From 139ea93512aeead8a4aee3910a3de86eb109a838 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Fri, 5 Nov 2021 15:52:31 +0100 Subject: VideoCore: implement channels on gpu caches. --- src/video_core/texture_cache/image_base.h | 3 + src/video_core/texture_cache/texture_cache.h | 209 ++++++++++++++-------- src/video_core/texture_cache/texture_cache_base.h | 73 ++++++-- 3 files changed, 195 insertions(+), 90 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/video_core/texture_cache/image_base.h b/src/video_core/texture_cache/image_base.h index 1f85ec9da..620565684 100644 --- a/src/video_core/texture_cache/image_base.h +++ b/src/video_core/texture_cache/image_base.h @@ -88,6 +88,9 @@ struct ImageBase { u32 scale_rating = 0; u64 scale_tick = 0; bool has_scaled = false; + + size_t channel = 0; + ImageFlagBits flags = ImageFlagBits::CpuModified; GPUVAddr gpu_addr = 0; diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 1dbe01bc0..2731aead0 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -7,6 +7,7 @@ #include "common/alignment.h" #include "common/settings.h" +#include "video_core/control/channel_state.h" #include "video_core/dirty_flags.h" #include "video_core/engines/kepler_compute.h" #include "video_core/texture_cache/image_view_base.h" @@ -29,12 +30,8 @@ using VideoCore::Surface::SurfaceType; using namespace Common::Literals; template -TextureCache

::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface& rasterizer_, - Tegra::Engines::Maxwell3D& maxwell3d_, - Tegra::Engines::KeplerCompute& kepler_compute_, - Tegra::MemoryManager& gpu_memory_) - : runtime{runtime_}, rasterizer{rasterizer_}, maxwell3d{maxwell3d_}, - kepler_compute{kepler_compute_}, gpu_memory{gpu_memory_} { +TextureCache

::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface& rasterizer_) + : runtime{runtime_}, rasterizer{rasterizer_} { // Configure null sampler TSCEntry sampler_descriptor{}; sampler_descriptor.min_filter.Assign(Tegra::Texture::TextureFilter::Linear); @@ -42,6 +39,13 @@ TextureCache

::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface& sampler_descriptor.mipmap_filter.Assign(Tegra::Texture::TextureMipmapFilter::Linear); sampler_descriptor.cubemap_anisotropy.Assign(1); + // Setup channels + current_channel_id = UNSET_CHANNEL; + state = nullptr; + maxwell3d = nullptr; + kepler_compute = nullptr; + gpu_memory = nullptr; + // Make sure the first index is reserved for the null resources // This way the null resource becomes a compile time constant void(slot_images.insert(NullImageParams{})); @@ -93,7 +97,7 @@ void TextureCache

::RunGarbageCollector() { const auto copies = FullDownloadCopies(image.info); image.DownloadMemory(map, copies); runtime.Finish(); - SwizzleImage(gpu_memory, image.gpu_addr, image.info, copies, map.mapped_span); + SwizzleImage(*gpu_memory, image.gpu_addr, image.info, copies, map.mapped_span); } if (True(image.flags & ImageFlagBits::Tracked)) { UntrackImage(image, image_id); @@ -152,22 +156,23 @@ void TextureCache

::MarkModification(ImageId id) noexcept { template template void TextureCache

::FillGraphicsImageViews(std::span views) { - FillImageViews(graphics_image_table, graphics_image_view_ids, views); + FillImageViews(state->graphics_image_table, state->graphics_image_view_ids, + views); } template void TextureCache

::FillComputeImageViews(std::span views) { - FillImageViews(compute_image_table, compute_image_view_ids, views); + FillImageViews(state->compute_image_table, state->compute_image_view_ids, views); } template typename P::Sampler* TextureCache

::GetGraphicsSampler(u32 index) { - if (index > graphics_sampler_table.Limit()) { + if (index > state->graphics_sampler_table.Limit()) { LOG_DEBUG(HW_GPU, "Invalid sampler index={}", index); return &slot_samplers[NULL_SAMPLER_ID]; } - const auto [descriptor, is_new] = graphics_sampler_table.Read(index); - SamplerId& id = graphics_sampler_ids[index]; + const auto [descriptor, is_new] = state->graphics_sampler_table.Read(index); + SamplerId& id = state->graphics_sampler_ids[index]; if (is_new) { id = FindSampler(descriptor); } @@ -176,12 +181,12 @@ typename P::Sampler* TextureCache

::GetGraphicsSampler(u32 index) { template typename P::Sampler* TextureCache

::GetComputeSampler(u32 index) { - if (index > compute_sampler_table.Limit()) { + if (index > state->compute_sampler_table.Limit()) { LOG_DEBUG(HW_GPU, "Invalid sampler index={}", index); return &slot_samplers[NULL_SAMPLER_ID]; } - const auto [descriptor, is_new] = compute_sampler_table.Read(index); - SamplerId& id = compute_sampler_ids[index]; + const auto [descriptor, is_new] = state->compute_sampler_table.Read(index); + SamplerId& id = state->compute_sampler_ids[index]; if (is_new) { id = FindSampler(descriptor); } @@ -191,34 +196,34 @@ typename P::Sampler* TextureCache

::GetComputeSampler(u32 index) { template void TextureCache

::SynchronizeGraphicsDescriptors() { using SamplerIndex = Tegra::Engines::Maxwell3D::Regs::SamplerIndex; - const bool linked_tsc = maxwell3d.regs.sampler_index == SamplerIndex::ViaHeaderIndex; - const u32 tic_limit = maxwell3d.regs.tic.limit; - const u32 tsc_limit = linked_tsc ? tic_limit : maxwell3d.regs.tsc.limit; - if (graphics_sampler_table.Synchornize(maxwell3d.regs.tsc.Address(), tsc_limit)) { - graphics_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID); + const bool linked_tsc = maxwell3d->regs.sampler_index == SamplerIndex::ViaHeaderIndex; + const u32 tic_limit = maxwell3d->regs.tic.limit; + const u32 tsc_limit = linked_tsc ? tic_limit : maxwell3d->regs.tsc.limit; + if (state->graphics_sampler_table.Synchornize(maxwell3d->regs.tsc.Address(), tsc_limit)) { + state->graphics_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID); } - if (graphics_image_table.Synchornize(maxwell3d.regs.tic.Address(), tic_limit)) { - graphics_image_view_ids.resize(tic_limit + 1, CORRUPT_ID); + if (state->graphics_image_table.Synchornize(maxwell3d->regs.tic.Address(), tic_limit)) { + state->graphics_image_view_ids.resize(tic_limit + 1, CORRUPT_ID); } } template void TextureCache

::SynchronizeComputeDescriptors() { - const bool linked_tsc = kepler_compute.launch_description.linked_tsc; - const u32 tic_limit = kepler_compute.regs.tic.limit; - const u32 tsc_limit = linked_tsc ? tic_limit : kepler_compute.regs.tsc.limit; - const GPUVAddr tsc_gpu_addr = kepler_compute.regs.tsc.Address(); - if (compute_sampler_table.Synchornize(tsc_gpu_addr, tsc_limit)) { - compute_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID); + const bool linked_tsc = kepler_compute->launch_description.linked_tsc; + const u32 tic_limit = kepler_compute->regs.tic.limit; + const u32 tsc_limit = linked_tsc ? tic_limit : kepler_compute->regs.tsc.limit; + const GPUVAddr tsc_gpu_addr = kepler_compute->regs.tsc.Address(); + if (state->compute_sampler_table.Synchornize(tsc_gpu_addr, tsc_limit)) { + state->compute_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID); } - if (compute_image_table.Synchornize(kepler_compute.regs.tic.Address(), tic_limit)) { - compute_image_view_ids.resize(tic_limit + 1, CORRUPT_ID); + if (state->compute_image_table.Synchornize(kepler_compute->regs.tic.Address(), tic_limit)) { + state->compute_image_view_ids.resize(tic_limit + 1, CORRUPT_ID); } } template bool TextureCache

::RescaleRenderTargets(bool is_clear) { - auto& flags = maxwell3d.dirty.flags; + auto& flags = maxwell3d->dirty.flags; u32 scale_rating = 0; bool rescaled = false; std::array tmp_color_images{}; @@ -315,7 +320,7 @@ bool TextureCache

::RescaleRenderTargets(bool is_clear) { template void TextureCache

::UpdateRenderTargets(bool is_clear) { using namespace VideoCommon::Dirty; - auto& flags = maxwell3d.dirty.flags; + auto& flags = maxwell3d->dirty.flags; if (!flags[Dirty::RenderTargets]) { for (size_t index = 0; index < NUM_RT; ++index) { ImageViewId& color_buffer_id = render_targets.color_buffer_ids[index]; @@ -342,7 +347,7 @@ void TextureCache

::UpdateRenderTargets(bool is_clear) { PrepareImageView(depth_buffer_id, true, is_clear && IsFullClear(depth_buffer_id)); for (size_t index = 0; index < NUM_RT; ++index) { - render_targets.draw_buffers[index] = static_cast(maxwell3d.regs.rt_control.Map(index)); + render_targets.draw_buffers[index] = static_cast(maxwell3d->regs.rt_control.Map(index)); } u32 up_scale = 1; u32 down_shift = 0; @@ -351,8 +356,8 @@ void TextureCache

::UpdateRenderTargets(bool is_clear) { down_shift = Settings::values.resolution_info.down_shift; } render_targets.size = Extent2D{ - (maxwell3d.regs.render_area.width * up_scale) >> down_shift, - (maxwell3d.regs.render_area.height * up_scale) >> down_shift, + (maxwell3d->regs.render_area.width * up_scale) >> down_shift, + (maxwell3d->regs.render_area.height * up_scale) >> down_shift, }; flags[Dirty::DepthBiasGlobal] = true; @@ -458,7 +463,7 @@ void TextureCache

::DownloadMemory(VAddr cpu_addr, size_t size) { const auto copies = FullDownloadCopies(image.info); image.DownloadMemory(map, copies); runtime.Finish(); - SwizzleImage(gpu_memory, image.gpu_addr, image.info, copies, map.mapped_span); + SwizzleImage(*gpu_memory, image.gpu_addr, image.info, copies, map.mapped_span); } } @@ -655,7 +660,7 @@ void TextureCache

::PopAsyncFlushes() { for (const ImageId image_id : download_ids) { const ImageBase& image = slot_images[image_id]; const auto copies = FullDownloadCopies(image.info); - SwizzleImage(gpu_memory, image.gpu_addr, image.info, copies, download_span); + SwizzleImage(*gpu_memory, image.gpu_addr, image.info, copies, download_span); download_map.offset += image.unswizzled_size_bytes; download_span = download_span.subspan(image.unswizzled_size_bytes); } @@ -714,26 +719,26 @@ void TextureCache

::UploadImageContents(Image& image, StagingBuffer& staging) const GPUVAddr gpu_addr = image.gpu_addr; if (True(image.flags & ImageFlagBits::AcceleratedUpload)) { - gpu_memory.ReadBlockUnsafe(gpu_addr, mapped_span.data(), mapped_span.size_bytes()); + gpu_memory->ReadBlockUnsafe(gpu_addr, mapped_span.data(), mapped_span.size_bytes()); const auto uploads = FullUploadSwizzles(image.info); runtime.AccelerateImageUpload(image, staging, uploads); } else if (True(image.flags & ImageFlagBits::Converted)) { std::vector unswizzled_data(image.unswizzled_size_bytes); - auto copies = UnswizzleImage(gpu_memory, gpu_addr, image.info, unswizzled_data); + auto copies = UnswizzleImage(*gpu_memory, gpu_addr, image.info, unswizzled_data); ConvertImage(unswizzled_data, image.info, mapped_span, copies); image.UploadMemory(staging, copies); } else { - const auto copies = UnswizzleImage(gpu_memory, gpu_addr, image.info, mapped_span); + const auto copies = UnswizzleImage(*gpu_memory, gpu_addr, image.info, mapped_span); image.UploadMemory(staging, copies); } } template ImageViewId TextureCache

::FindImageView(const TICEntry& config) { - if (!IsValidEntry(gpu_memory, config)) { + if (!IsValidEntry(*gpu_memory, config)) { return NULL_IMAGE_VIEW_ID; } - const auto [pair, is_new] = image_views.try_emplace(config); + const auto [pair, is_new] = state->image_views.try_emplace(config); ImageViewId& image_view_id = pair->second; if (is_new) { image_view_id = CreateImageView(config); @@ -777,9 +782,9 @@ ImageId TextureCache

::FindOrInsertImage(const ImageInfo& info, GPUVAddr gpu_a template ImageId TextureCache

::FindImage(const ImageInfo& info, GPUVAddr gpu_addr, RelaxedOptions options) { - std::optional cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr); + std::optional cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr); if (!cpu_addr) { - cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr, CalculateGuestSizeInBytes(info)); + cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr, CalculateGuestSizeInBytes(info)); if (!cpu_addr) { return ImageId{}; } @@ -860,7 +865,7 @@ void TextureCache

::InvalidateScale(Image& image) { image.scale_tick = frame_tick + 1; } const std::span image_view_ids = image.image_view_ids; - auto& dirty = maxwell3d.dirty.flags; + auto& dirty = maxwell3d->dirty.flags; dirty[Dirty::RenderTargets] = true; dirty[Dirty::ZetaBuffer] = true; for (size_t rt = 0; rt < NUM_RT; ++rt) { @@ -881,11 +886,11 @@ void TextureCache

::InvalidateScale(Image& image) { image.image_view_ids.clear(); image.image_view_infos.clear(); if constexpr (ENABLE_VALIDATION) { - std::ranges::fill(graphics_image_view_ids, CORRUPT_ID); - std::ranges::fill(compute_image_view_ids, CORRUPT_ID); + std::ranges::fill(state->graphics_image_view_ids, CORRUPT_ID); + std::ranges::fill(state->compute_image_view_ids, CORRUPT_ID); } - graphics_image_table.Invalidate(); - compute_image_table.Invalidate(); + state->graphics_image_table.Invalidate(); + state->compute_image_table.Invalidate(); has_deleted_images = true; } @@ -929,10 +934,10 @@ bool TextureCache

::ScaleDown(Image& image) { template ImageId TextureCache

::InsertImage(const ImageInfo& info, GPUVAddr gpu_addr, RelaxedOptions options) { - std::optional cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr); + std::optional cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr); if (!cpu_addr) { const auto size = CalculateGuestSizeInBytes(info); - cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr, size); + cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr, size); if (!cpu_addr) { const VAddr fake_addr = ~(1ULL << 40ULL) + virtual_invalid_space; virtual_invalid_space += Common::AlignUp(size, 32); @@ -1050,7 +1055,7 @@ ImageId TextureCache

