLOAT: - return PixelFormat::Z32F; - case Tegra::DepthFormat::Z16_UNORM: - return PixelFormat::Z16; - case Tegra::DepthFormat::Z32_S8_X24_FLOAT: - return PixelFormat::Z32FS8; + case Tegra::DepthFormat::S8_UINT_Z24_UNORM: + return PixelFormat::S8_UINT_D24_UNORM; + case Tegra::DepthFormat::D24S8_UNORM: + return PixelFormat::D24_UNORM_S8_UINT; + case Tegra::DepthFormat::D32_FLOAT: + return PixelFormat::D32_FLOAT; + case Tegra::DepthFormat::D16_UNORM: + return PixelFormat::D16_UNORM; + case Tegra::DepthFormat::D32_FLOAT_S8X24_UINT: + return PixelFormat::D32_FLOAT_S8_UINT; default: - LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); - UNREACHABLE(); - return PixelFormat::S8Z24; + UNIMPLEMENTED_MSG("Unimplemented format={}", static_cast<u32>(format)); + return PixelFormat::S8_UINT_D24_UNORM; } } PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format) { switch (format) { - case Tegra::RenderTargetFormat::RGBA8_SRGB: - return PixelFormat::RGBA8_SRGB; - case Tegra::RenderTargetFormat::RGBA8_UNORM: - return PixelFormat::ABGR8U; - case Tegra::RenderTargetFormat::RGBA8_SNORM: - return PixelFormat::ABGR8S; - case Tegra::RenderTargetFormat::RGBA8_UINT: - return PixelFormat::ABGR8UI; - case Tegra::RenderTargetFormat::BGRA8_SRGB: - return PixelFormat::BGRA8_SRGB; - case Tegra::RenderTargetFormat::BGRA8_UNORM: - return PixelFormat::BGRA8; - case Tegra::RenderTargetFormat::RGB10_A2_UNORM: - return PixelFormat::A2B10G10R10U; - case Tegra::RenderTargetFormat::RGBA16_FLOAT: - return PixelFormat::RGBA16F; - case Tegra::RenderTargetFormat::RGBA16_UNORM: - return PixelFormat::RGBA16U; - case Tegra::RenderTargetFormat::RGBA16_SNORM: - return PixelFormat::RGBA16S; - case Tegra::RenderTargetFormat::RGBA16_UINT: - return PixelFormat::RGBA16UI; - case Tegra::RenderTargetFormat::RGBA32_FLOAT: - return PixelFormat::RGBA32F; - case Tegra::RenderTargetFormat::RG32_FLOAT: - return PixelFormat::RG32F; - case Tegra::RenderTargetFormat::R11G11B10_FLOAT: - return PixelFormat::R11FG11FB10F; - case Tegra::RenderTargetFormat::B5G6R5_UNORM: - return PixelFormat::B5G6R5U; - case Tegra::RenderTargetFormat::BGR5A1_UNORM: - return PixelFormat::A1B5G5R5U; - case Tegra::RenderTargetFormat::RGBA32_UINT: - return PixelFormat::RGBA32UI; - case Tegra::RenderTargetFormat::R8_UNORM: - return PixelFormat::R8U; - case Tegra::RenderTargetFormat::R8_UINT: - return PixelFormat::R8UI; - case Tegra::RenderTargetFormat::RG16_FLOAT: - return PixelFormat::RG16F; - case Tegra::RenderTargetFormat::RG16_UINT: - return PixelFormat::RG16UI; - case Tegra::RenderTargetFormat::RG16_SINT: - return PixelFormat::RG16I; - case Tegra::RenderTargetFormat::RG16_UNORM: - return PixelFormat::RG16; - case Tegra::RenderTargetFormat::RG16_SNORM: - return PixelFormat::RG16S; - case Tegra::RenderTargetFormat::RG8_UNORM: - return PixelFormat::RG8U; - case Tegra::RenderTargetFormat::RG8_SNORM: - return PixelFormat::RG8S; - case Tegra::RenderTargetFormat::RG8_UINT: - return PixelFormat::RG8UI; - case Tegra::RenderTargetFormat::R16_FLOAT: - return PixelFormat::R16F; + case Tegra::RenderTargetFormat::R32B32G32A32_FLOAT: + return PixelFormat::R32G32B32A32_FLOAT; + case Tegra::RenderTargetFormat::R32G32B32A32_SINT: + return PixelFormat::R32G32B32A32_SINT; + case Tegra::RenderTargetFormat::R32G32B32A32_UINT: + return PixelFormat::R32G32B32A32_UINT; + case Tegra::RenderTargetFormat::R16G16B16A16_UNORM: + return PixelFormat::R16G16B16A16_UNORM; + case Tegra::RenderTargetFormat::R16G16B16A16_SNORM: + return PixelFormat::R16G16B16A16_SNORM; + case Tegra::RenderTargetFormat::R16G16B16A16_SINT: + return PixelFormat::R16G16B16A16_SINT; + case Tegra::RenderTargetFormat::R16G16B16A16_UINT: + return PixelFormat::R16G16B16A16_UINT; + case Tegra::RenderTargetFormat::R16G16B16A16_FLOAT: + return PixelFormat::R16G16B16A16_FLOAT; + case Tegra::RenderTargetFormat::R32G32_FLOAT: + return PixelFormat::R32G32_FLOAT; + case Tegra::RenderTargetFormat::R32G32_SINT: + return PixelFormat::R32G32_SINT; + case Tegra::RenderTargetFormat::R32G32_UINT: + return PixelFormat::R32G32_UINT; + case Tegra::RenderTargetFormat::R16G16B16X16_FLOAT: + return PixelFormat::R16G16B16X16_FLOAT; + case Tegra::RenderTargetFormat::B8G8R8A8_UNORM: + return PixelFormat::B8G8R8A8_UNORM; + case Tegra::RenderTargetFormat::B8G8R8A8_SRGB: + return PixelFormat::B8G8R8A8_SRGB; + case Tegra::RenderTargetFormat::A2B10G10R10_UNORM: + return PixelFormat::A2B10G10R10_UNORM; + case Tegra::RenderTargetFormat::A2B10G10R10_UINT: + return PixelFormat::A2B10G10R10_UINT; + case Tegra::RenderTargetFormat::A8B8G8R8_UNORM: + return PixelFormat::A8B8G8R8_UNORM; + case Tegra::RenderTargetFormat::A8B8G8R8_SRGB: + return PixelFormat::A8B8G8R8_SRGB; + case Tegra::RenderTargetFormat::A8B8G8R8_SNORM: + return PixelFormat::A8B8G8R8_SNORM; + case Tegra::RenderTargetFormat::A8B8G8R8_SINT: + return PixelFormat::A8B8G8R8_SINT; + case Tegra::RenderTargetFormat::A8B8G8R8_UINT: + return PixelFormat::A8B8G8R8_UINT; + case Tegra::RenderTargetFormat::R16G16_UNORM: + return PixelFormat::R16G16_UNORM; + case Tegra::RenderTargetFormat::R16G16_SNORM: + return PixelFormat::R16G16_SNORM; + case Tegra::RenderTargetFormat::R16G16_SINT: + return PixelFormat::R16G16_SINT; + case Tegra::RenderTargetFormat::R16G16_UINT: + return PixelFormat::R16G16_UINT; + case Tegra::RenderTargetFormat::R16G16_FLOAT: + return PixelFormat::R16G16_FLOAT; + case Tegra::RenderTargetFormat::B10G11R11_FLOAT: + return PixelFormat::B10G11R11_FLOAT; + case Tegra::RenderTargetFormat::R32_SINT: + return PixelFormat::R32_SINT; + case Tegra::RenderTargetFormat::R32_UINT: + return PixelFormat::R32_UINT; + case Tegra::RenderTargetFormat::R32_FLOAT: + return PixelFormat::R32_FLOAT; + case Tegra::RenderTargetFormat::R5G6B5_UNORM: + return PixelFormat::R5G6B5_UNORM; + case Tegra::RenderTargetFormat::A1R5G5B5_UNORM: + return PixelFormat::A1R5G5B5_UNORM; + case Tegra::RenderTargetFormat::R8G8_UNORM: + return PixelFormat::R8G8_UNORM; + case Tegra::RenderTargetFormat::R8G8_SNORM: + return PixelFormat::R8G8_SNORM; + case Tegra::RenderTargetFormat::R8G8_SINT: + return PixelFormat::R8G8_SINT; + case Tegra::RenderTargetFormat::R8G8_UINT: + return PixelFormat::R8G8_UINT; case Tegra::RenderTargetFormat::R16_UNORM: - return PixelFormat::R16U; + return PixelFormat::R16_UNORM; case Tegra::RenderTargetFormat::R16_SNORM: - return PixelFormat::R16S; - case Tegra::RenderTargetFormat::R16_UINT: - return PixelFormat::R16UI; + return PixelFormat::R16_SNORM; case Tegra::RenderTargetFormat::R16_SINT: - return PixelFormat::R16I; - case Tegra::RenderTargetFormat::R32_FLOAT: - return PixelFormat::R32F; - case Tegra::RenderTargetFormat::R32_SINT: - return PixelFormat::R32I; - case Tegra::RenderTargetFormat::R32_UINT: - return PixelFormat::R32UI; - case Tegra::RenderTargetFormat::RG32_UINT: - return PixelFormat::RG32UI; - case Tegra::RenderTargetFormat::RGBX16_FLOAT: - return PixelFormat::RGBX16F; + return PixelFormat::R16_SINT; + case Tegra::RenderTargetFormat::R16_UINT: + return PixelFormat::R16_UINT; + case Tegra::RenderTargetFormat::R16_FLOAT: + return PixelFormat::R16_FLOAT; + case Tegra::RenderTargetFormat::R8_UNORM: + return PixelFormat::R8_UNORM; + case Tegra::RenderTargetFormat::R8_SNORM: + return PixelFormat::R8_SNORM; + case Tegra::RenderTargetFormat::R8_SINT: + return PixelFormat::R8_SINT; + case Tegra::RenderTargetFormat::R8_UINT: + return PixelFormat::R8_UINT; default: - LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); - UNREACHABLE(); - return