::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VA const ImageId new_image_id = slot_images.insert(runtime, new_info, gpu_addr, cpu_addr); Image& new_image = slot_images[new_image_id]; - if (!gpu_memory.IsContinousRange(new_image.gpu_addr, new_image.guest_size_bytes)) { + if (!gpu_memory->IsContinousRange(new_image.gpu_addr, new_image.guest_size_bytes)) { new_image.flags |= ImageFlagBits::Sparse; } @@ -1192,7 +1197,7 @@ SamplerId TextureCache

::FindSampler(const TSCEntry& config) { if (std::ranges::all_of(config.raw, [](u64 value) { return value == 0; })) { return NULL_SAMPLER_ID; } - const auto [pair, is_new] = samplers.try_emplace(config); + const auto [pair, is_new] = state->samplers.try_emplace(config); if (is_new) { pair->second = slot_samplers.insert(runtime, config); } @@ -1201,7 +1206,7 @@ SamplerId TextureCache

::FindSampler(const TSCEntry& config) { template ImageViewId TextureCache

::FindColorBuffer(size_t index, bool is_clear) { - const auto& regs = maxwell3d.regs; + const auto& regs = maxwell3d->regs; if (index >= regs.rt_control.count) { return ImageViewId{}; } @@ -1219,7 +1224,7 @@ ImageViewId TextureCache

::FindColorBuffer(size_t index, bool is_clear) { template ImageViewId TextureCache

::FindDepthBuffer(bool is_clear) { - const auto& regs = maxwell3d.regs; + const auto& regs = maxwell3d->regs; if (!regs.zeta_enable) { return ImageViewId{}; } @@ -1321,8 +1326,8 @@ void TextureCache

::ForEachImageInRegionGPU(GPUVAddr gpu_addr, size_t size, Fu static constexpr bool BOOL_BREAK = std::is_same_v; boost::container::small_vector images; ForEachGPUPage(gpu_addr, size, [this, &images, gpu_addr, size, func](u64 page) { - const auto it = gpu_page_table.find(page); - if (it == gpu_page_table.end()) { + const auto it = state->gpu_page_table.find(page); + if (it == state->gpu_page_table.end()) { if constexpr (BOOL_BREAK) { return false; } else { @@ -1403,9 +1408,9 @@ template void TextureCache

::ForEachSparseSegment(ImageBase& image, Func&& func) { using FuncReturn = typename std::invoke_result::type; static constexpr bool RETURNS_BOOL = std::is_same_v; - const auto segments = gpu_memory.GetSubmappedRange(image.gpu_addr, image.guest_size_bytes); + const auto segments = gpu_memory->GetSubmappedRange(image.gpu_addr, image.guest_size_bytes); for (const auto& [gpu_addr, size] : segments) { - std::optional cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr); + std::optional cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr); ASSERT(cpu_addr); if constexpr (RETURNS_BOOL) { if (func(gpu_addr, *cpu_addr, size)) { @@ -1449,7 +1454,7 @@ void TextureCache

::RegisterImage(ImageId image_id) { image.lru_index = lru_cache.Insert(image_id, frame_tick); ForEachGPUPage(image.gpu_addr, image.guest_size_bytes, - [this, image_id](u64 page) { gpu_page_table[page].push_back(image_id); }); + [this, image_id](u64 page) { state->gpu_page_table[page].push_back(image_id); }); if (False(image.flags & ImageFlagBits::Sparse)) { auto map_id = slot_map_views.insert(image.gpu_addr, image.cpu_addr, image.guest_size_bytes, image_id); @@ -1497,8 +1502,9 @@ void TextureCache

::UnregisterImage(ImageId image_id) { } image_ids.erase(vector_it); }; - ForEachGPUPage(image.gpu_addr, image.guest_size_bytes, - [this, &clear_page_table](u64 page) { clear_page_table(page, gpu_page_table); }); + ForEachGPUPage(image.gpu_addr, image.guest_size_bytes, [this, &clear_page_table](u64 page) { + clear_page_table(page, state->gpu_page_table); + }); if (False(image.flags & ImageFlagBits::Sparse)) { const auto map_id = image.map_view_id; ForEachCPUPage(image.cpu_addr, image.guest_size_bytes, [this, map_id](u64 page) { @@ -1631,7 +1637,7 @@ void TextureCache

::DeleteImage(ImageId image_id, bool immediate_delete) { ASSERT_MSG(False(image.flags & ImageFlagBits::Registered), "Image was not unregistered"); // Mark render targets as dirty - auto& dirty = maxwell3d.dirty.flags; + auto& dirty = maxwell3d->dirty.flags; dirty[Dirty::RenderTargets] = true; dirty[Dirty::ZetaBuffer] = true; for (size_t rt = 0; rt < NUM_RT; ++rt) { @@ -1681,22 +1687,24 @@ void TextureCache

::DeleteImage(ImageId image_id, bool immediate_delete) { if (alloc_images.empty()) { image_allocs_table.erase(alloc_it); } - if constexpr (ENABLE_VALIDATION) { - std::ranges::fill(graphics_image_view_ids, CORRUPT_ID); - std::ranges::fill(compute_image_view_ids, CORRUPT_ID); + for (auto& this_state : channel_storage) { + if constexpr (ENABLE_VALIDATION) { + std::ranges::fill(this_state.graphics_image_view_ids, CORRUPT_ID); + std::ranges::fill(this_state.compute_image_view_ids, CORRUPT_ID); + } + this_state.graphics_image_table.Invalidate(); + this_state.compute_image_table.Invalidate(); } - graphics_image_table.Invalidate(); - compute_image_table.Invalidate(); has_deleted_images = true; } template void TextureCache

::RemoveImageViewReferences(std::span removed_views) { - auto it = image_views.begin(); - while (it != image_views.end()) { + auto it = state->image_views.begin(); + while (it != state->image_views.end()) { const auto found = std::ranges::find(removed_views, it->second); if (found != removed_views.end()) { - it = image_views.erase(it); + it = state->image_views.erase(it); } else { ++it; } @@ -1943,7 +1951,7 @@ bool TextureCache

::IsFullClear(ImageViewId id) { const ImageViewBase& image_view = slot_image_views[id]; const ImageBase& image = slot_images[image_view.image_id]; const Extent3D size = image_view.size; - const auto& regs = maxwell3d.regs; + const auto& regs = maxwell3d->regs; const auto& scissor = regs.scissor_test[0]; if (image.info.resources.levels > 1 || image.info.resources.layers > 1) { // Images with multiple resources can't be cleared in a single call @@ -1958,4 +1966,61 @@ bool TextureCache

::IsFullClear(ImageViewId id) { scissor.max_y >= size.height; } +template +TextureCache

::ChannelInfo::ChannelInfo(Tegra::Control::ChannelState& state) noexcept + : maxwell3d{*state.maxwell_3d}, kepler_compute{*state.kepler_compute}, + gpu_memory{*state.memory_manager}, graphics_image_table{gpu_memory}, + graphics_sampler_table{gpu_memory}, compute_image_table{gpu_memory}, compute_sampler_table{ + gpu_memory} {} + +template +void TextureCache

::CreateChannel(struct Tegra::Control::ChannelState& channel) { + ASSERT(channel_map.find(channel.bind_id) == channel_map.end() && channel.bind_id >= 0); + auto new_id = [this, &channel]() { + if (!free_channel_ids.empty()) { + auto id = free_channel_ids.front(); + free_channel_ids.pop_front(); + new (&channel_storage[id]) ChannelInfo(channel); + return id; + } + channel_storage.emplace_back(channel); + return channel_storage.size() - 1; + }(); + channel_map.emplace(channel.bind_id, new_id); + if (current_channel_id != UNSET_CHANNEL) { + state = &channel_storage[current_channel_id]; + } +} + +/// Bind a channel for execution. +template +void TextureCache

::BindToChannel(s32 id) { + auto it = channel_map.find(id); + ASSERT(it != channel_map.end() && id >= 0); + current_channel_id = it->second; + state = &channel_storage[current_channel_id]; + maxwell3d = &state->maxwell3d; + kepler_compute = &state->kepler_compute; + gpu_memory = &state->gpu_memory; +} + +/// Erase channel's state. +template +void TextureCache

::EraseChannel(s32 id) { + const auto it = channel_map.find(id); + ASSERT(it != channel_map.end() && id >= 0); + const auto this_id = it->second; + free_channel_ids.push_back(this_id); + channel_map.erase(it); + if (this_id == current_channel_id) { + current_channel_id = UNSET_CHANNEL; + state = nullptr; + maxwell3d = nullptr; + kepler_compute = nullptr; + gpu_memory = nullptr; + } else if (current_channel_id != UNSET_CHANNEL) { + state = &channel_storage[current_channel_id]; + } +} + } // namespace VideoCommon diff --git a/src/video_core/texture_cache/texture_cache_base.h b/src/video_core/texture_cache/texture_cache_base.h index 7e6c6cef2..69efcb718 100644 --- a/src/video_core/texture_cache/texture_cache_base.h +++ b/src/video_core/texture_cache/texture_cache_base.h @@ -3,6 +3,8 @@ #pragma once +#include +#include #include #include #include @@ -26,6 +28,10 @@ #include "video_core/texture_cache/types.h" #include "video_core/textures/texture.h" +namespace Tegra::Control { +struct ChannelState; +} + namespace VideoCommon { using Tegra::Texture::SwizzleSource; @@ -58,6 +64,8 @@ class TextureCache { /// True when the API can provide info about the memory of the device. static constexpr bool HAS_DEVICE_MEMORY_INFO = P::HAS_DEVICE_MEMORY_INFO; + static constexpr size_t UNSET_CHANNEL{std::numeric_limits::max()}; + static constexpr s64 TARGET_THRESHOLD = 4_GiB; static constexpr s64 DEFAULT_EXPECTED_MEMORY = 1_GiB + 125_MiB; static constexpr s64 DEFAULT_CRITICAL_MEMORY = 1_GiB + 625_MiB; @@ -85,8 +93,7 @@ class TextureCache { }; public: - explicit TextureCache(Runtime&, VideoCore::RasterizerInterface&, Tegra::Engines::Maxwell3D&, - Tegra::Engines::KeplerCompute&, Tegra::MemoryManager&); + explicit TextureCache(Runtime&, VideoCore::RasterizerInterface&); /// Notify the cache that a new frame has been queued void TickFrame(); @@ -171,6 +178,15 @@ public: [[nodiscard]] bool IsRescaling(const ImageViewBase& image_view) const noexcept; + /// Create channel state. + void CreateChannel(struct Tegra::Control::ChannelState& channel); + + /// Bind a channel for execution. + void BindToChannel(s32 id); + + /// Erase channel's state. + void EraseChannel(s32 id); + std::mutex mutex; private: @@ -338,31 +354,52 @@ private: u64 GetScaledImageSizeBytes(ImageBase& image); Runtime& runtime; - VideoCore::RasterizerInterface& rasterizer; - Tegra::Engines::Maxwell3D& maxwell3d; - Tegra::Engines::KeplerCompute& kepler_compute; - Tegra::MemoryManager& gpu_memory; - DescriptorTable graphics_image_table{gpu_memory}; - DescriptorTable graphics_sampler_table{gpu_memory}; - std::vector graphics_sampler_ids; - std::vector graphics_image_view_ids; + struct ChannelInfo { + ChannelInfo() = delete; + ChannelInfo(struct Tegra::Control::ChannelState& state) noexcept; + ChannelInfo(const ChannelInfo& state) = delete; + ChannelInfo& operator=(const ChannelInfo&) = delete; + ChannelInfo(ChannelInfo&& other) noexcept = default; + ChannelInfo& operator=(ChannelInfo&& other) noexcept = default; + + Tegra::Engines::Maxwell3D& maxwell3d; + Tegra::Engines::KeplerCompute& kepler_compute; + Tegra::MemoryManager& gpu_memory; + + DescriptorTable graphics_image_table{gpu_memory}; + DescriptorTable graphics_sampler_table{gpu_memory}; + std::vector graphics_sampler_ids; + std::vector graphics_image_view_ids; - DescriptorTable compute_image_table{gpu_memory}; - DescriptorTable compute_sampler_table{gpu_memory}; - std::vector compute_sampler_ids; - std::vector compute_image_view_ids; + DescriptorTable compute_image_table{gpu_memory}; + DescriptorTable compute_sampler_table{gpu_memory}; + std::vector compute_sampler_ids; + std::vector compute_image_view_ids; + + std::unordered_map image_views; + std::unordered_map samplers; + + std::unordered_map, IdentityHash> gpu_page_table; + }; + + std::deque channel_storage; + std::deque free_channel_ids; + std::unordered_map channel_map; + + ChannelInfo* state; + size_t current_channel_id{UNSET_CHANNEL}; + VideoCore::RasterizerInterface& rasterizer; + Tegra::Engines::Maxwell3D* maxwell3d; + Tegra::Engines::KeplerCompute* kepler_compute; + Tegra::MemoryManager* gpu_memory; RenderTargets render_targets; - std::unordered_map image_views; - std::unordered_map samplers; std::unordered_map framebuffers; std::unordered_map, IdentityHash> page_table; - std::unordered_map, IdentityHash> gpu_page_table; std::unordered_map, IdentityHash> sparse_page_table; - std::unordered_map> sparse_views; VAddr virtual_invalid_space{}; -- cgit v1.2.3 From 3f8e7a55851a613becf715cbf3016a8e9f63d65f Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Sun, 7 Nov 2021 17:52:45 +0100 Subject: VideoCore: Fix channels with disk pipeline/shader cache. --- .../renderer_opengl/gl_compute_pipeline.cpp | 19 +++++++-------- .../renderer_opengl/gl_compute_pipeline.h | 16 ++++++++----- .../renderer_opengl/gl_graphics_pipeline.cpp | 28 +++++++++++----------- .../renderer_opengl/gl_graphics_pipeline.h | 16 ++++++++----- src/video_core/renderer_opengl/gl_rasterizer.cpp | 1 + src/video_core/renderer_opengl/gl_shader_cache.cpp | 11 ++++----- .../renderer_vulkan/vk_graphics_pipeline.cpp | 17 +++++++------ .../renderer_vulkan/vk_graphics_pipeline.h | 28 +++++++++++++--------- .../renderer_vulkan/vk_pipeline_cache.cpp | 8 +++---- src/video_core/renderer_vulkan/vk_rasterizer.cpp | 2 ++ src/video_core/texture_cache/texture_cache.h | 12 ++++++---- 11 files changed, 87 insertions(+), 71 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp index 1f0f156ed..26b51f442 100644 --- a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp @@ -28,12 +28,11 @@ bool ComputePipelineKey::operator==(const ComputePipelineKey& rhs) const noexcep } ComputePipeline::ComputePipeline(const Device& device, TextureCache& texture_cache_, - BufferCache& buffer_cache_, Tegra::MemoryManager& gpu_memory_, - Tegra::Engines::KeplerCompute& kepler_compute_, - ProgramManager& program_manager_, const Shader::Info& info_, - std::string code, std::vector code_v) - : texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, gpu_memory{gpu_memory_}, - kepler_compute{kepler_compute_}, program_manager{program_manager_}, info{info_} { + BufferCache& buffer_cache_, ProgramManager& program_manager_, + const Shader::Info& info_, std::string code, + std::vector code_v) + : texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, + program_manager{program_manager_}, info{info_} { switch (device.GetShaderBackend()) { case Settings::ShaderBackend::GLSL: source_program = CreateProgram(code, GL_COMPUTE_SHADER); @@ -86,7 +85,7 @@ void ComputePipeline::Configure() { GLsizei texture_binding{}; GLsizei image_binding{}; - const auto& qmd{kepler_compute.launch_description}; + const auto& qmd{kepler_compute->launch_description}; const auto& cbufs{qmd.const_buffer_config}; const bool via_header_index{qmd.linked_tsc != 0}; const auto read_handle{[&](const auto& desc, u32 index) { @@ -101,12 +100,12 @@ void ComputePipeline::Configure() { const u32 secondary_offset{desc.secondary_cbuf_offset + index_offset}; const GPUVAddr separate_addr{cbufs[desc.secondary_cbuf_index].Address() + secondary_offset}; - const u32 lhs_raw{gpu_memory.Read(addr)}; - const u32 rhs_raw{gpu_memory.Read(separate_addr)}; + const u32 lhs_raw{gpu_memory->Read(addr)}; + const u32 rhs_raw{gpu_memory->Read(separate_addr)}; return TexturePair(lhs_raw | rhs_raw, via_header_index); } } - return TexturePair(gpu_memory.Read(addr), via_header_index); + return TexturePair(gpu_memory->Read(addr), via_header_index); }}; const auto add_image{[&](const auto& desc, bool blacklist) { for (u32 index = 0; index < desc.count; ++index) { diff --git a/src/video_core/renderer_opengl/gl_compute_pipeline.h b/src/video_core/renderer_opengl/gl_compute_pipeline.h index 723f27f11..6534dec32 100644 --- a/src/video_core/renderer_opengl/gl_compute_pipeline.h +++ b/src/video_core/renderer_opengl/gl_compute_pipeline.h @@ -49,10 +49,8 @@ static_assert(std::is_trivially_constructible_v); class ComputePipeline { public: explicit ComputePipeline(const Device& device, TextureCache& texture_cache_, - BufferCache& buffer_cache_, Tegra::MemoryManager& gpu_memory_, - Tegra::Engines::KeplerCompute& kepler_compute_, - ProgramManager& program_manager_, const Shader::Info& info_, - std::string code, std::vector code_v); + BufferCache& buffer_cache_, ProgramManager& program_manager_, + const Shader::Info& info_, std::string code, std::vector code_v); void Configure(); @@ -60,11 +58,17 @@ public: return writes_global_memory; } + void SetEngine(Tegra::Engines::KeplerCompute* kepler_compute_, + Tegra::MemoryManager* gpu_memory_) { + kepler_compute = kepler_compute_; + gpu_memory = gpu_memory_; + } + private: TextureCache& texture_cache; BufferCache& buffer_cache; - Tegra::MemoryManager& gpu_memory; - Tegra::Engines::KeplerCompute& kepler_compute; + Tegra::MemoryManager* gpu_memory; + Tegra::Engines::KeplerCompute* kepler_compute; ProgramManager& program_manager; Shader::Info info; diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp index 67eae369d..c877d7792 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp @@ -169,15 +169,15 @@ ConfigureFuncPtr ConfigureFunc(const std::array& infos, u32 ena } } // Anonymous namespace -GraphicsPipeline::GraphicsPipeline( - const Device& device, TextureCache& texture_cache_, BufferCache& buffer_cache_, - Tegra::MemoryManager& gpu_memory_, Tegra::Engines::Maxwell3D& maxwell3d_, - ProgramManager& program_manager_, StateTracker& state_tracker_, ShaderWorker* thread_worker, - VideoCore::ShaderNotify* shader_notify, std::array sources, - std::array, 5> sources_spirv, const std::array& infos, - const GraphicsPipelineKey& key_) - : texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, - gpu_memory{gpu_memory_}, maxwell3d{maxwell3d_}, program_manager{program_manager_}, +GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_cache_, + BufferCache& buffer_cache_, ProgramManager& program_manager_, + StateTracker& state_tracker_, ShaderWorker* thread_worker, + VideoCore::ShaderNotify* shader_notify, + std::array sources, + std::array, 5> sources_spirv, + const std::array& infos, + const GraphicsPipelineKey& key_) + : texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, program_manager{program_manager_}, state_tracker{state_tracker_}, key{key_} { if (shader_notify) { shader_notify->MarkShaderBuilding(); @@ -285,7 +285,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { buffer_cache.runtime.SetBaseStorageBindings(base_storage_bindings); buffer_cache.runtime.SetEnableStorageBuffers(use_storage_buffers); - const auto& regs{maxwell3d.regs}; + const auto& regs{maxwell3d->regs}; const bool via_header_index{regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex}; const auto config_stage{[&](size_t stage) LAMBDA_FORCEINLINE { const Shader::Info& info{stage_infos[stage]}; @@ -299,7 +299,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { ++ssbo_index; } } - const auto& cbufs{maxwell3d.state.shader_stages[stage].const_buffers}; + const auto& cbufs{maxwell3d->state.shader_stages[stage].const_buffers}; const auto read_handle{[&](const auto& desc, u32 index) { ASSERT(cbufs[desc.cbuf_index].enabled); const u32 index_offset{index << desc.size_shift}; @@ -312,13 +312,13 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { const u32 second_offset{desc.secondary_cbuf_offset + index_offset}; const GPUVAddr separate_addr{cbufs[desc.secondary_cbuf_index].address + second_offset}; - const u32 lhs_raw{gpu_memory.Read(addr)}; - const u32 rhs_raw{gpu_memory.Read(separate_addr)}; + const u32 lhs_raw{gpu_memory->Read(addr)}; + const u32 rhs_raw{gpu_memory->Read(separate_addr)}; const u32 raw{lhs_raw | rhs_raw}; return TexturePair(raw, via_header_index); } } - return TexturePair(gpu_memory.Read(addr), via_header_index); + return TexturePair(gpu_memory->Read(addr), via_header_index); }}; const auto add_image{[&](const auto& desc, bool blacklist) LAMBDA_FORCEINLINE { for (u32 index = 0; index < desc.count; ++index) { diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.h b/src/video_core/renderer_opengl/gl_graphics_pipeline.h index 4ec15b966..a0f0e63cb 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.h +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.h @@ -71,10 +71,9 @@ static_assert(std::is_trivially_constructible_v); class GraphicsPipeline { public: explicit GraphicsPipeline(const Device& device, TextureCache& texture_cache_, - BufferCache& buffer_cache_, Tegra::MemoryManager& gpu_memory_, - Tegra::Engines::Maxwell3D& maxwell3d_, - ProgramManager& program_manager_, StateTracker& state_tracker_, - ShaderWorker* thread_worker, VideoCore::ShaderNotify* shader_notify, + BufferCache& buffer_cache_, ProgramManager& program_manager_, + StateTracker& state_tracker_, ShaderWorker* thread_worker, + VideoCore::ShaderNotify* shader_notify, std::array sources, std::array, 5> sources_spirv, const std::array& infos, @@ -107,6 +106,11 @@ public: }; } + void SetEngine(Tegra::Engines::Maxwell3D* maxwell3d_, Tegra::MemoryManager* gpu_memory_) { + maxwell3d = maxwell3d_; + gpu_memory = gpu_memory_; + } + private: template void ConfigureImpl(bool is_indexed); @@ -119,8 +123,8 @@ private: TextureCache& texture_cache; BufferCache& buffer_cache; - Tegra::MemoryManager& gpu_memory; - Tegra::Engines::Maxwell3D& maxwell3d; + Tegra::MemoryManager* gpu_memory; + Tegra::Engines::Maxwell3D* maxwell3d; ProgramManager& program_manager; StateTracker& state_tracker; const GraphicsPipelineKey key; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 8cfa0013f..4eb9bd116 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -216,6 +216,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) { return; } std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex}; + pipeline->SetEngine(maxwell3d, gpu_memory); pipeline->Configure(is_indexed); SyncState(); diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index 494581d0d..5a29a41d2 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -477,9 +477,9 @@ std::unique_ptr ShaderCache::CreateGraphicsPipeline( previous_program = &program; } auto* const thread_worker{build_in_parallel ? workers.get() : nullptr}; - return std::make_unique( - device, texture_cache, buffer_cache, *gpu_memory, *maxwell3d, program_manager, - state_tracker, thread_worker, &shader_notify, sources, sources_spirv, infos, key); + return std::make_unique(device, texture_cache, buffer_cache, program_manager, + state_tracker, thread_worker, &shader_notify, sources, + sources_spirv, infos, key); } catch (Shader::Exception& exception) { LOG_ERROR(Render_OpenGL, "{}", exception.what()); @@ -533,9 +533,8 @@ std::unique_ptr ShaderCache::CreateComputePipeline( break; } - return std::make_unique(device, texture_cache, buffer_cache, *gpu_memory, - *kepler_compute, program_manager, program.info, code, - code_spirv); + return std::make_unique(device, texture_cache, buffer_cache, program_manager, + program.info, code, code_spirv); } catch (Shader::Exception& exception) { LOG_ERROR(Render_OpenGL, "{}", exception.what()); return nullptr; diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp index 5aca8f038..1e993185f 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp @@ -215,15 +215,14 @@ ConfigureFuncPtr ConfigureFunc(const std::array& m } // Anonymous namespace GraphicsPipeline::GraphicsPipeline( - Tegra::Engines::Maxwell3D& maxwell3d_, Tegra::MemoryManager& gpu_memory_, Scheduler& scheduler_, - BufferCache& buffer_cache_, TextureCache& texture_cache_, + Scheduler& scheduler_, BufferCache& buffer_cache_, TextureCache& texture_cache_, VideoCore::ShaderNotify* shader_notify, const Device& device_, DescriptorPool& descriptor_pool, UpdateDescriptorQueue& update_descriptor_queue_, Common::ThreadWorker* worker_thread, PipelineStatistics* pipeline_statistics, RenderPassCache& render_pass_cache, const GraphicsPipelineCacheKey& key_, std::array stages, const std::array& infos) - : key{key_}, maxwell3d{maxwell3d_}, gpu_memory{gpu_memory_}, device{device_}, - texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, scheduler{scheduler_}, + : key{key_}, device{device_}, texture_cache{texture_cache_}, + buffer_cache{buffer_cache_}, scheduler{scheduler_}, update_descriptor_queue{update_descriptor_queue_}, spv_modules{std::move(stages)} { if (shader_notify) { shader_notify->MarkShaderBuilding(); @@ -288,7 +287,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { buffer_cache.SetUniformBuffersState(enabled_uniform_buffer_masks, &uniform_buffer_sizes); - const auto& regs{maxwell3d.regs}; + const auto& regs{maxwell3d->regs}; const bool via_header_index{regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex}; const auto config_stage{[&](size_t stage) LAMBDA_FORCEINLINE { const Shader::Info& info{stage_infos[stage]}; @@ -302,7 +301,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { ++ssbo_index; } } - const auto& cbufs{maxwell3d.state.shader_stages[stage].const_buffers}; + const auto& cbufs{maxwell3d->state.shader_stages[stage].const_buffers}; const auto read_handle{[&](const auto& desc, u32 index) { ASSERT(cbufs[desc.cbuf_index].enabled); const u32 index_offset{index << desc.size_shift}; @@ -315,13 +314,13 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { const u32 second_offset{desc.secondary_cbuf_offset + index_offset}; const GPUVAddr separate_addr{cbufs[desc.secondary_cbuf_index].address + second_offset}; - const u32 lhs_raw{gpu_memory.Read(addr)}; - const u32 rhs_raw{gpu_memory.Read(separate_addr)}; + const u32 lhs_raw{gpu_memory->Read(addr)}; + const u32 rhs_raw{gpu_memory->Read(separate_addr)}; const u32 raw{lhs_raw | rhs_raw}; return TexturePair(raw, via_header_index); } } - return TexturePair(gpu_memory.Read(addr), via_header_index); + return TexturePair(gpu_memory->Read(addr), via_header_index); }}; const auto add_image{[&](const auto& desc, bool blacklist) LAMBDA_FORCEINLINE { for (u32 index = 0; index < desc.count; ++index) { diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h index e8949a9ab..85602592b 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h @@ -69,15 +69,16 @@ class GraphicsPipeline { static constexpr size_t NUM_STAGES = Tegra::Engines::Maxwell3D::Regs::MaxShaderStage; public: - explicit GraphicsPipeline( - Tegra::Engines::Maxwell3D& maxwell3d, Tegra::MemoryManager& gpu_memory, - Scheduler& scheduler, BufferCache& buffer_cache, TextureCache& texture_cache, - VideoCore::ShaderNotify* shader_notify, const Device& device, - DescriptorPool& descriptor_pool, UpdateDescriptorQueue& update_descriptor_queue, - Common::ThreadWorker* worker_thread, PipelineStatistics* pipeline_statistics, - RenderPassCache& render_pass_cache, const GraphicsPipelineCacheKey& key, - std::array stages, - const std::array& infos); + explicit GraphicsPipeline(Scheduler& scheduler, BufferCache& buffer_cache, + TextureCache& texture_cache, VideoCore::ShaderNotify* shader_notify, + const Device& device, DescriptorPool& descriptor_pool, + UpdateDescriptorQueue& update_descriptor_queue, + Common::ThreadWorker* worker_thread, + PipelineStatistics* pipeline_statistics, + RenderPassCache& render_pass_cache, + const GraphicsPipelineCacheKey& key, + std::array stages, + const std::array& infos); GraphicsPipeline& operator=(GraphicsPipeline&&) noexcept = delete; GraphicsPipeline(GraphicsPipeline&&) noexcept = delete; @@ -109,6 +110,11 @@ public: return [](GraphicsPipeline* pl, bool is_indexed) { pl->ConfigureImpl(is_indexed); }; } + void SetEngine(Tegra::Engines::Maxwell3D* maxwell3d_, Tegra::MemoryManager* gpu_memory_) { + maxwell3d = maxwell3d_; + gpu_memory = gpu_memory_; + } + private: template void ConfigureImpl(bool is_indexed); @@ -120,8 +126,8 @@ private: void Validate(); const GraphicsPipelineCacheKey key; - Tegra::Engines::Maxwell3D& maxwell3d; - Tegra::MemoryManager& gpu_memory; + Tegra::Engines::Maxwell3D* maxwell3d; + Tegra::MemoryManager* gpu_memory; const Device& device; TextureCache& texture_cache; BufferCache& buffer_cache; diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp index b1e0b96c4..732e7b6f2 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp @@ -555,10 +555,10 @@ std::unique_ptr PipelineCache::CreateGraphicsPipeline( previous_stage = &program; } Common::ThreadWorker* const thread_worker{build_in_parallel ? &workers : nullptr}; - return std::make_unique( - *maxwell3d, *gpu_memory, scheduler, buffer_cache, texture_cache, &shader_notify, device, - descriptor_pool, update_descriptor_queue, thread_worker, statistics, render_pass_cache, key, - std::move(modules), infos); + return std::make_unique(scheduler, buffer_cache, texture_cache, + &shader_notify, device, descriptor_pool, + update_descriptor_queue, thread_worker, statistics, + render_pass_cache, key, std::move(modules), infos); } catch (const Shader::Exception& exception) { LOG_ERROR(Render_Vulkan, "{}", exception.what()); diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index bf750452f..7e0805544 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -190,6 +190,8 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) { return; } std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex}; + // update engine as channel may be different. + pipeline->SetEngine(maxwell3d, gpu_memory); pipeline->Configure(is_indexed); BeginTransformFeedback(); diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 2731aead0..e8b0b0a3b 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -885,12 +885,14 @@ void TextureCache