PixelFormat::RGBA8_SRGB; + UNIMPLEMENTED_MSG("Unimplemented format={}", static_cast<int>(format)); + return PixelFormat::A8B8G8R8_UNORM; } } PixelFormat PixelFormatFromGPUPixelFormat(Tegra::FramebufferConfig::PixelFormat format) { switch (format) { - case Tegra::FramebufferConfig::PixelFormat::ABGR8: - return PixelFormat::ABGR8U; - case Tegra::FramebufferConfig::PixelFormat::RGB565: - return PixelFormat::B5G6R5U; - case Tegra::FramebufferConfig::PixelFormat::BGRA8: - return PixelFormat::BGRA8; + case Tegra::FramebufferConfig::PixelFormat::A8B8G8R8_UNORM: + return PixelFormat::A8B8G8R8_UNORM; + case Tegra::FramebufferConfig::PixelFormat::RGB565_UNORM: + return PixelFormat::R5G6B5_UNORM; + case Tegra::FramebufferConfig::PixelFormat::B8G8R8A8_UNORM: + return PixelFormat::B8G8R8A8_UNORM; default: UNIMPLEMENTED_MSG("Unimplemented format={}", static_cast<u32>(format)); - return PixelFormat::ABGR8U; + return PixelFormat::A8B8G8R8_UNORM; } } @@ -212,27 +226,27 @@ SurfaceType GetFormatType(PixelFormat pixel_format) { bool IsPixelFormatASTC(PixelFormat format) { switch (format) { - case PixelFormat::ASTC_2D_4X4: - case PixelFormat::ASTC_2D_5X4: - case PixelFormat::ASTC_2D_5X5: - case PixelFormat::ASTC_2D_8X8: - case PixelFormat::ASTC_2D_8X5: + case PixelFormat::ASTC_2D_4X4_UNORM: + case PixelFormat::ASTC_2D_5X4_UNORM: + case PixelFormat::ASTC_2D_5X5_UNORM: + case PixelFormat::ASTC_2D_8X8_UNORM: + case PixelFormat::ASTC_2D_8X5_UNORM: case PixelFormat::ASTC_2D_4X4_SRGB: case PixelFormat::ASTC_2D_5X4_SRGB: case PixelFormat::ASTC_2D_5X5_SRGB: case PixelFormat::ASTC_2D_8X8_SRGB: case PixelFormat::ASTC_2D_8X5_SRGB: - case PixelFormat::ASTC_2D_10X8: + case PixelFormat::ASTC_2D_10X8_UNORM: case PixelFormat::ASTC_2D_10X8_SRGB: - case PixelFormat::ASTC_2D_6X6: + case PixelFormat::ASTC_2D_6X6_UNORM: case PixelFormat::ASTC_2D_6X6_SRGB: - case PixelFormat::ASTC_2D_10X10: + case PixelFormat::ASTC_2D_10X10_UNORM: case PixelFormat::ASTC_2D_10X10_SRGB: - case PixelFormat::ASTC_2D_12X12: + case PixelFormat::ASTC_2D_12X12_UNORM: case PixelFormat::ASTC_2D_12X12_SRGB: - case PixelFormat::ASTC_2D_8X6: + case PixelFormat::ASTC_2D_8X6_UNORM: case PixelFormat::ASTC_2D_8X6_SRGB: - case PixelFormat::ASTC_2D_6X5: + case PixelFormat::ASTC_2D_6X5_UNORM: case PixelFormat::ASTC_2D_6X5_SRGB: return true; default: @@ -242,12 +256,12 @@ bool IsPixelFormatASTC(PixelFormat format) { bool IsPixelFormatSRGB(PixelFormat format) { switch (format) { - case PixelFormat::RGBA8_SRGB: - case PixelFormat::BGRA8_SRGB: - case PixelFormat::DXT1_SRGB: - case PixelFormat::DXT23_SRGB: - case PixelFormat::DXT45_SRGB: - case PixelFormat::BC7U_SRGB: + case PixelFormat::A8B8G8R8_SRGB: + case PixelFormat::B8G8R8A8_SRGB: + case PixelFormat::BC1_RGBA_SRGB: + case PixelFormat::BC2_SRGB: + case PixelFormat::BC3_SRGB: + case PixelFormat::BC7_SRGB: case PixelFormat::ASTC_2D_4X4_SRGB: case PixelFormat::ASTC_2D_8X8_SRGB: case PixelFormat::ASTC_2D_8X5_SRGB: @@ -269,25 +283,4 @@ std::pair<u32, u32> GetASTCBlockSize(PixelFormat format) { return {GetDefaultBlockWidth(format), GetDefaultBlockHeight(format)}; } -bool IsFormatBCn(PixelFormat format) { - switch (format) { - case PixelFormat::DXT1: - case PixelFormat::DXT23: - case PixelFormat::DXT45: - case PixelFormat::DXN1: - case PixelFormat::DXN2SNORM: - case PixelFormat::DXN2UNORM: - case PixelFormat::BC7U: - case PixelFormat::BC6H_UF16: - case PixelFormat::BC6H_SF16: - case PixelFormat::DXT1_SRGB: - case PixelFormat::DXT23_SRGB: - case PixelFormat::DXT45_SRGB: - case PixelFormat::BC7U_SRGB: - return true; - default: - return false; - } -} - } // namespace VideoCore::Surface diff --git a/src/video_core/surface.h b/src/video_core/surface.h index 6da6a1b97..cfd12fa61 100644 --- a/src/video_core/surface.h +++ b/src/video_core/surface.h @@ -15,94 +15,105 @@ namespace VideoCore::Surface { enum class PixelFormat { - ABGR8U = 0, - ABGR8S = 1, - ABGR8UI = 2, - B5G6R5U = 3, - A2B10G10R10U = 4, - A1B5G5R5U = 5, - R8U = 6, - R8UI = 7, - RGBA16F = 8, - RGBA16U = 9, - RGBA16S = 10, - RGBA16UI = 11, - R11FG11FB10F = 12, - RGBA32UI = 13, - DXT1 = 14, - DXT23 = 15, - DXT45 = 16, - DXN1 = 17, // This is also known as BC4 - DXN2UNORM = 18, - DXN2SNORM = 19, - BC7U = 20, - BC6H_UF16 = 21, - BC6H_SF16 = 22, - ASTC_2D_4X4 = 23, - BGRA8 = 24, - RGBA32F = 25, - RG32F = 26, - R32F = 27, - R16F = 28, - R16U = 29, - R16S = 30, - R16UI = 31, - R16I = 32, - RG16 = 33, - RG16F = 34, - RG16UI = 35, - RG16I = 36, - RG16S = 37, - RGB32F = 38, - RGBA8_SRGB = 39, - RG8U = 40, - RG8S = 41, - RG8UI = 42, - RG32UI = 43, - RGBX16F = 44, - R32UI = 45, - R32I = 46, - ASTC_2D_8X8 = 47, - ASTC_2D_8X5 = 48, - ASTC_2D_5X4 = 49, - BGRA8_SRGB = 50, - DXT1_SRGB = 51, - DXT23_SRGB = 52, - DXT45_SRGB = 53, - BC7U_SRGB = 54, - R4G4B4A4U = 55, - ASTC_2D_4X4_SRGB = 56, - ASTC_2D_8X8_SRGB = 57, - ASTC_2D_8X5_SRGB = 58, - ASTC_2D_5X4_SRGB = 59, - ASTC_2D_5X5 = 60, - ASTC_2D_5X5_SRGB = 61, - ASTC_2D_10X8 = 62, - ASTC_2D_10X8_SRGB = 63, - ASTC_2D_6X6 = 64, - ASTC_2D_6X6_SRGB = 65, - ASTC_2D_10X10 = 66, - ASTC_2D_10X10_SRGB = 67, - ASTC_2D_12X12 = 68, - ASTC_2D_12X12_SRGB = 69, - ASTC_2D_8X6 = 70, - ASTC_2D_8X6_SRGB = 71, - ASTC_2D_6X5 = 72, - ASTC_2D_6X5_SRGB = 73, - E5B9G9R9F = 74, + A8B8G8R8_UNORM, + A8B8G8R8_SNORM, + A8B8G8R8_SINT, + A8B8G8R8_UINT, + R5G6B5_UNORM, + B5G6R5_UNORM, + A1R5G5B5_UNORM, + A2B10G10R10_UNORM, + A2B10G10R10_UINT, + A1B5G5R5_UNORM, + R8_UNORM, + R8_SNORM, + R8_SINT, + R8_UINT, + R16G16B16A16_FLOAT, + R16G16B16A16_UNORM, + R16G16B16A16_SNORM, + R16G16B16A16_SINT, + R16G16B16A16_UINT, + B10G11R11_FLOAT, + R32G32B32A32_UINT, + BC1_RGBA_UNORM, + BC2_UNORM, + BC3_UNORM, + BC4_UNORM, + BC4_SNORM, + BC5_UNORM, + BC5_SNORM, + BC7_UNORM, + BC6H_UFLOAT, + BC6H_SFLOAT, + ASTC_2D_4X4_UNORM, + B8G8R8A8_UNORM, + R32G32B32A32_FLOAT, + R32G32B32A32_SINT, + R32G32_FLOAT, + R32G32_SINT, + R32_FLOAT, + R16_FLOAT, + R16_UNORM, + R16_SNORM, + R16_UINT, + R16_SINT, + R16G16_UNORM, + R16G16_FLOAT, + R16G16_UINT, + R16G16_SINT, + R16G16_SNORM, + R32G32B32_FLOAT, + A8B8G8R8_SRGB, + R8G8_UNORM, + R8G8_SNORM, + R8G8_SINT, + R8G8_UINT, + R32G32_UINT, + R16G16B16X16_FLOAT, + R32_UINT, + R32_SINT, + ASTC_2D_8X8_UNORM, + ASTC_2D_8X5_UNORM, + ASTC_2D_5X4_UNORM, + B8G8R8A8_SRGB, + BC1_RGBA_SRGB, + BC2_SRGB, + BC3_SRGB, + BC7_SRGB, + A4B4G4R4_UNORM, + ASTC_2D_4X4_SRGB, + ASTC_2D_8X8_SRGB, + ASTC_2D_8X5_SRGB, + ASTC_2D_5X4_SRGB, + ASTC_2D_5X5_UNORM, + ASTC_2D_5X5_SRGB, + ASTC_2D_10X8_UNORM, + ASTC_2D_10X8_SRGB, + ASTC_2D_6X6_UNORM, + ASTC_2D_6X6_SRGB, + ASTC_2D_10X10_UNORM, + ASTC_2D_10X10_SRGB, + ASTC_2D_12X12_UNORM, + ASTC_2D_12X12_SRGB, + ASTC_2D_8X6_UNORM, + ASTC_2D_8X6_SRGB, + ASTC_2D_6X5_UNORM, + ASTC_2D_6X5_SRGB, + E5B9G9R9_FLOAT, MaxColorFormat, // Depth formats - Z32F = 75, - Z16 = 76, + D32_FLOAT = MaxColorFormat, + D16_UNORM, MaxDepthFormat, // DepthStencil formats - Z24S8 = 77, - S8Z24 = 78, - Z32FS8 = 79, + D24_UNORM_S8_UINT = MaxDepthFormat, + S8_UINT_D24_UNORM, + D32_FLOAT_S8_UINT, MaxDepthStencilFormat, @@ -130,86 +141,97 @@ enum class SurfaceTarget { }; constexpr std::array<u32, MaxPixelFormat> compression_factor_shift_table = {{ - 0, // ABGR8U - 0, // ABGR8S - 0, // ABGR8UI - 0, // B5G6R5U - 0, // A2B10G10R10U - 0, // A1B5G5R5U - 0, // R8U - 0, // R8UI - 0, // RGBA16F - 0, // RGBA16U - 0, // RGBA16S - 0, // RGBA16UI - 0, // R11FG11FB10F - 0, // RGBA32UI - 2, // DXT1 - 2, // DXT23 - 2, // DXT45 - 2, // DXN1 - 2, // DXN2UNORM - 2, // DXN2SNORM - 2, // BC7U - 2, // BC6H_UF16 - 2, // BC6H_SF16 - 2, // ASTC_2D_4X4 - 0, // BGRA8 - 0, // RGBA32F - 0, // RG32F - 0, // R32F - 0, // R16F - 0, // R16U - 0, // R16S - 0, // R16UI - 0, // R16I - 0, // RG16 - 0, // RG16F - 0, // RG16UI - 0, // RG16I - 0, // RG16S - 0, // RGB32F - 0, // RGBA8_SRGB - 0, // RG8U - 0, // RG8S - 0, // RG8UI - 0, // RG32UI - 0, // RGBX16F - 0, // R32UI - 0, // R32I - 2, // ASTC_2D_8X8 - 2, // ASTC_2D_8X5 - 2, // ASTC_2D_5X4 - 0, // BGRA8_SRGB - 2, // DXT1_SRGB - 2, // DXT23_SRGB - 2, // DXT45_SRGB - 2, // BC7U_SRGB - 0, // R4G4B4A4U + 0, // A8B8G8R8_UNORM + 0, // A8B8G8R8_SNORM + 0, // A8B8G8R8_SINT + 0, // A8B8G8R8_UINT + 0, // R5G6B5_UNORM + 0, // B5G6R5_UNORM + 0, // A1R5G5B5_UNORM + 0, // A2B10G10R10_UNORM + 0, // A2B10G10R10_UINT + 0, // A1B5G5R5_UNORM + 0, // R8_UNORM + 0, // R8_SNORM + 0, // R8_SINT + 0, // R8_UINT + 0, // R16G16B16A16_FLOAT + 0, // R16G16B16A16_UNORM + 0, // R16G16B16A16_SNORM + 0, // R16G16B16A16_SINT + 0, // R16G16B16A16_UINT + 0, // B10G11R11_FLOAT + 0, // R32G32B32A32_UINT + 2, // BC1_RGBA_UNORM + 2, // BC2_UNORM + 2, // BC3_UNORM + 2, // BC4_UNORM + 2, // BC4_SNORM + 2, // BC5_UNORM + 2, // BC5_SNORM + 2, // BC7_UNORM + 2, // BC6H_UFLOAT + 2, // BC6H_SFLOAT + 2, // ASTC_2D_4X4_UNORM + 0, // B8G8R8A8_UNORM + 0, // R32G32B32A32_FLOAT + 0, // R32G32B32A32_SINT + 0, // R32G32_FLOAT + 0, // R32G32_SINT + 0, // R32_FLOAT + 0, // R16_FLOAT + 0, // R16_UNORM + 0, // R16_SNORM + 0, // R16_UINT + 0, // R16_SINT + 0, // R16G16_UNORM + 0, // R16G16_FLOAT + 0, // R16G16_UINT + 0, // R16G16_SINT + 0, // R16G16_SNORM + 0, // R32G32B32_FLOAT + 0, // A8B8G8R8_SRGB + 0, // R8G8_UNORM + 0, // R8G8_SNORM + 0, // R8G8_SINT + 0, // R8G8_UINT + 0, // R32G32_UINT + 0, // R16G16B16X16_FLOAT + 0, // R32_UINT + 0, // R32_SINT + 2, // ASTC_2D_8X8_UNORM + 2, // ASTC_2D_8X5_UNORM + 2, // ASTC_2D_5X4_UNORM + 0, // B8G8R8A8_SRGB + 2, // BC1_RGBA_SRGB + 2, // BC2_SRGB + 2, // BC3_SRGB + 2, // BC7_SRGB + 0, // A4B4G4R4_UNORM 2, // ASTC_2D_4X4_SRGB 2, // ASTC_2D_8X8_SRGB 2, // ASTC_2D_8X5_SRGB 2, // ASTC_2D_5X4_SRGB - 2, // ASTC_2D_5X5 + 2, // ASTC_2D_5X5_UNORM 2, // ASTC_2D_5X5_SRGB - 2, // ASTC_2D_10X8 + 2, // ASTC_2D_10X8_UNORM 2, // ASTC_2D_10X8_SRGB - 2, // ASTC_2D_6X6 + 2, // ASTC_2D_6X6_UNORM 2, // ASTC_2D_6X6_SRGB - 2, // ASTC_2D_10X10 + 2, // ASTC_2D_10X10_UNORM 2, // ASTC_2D_10X10_SRGB - 2, // ASTC_2D_12X12 + 2, // ASTC_2D_12X12_UNORM 2, // ASTC_2D_12X12_SRGB - 2, // ASTC_2D_8X6 + 2, // ASTC_2D_8X6_UNORM 2, // ASTC_2D_8X6_SRGB - 2, // ASTC_2D_6X5 + 2, // ASTC_2D_6X5_UNORM 2, // ASTC_2D_6X5_SRGB - 0, // E5B9G9R9F - 0, // Z32F - 0, // Z16 - 0, // Z24S8 - 0, // S8Z24 - 0, // Z32FS8 + 0, // E5B9G9R9_FLOAT + 0, // D32_FLOAT + 0, // D16_UNORM + 0, // D24_UNORM_S8_UINT + 0, // S8_UINT_D24_UNORM + 0, // D32_FLOAT_S8_UINT }}; /** @@ -229,86 +251,97 @@ inline constexpr u32 GetCompressionFactor(PixelFormat format) { } constexpr std::array<u32, MaxPixelFormat> block_width_table = {{ - 1, // ABGR8U - 1, // ABGR8S - 1, // ABGR8UI - 1, // B5G6R5U - 1, // A2B10G10R10U - 1, // A1B5G5R5U - 1, // R8U - 1, // R8UI - 1, // RGBA16F - 1, // RGBA16U - 1, // RGBA16S - 1, // RGBA16UI - 1, // R11FG11FB10F - 1, // RGBA32UI - 4, // DXT1 - 4, // DXT23 - 4, // DXT45 - 4, // DXN1 - 4, // DXN2UNORM - 4, // DXN2SNORM - 4, // BC7U - 4, // BC6H_UF16 - 4, // BC6H_SF16 - 4, // ASTC_2D_4X4 - 1, // BGRA8 - 1, // RGBA32F - 1, // RG32F - 1, // R32F - 1, // R16F - 1, // R16U - 1, // R16S - 1, // R16UI - 1, // R16I - 1, // RG16 - 1, // RG16F - 1, // RG16UI - 1, // RG16I - 1, // RG16S - 1, // RGB32F - 1, // RGBA8_SRGB - 1, // RG8U - 1, // RG8S - 1, // RG8UI - 1, // RG32UI - 1, // RGBX16F - 1, // R32UI - 1, // R32I - 8, // ASTC_2D_8X8 - 8, // ASTC_2D_8X5 - 5, // ASTC_2D_5X4 - 1, // BGRA8_SRGB - 4, // DXT1_SRGB - 4, // DXT23_SRGB - 4, // DXT45_SRGB - 4, // BC7U_SRGB - 1, // R4G4B4A4U + 1, // A8B8G8R8_UNORM + 1, // A8B8G8R8_SNORM + 1, // A8B8G8R8_SINT + 1, // A8B8G8R8_UINT + 1, // R5G6B5_UNORM + 1, // B5G6R5_UNORM + 1, // A1R5G5B5_UNORM + 1, // A2B10G10R10_UNORM + 1, // A2B10G10R10_UINT + 1, // A1B5G5R5_UNORM + 1, // R8_UNORM + 1, // R8_SNORM + 1, // R8_SINT + 1, // R8_UINT + 1, // R16G16B16A16_FLOAT + 1, // R16G16B16A16_UNORM + 1, // R16G16B16A16_SNORM + 1, // R16G16B16A16_SINT + 1, // R16G16B16A16_UINT + 1, // B10G11R11_FLOAT + 1, // R32G32B32A32_UINT + 4, // BC1_RGBA_UNORM + 4, // BC2_UNORM + 4, // BC3_UNORM + 4, // BC4_UNORM + 4, // BC4_SNORM + 4, // BC5_UNORM + 4, // BC5_SNORM + 4, // BC7_UNORM + 4, // BC6H_UFLOAT + 4, // BC6H_SFLOAT + 4, // ASTC_2D_4X4_UNORM + 1, // B8G8R8A8_UNORM + 1, // R32G32B32A32_FLOAT + 1, // R32G32B32A32_SINT + 1, // R32G32_FLOAT + 1, // R32G32_SINT + 1, // R32_FLOAT + 1, // R16_FLOAT + 1, // R16_UNORM + 1, // R16_SNORM + 1, // R16_UINT + 1, // R16_SINT + 1, // R16G16_UNORM + 1, // R16G16_FLOAT + 1, // R16G16_UINT + 1, // R16G16_SINT + 1, // R16G16_SNORM + 1, // R32G32B32_FLOAT + 1, // A8B8G8R8_SRGB + 1, // R8G8_UNORM + 1, // R8G8_SNORM + 1, // R8G8_SINT + 1, // R8G8_UINT + 1, // R32G32_UINT + 1, // R16G16B16X16_FLOAT + 1, // R32_UINT + 1, // R32_SINT + 8, // ASTC_2D_8X8_UNORM + 8, // ASTC_2D_8X5_UNORM + 5, // ASTC_2D_5X4_UNORM + 1, // B8G8R8A8_SRGB + 4, // BC1_RGBA_SRGB + 4, // BC2_SRGB + 4, // BC3_SRGB + 4, // BC7_SRGB + 1, // A4B4G4R4_UNORM 4, // ASTC_2D_4X4_SRGB 8, // ASTC_2D_8X8_SRGB 8, // ASTC_2D_8X5_SRGB 5, // ASTC_2D_5X4_SRGB - 5, // ASTC_2D_5X5 + 5, // ASTC_2D_5X5_UNORM 5, // ASTC_2D_5X5_SRGB - 10, // ASTC_2D_10X8 + 10, // ASTC_2D_10X8_UNORM 10, // ASTC_2D_10X8_SRGB - 6, // ASTC_2D_6X6 + 6, // ASTC_2D_6X6_UNORM 6, // ASTC_2D_6X6_SRGB - 10, // ASTC_2D_10X10 + 10, // ASTC_2D_10X10_UNORM 10, // ASTC_2D_10X10_SRGB - 12, // ASTC_2D_12X12 + 12, // ASTC_2D_12X12_UNORM 12, // ASTC_2D_12X12_SRGB - 8, // ASTC_2D_8X6 + 8, // ASTC_2D_8X6_UNORM 8, // ASTC_2D_8X6_SRGB - 6, // ASTC_2D_6X5 + 6, // ASTC_2D_6X5_UNORM 6, // ASTC_2D_6X5_SRGB - 1, // E5B9G9R9F - 1, // Z32F - 1, // Z16 - 1, // Z24S8 - 1, // S8Z24 - 1, // Z32FS8 + 1, // E5B9G9R9_FLOAT + 1, // D32_FLOAT + 1, // D16_UNORM + 1, // D24_UNORM_S8_UINT + 1, // S8_UINT_D24_UNORM + 1, // D32_FLOAT_S8_UINT }}; static constexpr u32 GetDefaultBlockWidth(PixelFormat format) { @@ -320,86 +353,97 @@ static constexpr u32 GetDefaultBlockWidth(PixelFormat format) { } constexpr std::array<u32, MaxPixelFormat> block_height_table = {{ - 1, // ABGR8U - 1, // ABGR8S - 1, // ABGR8UI - 1, // B5G6R5U - 1, // A2B10G10R10U - 1, // A1B5G5R5U - 1, // R8U - 1, // R8UI - 1, // RGBA16F - 1, // RGBA16U - 1, // RGBA16S - 1, // RGBA16UI - 1, // R11FG11FB10F - 1, // RGBA32UI - 4, // DXT1 - 4, // DXT23 - 4, // DXT45 - 4, // DXN1 - 4, // DXN2UNORM - 4, // DXN2SNORM - 4, // BC7U - 4, // BC6H_UF16 - 4, // BC6H_SF16 - 4, // ASTC_2D_4X4 - 1, // BGRA8 - 1, // RGBA32F - 1, // RG32F - 1, // R32F - 1, // R16F - 1, // R16U - 1, // R16S - 1, // R16UI - 1, // R16I - 1, // RG16 - 1, // RG16F - 1, // RG16UI - 1, // RG16I - 1, // RG16S - 1, // RGB32F - 1, // RGBA8_SRGB - 1, // RG8U - 1, // RG8S - 1, // RG8UI - 1, // RG32UI - 1, // RGBX16F - 1, // R32UI - 1, // R32I - 8, // ASTC_2D_8X8 - 5, // ASTC_2D_8X5 - 4, // ASTC_2D_5X4 - 1, // BGRA8_SRGB - 4, // DXT1_SRGB - 4, // DXT23_SRGB - 4, // DXT45_SRGB - 4, // BC7U_SRGB - 1, // R4G4B4A4U + 1, // A8B8G8R8_UNORM + 1, // A8B8G8R8_SNORM + 1, // A8B8G8R8_SINT + 1, // A8B8G8R8_UINT + 1, // R5G6B5_UNORM + 1, // B5G6R5_UNORM + 1, // A1R5G5B5_UNORM + 1, // A2B10G10R10_UNORM + 1, // A2B10G10R10_UINT + 1, // A1B5G5R5_UNORM + 1, // R8_UNORM + 1, // R8_SNORM + 1, // R8_SINT + 1, // R8_UINT + 1, // R16G16B16A16_FLOAT + 1, // R16G16B16A16_UNORM + 1, // R16G16B16A16_SNORM + 1, // R16G16B16A16_SINT + 1, // R16G16B16A16_UINT + 1, // B10G11R11_FLOAT + 1, // R32G32B32A32_UINT + 4, // BC1_RGBA_UNORM + 4, // BC2_UNORM + 4, // BC3_UNORM + 4, // BC4_UNORM + 4, // BC4_SNORM + 4, // BC5_UNORM + 4, // BC5_SNORM + 4, // BC7_UNORM + 4, // BC6H_UFLOAT + 4, // BC6H_SFLOAT + 4, // ASTC_2D_4X4_UNORM + 1, // B8G8R8A8_UNORM + 1, // R32G32B32A32_FLOAT + 1, // R32G32B32A32_SINT + 1, // R32G32_FLOAT + 1, // R32G32_SINT + 1, // R32_FLOAT + 1, // R16_FLOAT + 1, // R16_UNORM + 1, // R16_SNORM + 1, // R16_UINT + 1, // R16_SINT + 1, // R16G16_UNORM + 1, // R16G16_FLOAT + 1, // R16G16_UINT + 1, // R16G16_SINT + 1, // R16G16_SNORM + 1, // R32G32B32_FLOAT + 1, // A8B8G8R8_SRGB + 1, // R8G8_UNORM + 1, // R8G8_SNORM + 1, // R8G8_SINT + 1, // R8G8_UINT + 1, // R32G32_UINT + 1, // R16G16B16X16_FLOAT + 1, // R32_UINT + 1, // R32_SINT + 8, // ASTC_2D_8X8_UNORM + 5, // ASTC_2D_8X5_UNORM + 4, // ASTC_2D_5X4_UNORM + 1, // B8G8R8A8_SRGB + 4, // BC1_RGBA_SRGB + 4, // BC2_SRGB + 4, // BC3_SRGB + 4, // BC7_SRGB + 1, // A4B4G4R4_UNORM 4, // ASTC_2D_4X4_SRGB 8, // ASTC_2D_8X8_SRGB 5, // ASTC_2D_8X5_SRGB 4, // ASTC_2D_5X4_SRGB - 5, // ASTC_2D_5X5 + 5, // ASTC_2D_5X5_UNORM 5, // ASTC_2D_5X5_SRGB - 8, // ASTC_2D_10X8 + 8, // ASTC_2D_10X8_UNORM 8, // ASTC_2D_10X8_SRGB - 6, // ASTC_2D_6X6 + 6, // ASTC_2D_6X6_UNORM 6, // ASTC_2D_6X6_SRGB - 10, // ASTC_2D_10X10 + 10, // ASTC_2D_10X10_UNORM 10, // ASTC_2D_10X10_SRGB - 12, // ASTC_2D_12X12 + 12, // ASTC_2D_12X12_UNORM 12, // ASTC_2D_12X12_SRGB - 6, // ASTC_2D_8X6 + 6, // ASTC_2D_8X6_UNORM 6, // ASTC_2D_8X6_SRGB - 5, // ASTC_2D_6X5 + 5, // ASTC_2D_6X5_UNORM 5, // ASTC_2D_6X5_SRGB - 1, // E5B9G9R9F - 1, // Z32F - 1, // Z16 - 1, // Z24S8 - 1, // S8Z24 - 1, // Z32FS8 + 1, // E5B9G9R9_FLOAT + 1, // D32_FLOAT + 1, // D16_UNORM + 1, // D24_UNORM_S8_UINT + 1, // S8_UINT_D24_UNORM + 1, // D32_FLOAT_S8_UINT }}; static constexpr u32 GetDefaultBlockHeight(PixelFormat format) { @@ -411,86 +455,97 @@ static constexpr u32 GetDefaultBlockHeight(PixelFormat format) { } constexpr std::array<u32, MaxPixelFormat> bpp_table = {{ - 32, // ABGR8U - 32, // ABGR8S - 32, // ABGR8UI - 16, // B5G6R5U - 32, // A2B10G10R10U - 16, // A1B5G5R5U - 8, // R8U - 8, // R8UI - 64, // RGBA16F - 64, // RGBA16U - 64, // RGBA16S - 64, // RGBA16UI - 32, // R11FG11FB10F - 128, // RGBA32UI - 64, // DXT1 - 128, // DXT23 - 128, // DXT45 - 64, // DXN1 - 128, // DXN2UNORM - 128, // DXN2SNORM - 128, // BC7U - 128, // BC6H_UF16 - 128, // BC6H_SF16 - 128, // ASTC_2D_4X4 - 32, // BGRA8 - 128, // RGBA32F - 64, // RG32F - 32, // R32F - 16, // R16F - 16, // R16U - 16, // R16S - 16, // R16UI - 16, // R16I - 32, // RG16 - 32, // RG16F - 32, // RG16UI - 32, // RG16I - 32, // RG16S - 96, // RGB32F - 32, // RGBA8_SRGB - 16, // RG8U - 16, // RG8S - 16, // RG8UI - 64, // RG32UI - 64, // RGBX16F - 32, // R32UI - 32, // R32I - 128, // ASTC_2D_8X8 - 128, // ASTC_2D_8X5 - 128, // ASTC_2D_5X4 - 32, // BGRA8_SRGB - 64, // DXT1_SRGB - 128, // DXT23_SRGB - 128, // DXT45_SRGB - 128, // BC7U - 16, // R4G4B4A4U + 32, // A8B8G8R8_UNORM + 32, // A8B8G8R8_SNORM + 32, // A8B8G8R8_SINT + 32, // A8B8G8R8_UINT + 16, // R5G6B5_UNORM + 16, // B5G6R5_UNORM + 16, // A1R5G5B5_UNORM + 32, // A2B10G10R10_UNORM + 32, // A2B10G10R10_UINT + 16, // A1B5G5R5_UNORM + 8, // R8_UNORM + 8, // R8_SNORM + 8, // R8_SINT + 8, // R8_UINT + 64, // R16G16B16A16_FLOAT + 64, // R16G16B16A16_UNORM + 64, // R16G16B16A16_SNORM + 64, // R16G16B16A16_SINT + 64, // R16G16B16A16_UINT + 32, // B10G11R11_FLOAT + 128, // R32G32B32A32_UINT + 64, // BC1_RGBA_UNORM + 128, // BC2_UNORM + 128, // BC3_UNORM + 64, // BC4_UNORM + 64, // BC4_SNORM + 128, // BC5_UNORM + 128, // BC5_SNORM + 128, // BC7_UNORM + 128, // BC6H_UFLOAT + 128, // BC6H_SFLOAT + 128, // ASTC_2D_4X4_UNORM + 32, // B8G8R8A8_UNORM + 128, // R32G32B32A32_FLOAT + 128, // R32G32B32A32_SINT + 64, // R32G32_FLOAT + 64, // R32G32_SINT + 32, // R32_FLOAT + 16, // R16_FLOAT + 16, // R16_UNORM + 16, // R16_SNORM + 16, // R16_UINT + 16, // R16_SINT + 32, // R16G16_UNORM + 32, // R16G16_FLOAT + 32, // R16G16_UINT + 32, // R16G16_SINT + 32, // R16G16_SNORM + 96, // R32G32B32_FLOAT + 32, // A8B8G8R8_SRGB + 16, // R8G8_UNORM + 16, // R8G8_SNORM + 16, // R8G8_SINT + 16, // R8G8_UINT + 64, // R32G32_UINT + 64, // R16G16B16X16_FLOAT + 32, // R32_UINT + 32, // R32_SINT + 128, // ASTC_2D_8X8_UNORM + 128, // ASTC_2D_8X5_UNORM + 128, // ASTC_2D_5X4_UNORM + 32, // B8G8R8A8_SRGB + 64, // BC1_RGBA_SRGB + 128, // BC2_SRGB + 128, // BC3_SRGB + 128, // BC7_UNORM + 16, // A4B4G4R4_UNORM 128, // ASTC_2D_4X4_SRGB 128, // ASTC_2D_8X8_SRGB 128, // ASTC_2D_8X5_SRGB 128, // ASTC_2D_5X4_SRGB - 128, // ASTC_2D_5X5 + 128, // ASTC_2D_5X5_UNORM 128, // ASTC_2D_5X5_SRGB - 128, // ASTC_2D_10X8 + 128, // ASTC_2D_10X8_UNORM 128, // ASTC_2D_10X8_SRGB - 128, // ASTC_2D_6X6 + 128, // ASTC_2D_6X6_UNORM 128, // ASTC_2D_6X6_SRGB - 128, // ASTC_2D_10X10 + 128, // ASTC_2D_10X10_UNORM 128, // ASTC_2D_10X10_SRGB - 128, // ASTC_2D_12X12 + 128, // ASTC_2D_12X12_UNORM 128, // ASTC_2D_12X12_SRGB - 128, // ASTC_2D_8X6 + 128, // ASTC_2D_8X6_UNORM 128, // ASTC_2D_8X6_SRGB - 128, // ASTC_2D_6X5 + 128, // ASTC_2D_6X5_UNORM 128, // ASTC_2D_6X5_SRGB - 32, // E5B9G9R9F - 32, // Z32F - 16, // Z16 - 32, // Z24S8 - 32, // S8Z24 - 64, // Z32FS8 + 32, // E5B9G9R9_FLOAT + 32, // D32_FLOAT + 16, // D16_UNORM + 32, // D24_UNORM_S8_UINT + 32, // S8_UINT_D24_UNORM + 64, // D32_FLOAT_S8_UINT }}; static constexpr u32 GetFormatBpp(PixelFormat format) { @@ -529,7 +584,4 @@ bool IsPixelFormatSRGB(PixelFormat format); std::pair<u32, u32> GetASTCBlockSize(PixelFormat format); -/// Returns true if the specified PixelFormat is a BCn format, e.g. DXT or DXN -bool IsFormatBCn(PixelFormat format); - } // namespace VideoCore::Surface diff --git a/src/video_core/texture_cache/format_lookup_table.cpp b/src/video_core/texture_cache/format_lookup_table.cpp index f476f03b0..7d5a75648 100644 --- a/src/video_core/texture_cache/format_lookup_table.cpp +++ b/src/video_core/texture_cache/format_lookup_table.cpp @@ -19,8 +19,6 @@ constexpr auto SNORM = ComponentType::SNORM; constexpr auto UNORM = ComponentType::UNORM; constexpr auto SINT = ComponentType::SINT; constexpr auto UINT = ComponentType::UINT; -constexpr auto SNORM_FORCE_FP16 = ComponentType::SNORM_FORCE_FP16; -constexpr auto UNORM_FORCE_FP16 = ComponentType::UNORM_FORCE_FP16; constexpr auto FLOAT = ComponentType::FLOAT; constexpr bool C = false; // Normal color constexpr bool S = true; // Srgb @@ -41,119 +39,126 @@ struct Table { ComponentType alpha_component; bool is_srgb; }; -constexpr std::array<Table, 78> DefinitionTable = {{ - {TextureFormat::A8R8G8B8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ABGR8U}, - {TextureFormat::A8R8G8B8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::ABGR8S}, - {TextureFormat::A8R8G8B8, C, UINT, UINT, UINT, UINT, PixelFormat::ABGR8UI}, - {TextureFormat::A8R8G8B8, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::RGBA8_SRGB}, +constexpr std::array<Table, 86> DefinitionTable = {{ + {TextureFormat::A8R8G8B8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::A8B8G8R8_UNORM}, + {TextureFormat::A8R8G8B8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::A8B8G8R8_SNORM}, + {TextureFormat::A8R8G8B8, C, UINT, UINT, UINT, UINT, PixelFormat::A8B8G8R8_UINT}, + {TextureFormat::A8R8G8B8, C, SINT, SINT, SINT, SINT, PixelFormat::A8B8G8R8_SINT}, + {TextureFormat::A8R8G8B8, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::A8B8G8R8_SRGB}, - {TextureFormat::B5G6R5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::B5G6R5U}, + {TextureFormat::B5G6R5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::B5G6R5_UNORM}, - {TextureFormat::A2B10G10R10, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::A2B10G10R10U}, + {TextureFormat::A2B10G10R10, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::A2B10G10R10_UNORM}, + {TextureFormat::A2B10G10R10, C, UINT, UINT, UINT, UINT, PixelFormat::A2B10G10R10_UINT}, - {TextureFormat::A1B5G5R5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::A1B5G5R5U}, + {TextureFormat::A1B5G5R5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::A1B5G5R5_UNORM}, - {TextureFormat::A4B4G4R4, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::R4G4B4A4U}, + {TextureFormat::A4B4G4R4, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::A4B4G4R4_UNORM}, - {TextureFormat::R8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::R8U}, - {TextureFormat::R8, C, UINT, UINT, UINT, UINT, PixelFormat::R8UI}, + {TextureFormat::R8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::R8_UNORM}, + {TextureFormat::R8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::R8_SNORM}, + {TextureFormat::R8, C, UINT, UINT, UINT, UINT, PixelFormat::R8_UINT}, + {TextureFormat::R8, C, SINT, SINT, SINT, SINT, PixelFormat::R8_SINT}, - {TextureFormat::G8R8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RG8U}, - {TextureFormat::G8R8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RG8S}, - {TextureFormat::G8R8, C, UINT, UINT, UINT, UINT, PixelFormat::RG8UI}, + {TextureFormat::R8G8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::R8G8_UNORM}, + {TextureFormat::R8G8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::R8G8_SNORM}, + {TextureFormat::R8G8, C, UINT, UINT, UINT, UINT, PixelFormat::R8G8_UINT}, + {TextureFormat::R8G8, C, SINT, SINT, SINT, SINT, PixelFormat::R8G8_SINT}, - {TextureFormat::R16_G16_B16_A16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RGBA16S}, - {TextureFormat::R16_G16_B16_A16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RGBA16U}, - {TextureFormat::R16_G16_B16_A16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::RGBA16F}, - {TextureFormat::R16_G16_B16_A16, C, UINT, UINT, UINT, UINT, PixelFormat::RGBA16UI}, + {TextureFormat::R16G16B16A16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::R16G16B16A16_SNORM}, + {TextureFormat::R16G16B16A16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::R16G16B16A16_UNORM}, + {TextureFormat::R16G16B16A16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R16G16B16A16_FLOAT}, + {TextureFormat::R16G16B16A16, C, UINT, UINT, UINT, UINT, PixelFormat::R16G16B16A16_UINT}, + {TextureFormat::R16G16B16A16, C, SINT, SINT, SINT, SINT, PixelFormat::R16G16B16A16_SINT}, - {TextureFormat::R16_G16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::RG16F}, - {TextureFormat::R16_G16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RG16}, - {TextureFormat::R16_G16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RG16S}, - {TextureFormat::R16_G16, C, UINT, UINT, UINT, UINT, PixelFormat::RG16UI}, - {TextureFormat::R16_G16, C, SINT, SINT, SINT, SINT, PixelFormat::RG16I}, + {TextureFormat::R16G16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R16G16_FLOAT}, + {TextureFormat::R16G16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::R16G16_UNORM}, + {TextureFormat::R16G16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::R16G16_SNORM}, + {TextureFormat::R16G16, C, UINT, UINT, UINT, UINT, PixelFormat::R16G16_UINT}, + {TextureFormat::R16G16, C, SINT, SINT, SINT, SINT, PixelFormat::R16G16_SINT}, - {TextureFormat::R16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R16F}, - {TextureFormat::R16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::R16U}, - {TextureFormat::R16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::R16S}, - {TextureFormat::R16, C, UINT, UINT, UINT, UINT, PixelFormat::R16UI}, - {TextureFormat::R16, C, SINT, SINT, SINT, SINT, PixelFormat::R16I}, + {TextureFormat::R16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R16_FLOAT}, + {TextureFormat::R16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::R16_UNORM}, + {TextureFormat::R16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::R16_SNORM}, + {TextureFormat::R16, C, UINT, UINT, UINT, UINT, PixelFormat::R16_UINT}, + {TextureFormat::R16, C, SINT, SINT, SINT, SINT, PixelFormat::R16_SINT}, - {TextureFormat::BF10GF11RF11, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R11FG11FB10F}, + {TextureFormat::B10G11R11, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::B10G11R11_FLOAT}, - {TextureFormat::R32_G32_B32_A32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::RGBA32F}, - {TextureFormat::R32_G32_B32_A32, C, UINT, UINT, UINT, UINT, PixelFormat::RGBA32UI}, + {TextureFormat::R32G32B32A32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R32G32B32A32_FLOAT}, + {TextureFormat::R32G32B32A32, C, UINT, UINT, UINT, UINT, PixelFormat::R32G32B32A32_UINT}, + {TextureFormat::R32G32B32A32, C, SINT, SINT, SINT, SINT, PixelFormat::R32G32B32A32_SINT}, - {TextureFormat::R32_G32_B32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::RGB32F}, + {TextureFormat::R32G32B32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R32G32B32_FLOAT}, - {TextureFormat::R32_G32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::RG32F}, - {TextureFormat::R32_G32, C, UINT, UINT, UINT, UINT, PixelFormat::RG32UI}, + {TextureFormat::R32G32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R32G32_FLOAT}, + {TextureFormat::R32G32, C, UINT, UINT, UINT, UINT, PixelFormat::R32G32_UINT}, + {TextureFormat::R32G32, C, SINT, SINT, SINT, SINT, PixelFormat::R32G32_SINT}, - {TextureFormat::R32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R32F}, - {TextureFormat::R32, C, UINT, UINT, UINT, UINT, PixelFormat::R32UI}, - {TextureFormat::R32, C, SINT, SINT, SINT, SINT, PixelFormat::R32I}, + {TextureFormat::R32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R32_FLOAT}, + {TextureFormat::R32, C, UINT, UINT, UINT, UINT, PixelFormat::R32_UINT}, + {TextureFormat::R32, C, SINT, SINT, SINT, SINT, PixelFormat::R32_SINT}, - {TextureFormat::E5B9G9R9_SHAREDEXP, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::E5B9G9R9F}, + {TextureFormat::E5B9G9R9, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::E5B9G9R9_FLOAT}, - {TextureFormat::ZF32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::Z32F}, - {TextureFormat::Z16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::Z16}, - {TextureFormat::S8Z24, C, UINT, UNORM, UNORM, UNORM, PixelFormat::S8Z24}, - {TextureFormat::G24R8, C, UINT, UNORM, UNORM, UNORM, PixelFormat::S8Z24}, - {TextureFormat::ZF32_X24S8, C, FLOAT, UINT, UNORM, UNORM, PixelFormat::Z32FS8}, + {TextureFormat::D32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::D32_FLOAT}, + {TextureFormat::D16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::D16_UNORM}, + {TextureFormat::S8D24, C, UINT, UNORM, UNORM, UNORM, PixelFormat::S8_UINT_D24_UNORM}, + {TextureFormat::R8G24, C, UINT, UNORM, UNORM, UNORM, PixelFormat::S8_UINT_D24_UNORM}, + {TextureFormat::D32S8, C, FLOAT, UINT, UNORM, UNORM, PixelFormat::D32_FLOAT_S8_UINT}, - {TextureFormat::DXT1, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::DXT1}, - {TextureFormat::DXT1, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::DXT1_SRGB}, + {TextureFormat::BC1_RGBA, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC1_RGBA_UNORM}, + {TextureFormat::BC1_RGBA, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC1_RGBA_SRGB}, - {TextureFormat::DXT23, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::DXT23}, - {TextureFormat::DXT23, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::DXT23_SRGB}, + {TextureFormat::BC2, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC2_UNORM}, + {TextureFormat::BC2, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC2_SRGB}, - {TextureFormat::DXT45, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::DXT45}, - {TextureFormat::DXT45, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::DXT45_SRGB}, + {TextureFormat::BC3, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC3_UNORM}, + {TextureFormat::BC3, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC3_SRGB}, - // TODO: Use a different pixel format for SNORM - {TextureFormat::DXN1, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::DXN1}, - {TextureFormat::DXN1, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::DXN1}, + {TextureFormat::BC4, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC4_UNORM}, + {TextureFormat::BC4, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::BC4_SNORM}, - {TextureFormat::DXN2, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::DXN2UNORM}, - {TextureFormat::DXN2, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::DXN2SNORM}, + {TextureFormat::BC5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC5_UNORM}, + {TextureFormat::BC5, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::BC5_SNORM}, - {TextureFormat::BC7U, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC7U}, - {TextureFormat::BC7U, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC7U_SRGB}, + {TextureFormat::BC7, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC7_UNORM}, + {TextureFormat::BC7, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::BC7_SRGB}, - {TextureFormat::BC6H_SF16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::BC6H_SF16}, - {TextureFormat::BC6H_UF16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::BC6H_UF16}, + {TextureFormat::BC6H_SFLOAT, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::BC6H_SFLOAT}, + {TextureFormat::BC6H_UFLOAT, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::BC6H_UFLOAT}, - {TextureFormat::ASTC_2D_4X4, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_4X4}, + {TextureFormat::ASTC_2D_4X4, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_4X4_UNORM}, {TextureFormat::ASTC_2D_4X4, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_4X4_SRGB}, - {TextureFormat::ASTC_2D_5X4, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_5X4}, + {TextureFormat::ASTC_2D_5X4, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_5X4_UNORM}, {TextureFormat::ASTC_2D_5X4, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_5X4_SRGB}, - {TextureFormat::ASTC_2D_5X5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_5X5}, + {TextureFormat::ASTC_2D_5X5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_5X5_UNORM}, {TextureFormat::ASTC_2D_5X5, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_5X5_SRGB}, - {TextureFormat::ASTC_2D_8X8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_8X8}, + {TextureFormat::ASTC_2D_8X8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_8X8_UNORM}, {TextureFormat::ASTC_2D_8X8, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_8X8_SRGB}, - {TextureFormat::ASTC_2D_8X5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_8X5}, + {TextureFormat::ASTC_2D_8X5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_8X5_UNORM}, {TextureFormat::ASTC_2D_8X5, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_8X5_SRGB}, - {TextureFormat::ASTC_2D_10X8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_10X8}, + {TextureFormat::ASTC_2D_10X8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_10X8_UNORM}, {TextureFormat::ASTC_2D_10X8, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_10X8_SRGB}, - {TextureFormat::ASTC_2D_6X6, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_6X6}, + {TextureFormat::ASTC_2D_6X6, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_6X6_UNORM}, {TextureFormat::ASTC_2D_6X6, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_6X6_SRGB}, - {TextureFormat::ASTC_2D_10X10, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_10X10}, + {TextureFormat::ASTC_2D_10X10, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_10X10_UNORM}, {TextureFormat::ASTC_2D_10X10, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_10X10_SRGB}, - {TextureFormat::ASTC_2D_12X12, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_12X12}, + {TextureFormat::ASTC_2D_12X12, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_12X12_UNORM}, {TextureFormat::ASTC_2D_12X12, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_12X12_SRGB}, - {TextureFormat::ASTC_2D_8X6, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_8X6}, + {TextureFormat::ASTC_2D_8X6, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_8X6_UNORM}, {TextureFormat::ASTC_2D_8X6, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_8X6_SRGB}, - {TextureFormat::ASTC_2D_6X5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_6X5}, + {TextureFormat::ASTC_2D_6X5, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_6X5_UNORM}, {TextureFormat::ASTC_2D_6X5, S, UNORM, UNORM, UNORM, UNORM, PixelFormat::ASTC_2D_6X5_SRGB}, }}; @@ -184,7 +189,7 @@ PixelFormat FormatLookupTable::GetPixelFormat(TextureFormat format, bool is_srgb static_cast<int>(format), is_srgb, static_cast<int>(red_component), static_cast<int>(green_component), static_cast<int>(blue_component), static_cast<int>(alpha_component)); - return PixelFormat::ABGR8U; + return PixelFormat::A8B8G8R8_UNORM; } void FormatLookupTable::Set(TextureFormat format, bool is_srgb, ComponentType red_component, diff --git a/src/video_core/texture_cache/surface_base.cpp b/src/video_core/texture_cache/surface_base.cpp index 0caf3b4f0..dfcf36e0b 100644 --- a/src/video_core/texture_cache/surface_base.cpp +++ b/src/video_core/texture_cache/surface_base.cpp @@ -228,7 +228,7 @@ void SurfaceBaseImpl::LoadBuffer(Tegra::MemoryManager& memory_manager, } } - if (!is_converted && params.pixel_format != PixelFormat::S8Z24) { + if (!is_converted && params.pixel_format != PixelFormat::S8_UINT_D24_UNORM) { return; } diff --git a/src/video_core/texture_cache/surface_params.cpp b/src/video_core/texture_cache/surface_params.cpp index 0ecf42368..e614a92df 100644 --- a/src/video_core/texture_cache/surface_params.cpp +++ b/src/video_core/texture_cache/surface_params.cpp @@ -74,21 +74,21 @@ SurfaceParams SurfaceParams::CreateForTexture(const FormatLookupTable& lookup_ta SurfaceParams params; params.is_tiled = tic.IsTiled(); params.srgb_conversion = tic.IsSrgbConversionEnabled(); - params.block_width = params.is_tiled ? tic.BlockWidth() : 0, - params.block_height = params.is_tiled ? tic.BlockHeight() : 0, - params.block_depth = params.is_tiled ? tic.BlockDepth() : 0, + params.block_width = params.is_tiled ? tic.BlockWidth() : 0; + params.block_height = params.is_tiled ? tic.BlockHeight() : 0; + params.block_depth = params.is_tiled ? tic.BlockDepth() : 0; params.tile_width_spacing = params.is_tiled ? (1 << tic.tile_width_spacing.Value()) : 1; params.pixel_format = lookup_table.GetPixelFormat( tic.format, params.srgb_conversion, tic.r_type, tic.g_type, tic.b_type, tic.a_type); params.type = GetFormatType(params.pixel_format); if (entry.is_shadow && params.type == SurfaceType::ColorTexture) { switch (params.pixel_format) { - case PixelFormat::R16U: - case PixelFormat::R16F: - params.pixel_format = PixelFormat::Z16; + case PixelFormat::R16_UNORM: + case PixelFormat::R16_FLOAT: + params.pixel_format = PixelFormat::D16_UNORM; break; - case PixelFormat::R32F: - params.pixel_format = PixelFormat::Z32F; + case PixelFormat::R32_FLOAT: + params.pixel_format = PixelFormat::D32_FLOAT; break; default: UNIMPLEMENTED_MSG("Unimplemented shadow convert format: {}", @@ -129,14 +129,13 @@ SurfaceParams SurfaceParams::CreateForImage(const FormatLookupTable& lookup_tabl SurfaceParams params; params.is_tiled = tic.IsTiled(); params.srgb_conversion = tic.IsSrgbConversionEnabled(); - params.block_width = params.is_tiled ? tic.BlockWidth() : 0, - params.block_height = params.is_tiled ? tic.BlockHeight() : 0, - params.block_depth = params.is_tiled ? tic.BlockDepth() : 0, + params.block_width = params.is_tiled ? tic.BlockWidth() : 0; + params.block_height = params.is_tiled ? tic.BlockHeight() : 0; + params.block_depth = params.is_tiled ? tic.BlockDepth() : 0; params.tile_width_spacing = params.is_tiled ? (1 << tic.tile_width_spacing.Value()) : 1; params.pixel_format = lookup_table.GetPixelFormat( tic.format, params.srgb_conversion, tic.r_type, tic.g_type, tic.b_type, tic.a_type); params.type = GetFormatType(params.pixel_format); - params.type = GetFormatType(params.pixel_format); params.target = ImageTypeToSurfaceTarget(entry.type); // TODO: on 1DBuffer we should use the tic info. if (tic.IsBuffer()) { @@ -166,27 +165,30 @@ SurfaceParams SurfaceParams::CreateForImage(const FormatLookupTable& lookup_tabl SurfaceParams SurfaceParams::CreateForDepthBuffer(Core::System& system) { const auto& regs = system.GPU().Maxwell3D().regs; - SurfaceParams params; - params.is_tiled = regs.zeta.memory_layout.type == - Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear; - params.srgb_conversion = false; - params.block_width = std::min(regs.zeta.memory_layout.block_width.Value(), 5U); - params.block_height = std::min(regs.zeta.memory_layout.block_height.Value(), 5U); - params.block_depth = std::min(regs.zeta.memory_layout.block_depth.Value(), 5U); - params.tile_width_spacing = 1; - params.pixel_format = PixelFormatFromDepthFormat(regs.zeta.format); - params.type = GetFormatType(params.pixel_format); - params.width = regs.zeta_width; - params.height = regs.zeta_height; - params.pitch = 0; - params.num_levels = 1; - params.emulated_levels = 1; - const bool is_layered = regs.zeta_layers > 1 && params.block_depth == 0; - params.is_layered = is_layered; - params.target = is_layered ? SurfaceTarget::Texture2DArray : SurfaceTarget::Texture2D; - params.depth = is_layered ? regs.zeta_layers.Value() : 1U; - return params; + const auto block_depth = std::min(regs.zeta.memory_layout.block_depth.Value(), 5U); + const bool is_layered = regs.zeta_layers > 1 && block_depth == 0; + const auto pixel_format = PixelFormatFromDepthFormat(regs.zeta.format); + + return { + .is_tiled = regs.zeta.memory_layout.type == + Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear, + .srgb_conversion = false, + .is_layered = is_layered, + .block_width = std::min(regs.zeta.memory_layout.block_width.Value(), 5U), + .block_height = std::min(regs.zeta.memory_layout.block_height.Value(), 5U), + .block_depth = block_depth, + .tile_width_spacing = 1, + .width = regs.zeta_width, + .height = regs.zeta_height, + .depth = is_layered ? regs.zeta_layers.Value() : 1U, + .pitch = 0, + .num_levels = 1, + .emulated_levels = 1, + .pixel_format = pixel_format, + .type = GetFormatType(pixel_format), + .target = is_layered ? SurfaceTarget::Texture2DArray : SurfaceTarget::Texture2D, + }; } SurfaceParams SurfaceParams::CreateForFramebuffer(Core::System& system, std::size_t index) { @@ -194,8 +196,8 @@ SurfaceParams SurfaceParams::CreateForFramebuffer(Core::System& system, std::siz SurfaceParams params; params.is_tiled = config.memory_layout.type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear; - params.srgb_conversion = config.format == Tegra::RenderTargetFormat::BGRA8_SRGB || - config.format == Tegra::RenderTargetFormat::RGBA8_SRGB; + params.srgb_conversion = config.format == Tegra::RenderTargetFormat::B8G8R8A8_SRGB || + config.format == Tegra::RenderTargetFormat::A8B8G8R8_SRGB; params.block_width = config.memory_layout.block_width; params.block_height = config.memory_layout.block_height; params.block_depth = config.memory_layout.block_depth; @@ -232,24 +234,29 @@ SurfaceParams SurfaceParams::CreateForFramebuffer(Core::System& system, std::siz SurfaceParams SurfaceParams::CreateForFermiCopySurface( const Tegra::Engines::Fermi2D::Regs::Surface& config) { - SurfaceParams params{}; - params.is_tiled = !config.linear; - params.srgb_conversion = config.format == Tegra::RenderTargetFormat::BGRA8_SRGB || - config.format == Tegra::RenderTargetFormat::RGBA8_SRGB; - params.block_width = params.is_tiled ? std::min(config.BlockWidth(), 5U) : 0, - params.block_height = params.is_tiled ? std::min(config.BlockHeight(), 5U) : 0, - params.block_depth = params.is_tiled ? std::min(config.BlockDepth(), 5U) : 0, - params.tile_width_spacing = 1; - params.pixel_format = PixelFormatFromRenderTargetFormat(config.format); - params.type = GetFormatType(params.pixel_format); - params.width = config.width; - params.height = config.height; - params.pitch = config.pitch; - // TODO(Rodrigo): Try to guess texture arrays from parameters - params.target = SurfaceTarget::Texture2D; - params.depth = 1; - params.num_levels = 1; - params.emulated_levels = 1; + const bool is_tiled = !config.linear; + const auto pixel_format = PixelFormatFromRenderTargetFormat(config.format); + + SurfaceParams params{ + .is_tiled = is_tiled, + .srgb_conversion = config.format == Tegra::RenderTargetFormat::B8G8R8A8_SRGB || + config.format == Tegra::RenderTargetFormat::A8B8G8R8_SRGB, + .block_width = is_tiled ? std::min(config.BlockWidth(), 5U) : 0U, + .block_height = is_tiled ? std::min(config.BlockHeight(), 5U) : 0U, + .block_depth = is_tiled ? std::min(config.BlockDepth(), 5U) : 0U, + .tile_width_spacing = 1, + .width = config.width, + .height = config.height, + .depth = 1, + .pitch = config.pitch, + .num_levels = 1, + .emulated_levels = 1, + .pixel_format = pixel_format, + .type = GetFormatType(pixel_format), + // TODO(Rodrigo): Try to guess texture arrays from parameters + .target = SurfaceTarget::Texture2D, + }; + params.is_layered = params.IsLayered(); return params; } diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index cdcddb225..96c4e4cc2 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -373,9 +373,9 @@ protected: siblings_table[static_cast<std::size_t>(b)] = a; }; std::fill(siblings_table.begin(), siblings_table.end(), PixelFormat::Invalid); - make_siblings(PixelFormat::Z16, PixelFormat::R16U); - make_siblings(PixelFormat::Z32F, PixelFormat::R32F); - make_siblings(PixelFormat::Z32FS8, PixelFormat::RG32F); + make_siblings(PixelFormat::D16_UNORM, PixelFormat::R16_UNORM); + make_siblings(PixelFormat::D32_FLOAT, PixelFormat::R32_FLOAT); + make_siblings(PixelFormat::D32_FLOAT_S8_UINT, PixelFormat::R32G32_FLOAT); sampled_textures.reserve(64); } @@ -1031,7 +1031,7 @@ private: params.pitch = 4; params.num_levels = 1; params.emulated_levels = 1; - params.pixel_format = VideoCore::Surface::PixelFormat::R8U; + params.pixel_format = VideoCore::Surface::PixelFormat::R8_UNORM; params.type = VideoCore::Surface::SurfaceType::ColorTexture; auto surface = CreateSurface(0ULL, params); invalid_memory.resize(surface->GetHostSizeInBytes(), 0U); diff --git a/src/video_core/textures/convert.cpp b/src/video_core/textures/convert.cpp index f3efa7eb0..962921483 100644 --- a/src/video_core/textures/convert.cpp +++ b/src/video_core/textures/convert.cpp @@ -35,7 +35,7 @@ void SwapS8Z24ToZ24S8(u8* data, u32 width, u32 height) { S8Z24 s8z24_pixel{}; Z24S8 z24s8_pixel{}; constexpr auto bpp{ - VideoCore::Surface::GetBytesPerPixel(VideoCore::Surface::PixelFormat::S8Z24)}; + VideoCore::Surface::GetBytesPerPixel(VideoCore::Surface::PixelFormat::S8_UINT_D24_UNORM)}; for (std::size_t y = 0; y < height; ++y) { for (std::size_t x = 0; x < width; ++x) { const std::size_t offset{bpp * (y * width + x)}; @@ -73,7 +73,7 @@ void ConvertFromGuestToHost(u8* in_data, u8* out_data, PixelFormat pixel_format, in_data, width, height, depth, block_width, block_height); std::copy(rgba8_data.begin(), rgba8_data.end(), out_data); - } else if (convert_s8z24 && pixel_format == PixelFormat::S8Z24) { + } else if (convert_s8z24 && pixel_format == PixelFormat::S8_UINT_D24_UNORM) { Tegra::Texture::ConvertS8Z24ToZ24S8(in_data, width, height); } } @@ -85,7 +85,7 @@ void ConvertFromHostToGuest(u8* data, PixelFormat pixel_format, u32 width, u32 h static_cast<u32>(pixel_format)); UNREACHABLE(); - } else if (convert_s8z24 && pixel_format == PixelFormat::S8Z24) { + } else if (convert_s8z24 && pixel_format == PixelFormat::S8_UINT_D24_UNORM) { Tegra::Texture::ConvertZ24S8ToS8Z24(data, width, height); } } diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 98beabef1..474ae620a 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -184,53 +184,6 @@ void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel, } } -u32 BytesPerPixel(TextureFormat format) { - switch (format) { - case TextureFormat::DXT1: - case TextureFormat::DXN1: - // In this case a 'pixel' actually refers to a 4x4 tile. - return 8; - case TextureFormat::DXT23: - case TextureFormat::DXT45: - case TextureFormat::DXN2: - case TextureFormat::BC7U: - case TextureFormat::BC6H_UF16: - case TextureFormat::BC6H_SF16: - // In this case a 'pixel' actually refers to a 4x4 tile. - return 16; - case TextureFormat::R32_G32_B32: - return 12; - case TextureFormat::ASTC_2D_4X4: - case TextureFormat::ASTC_2D_5X4: - case TextureFormat::ASTC_2D_8X8: - case TextureFormat::ASTC_2D_8X5: - case TextureFormat::ASTC_2D_10X8: - case TextureFormat::ASTC_2D_5X5: - case TextureFormat::A8R8G8B8: - case TextureFormat::A2B10G10R10: - case TextureFormat::BF10GF11RF11: - case TextureFormat::R32: - case TextureFormat::R16_G16: - return 4; - case TextureFormat::A1B5G5R5: - case TextureFormat::B5G6R5: - case TextureFormat::G8R8: - case TextureFormat::R16: - return 2; - case TextureFormat::R8: - return 1; - case TextureFormat::R16_G16_B16_A16: - return 8; - case TextureFormat::R32_G32_B32_A32: - return 16; - case TextureFormat::R32_G32: - return 8; - default: - UNIMPLEMENTED_MSG("Format not implemented"); - return 1; - } -} - void UnswizzleTexture(u8* const unswizzled_data, u8* address, u32 tile_size_x, u32 tile_size_y, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth, u32 width_spacing) { @@ -348,48 +301,6 @@ void SwizzleKepler(const u32 width, const u32 height, const u32 dst_x, const u32 } } -std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat format, u32 width, - u32 height) { - std::vector<u8> rgba_data; - - // TODO(Subv): Implement. - switch (format) { - case TextureFormat::DXT1: - case TextureFormat::DXT23: - case TextureFormat::DXT45: - case TextureFormat::DXN1: - case TextureFormat::DXN2: - case TextureFormat::BC7U: - case TextureFormat::BC6H_UF16: - case TextureFormat::BC6H_SF16: - case TextureFormat::ASTC_2D_4X4: - case TextureFormat::ASTC_2D_8X8: - case TextureFormat::ASTC_2D_5X5: - case TextureFormat::ASTC_2D_10X8: - case TextureFormat::A8R8G8B8: - case TextureFormat::A2B10G10R10: - case TextureFormat::A1B5G5R5: - case TextureFormat::B5G6R5: - case TextureFormat::R8: - case TextureFormat::G8R8: - case TextureFormat::BF10GF11RF11: - case TextureFormat::R32_G32_B32_A32: - case TextureFormat::R32_G32: - case TextureFormat::R32: - case TextureFormat::R16: - case TextureFormat::R16_G16: - case TextureFormat::R32_G32_B32: - // TODO(Subv): For the time being just forward the same data without any decoding. - rgba_data = texture_data; - break; - default: - UNIMPLEMENTED_MSG("Format not implemented"); - break; - } - - return rgba_data; -} - std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth) { if (tiled) { diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h index 232b696b3..d6fe35d37 100644 --- a/src/video_core/textures/decoders.h +++ b/src/video_core/textures/decoders.h @@ -38,10 +38,6 @@ void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel, u32 out_bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, bool unswizzle, u32 block_height, u32 block_depth, u32 width_spacing); -/// Decodes an unswizzled texture into a A8R8G8B8 texture. -std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat format, u32 width, - u32 height); - /// This function calculates the correct size of a texture depending if it's tiled or not. std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth); diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h index eba05aced..0574fef12 100644 --- a/src/video_core/textures/texture.h +++ b/src/video_core/textures/texture.h @@ -12,10 +12,10 @@ namespace Tegra::Texture { enum class TextureFormat : u32 { - R32_G32_B32_A32 = 0x01, - R32_G32_B32 = 0x02, - R16_G16_B16_A16 = 0x03, - R32_G32 = 0x04, + R32G32B32A32 = 0x01, + R32G32B32 = 0x02, + R16G16B16A16 = 0x03, + R32G32 = 0x04, R32_B24G8 = 0x05, ETC2_RGB = 0x06, X8B8G8R8 = 0x07, @@ -23,19 +23,19 @@ enum class TextureFormat : u32 { A2B10G10R10 = 0x09, ETC2_RGB_PTA = 0x0a, ETC2_RGBA = 0x0b, - R16_G16 = 0x0c, - G8R24 = 0x0d, - G24R8 = 0x0e, + R16G16 = 0x0c, + R24G8 = 0x0d, + R8G24 = 0x0e, R32 = 0x0f, - BC6H_SF16 = 0x10, - BC6H_UF16 = 0x11, + BC6H_SFLOAT = 0x10, + BC6H_UFLOAT = 0x11, A4B4G4R4 = 0x12, A5B5G5R1 = 0x13, A1B5G5R5 = 0x14, B5G6R5 = 0x15, B6G5R5 = 0x16, - BC7U = 0x17, - G8R8 = 0x18, + BC7 = 0x17, + R8G8 = 0x18, EAC = 0x19, EACX2 = 0x1a, R16 = 0x1b, @@ -43,23 +43,23 @@ enum class TextureFormat : u32 { R8 = 0x1d, G4R4 = 0x1e, R1 = 0x1f, - E5B9G9R9_SHAREDEXP = 0x20, - BF10GF11RF11 = 0x21, + E5B9G9R9 = 0x20, + B10G11R11 = 0x21, G8B8G8R8 = 0x22, B8G8R8G8 = 0x23, - DXT1 = 0x24, - DXT23 = 0x25, - DXT45 = 0x26, - DXN1 = 0x27, - DXN2 = 0x28, - S8Z24 = 0x29, + BC1_RGBA = 0x24, + BC2 = 0x25, + BC3 = 0x26, + BC4 = 0x27, + BC5 = 0x28, + S8D24 = 0x29, X8Z24 = 0x2a, - Z24S8 = 0x2b, + D24S8 = 0x2b, X4V4Z24__COV4R4V = 0x2c, X4V4Z24__COV8R8V = 0x2d, V8Z24__COV4R12V = 0x2e, - ZF32 = 0x2f, - ZF32_X24S8 = 0x30, + D32 = 0x2f, + D32S8 = 0x30, X8Z24_X20V4S8__COV4R4V = 0x31, X8Z24_X20V4S8__COV8R8V = 0x32, ZF32_X20V4X8__COV4R4V = 0x33, @@ -69,7 +69,7 @@ enum class TextureFormat : u32 { X8Z24_X16V8S8__COV4R12V = 0x37, ZF32_X16V8X8__COV4R12V = 0x38, ZF32_X16V8S8__COV4R12V = 0x39, - Z16 = 0x3a, + D16 = 0x3a, V8Z24__COV8R24V = 0x3b, X8Z24_X16V8S8__COV8R24V = 0x3c, ZF32_X16V8X8__COV8R24V = 0x3d, @@ -375,7 +375,4 @@ struct FullTextureInfo { TSCEntry tsc; }; -/// Returns the number of bytes per pixel of the input texture format. -u32 BytesPerPixel(TextureFormat format); - } // namespace Tegra::Texture |