::InvalidateScale(Image& image) { } image.image_view_ids.clear(); image.image_view_infos.clear(); - if constexpr (ENABLE_VALIDATION) { - std::ranges::fill(state->graphics_image_view_ids, CORRUPT_ID); - std::ranges::fill(state->compute_image_view_ids, CORRUPT_ID); + for (auto& this_state : channel_storage) { + if constexpr (ENABLE_VALIDATION) { + std::ranges::fill(this_state.graphics_image_view_ids, CORRUPT_ID); + std::ranges::fill(this_state.compute_image_view_ids, CORRUPT_ID); + } + this_state.graphics_image_table.Invalidate(); + this_state.compute_image_table.Invalidate(); } - state->graphics_image_table.Invalidate(); - state->compute_image_table.Invalidate(); has_deleted_images = true; } -- cgit v1.2.3 From 6fc4012396e98a1a6ac455791b314d2280a12a51 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Tue, 16 Nov 2021 00:01:40 +0100 Subject: VideoCore: Extra Fixes. --- src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp | 4 +++- src/video_core/engines/puller.cpp | 2 +- src/video_core/texture_cache/util.cpp | 2 +- 3 files changed, 5 insertions(+), 3 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp index a859a7abd..54074af75 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp @@ -143,7 +143,7 @@ NvResult nvhost_ctrl::IocCtrlEventWait(const std::vector& input, std::vector } }(); - must_unmark_fail = true; + must_unmark_fail = false; const auto check_failing = [&]() { if (events[slot].fails > 2) { @@ -164,6 +164,7 @@ NvResult nvhost_ctrl::IocCtrlEventWait(const std::vector& input, std::vector if (params.timeout == 0) { if (check_failing()) { + events[slot].fails = 0; return NvResult::Success; } return NvResult::Timeout; @@ -180,6 +181,7 @@ NvResult nvhost_ctrl::IocCtrlEventWait(const std::vector& input, std::vector } if (check_failing()) { + event.fails = 0; return NvResult::Success; } diff --git a/src/video_core/engines/puller.cpp b/src/video_core/engines/puller.cpp index 37f2ced18..3866c8746 100644 --- a/src/video_core/engines/puller.cpp +++ b/src/video_core/engines/puller.cpp @@ -118,7 +118,7 @@ void Puller::ProcessSemaphoreTriggerMethod() { } void Puller::ProcessSemaphoreRelease() { - memory_manager.Write(regs.semaphore_address.SemaphoreAddress(), regs.semaphore_release); + rasterizer->SignalSemaphore(regs.semaphore_address.SemaphoreAddress(), regs.semaphore_release); } void Puller::ProcessSemaphoreAcquire() { diff --git a/src/video_core/texture_cache/util.cpp b/src/video_core/texture_cache/util.cpp index 1820823b2..bea1c27d0 100644 --- a/src/video_core/texture_cache/util.cpp +++ b/src/video_core/texture_cache/util.cpp @@ -755,7 +755,7 @@ bool IsValidEntry(const Tegra::MemoryManager& gpu_memory, const TICEntry& config if (address == 0) { return false; } - if (address > (1ULL << 48)) { + if (address >= (1ULL << 40)) { return false; } if (gpu_memory.GpuToCpuAddress(address).has_value()) { -- cgit v1.2.3 From bb74973bba6005cee5f1409b3b5a6c572da3cf66 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Wed, 17 Nov 2021 05:44:01 +0100 Subject: General: Rebase fixes. --- src/video_core/texture_cache/texture_cache.h | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index e8b0b0a3b..0ec999d63 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -885,14 +885,13 @@ void TextureCache

::InvalidateScale(Image& image) { } image.image_view_ids.clear(); image.image_view_infos.clear(); - for (auto& this_state : channel_storage) { - if constexpr (ENABLE_VALIDATION) { - std::ranges::fill(this_state.graphics_image_view_ids, CORRUPT_ID); - std::ranges::fill(this_state.compute_image_view_ids, CORRUPT_ID); - } - this_state.graphics_image_table.Invalidate(); - this_state.compute_image_table.Invalidate(); + auto& channel_info = channel_storage[image.channel]; + if constexpr (ENABLE_VALIDATION) { + std::ranges::fill(channel_info.graphics_image_view_ids, CORRUPT_ID); + std::ranges::fill(channel_info.compute_image_view_ids, CORRUPT_ID); } + channel_info.graphics_image_table.Invalidate(); + channel_info.compute_image_table.Invalidate(); has_deleted_images = true; } -- cgit v1.2.3 From e462191482c6507daed67802c6c1d2c50f10c96e Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Fri, 17 Dec 2021 16:45:06 +0100 Subject: Refactor VideoCore to use AS sepparate from Channel. --- src/common/hash.h | 7 ++ src/video_core/CMakeLists.txt | 1 + src/video_core/control/channel_state_cache.cpp | 8 +- src/video_core/control/channel_state_cache.h | 28 ++++- src/video_core/control/channel_state_cache.inc | 23 +++- src/video_core/memory_manager.cpp | 5 +- src/video_core/memory_manager.h | 9 ++ src/video_core/texture_cache/texture_cache.cpp | 16 +++ src/video_core/texture_cache/texture_cache.h | 130 ++++++++-------------- src/video_core/texture_cache/texture_cache_base.h | 96 +++++++--------- 10 files changed, 171 insertions(+), 152 deletions(-) create mode 100644 src/video_core/texture_cache/texture_cache.cpp (limited to 'src/video_core/texture_cache') diff --git a/src/common/hash.h b/src/common/hash.h index b6f3e6d6f..e8fe78b07 100644 --- a/src/common/hash.h +++ b/src/common/hash.h @@ -18,4 +18,11 @@ struct PairHash { } }; +template +struct IdentityHash { + [[nodiscard]] size_t operator()(T value) const noexcept { + return static_cast(value); + } +}; + } // namespace Common diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index e216c51a2..35faa70a0 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -203,6 +203,7 @@ add_library(video_core STATIC texture_cache/render_targets.h texture_cache/samples_helper.h texture_cache/slot_vector.h + texture_cache/texture_cache.cpp texture_cache/texture_cache.h texture_cache/texture_cache_base.h texture_cache/types.h diff --git a/src/video_core/control/channel_state_cache.cpp b/src/video_core/control/channel_state_cache.cpp index f72a97b2f..ec7ba907c 100644 --- a/src/video_core/control/channel_state_cache.cpp +++ b/src/video_core/control/channel_state_cache.cpp @@ -1,5 +1,11 @@ #include "video_core/control/channel_state_cache.inc" namespace VideoCommon { + +ChannelInfo::ChannelInfo(Tegra::Control::ChannelState& channel_state) + : maxwell3d{*channel_state.maxwell_3d}, kepler_compute{*channel_state.kepler_compute}, + gpu_memory{*channel_state.memory_manager} {} + template class VideoCommon::ChannelSetupCaches; -} + +} // namespace VideoCommon diff --git a/src/video_core/control/channel_state_cache.h b/src/video_core/control/channel_state_cache.h index c8298c003..c51040c83 100644 --- a/src/video_core/control/channel_state_cache.h +++ b/src/video_core/control/channel_state_cache.h @@ -2,6 +2,7 @@ #include #include +#include #include #include "common/common_types.h" @@ -41,9 +42,10 @@ template class ChannelSetupCaches { public: /// Operations for seting the channel of execution. + virtual ~ChannelSetupCaches(); /// Create channel state. - void CreateChannel(Tegra::Control::ChannelState& channel); + virtual void CreateChannel(Tegra::Control::ChannelState& channel); /// Bind a channel for execution. void BindToChannel(s32 id); @@ -51,18 +53,34 @@ public: /// Erase channel's state. void EraseChannel(s32 id); + Tegra::MemoryManager* GetFromID(size_t id) const { + std::unique_lock lk(config_mutex); + const auto ref = address_spaces.find(id); + return ref->second.gpu_memory; + } + protected: static constexpr size_t UNSET_CHANNEL{std::numeric_limits::max()}; - std::deque

channel_storage; - std::deque free_channel_ids; - std::unordered_map channel_map; - P* channel_state; size_t current_channel_id{UNSET_CHANNEL}; + size_t current_address_space{}; Tegra::Engines::Maxwell3D* maxwell3d; Tegra::Engines::KeplerCompute* kepler_compute; Tegra::MemoryManager* gpu_memory; + + std::deque

channel_storage; + std::deque free_channel_ids; + std::unordered_map channel_map; + struct AddresSpaceRef { + size_t ref_count; + size_t storage_id; + Tegra::MemoryManager* gpu_memory; + }; + std::unordered_map address_spaces; + mutable std::mutex config_mutex; + + virtual void OnGPUASRegister([[maybe_unused]] size_t map_id) {} }; } // namespace VideoCommon diff --git a/src/video_core/control/channel_state_cache.inc b/src/video_core/control/channel_state_cache.inc index 3eb73af9f..185eabc35 100644 --- a/src/video_core/control/channel_state_cache.inc +++ b/src/video_core/control/channel_state_cache.inc @@ -6,18 +6,18 @@ namespace VideoCommon { -ChannelInfo::ChannelInfo(Tegra::Control::ChannelState& channel_state) - : maxwell3d{*channel_state.maxwell_3d}, kepler_compute{*channel_state.kepler_compute}, - gpu_memory{*channel_state.memory_manager} {} +template +ChannelSetupCaches

::~ChannelSetupCaches() = default; template void ChannelSetupCaches

::CreateChannel(struct Tegra::Control::ChannelState& channel) { + std::unique_lock lk(config_mutex); ASSERT(channel_map.find(channel.bind_id) == channel_map.end() && channel.bind_id >= 0); auto new_id = [this, &channel]() { if (!free_channel_ids.empty()) { auto id = free_channel_ids.front(); free_channel_ids.pop_front(); - new (&channel_storage[id]) ChannelInfo(channel); + new (&channel_storage[id]) P(channel); return id; } channel_storage.emplace_back(channel); @@ -27,11 +27,24 @@ void ChannelSetupCaches

::CreateChannel(struct Tegra::Control::ChannelState& c if (current_channel_id != UNSET_CHANNEL) { channel_state = &channel_storage[current_channel_id]; } + auto as_it = address_spaces.find(channel.memory_manager->GetID()); + if (as_it != address_spaces.end()) { + as_it->second.ref_count++; + return; + } + AddresSpaceRef new_gpu_mem_ref{ + .ref_count = 1, + .storage_id = address_spaces.size(), + .gpu_memory = channel.memory_manager.get(), + }; + address_spaces.emplace(channel.memory_manager->GetID(), new_gpu_mem_ref); + OnGPUASRegister(channel.memory_manager->GetID()); } /// Bind a channel for execution. template void ChannelSetupCaches

::BindToChannel(s32 id) { + std::unique_lock lk(config_mutex); auto it = channel_map.find(id); ASSERT(it != channel_map.end() && id >= 0); current_channel_id = it->second; @@ -39,11 +52,13 @@ void ChannelSetupCaches

::BindToChannel(s32 id) { maxwell3d = &channel_state->maxwell3d; kepler_compute = &channel_state->kepler_compute; gpu_memory = &channel_state->gpu_memory; + current_address_space = gpu_memory->GetID(); } /// Erase channel's channel_state. template void ChannelSetupCaches

::EraseChannel(s32 id) { + std::unique_lock lk(config_mutex); const auto it = channel_map.find(id); ASSERT(it != channel_map.end() && id >= 0); const auto this_id = it->second; diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index fc68bcc73..d4c0dca78 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -16,9 +16,12 @@ namespace Tegra { +std::atomic MemoryManager::unique_identifier_generator{}; + MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64 page_bits_) : system{system_}, address_space_bits{address_space_bits_}, page_bits{page_bits_}, entries{}, - page_table{address_space_bits, address_space_bits + page_bits - 38, page_bits} { + page_table{address_space_bits, address_space_bits + page_bits - 38, page_bits}, + unique_identifier{unique_identifier_generator.fetch_add(1, std::memory_order_acq_rel)} { address_space_size = 1ULL << address_space_bits; allocate_start = address_space_bits > 32 ? 1ULL << 32 : 0; page_size = 1ULL << page_bits; diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index b8878476a..56604ef3e 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -3,6 +3,7 @@ #pragma once +#include #include #include #include @@ -26,6 +27,10 @@ public: u64 page_bits_ = 16); ~MemoryManager(); + size_t GetID() const { + return unique_identifier; + } + /// Binds a renderer to the memory manager. void BindRasterizer(VideoCore::RasterizerInterface* rasterizer); @@ -140,6 +145,10 @@ private: void SetEntry(size_t position, EntryType entry); Common::MultiLevelPageTable page_table; + + const size_t unique_identifier; + + static std::atomic unique_identifier_generator; }; } // namespace Tegra diff --git a/src/video_core/texture_cache/texture_cache.cpp b/src/video_core/texture_cache/texture_cache.cpp new file mode 100644 index 000000000..bc905a1a4 --- /dev/null +++ b/src/video_core/texture_cache/texture_cache.cpp @@ -0,0 +1,16 @@ +// Copyright 2021 yuzu Emulator Project +// Licensed under GPLv3 or any later version +// Refer to the license.txt file included. + +#include "video_core/control/channel_state_cache.inc" +#include "video_core/texture_cache/texture_cache_base.h" + +namespace VideoCommon { + +TextureCacheChannelInfo::TextureCacheChannelInfo(Tegra::Control::ChannelState& state) noexcept + : ChannelInfo(state), graphics_image_table{gpu_memory}, graphics_sampler_table{gpu_memory}, + compute_image_table{gpu_memory}, compute_sampler_table{gpu_memory} {} + +template class VideoCommon::ChannelSetupCaches; + +} // namespace VideoCommon diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 0ec999d63..89c5faf88 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -1,5 +1,7 @@ -// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project -// SPDX-License-Identifier: GPL-2.0-or-later +// SPDX-FileCopyrightText: 2021 yuzu emulator team +// (https://github.com/skyline-emu/) +// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 +// or any later version Refer to the license.txt file included. #pragma once @@ -41,10 +43,6 @@ TextureCache

::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface& // Setup channels current_channel_id = UNSET_CHANNEL; - state = nullptr; - maxwell3d = nullptr; - kepler_compute = nullptr; - gpu_memory = nullptr; // Make sure the first index is reserved for the null resources // This way the null resource becomes a compile time constant @@ -156,23 +154,24 @@ void TextureCache

::MarkModification(ImageId id) noexcept { template template void TextureCache

::FillGraphicsImageViews(std::span views) { - FillImageViews(state->graphics_image_table, state->graphics_image_view_ids, - views); + FillImageViews(channel_state->graphics_image_table, + channel_state->graphics_image_view_ids, views); } template void TextureCache

::FillComputeImageViews(std::span views) { - FillImageViews(state->compute_image_table, state->compute_image_view_ids, views); + FillImageViews(channel_state->compute_image_table, channel_state->compute_image_view_ids, + views); } template typename P::Sampler* TextureCache

::GetGraphicsSampler(u32 index) { - if (index > state->graphics_sampler_table.Limit()) { + if (index > channel_state->graphics_sampler_table.Limit()) { LOG_DEBUG(HW_GPU, "Invalid sampler index={}", index); return &slot_samplers[NULL_SAMPLER_ID]; } - const auto [descriptor, is_new] = state->graphics_sampler_table.Read(index); - SamplerId& id = state->graphics_sampler_ids[index]; + const auto [descriptor, is_new] = channel_state->graphics_sampler_table.Read(index); + SamplerId& id = channel_state->graphics_sampler_ids[index]; if (is_new) { id = FindSampler(descriptor); } @@ -181,12 +180,12 @@ typename P::Sampler* TextureCache

::GetGraphicsSampler(u32 index) { template typename P::Sampler* TextureCache

::GetComputeSampler(u32 index) { - if (index > state->compute_sampler_table.Limit()) { + if (index > channel_state->compute_sampler_table.Limit()) { LOG_DEBUG(HW_GPU, "Invalid sampler index={}", index); return &slot_samplers[NULL_SAMPLER_ID]; } - const auto [descriptor, is_new] = state->compute_sampler_table.Read(index); - SamplerId& id = state->compute_sampler_ids[index]; + const auto [descriptor, is_new] = channel_state->compute_sampler_table.Read(index); + SamplerId& id = channel_state->compute_sampler_ids[index]; if (is_new) { id = FindSampler(descriptor); } @@ -199,11 +198,12 @@ void TextureCache

::SynchronizeGraphicsDescriptors() { const bool linked_tsc = maxwell3d->regs.sampler_index == SamplerIndex::ViaHeaderIndex; const u32 tic_limit = maxwell3d->regs.tic.limit; const u32 tsc_limit = linked_tsc ? tic_limit : maxwell3d->regs.tsc.limit; - if (state->graphics_sampler_table.Synchornize(maxwell3d->regs.tsc.Address(), tsc_limit)) { - state->graphics_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID); + if (channel_state->graphics_sampler_table.Synchornize(maxwell3d->regs.tsc.Address(), + tsc_limit)) { + channel_state->graphics_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID); } - if (state->graphics_image_table.Synchornize(maxwell3d->regs.tic.Address(), tic_limit)) { - state->graphics_image_view_ids.resize(tic_limit + 1, CORRUPT_ID); + if (channel_state->graphics_image_table.Synchornize(maxwell3d->regs.tic.Address(), tic_limit)) { + channel_state->graphics_image_view_ids.resize(tic_limit + 1, CORRUPT_ID); } } @@ -213,11 +213,12 @@ void TextureCache

::SynchronizeComputeDescriptors() { const u32 tic_limit = kepler_compute->regs.tic.limit; const u32 tsc_limit = linked_tsc ? tic_limit : kepler_compute->regs.tsc.limit; const GPUVAddr tsc_gpu_addr = kepler_compute->regs.tsc.Address(); - if (state->compute_sampler_table.Synchornize(tsc_gpu_addr, tsc_limit)) { - state->compute_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID); + if (channel_state->compute_sampler_table.Synchornize(tsc_gpu_addr, tsc_limit)) { + channel_state->compute_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID); } - if (state->compute_image_table.Synchornize(kepler_compute->regs.tic.Address(), tic_limit)) { - state->compute_image_view_ids.resize(tic_limit + 1, CORRUPT_ID); + if (channel_state->compute_image_table.Synchornize(kepler_compute->regs.tic.Address(), + tic_limit)) { + channel_state->compute_image_view_ids.resize(tic_limit + 1, CORRUPT_ID); } } @@ -738,7 +739,7 @@ ImageViewId TextureCache

::FindImageView(const TICEntry& config) { if (!IsValidEntry(*gpu_memory, config)) { return NULL_IMAGE_VIEW_ID; } - const auto [pair, is_new] = state->image_views.try_emplace(config); + const auto [pair, is_new] = channel_state->image_views.try_emplace(config); ImageViewId& image_view_id = pair->second; if (is_new) { image_view_id = CreateImageView(config); @@ -1198,7 +1199,7 @@ SamplerId TextureCache

::FindSampler(const TSCEntry& config) { if (std::ranges::all_of(config.raw, [](u64 value) { return value == 0; })) { return NULL_SAMPLER_ID; } - const auto [pair, is_new] = state->samplers.try_emplace(config); + const auto [pair, is_new] = channel_state->samplers.try_emplace(config); if (is_new) { pair->second = slot_samplers.insert(runtime, config); } @@ -1327,8 +1328,8 @@ void TextureCache

::ForEachImageInRegionGPU(GPUVAddr gpu_addr, size_t size, Fu static constexpr bool BOOL_BREAK = std::is_same_v; boost::container::small_vector images; ForEachGPUPage(gpu_addr, size, [this, &images, gpu_addr, size, func](u64 page) { - const auto it = state->gpu_page_table.find(page); - if (it == state->gpu_page_table.end()) { + const auto it = channel_state->gpu_page_table->find(page); + if (it == channel_state->gpu_page_table->end()) { if constexpr (BOOL_BREAK) { return false; } else { @@ -1454,8 +1455,9 @@ void TextureCache

::RegisterImage(ImageId image_id) { } image.lru_index = lru_cache.Insert(image_id, frame_tick); - ForEachGPUPage(image.gpu_addr, image.guest_size_bytes, - [this, image_id](u64 page) { state->gpu_page_table[page].push_back(image_id); }); + ForEachGPUPage(image.gpu_addr, image.guest_size_bytes, [this, image_id](u64 page) { + (*channel_state->gpu_page_table)[page].push_back(image_id); + }); if (False(image.flags & ImageFlagBits::Sparse)) { auto map_id = slot_map_views.insert(image.gpu_addr, image.cpu_addr, image.guest_size_bytes, image_id); @@ -1486,9 +1488,9 @@ void TextureCache

::UnregisterImage(ImageId image_id) { image.flags &= ~ImageFlagBits::BadOverlap; lru_cache.Free(image.lru_index); const auto& clear_page_table = - [this, image_id]( - u64 page, - std::unordered_map, IdentityHash>& selected_page_table) { + [this, image_id](u64 page, + std::unordered_map, Common::IdentityHash>& + selected_page_table) { const auto page_it = selected_page_table.find(page); if (page_it == selected_page_table.end()) { ASSERT_MSG(false, "Unregistering unregistered page=0x{:x}", page << YUZU_PAGEBITS); @@ -1504,7 +1506,7 @@ void TextureCache

::UnregisterImage(ImageId image_id) { image_ids.erase(vector_it); }; ForEachGPUPage(image.gpu_addr, image.guest_size_bytes, [this, &clear_page_table](u64 page) { - clear_page_table(page, state->gpu_page_table); + clear_page_table(page, (*channel_state->gpu_page_table)); }); if (False(image.flags & ImageFlagBits::Sparse)) { const auto map_id = image.map_view_id; @@ -1701,11 +1703,11 @@ void TextureCache

::DeleteImage(ImageId image_id, bool immediate_delete) { template void TextureCache

::RemoveImageViewReferences(std::span removed_views) { - auto it = state->image_views.begin(); - while (it != state->image_views.end()) { + auto it = channel_state->image_views.begin(); + while (it != channel_state->image_views.end()) { const auto found = std::ranges::find(removed_views, it->second); if (found != removed_views.end()) { - it = state->image_views.erase(it); + it = channel_state->image_views.erase(it); } else { ++it; } @@ -1967,61 +1969,19 @@ bool TextureCache

::IsFullClear(ImageViewId id) { scissor.max_y >= size.height; } -template -TextureCache

::ChannelInfo::ChannelInfo(Tegra::Control::ChannelState& state) noexcept - : maxwell3d{*state.maxwell_3d}, kepler_compute{*state.kepler_compute}, - gpu_memory{*state.memory_manager}, graphics_image_table{gpu_memory}, - graphics_sampler_table{gpu_memory}, compute_image_table{gpu_memory}, compute_sampler_table{ - gpu_memory} {} - template void TextureCache

::CreateChannel(struct Tegra::Control::ChannelState& channel) { - ASSERT(channel_map.find(channel.bind_id) == channel_map.end() && channel.bind_id >= 0); - auto new_id = [this, &channel]() { - if (!free_channel_ids.empty()) { - auto id = free_channel_ids.front(); - free_channel_ids.pop_front(); - new (&channel_storage[id]) ChannelInfo(channel); - return id; - } - channel_storage.emplace_back(channel); - return channel_storage.size() - 1; - }(); - channel_map.emplace(channel.bind_id, new_id); - if (current_channel_id != UNSET_CHANNEL) { - state = &channel_storage[current_channel_id]; - } + VideoCommon::ChannelSetupCaches::CreateChannel(channel); + const auto it = channel_map.find(channel.bind_id); + auto* this_state = &channel_storage[it->second]; + const auto& this_as_ref = address_spaces[channel.memory_manager->GetID()]; + this_state->gpu_page_table = &gpu_page_table_storage[this_as_ref.storage_id]; } /// Bind a channel for execution. template -void TextureCache

::BindToChannel(s32 id) { - auto it = channel_map.find(id); - ASSERT(it != channel_map.end() && id >= 0); - current_channel_id = it->second; - state = &channel_storage[current_channel_id]; - maxwell3d = &state->maxwell3d; - kepler_compute = &state->kepler_compute; - gpu_memory = &state->gpu_memory; -} - -/// Erase channel's state. -template -void TextureCache

::EraseChannel(s32 id) { - const auto it = channel_map.find(id); - ASSERT(it != channel_map.end() && id >= 0); - const auto this_id = it->second; - free_channel_ids.push_back(this_id); - channel_map.erase(it); - if (this_id == current_channel_id) { - current_channel_id = UNSET_CHANNEL; - state = nullptr; - maxwell3d = nullptr; - kepler_compute = nullptr; - gpu_memory = nullptr; - } else if (current_channel_id != UNSET_CHANNEL) { - state = &channel_storage[current_channel_id]; - } +void TextureCache

::OnGPUASRegister([[maybe_unused]] size_t map_id) { + gpu_page_table_storage.emplace_back(); } } // namespace VideoCommon diff --git a/src/video_core/texture_cache/texture_cache_base.h b/src/video_core/texture_cache/texture_cache_base.h index 69efcb718..b24968b03 100644 --- a/src/video_core/texture_cache/texture_cache_base.h +++ b/src/video_core/texture_cache/texture_cache_base.h @@ -1,5 +1,7 @@ -// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project -// SPDX-License-Identifier: GPL-2.0-or-later +// SPDX-FileCopyrightText: 2021 yuzu emulator team +// (https://github.com/skyline-emu/) +// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 +// or any later version Refer to the license.txt file included. #pragma once @@ -13,9 +15,11 @@ #include #include "common/common_types.h" +#include "common/hash.h" #include "common/literals.h" #include "common/lru_cache.h" #include "video_core/compatible_formats.h" +#include "video_core/control/channel_state_cache.h" #include "video_core/delayed_destruction_ring.h" #include "video_core/engines/fermi_2d.h" #include "video_core/surface.h" @@ -50,8 +54,35 @@ struct ImageViewInOut { ImageViewId id{}; }; +using TextureCacheGPUMap = std::unordered_map, Common::IdentityHash>; + +class TextureCacheChannelInfo : public ChannelInfo { +public: + TextureCacheChannelInfo() = delete; + TextureCacheChannelInfo(Tegra::Control::ChannelState& state) noexcept; + TextureCacheChannelInfo(const TextureCacheChannelInfo& state) = delete; + TextureCacheChannelInfo& operator=(const TextureCacheChannelInfo&) = delete; + TextureCacheChannelInfo(TextureCacheChannelInfo&& other) noexcept = default; + TextureCacheChannelInfo& operator=(TextureCacheChannelInfo&& other) noexcept = default; + + DescriptorTable graphics_image_table{gpu_memory}; + DescriptorTable graphics_sampler_table{gpu_memory}; + std::vector graphics_sampler_ids; + std::vector graphics_image_view_ids; + + DescriptorTable compute_image_table{gpu_memory}; + DescriptorTable compute_sampler_table{gpu_memory}; + std::vector compute_sampler_ids; + std::vector compute_image_view_ids; + + std::unordered_map image_views; + std::unordered_map samplers; + + TextureCacheGPUMap* gpu_page_table; +}; + template -class TextureCache { +class TextureCache : public VideoCommon::ChannelSetupCaches { /// Address shift for caching images into a hash table static constexpr u64 YUZU_PAGEBITS = 20; @@ -85,13 +116,6 @@ class TextureCache { PixelFormat src_format; }; - template - struct IdentityHash { - [[nodiscard]] size_t operator()(T value) const noexcept { - return static_cast(value); - } - }; - public: explicit TextureCache(Runtime&, VideoCore::RasterizerInterface&); @@ -179,13 +203,7 @@ public: [[nodiscard]] bool IsRescaling(const ImageViewBase& image_view) const noexcept; /// Create channel state. - void CreateChannel(struct Tegra::Control::ChannelState& channel); - - /// Bind a channel for execution. - void BindToChannel(s32 id); - - /// Erase channel's state. - void EraseChannel(s32 id); + void CreateChannel(Tegra::Control::ChannelState& channel) final override; std::mutex mutex; @@ -221,6 +239,8 @@ private: } } + void OnGPUASRegister(size_t map_id) final override; + /// Runs the Garbage Collector. void RunGarbageCollector(); @@ -355,51 +375,15 @@ private: Runtime& runtime; - struct ChannelInfo { - ChannelInfo() = delete; - ChannelInfo(struct Tegra::Control::ChannelState& state) noexcept; - ChannelInfo(const ChannelInfo& state) = delete; - ChannelInfo& operator=(const ChannelInfo&) = delete; - ChannelInfo(ChannelInfo&& other) noexcept = default; - ChannelInfo& operator=(ChannelInfo&& other) noexcept = default; - - Tegra::Engines::Maxwell3D& maxwell3d; - Tegra::Engines::KeplerCompute& kepler_compute; - Tegra::MemoryManager& gpu_memory; - - DescriptorTable graphics_image_table{gpu_memory}; - DescriptorTable graphics_sampler_table{gpu_memory}; - std::vector graphics_sampler_ids; - std::vector graphics_image_view_ids; - - DescriptorTable compute_image_table{gpu_memory}; - DescriptorTable compute_sampler_table{gpu_memory}; - std::vector compute_sampler_ids; - std::vector compute_image_view_ids; - - std::unordered_map image_views; - std::unordered_map samplers; - - std::unordered_map, IdentityHash> gpu_page_table; - }; - - std::deque channel_storage; - std::deque free_channel_ids; - std::unordered_map channel_map; - - ChannelInfo* state; - size_t current_channel_id{UNSET_CHANNEL}; VideoCore::RasterizerInterface& rasterizer; - Tegra::Engines::Maxwell3D* maxwell3d; - Tegra::Engines::KeplerCompute* kepler_compute; - Tegra::MemoryManager* gpu_memory; + std::deque gpu_page_table_storage; RenderTargets render_targets; std::unordered_map framebuffers; - std::unordered_map, IdentityHash> page_table; - std::unordered_map, IdentityHash> sparse_page_table; + std::unordered_map, Common::IdentityHash> page_table; + std::unordered_map, Common::IdentityHash> sparse_page_table; std::unordered_map> sparse_views; VAddr virtual_invalid_space{}; -- cgit v1.2.3 From 9cf4c8831d6a8b0d9c6e2a48e89b667fd73accee Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Thu, 23 Dec 2021 01:40:45 +0100 Subject: Texture cache: Fix dangling references on multichannel. --- src/video_core/control/channel_state_cache.h | 1 + src/video_core/control/channel_state_cache.inc | 17 ++++++---- src/video_core/texture_cache/texture_cache.h | 45 ++++++++++++++------------ 3 files changed, 36 insertions(+), 27 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/video_core/control/channel_state_cache.h b/src/video_core/control/channel_state_cache.h index c51040c83..9b1d7362b 100644 --- a/src/video_core/control/channel_state_cache.h +++ b/src/video_core/control/channel_state_cache.h @@ -72,6 +72,7 @@ protected: std::deque

channel_storage; std::deque free_channel_ids; std::unordered_map channel_map; + std::vector active_channel_ids; struct AddresSpaceRef { size_t ref_count; size_t storage_id; diff --git a/src/video_core/control/channel_state_cache.inc b/src/video_core/control/channel_state_cache.inc index 185eabc35..d3ae758b2 100644 --- a/src/video_core/control/channel_state_cache.inc +++ b/src/video_core/control/channel_state_cache.inc @@ -1,3 +1,6 @@ + +#include + #include "video_core/control/channel_state.h" #include "video_core/control/channel_state_cache.h" #include "video_core/engines/kepler_compute.h" @@ -27,15 +30,16 @@ void ChannelSetupCaches

::CreateChannel(struct Tegra::Control::ChannelState& c if (current_channel_id != UNSET_CHANNEL) { channel_state = &channel_storage[current_channel_id]; } + active_channel_ids.push_back(new_id); auto as_it = address_spaces.find(channel.memory_manager->GetID()); if (as_it != address_spaces.end()) { - as_it->second.ref_count++; - return; + as_it->second.ref_count++; + return; } AddresSpaceRef new_gpu_mem_ref{ - .ref_count = 1, - .storage_id = address_spaces.size(), - .gpu_memory = channel.memory_manager.get(), + .ref_count = 1, + .storage_id = address_spaces.size(), + .gpu_memory = channel.memory_manager.get(), }; address_spaces.emplace(channel.memory_manager->GetID(), new_gpu_mem_ref); OnGPUASRegister(channel.memory_manager->GetID()); @@ -73,7 +77,8 @@ void ChannelSetupCaches

::EraseChannel(s32 id) { } else if (current_channel_id != UNSET_CHANNEL) { channel_state = &channel_storage[current_channel_id]; } + active_channel_ids.erase( + std::find(active_channel_ids.begin(), active_channel_ids.end(), this_id)); } - } // namespace VideoCommon diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 89c5faf88..5ef07d18f 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -41,9 +41,6 @@ TextureCache

::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface& sampler_descriptor.mipmap_filter.Assign(Tegra::Texture::TextureMipmapFilter::Linear); sampler_descriptor.cubemap_anisotropy.Assign(1); - // Setup channels - current_channel_id = UNSET_CHANNEL; - // Make sure the first index is reserved for the null resources // This way the null resource becomes a compile time constant void(slot_images.insert(NullImageParams{})); @@ -886,13 +883,15 @@ void TextureCache

::InvalidateScale(Image& image) { } image.image_view_ids.clear(); image.image_view_infos.clear(); - auto& channel_info = channel_storage[image.channel]; - if constexpr (ENABLE_VALIDATION) { - std::ranges::fill(channel_info.graphics_image_view_ids, CORRUPT_ID); - std::ranges::fill(channel_info.compute_image_view_ids, CORRUPT_ID); + for (size_t c : active_channel_ids) { + auto& channel_info = channel_storage[c]; + if constexpr (ENABLE_VALIDATION) { + std::ranges::fill(channel_info.graphics_image_view_ids, CORRUPT_ID); + std::ranges::fill(channel_info.compute_image_view_ids, CORRUPT_ID); + } + channel_info.graphics_image_table.Invalidate(); + channel_info.compute_image_table.Invalidate(); } - channel_info.graphics_image_table.Invalidate(); - channel_info.compute_image_table.Invalidate(); has_deleted_images = true; } @@ -1690,26 +1689,30 @@ void TextureCache

::DeleteImage(ImageId image_id, bool immediate_delete) { if (alloc_images.empty()) { image_allocs_table.erase(alloc_it); } - for (auto& this_state : channel_storage) { + for (size_t c : active_channel_ids) { + auto& channel_info = channel_storage[c]; if constexpr (ENABLE_VALIDATION) { - std::ranges::fill(this_state.graphics_image_view_ids, CORRUPT_ID); - std::ranges::fill(this_state.compute_image_view_ids, CORRUPT_ID); + std::ranges::fill(channel_info.graphics_image_view_ids, CORRUPT_ID); + std::ranges::fill(channel_info.compute_image_view_ids, CORRUPT_ID); } - this_state.graphics_image_table.Invalidate(); - this_state.compute_image_table.Invalidate(); + channel_info.graphics_image_table.Invalidate(); + channel_info.compute_image_table.Invalidate(); } has_deleted_images = true; } template void TextureCache

::RemoveImageViewReferences(std::span removed_views) { - auto it = channel_state->image_views.begin(); - while (it != channel_state->image_views.end()) { - const auto found = std::ranges::find(removed_views, it->second); - if (found != removed_views.end()) { - it = channel_state->image_views.erase(it); - } else { - ++it; + for (size_t c : active_channel_ids) { + auto& channel_info = channel_storage[c]; + auto it = channel_info.image_views.begin(); + while (it != channel_info.image_views.end()) { + const auto found = std::ranges::find(removed_views, it->second); + if (found != removed_views.end()) { + it = channel_info.image_views.erase(it); + } else { + ++it; + } } } } -- cgit v1.2.3 From f350c3d74ea7880fc6d21f7f638b0d4a70a3246b Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Sat, 1 Jan 2022 22:03:37 +0100 Subject: Texture cache: Fix the remaining issues with memory mnagement and unmapping. --- .../hle/service/nvdrv/devices/nvhost_as_gpu.cpp | 3 +++ src/video_core/control/channel_state_cache.h | 10 ++++++++ src/video_core/gpu.cpp | 8 +++++++ src/video_core/gpu.h | 2 ++ src/video_core/memory_manager.cpp | 11 ++++++++- src/video_core/rasterizer_interface.h | 2 +- src/video_core/renderer_opengl/gl_rasterizer.cpp | 4 ++-- src/video_core/renderer_opengl/gl_rasterizer.h | 2 +- src/video_core/renderer_vulkan/vk_rasterizer.cpp | 4 ++-- src/video_core/renderer_vulkan/vk_rasterizer.h | 2 +- src/video_core/texture_cache/texture_cache.h | 27 +++++++++++++++++----- src/video_core/texture_cache/texture_cache_base.h | 4 ++-- 12 files changed, 63 insertions(+), 16 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp index 344ddfc90..db2a6c3b2 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp @@ -16,6 +16,7 @@ #include "core/hle/service/nvdrv/devices/nvhost_gpu.h" #include "core/hle/service/nvdrv/nvdrv.h" #include "video_core/control/channel_state.h" +#include "video_core/gpu.h" #include "video_core/memory_manager.h" #include "video_core/rasterizer_interface.h" @@ -24,6 +25,7 @@ namespace Service::Nvidia::Devices { nvhost_as_gpu::nvhost_as_gpu(Core::System& system_, Module& module_, NvCore::Container& core) : nvdevice{system_}, module{module_}, container{core}, nvmap{core.GetNvMapFile()}, vm{}, gmmu{} {} + nvhost_as_gpu::~nvhost_as_gpu() = default; NvResult nvhost_as_gpu::Ioctl1(DeviceFD fd, Ioctl command, const std::vector& input, @@ -132,6 +134,7 @@ NvResult nvhost_as_gpu::AllocAsEx(const std::vector& input, std::vector& vm.big_page_allocator = std::make_unique(start_big_pages, end_big_pages); gmmu = std::make_shared(system, 40, VM::PAGE_SIZE_BITS); + system.GPU().InitAddressSpace(*gmmu); vm.initialised = true; return NvResult::Success; diff --git a/src/video_core/control/channel_state_cache.h b/src/video_core/control/channel_state_cache.h index 9b1d7362b..31d80e8b7 100644 --- a/src/video_core/control/channel_state_cache.h +++ b/src/video_core/control/channel_state_cache.h @@ -3,6 +3,7 @@ #include #include #include +#include #include #include "common/common_types.h" @@ -59,6 +60,15 @@ public: return ref->second.gpu_memory; } + std::optional getStorageID(size_t id) const { + std::unique_lock lk(config_mutex); + const auto ref = address_spaces.find(id); + if (ref == address_spaces.end()) { + return std::nullopt; + } + return ref->second.storage_id; + } + protected: static constexpr size_t UNSET_CHANNEL{std::numeric_limits::max()}; diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 8c0ff0094..eebd7f3ff 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -73,6 +73,10 @@ struct GPU::Impl { rasterizer->InitializeChannel(to_init); } + void InitAddressSpace(Tegra::MemoryManager& memory_manager) { + memory_manager.BindRasterizer(rasterizer); + } + void ReleaseChannel(Control::ChannelState& to_release) { UNIMPLEMENTED(); } @@ -452,6 +456,10 @@ void GPU::ReleaseChannel(Control::ChannelState& to_release) { impl->ReleaseChannel(to_release); } +void GPU::InitAddressSpace(Tegra::MemoryManager& memory_manager) { + impl->InitAddressSpace(memory_manager); +} + void GPU::BindRenderer(std::unique_ptr renderer) { impl->BindRenderer(std::move(renderer)); } diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h index 74d55e074..7e84b0d2f 100644 --- a/src/video_core/gpu.h +++ b/src/video_core/gpu.h @@ -118,6 +118,8 @@ public: void ReleaseChannel(Control::ChannelState& to_release); + void InitAddressSpace(Tegra::MemoryManager& memory_manager); + /// Request a host GPU memory flush from the CPU. [[nodiscard]] u64 RequestFlush(VAddr addr, std::size_t size); diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index d4c0dca78..b36067613 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -59,10 +59,19 @@ GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cp } for (u64 offset{}; offset < size; offset += page_size) { const GPUVAddr current_gpu_addr = gpu_addr + offset; + [[maybe_unused]] const auto current_entry_type = GetEntry(current_gpu_addr); SetEntry(current_gpu_addr, entry_type); + if (current_entry_type != entry_type) { + rasterizer->ModifyGPUMemory(unique_identifier, gpu_addr, page_size); + } if constexpr (entry_type == EntryType::Mapped) { const VAddr current_cpu_addr = cpu_addr + offset; const auto index = PageEntryIndex(current_gpu_addr); + const u32 sub_value = static_cast(current_cpu_addr >> 12ULL); + if (current_entry_type == entry_type && sub_value != page_table[index]) { + rasterizer->InvalidateRegion(static_cast(page_table[index]) << 12ULL, + page_size); + } page_table[index] = static_cast(current_cpu_addr >> 12ULL); } remaining_size -= page_size; @@ -168,7 +177,7 @@ std::optional MemoryManager::GpuToCpuAddress(GPUVAddr addr, std::size_t s const size_t page_last{(addr + size + page_size - 1) >> page_bits}; while (page_index < page_last) { const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; - if (page_addr && *page_addr != 0) { + if (page_addr) { return page_addr; } ++page_index; diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h index 8dacb2626..5362aafb6 100644 --- a/src/video_core/rasterizer_interface.h +++ b/src/video_core/rasterizer_interface.h @@ -95,7 +95,7 @@ public: virtual void UnmapMemory(VAddr addr, u64 size) = 0; /// Remap GPU memory range. This means underneath backing memory changed - virtual void ModifyGPUMemory(GPUVAddr addr, u64 size) = 0; + virtual void ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) = 0; /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory /// and invalidated diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 4eb9bd116..f745fbf56 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -379,10 +379,10 @@ void RasterizerOpenGL::UnmapMemory(VAddr addr, u64 size) { shader_cache.OnCPUWrite(addr, size); } -void RasterizerOpenGL::ModifyGPUMemory(GPUVAddr addr, u64 size) { +void RasterizerOpenGL::ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) { { std::scoped_lock lock{texture_cache.mutex}; - texture_cache.UnmapGPUMemory(addr, size); + texture_cache.UnmapGPUMemory(as_id, addr, size); } } diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 39b20cfb2..d469075a1 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -82,7 +82,7 @@ public: void OnCPUWrite(VAddr addr, u64 size) override; void SyncGuestHost() override; void UnmapMemory(VAddr addr, u64 size) override; - void ModifyGPUMemory(GPUVAddr addr, u64 size) override; + void ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) override; void SignalSemaphore(GPUVAddr addr, u32 value) override; void SignalSyncPoint(u32 value) override; void SignalReference() override; diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index 7e0805544..50fdf5e18 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -441,10 +441,10 @@ void RasterizerVulkan::UnmapMemory(VAddr addr, u64 size) { pipeline_cache.OnCPUWrite(addr, size); } -void RasterizerVulkan::ModifyGPUMemory(GPUVAddr addr, u64 size) { +void RasterizerVulkan::ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) { { std::scoped_lock lock{texture_cache.mutex}; - texture_cache.UnmapGPUMemory(addr, size); + texture_cache.UnmapGPUMemory(as_id, addr, size); } } diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h index 642fe6576..c836158b8 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.h +++ b/src/video_core/renderer_vulkan/vk_rasterizer.h @@ -78,7 +78,7 @@ public: void OnCPUWrite(VAddr addr, u64 size) override; void SyncGuestHost() override; void UnmapMemory(VAddr addr, u64 size) override; - void ModifyGPUMemory(GPUVAddr addr, u64 size) override; + void ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) override; void SignalSemaphore(GPUVAddr addr, u32 value) override; void SignalSyncPoint(u32 value) override; void SignalReference() override; diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 5ef07d18f..a483892e4 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -480,12 +480,20 @@ void TextureCache

::UnmapMemory(VAddr cpu_addr, size_t size) { } template -void TextureCache

::UnmapGPUMemory(GPUVAddr gpu_addr, size_t size) { +void TextureCache

::UnmapGPUMemory(size_t as_id, GPUVAddr gpu_addr, size_t size) { std::vector deleted_images; - ForEachImageInRegionGPU(gpu_addr, size, + ForEachImageInRegionGPU(as_id, gpu_addr, size, [&](ImageId id, Image&) { deleted_images.push_back(id); }); for (const ImageId id : deleted_images) { Image& image = slot_images[id]; + if (True(image.flags & ImageFlagBits::CpuModified)) { + return; + } + image.flags |= ImageFlagBits::CpuModified; + if (True(image.flags & ImageFlagBits::Tracked)) { + UntrackImage(image, id); + } + /* if (True(image.flags & ImageFlagBits::Remapped)) { continue; } @@ -493,6 +501,7 @@ void TextureCache

::UnmapGPUMemory(GPUVAddr gpu_addr, size_t size) { if (True(image.flags & ImageFlagBits::Tracked)) { UntrackImage(image, id); } + */ } } @@ -1322,13 +1331,19 @@ void TextureCache

::ForEachImageInRegion(VAddr cpu_addr, size_t size, Func&& f template template -void TextureCache

::ForEachImageInRegionGPU(GPUVAddr gpu_addr, size_t size, Func&& func) { +void TextureCache

::ForEachImageInRegionGPU(size_t as_id, GPUVAddr gpu_addr, size_t size, + Func&& func) { using FuncReturn = typename std::invoke_result::type; static constexpr bool BOOL_BREAK = std::is_same_v; boost::container::small_vector images; - ForEachGPUPage(gpu_addr, size, [this, &images, gpu_addr, size, func](u64 page) { - const auto it = channel_state->gpu_page_table->find(page); - if (it == channel_state->gpu_page_table->end()) { + auto storage_id = getStorageID(as_id); + if (!storage_id) { + return; + } + auto& gpu_page_table = gpu_page_table_storage[*storage_id]; + ForEachGPUPage(gpu_addr, size, [this, gpu_page_table, &images, gpu_addr, size, func](u64 page) { + const auto it = gpu_page_table.find(page); + if (it == gpu_page_table.end()) { if constexpr (BOOL_BREAK) { return false; } else { diff --git a/src/video_core/texture_cache/texture_cache_base.h b/src/video_core/texture_cache/texture_cache_base.h index b24968b03..2f4db5047 100644 --- a/src/video_core/texture_cache/texture_cache_base.h +++ b/src/video_core/texture_cache/texture_cache_base.h @@ -173,7 +173,7 @@ public: void UnmapMemory(VAddr cpu_addr, size_t size); /// Remove images in a region - void UnmapGPUMemory(GPUVAddr gpu_addr, size_t size); + void UnmapGPUMemory(size_t as_id, GPUVAddr gpu_addr, size_t size); /// Blit an image with the given parameters void BlitImage(const Tegra::Engines::Fermi2D::Surface& dst, @@ -309,7 +309,7 @@ private: void ForEachImageInRegion(VAddr cpu_addr, size_t size, Func&& func); template - void ForEachImageInRegionGPU(GPUVAddr gpu_addr, size_t size, Func&& func); + void ForEachImageInRegionGPU(size_t as_id, GPUVAddr gpu_addr, size_t size, Func&& func); template void ForEachSparseImageInRegion(GPUVAddr gpu_addr, size_t size, Func&& func); -- cgit v1.2.3 From e44ac8b821833672f7c99ad22dbe57ac9403279d Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Thu, 6 Jan 2022 19:47:15 +0100 Subject: Texture Cache: Fix GC and GPU Modified on Joins. --- src/video_core/texture_cache/texture_cache.h | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index a483892e4..66de41f04 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -1758,6 +1758,7 @@ void TextureCache

::SynchronizeAliases(ImageId image_id) { boost::container::small_vector aliased_images; Image& image = slot_images[image_id]; bool any_rescaled = True(image.flags & ImageFlagBits::Rescaled); + bool any_modified = True(image.flags & ImageFlagBits::GpuModified); u64 most_recent_tick = image.modification_tick; for (const AliasedImage& aliased : image.aliased_images) { ImageBase& aliased_image = slot_images[aliased.id]; @@ -1765,9 +1766,7 @@ void TextureCache

::SynchronizeAliases(ImageId image_id) { most_recent_tick = std::max(most_recent_tick, aliased_image.modification_tick); aliased_images.push_back(&aliased); any_rescaled |= True(aliased_image.flags & ImageFlagBits::Rescaled); - if (True(aliased_image.flags & ImageFlagBits::GpuModified)) { - image.flags |= ImageFlagBits::GpuModified; - } + any_modified |= True(aliased_image.flags & ImageFlagBits::GpuModified); } } if (aliased_images.empty()) { @@ -1782,6 +1781,9 @@ void TextureCache

::SynchronizeAliases(ImageId image_id) { } } image.modification_tick = most_recent_tick; + if (any_modified) { + image.flags |= ImageFlagBits::GpuModified; + } std::ranges::sort(aliased_images, [this](const AliasedImage* lhs, const AliasedImage* rhs) { const ImageBase& lhs_image = slot_images[lhs->id]; const ImageBase& rhs_image = slot_images[rhs->id]; -- cgit v1.2.3 From f5fd6b5c8674fcf64a3e70809ee0a34d3a95beb6 Mon Sep 17 00:00:00 2001 From: bunnei Date: Sun, 14 Aug 2022 02:36:36 -0700 Subject: DMA & InlineToMemory Engines Rework. --- src/common/algorithm.h | 8 + src/video_core/buffer_cache/buffer_cache.h | 4 +- src/video_core/engines/engine_upload.cpp | 46 ++++- src/video_core/engines/engine_upload.h | 6 +- src/video_core/engines/kepler_compute.cpp | 13 +- src/video_core/engines/kepler_memory.cpp | 13 +- src/video_core/engines/maxwell_3d.cpp | 5 +- src/video_core/engines/maxwell_dma.cpp | 91 ++++++--- src/video_core/engines/maxwell_dma.h | 6 + src/video_core/host1x/vic.cpp | 5 +- src/video_core/memory_manager.cpp | 91 +++++++++ src/video_core/memory_manager.h | 6 + src/video_core/rasterizer_interface.h | 2 +- src/video_core/renderer_opengl/gl_rasterizer.cpp | 2 +- src/video_core/renderer_opengl/gl_rasterizer.h | 2 +- src/video_core/renderer_vulkan/vk_compute_pass.cpp | 2 - src/video_core/renderer_vulkan/vk_rasterizer.cpp | 2 +- src/video_core/renderer_vulkan/vk_rasterizer.h | 2 +- src/video_core/texture_cache/util.cpp | 1 - src/video_core/textures/decoders.cpp | 225 +++++++-------------- src/video_core/textures/decoders.h | 33 +-- 21 files changed, 323 insertions(+), 242 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/common/algorithm.h b/src/common/algorithm.h index 9ddfd637b..055dca142 100644 --- a/src/common/algorithm.h +++ b/src/common/algorithm.h @@ -24,4 +24,12 @@ template > return first != last && !comp(value, *first) ? first : last; } +template +T FoldRight(T initial_value, Func&& func, Args&&... args) { + T value{initial_value}; + const auto high_func = [&value, &func](T x) { value = func(value, x); }; + (std::invoke(high_func, std::forward(args)), ...); + return value; +} + } // namespace Common diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index e55cac0d6..359c11d6f 100644 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -126,7 +126,7 @@ public: void DownloadMemory(VAddr cpu_addr, u64 size); - bool InlineMemory(VAddr dest_address, size_t copy_size, std::span inlined_buffer); + bool InlineMemory(VAddr dest_address, size_t copy_size, std::span inlined_buffer); void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size); @@ -1685,7 +1685,7 @@ void BufferCache

::MappedUploadMemory(Buffer& buffer, u64 total_size_bytes, template bool BufferCache

::InlineMemory(VAddr dest_address, size_t copy_size, - std::span inlined_buffer) { + std::span inlined_buffer) { const bool is_dirty = IsRegionRegistered(dest_address, copy_size); if (!is_dirty) { return false; diff --git a/src/video_core/engines/engine_upload.cpp b/src/video_core/engines/engine_upload.cpp index 6ff5b1eca..a34819234 100644 --- a/src/video_core/engines/engine_upload.cpp +++ b/src/video_core/engines/engine_upload.cpp @@ -3,6 +3,7 @@ #include +#include "common/algorithm.h" #include "common/assert.h" #include "video_core/engines/engine_upload.h" #include "video_core/memory_manager.h" @@ -34,21 +35,48 @@ void State::ProcessData(const u32 data, const bool is_last_call) { if (!is_last_call) { return; } + ProcessData(inner_buffer); +} + +void State::ProcessData(const u32* data, size_t num_data) { + std::span read_buffer(reinterpret_cast(data), num_data * sizeof(u32)); + ProcessData(read_buffer); +} + +void State::ProcessData(std::span read_buffer) { const GPUVAddr address{regs.dest.Address()}; if (is_linear) { - rasterizer->AccelerateInlineToMemory(address, copy_size, inner_buffer); + if (regs.line_count == 1) { + rasterizer->AccelerateInlineToMemory(address, copy_size, read_buffer); + } else { + for (u32 line = 0; line < regs.line_count; ++line) { + const GPUVAddr dest_line = address + static_cast(line) * regs.dest.pitch; + memory_manager.WriteBlockUnsafe( + dest_line, read_buffer.data() + static_cast(line) * regs.line_length_in, + regs.line_length_in); + } + memory_manager.InvalidateRegion(address, regs.dest.pitch * regs.line_count); + } } else { - UNIMPLEMENTED_IF(regs.dest.z != 0); - UNIMPLEMENTED_IF(regs.dest.depth != 1); - UNIMPLEMENTED_IF(regs.dest.BlockWidth() != 0); - UNIMPLEMENTED_IF(regs.dest.BlockDepth() != 0); + u32 width = regs.dest.width; + u32 x_elements = regs.line_length_in; + u32 x_offset = regs.dest.x; + const u32 bpp_shift = Common::FoldRight( + 4U, [](u32 x, u32 y) { return std::min(x, static_cast(std::countr_zero(y))); }, + width, x_elements, x_offset, static_cast(address)); + width >>= bpp_shift; + x_elements >>= bpp_shift; + x_offset >>= bpp_shift; + const u32 bytes_per_pixel = 1U << bpp_shift; const std::size_t dst_size = Tegra::Texture::CalculateSize( - true, 1, regs.dest.width, regs.dest.height, 1, regs.dest.BlockHeight(), 0); + true, bytes_per_pixel, width, regs.dest.height, regs.dest.depth, + regs.dest.BlockHeight(), regs.dest.BlockDepth()); tmp_buffer.resize(dst_size); memory_manager.ReadBlock(address, tmp_buffer.data(), dst_size); - Tegra::Texture::SwizzleKepler(regs.dest.width, regs.dest.height, regs.dest.x, regs.dest.y, - regs.dest.BlockHeight(), copy_size, inner_buffer.data(), - tmp_buffer.data()); + Tegra::Texture::SwizzleSubrect(tmp_buffer, read_buffer, bytes_per_pixel, width, + regs.dest.height, regs.dest.depth, x_offset, regs.dest.y, + x_elements, regs.line_count, regs.dest.BlockHeight(), + regs.dest.BlockDepth(), regs.line_length_in); memory_manager.WriteBlock(address, tmp_buffer.data(), dst_size); } } diff --git a/src/video_core/engines/engine_upload.h b/src/video_core/engines/engine_upload.h index 94ff3314a..f08f6e36a 100644 --- a/src/video_core/engines/engine_upload.h +++ b/src/video_core/engines/engine_upload.h @@ -3,6 +3,7 @@ #pragma once +#include #include #include "common/bit_field.h" #include "common/common_types.h" @@ -33,7 +34,7 @@ struct Registers { u32 width; u32 height; u32 depth; - u32 z; + u32 layer; u32 x; u32 y; @@ -62,11 +63,14 @@ public: void ProcessExec(bool is_linear_); void ProcessData(u32 data, bool is_last_call); + void ProcessData(const u32* data, size_t num_data); /// Binds a rasterizer to this engine. void BindRasterizer(VideoCore::RasterizerInterface* rasterizer); private: + void ProcessData(std::span read_buffer); + u32 write_offset = 0; u32 copy_size = 0; std::vector inner_buffer; diff --git a/src/video_core/engines/kepler_compute.cpp b/src/video_core/engines/kepler_compute.cpp index 5db254d94..7c50bdbe0 100644 --- a/src/video_core/engines/kepler_compute.cpp +++ b/src/video_core/engines/kepler_compute.cpp @@ -36,8 +36,6 @@ void KeplerCompute::CallMethod(u32 method, u32 method_argument, bool is_last_cal } case KEPLER_COMPUTE_REG_INDEX(data_upload): { upload_state.ProcessData(method_argument, is_last_call); - if (is_last_call) { - } break; } case KEPLER_COMPUTE_REG_INDEX(launch): @@ -50,8 +48,15 @@ void KeplerCompute::CallMethod(u32 method, u32 method_argument, bool is_last_cal void KeplerCompute::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) { - for (std::size_t i = 0; i < amount; i++) { - CallMethod(method, base_start[i], methods_pending - static_cast(i) <= 1); + switch (method) { + case KEPLER_COMPUTE_REG_INDEX(data_upload): + upload_state.ProcessData(base_start, static_cast(amount)); + return; + default: + for (std::size_t i = 0; i < amount; i++) { + CallMethod(method, base_start[i], methods_pending - static_cast(i) <= 1); + } + break; } } diff --git a/src/video_core/engines/kepler_memory.cpp b/src/video_core/engines/kepler_memory.cpp index e2b029542..a3fbab1e5 100644 --- a/src/video_core/engines/kepler_memory.cpp +++ b/src/video_core/engines/kepler_memory.cpp @@ -33,8 +33,6 @@ void KeplerMemory::CallMethod(u32 method, u32 method_argument, bool is_last_call } case KEPLERMEMORY_REG_INDEX(data): { upload_state.ProcessData(method_argument, is_last_call); - if (is_last_call) { - } break; } } @@ -42,8 +40,15 @@ void KeplerMemory::CallMethod(u32 method, u32 method_argument, bool is_last_call void KeplerMemory::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) { - for (std::size_t i = 0; i < amount; i++) { - CallMethod(method, base_start[i], methods_pending - static_cast(i) <= 1); + switch (method) { + case KEPLERMEMORY_REG_INDEX(data): + upload_state.ProcessData(base_start, static_cast(amount)); + return; + default: + for (std::size_t i = 0; i < amount; i++) { + CallMethod(method, base_start[i], methods_pending - static_cast(i) <= 1); + } + break; } } diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index add1ccebe..632052c53 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -239,8 +239,6 @@ void Maxwell3D::ProcessMethodCall(u32 method, u32 argument, u32 nonshadow_argume return upload_state.ProcessExec(regs.exec_upload.linear != 0); case MAXWELL3D_REG_INDEX(data_upload): upload_state.ProcessData(argument, is_last_call); - if (is_last_call) { - } return; case MAXWELL3D_REG_INDEX(fragment_barrier): return rasterizer->FragmentBarrier(); @@ -316,6 +314,9 @@ void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, case MAXWELL3D_REG_INDEX(const_buffer.cb_data) + 15: ProcessCBMultiData(base_start, amount); break; + case MAXWELL3D_REG_INDEX(data_upload): + upload_state.ProcessData(base_start, static_cast(amount)); + return; default: for (std::size_t i = 0; i < amount; i++) { CallMethod(method, base_start[i], methods_pending - static_cast(i) <= 1); diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp index 0efe58282..a12a95ce2 100644 --- a/src/video_core/engines/maxwell_dma.cpp +++ b/src/video_core/engines/maxwell_dma.cpp @@ -1,6 +1,7 @@ // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later +#include "common/algorithm.h" #include "common/assert.h" #include "common/logging/log.h" #include "common/microprofile.h" @@ -54,8 +55,6 @@ void MaxwellDMA::Launch() { const LaunchDMA& launch = regs.launch_dma; ASSERT(launch.interrupt_type == LaunchDMA::InterruptType::NONE); ASSERT(launch.data_transfer_type == LaunchDMA::DataTransferType::NON_PIPELINED); - ASSERT(regs.dst_params.origin.x == 0); - ASSERT(regs.dst_params.origin.y == 0); const bool is_src_pitch = launch.src_memory_layout == LaunchDMA::MemoryLayout::PITCH; const bool is_dst_pitch = launch.dst_memory_layout == LaunchDMA::MemoryLayout::PITCH; @@ -121,12 +120,13 @@ void MaxwellDMA::CopyPitchToPitch() { void MaxwellDMA::CopyBlockLinearToPitch() { UNIMPLEMENTED_IF(regs.src_params.block_size.width != 0); - UNIMPLEMENTED_IF(regs.src_params.block_size.depth != 0); UNIMPLEMENTED_IF(regs.src_params.layer != 0); + const bool is_remapping = regs.launch_dma.remap_enable != 0; + // Optimized path for micro copies. const size_t dst_size = static_cast(regs.pitch_out) * regs.line_count; - if (dst_size < GOB_SIZE && regs.pitch_out <= GOB_SIZE_X && + if (!is_remapping && dst_size < GOB_SIZE && regs.pitch_out <= GOB_SIZE_X && regs.src_params.height > GOB_SIZE_Y) { FastCopyBlockLinearToPitch(); return; @@ -134,10 +134,27 @@ void MaxwellDMA::CopyBlockLinearToPitch() { // Deswizzle the input and copy it over. UNIMPLEMENTED_IF(regs.launch_dma.remap_enable != 0); - const u32 bytes_per_pixel = - regs.launch_dma.remap_enable ? regs.pitch_out / regs.line_length_in : 1; const Parameters& src_params = regs.src_params; - const u32 width = src_params.width; + + const u32 num_remap_components = regs.remap_const.num_dst_components_minus_one + 1; + const u32 remap_components_size = regs.remap_const.component_size_minus_one + 1; + + const u32 base_bpp = !is_remapping ? 1U : num_remap_components * remap_components_size; + + u32 width = src_params.width; + u32 x_elements = regs.line_length_in; + u32 x_offset = src_params.origin.x; + u32 bpp_shift = 0U; + if (!is_remapping) { + bpp_shift = Common::FoldRight( + 4U, [](u32 x, u32 y) { return std::min(x, static_cast(std::countr_zero(y))); }, + width, x_elements, x_offset, static_cast(regs.offset_in)); + width >>= bpp_shift; + x_elements >>= bpp_shift; + x_offset >>= bpp_shift; + } + + const u32 bytes_per_pixel = base_bpp << bpp_shift; const u32 height = src_params.height; const u32 depth = src_params.depth; const u32 block_height = src_params.block_size.height; @@ -155,30 +172,46 @@ void MaxwellDMA::CopyBlockLinearToPitch() { memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size); memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size); - UnswizzleSubrect(regs.line_length_in, regs.line_count, regs.pitch_out, width, bytes_per_pixel, - block_height, src_params.origin.x, src_params.origin.y, write_buffer.data(), - read_buffer.data()); + UnswizzleSubrect(write_buffer, read_buffer, bytes_per_pixel, width, height, depth, x_offset, + src_params.origin.y, x_elements, regs.line_count, block_height, block_depth, + regs.pitch_out); memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size); } void MaxwellDMA::CopyPitchToBlockLinear() { UNIMPLEMENTED_IF_MSG(regs.dst_params.block_size.width != 0, "Block width is not one"); + UNIMPLEMENTED_IF(regs.dst_params.layer != 0); UNIMPLEMENTED_IF(regs.launch_dma.remap_enable != 0); + const bool is_remapping = regs.launch_dma.remap_enable != 0; + const u32 num_remap_components = regs.remap_const.num_dst_components_minus_one + 1; + const u32 remap_components_size = regs.remap_const.component_size_minus_one + 1; + const auto& dst_params = regs.dst_params; - const u32 bytes_per_pixel = - regs.launch_dma.remap_enable ? regs.pitch_in / regs.line_length_in : 1; - const u32 width = dst_params.width; + + const u32 base_bpp = !is_remapping ? 1U : num_remap_components * remap_components_size; + + u32 width = dst_params.width; + u32 x_elements = regs.line_length_in; + u32 x_offset = dst_params.origin.x; + u32 bpp_shift = 0U; + if (!is_remapping) { + bpp_shift = Common::FoldRight( + 4U, [](u32 x, u32 y) { return std::min(x, static_cast(std::countr_zero(y))); }, + width, x_elements, x_offset, static_cast(regs.offset_out)); + width >>= bpp_shift; + x_elements >>= bpp_shift; + x_offset >>= bpp_shift; + } + + const u32 bytes_per_pixel = base_bpp << bpp_shift; const u32 height = dst_params.height; const u32 depth = dst_params.depth; const u32 block_height = dst_params.block_size.height; const u32 block_depth = dst_params.block_size.depth; const size_t dst_size = CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth); - const size_t dst_layer_size = - CalculateSize(true, bytes_per_pixel, width, height, 1, block_height, block_depth); - const size_t src_size = static_cast(regs.pitch_in) * regs.line_count; if (read_buffer.size() < src_size) { @@ -188,32 +221,23 @@ void MaxwellDMA::CopyPitchToBlockLinear() { write_buffer.resize(dst_size); } + memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size); if (Settings::IsGPULevelExtreme()) { - memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size); memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size); } else { - memory_manager.ReadBlockUnsafe(regs.offset_in, read_buffer.data(), src_size); memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size); } // If the input is linear and the output is tiled, swizzle the input and copy it over. - if (regs.dst_params.block_size.depth > 0) { - ASSERT(dst_params.layer == 0); - SwizzleSliceToVoxel(regs.line_length_in, regs.line_count, regs.pitch_in, width, height, - bytes_per_pixel, block_height, block_depth, dst_params.origin.x, - dst_params.origin.y, write_buffer.data(), read_buffer.data()); - } else { - SwizzleSubrect(regs.line_length_in, regs.line_count, regs.pitch_in, width, bytes_per_pixel, - write_buffer.data() + dst_layer_size * dst_params.layer, read_buffer.data(), - block_height, dst_params.origin.x, dst_params.origin.y); - } + SwizzleSubrect(write_buffer, read_buffer, bytes_per_pixel, width, height, depth, x_offset, + dst_params.origin.y, x_elements, regs.line_count, block_height, block_depth, + regs.pitch_in); memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size); } void MaxwellDMA::FastCopyBlockLinearToPitch() { - const u32 bytes_per_pixel = - regs.launch_dma.remap_enable ? regs.pitch_out / regs.line_length_in : 1; + const u32 bytes_per_pixel = 1U; const size_t src_size = GOB_SIZE; const size_t dst_size = static_cast(regs.pitch_out) * regs.line_count; u32 pos_x = regs.src_params.origin.x; @@ -239,9 +263,10 @@ void MaxwellDMA::FastCopyBlockLinearToPitch() { memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size); } - UnswizzleSubrect(regs.line_length_in, regs.line_count, regs.pitch_out, regs.src_params.width, - bytes_per_pixel, regs.src_params.block_size.height, pos_x, pos_y, - write_buffer.data(), read_buffer.data()); + UnswizzleSubrect(write_buffer, read_buffer, bytes_per_pixel, regs.src_params.width, + regs.src_params.height, 1, pos_x, pos_y, regs.line_length_in, regs.line_count, + regs.src_params.block_size.height, regs.src_params.block_size.depth, + regs.pitch_out); memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size); } diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h index 074bac92c..9c5d567a6 100644 --- a/src/video_core/engines/maxwell_dma.h +++ b/src/video_core/engines/maxwell_dma.h @@ -189,10 +189,16 @@ public: BitField<4, 3, Swizzle> dst_y; BitField<8, 3, Swizzle> dst_z; BitField<12, 3, Swizzle> dst_w; + BitField<0, 12, u32> dst_components_raw; BitField<16, 2, u32> component_size_minus_one; BitField<20, 2, u32> num_src_components_minus_one; BitField<24, 2, u32> num_dst_components_minus_one; }; + + Swizzle GetComponent(size_t i) { + const u32 raw = dst_components_raw; + return static_cast((raw >> (i * 3)) & 0x7); + } }; static_assert(sizeof(RemapConst) == 12); diff --git a/src/video_core/host1x/vic.cpp b/src/video_core/host1x/vic.cpp index 5d8039841..b9ac41529 100644 --- a/src/video_core/host1x/vic.cpp +++ b/src/video_core/host1x/vic.cpp @@ -156,8 +156,9 @@ void Vic::WriteRGBFrame(const AVFrame* frame, const VicConfig& config) { const u32 block_height = static_cast(config.block_linear_height_log2); const auto size = Texture::CalculateSize(true, 4, width, height, 1, block_height, 0); luma_buffer.resize(size); - Texture::SwizzleSubrect(width, height, width * 4, width, 4, luma_buffer.data(), - converted_frame_buf_addr, block_height, 0, 0); + std::span frame_buff(converted_frame_buf_addr, 4 * width * height); + Texture::SwizzleSubrect(luma_buffer, frame_buff, 4, width, height, 1, + 0, 0, width, height, block_height, 0, width * 4); host1x.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(), size); } else { diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index 4e52ce0fd..4a692448e 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -462,6 +462,97 @@ void MemoryManager::FlushRegion(GPUVAddr gpu_addr, size_t size) const { MemoryOperation(gpu_addr, size, mapped_big, do_nothing, flush_short_pages); } +bool MemoryManager::IsMemoryDirty(GPUVAddr gpu_addr, size_t size) const { + bool result = false; + auto do_nothing = [&]([[maybe_unused]] std::size_t page_index, + [[maybe_unused]] std::size_t offset, + [[maybe_unused]] std::size_t copy_amount) { return false; }; + + auto mapped_normal = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) { + const VAddr cpu_addr_base = + (static_cast(page_table[page_index]) << cpu_page_bits) + offset; + result |= rasterizer->MustFlushRegion(cpu_addr_base, copy_amount); + return result; + }; + auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) { + const VAddr cpu_addr_base = + (static_cast(big_page_table_cpu[page_index]) << cpu_page_bits) + offset; + result |= rasterizer->MustFlushRegion(cpu_addr_base, copy_amount); + return result; + }; + auto check_short_pages = [&](std::size_t page_index, std::size_t offset, + std::size_t copy_amount) { + GPUVAddr base = (page_index << big_page_bits) + offset; + MemoryOperation(base, copy_amount, mapped_normal, do_nothing, do_nothing); + return result; + }; + MemoryOperation(gpu_addr, size, mapped_big, do_nothing, check_short_pages); + return result; +} + +size_t MemoryManager::MaxContinousRange(GPUVAddr gpu_addr, size_t size) const { + std::optional old_page_addr{}; + size_t range_so_far = 0; + bool result{false}; + auto fail = [&]([[maybe_unused]] std::size_t page_index, [[maybe_unused]] std::size_t offset, + std::size_t copy_amount) { + result = true; + return true; + }; + auto short_check = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) { + const VAddr cpu_addr_base = + (static_cast(page_table[page_index]) << cpu_page_bits) + offset; + if (old_page_addr && *old_page_addr != cpu_addr_base) { + result = true; + return true; + } + range_so_far += copy_amount; + old_page_addr = {cpu_addr_base + copy_amount}; + return false; + }; + auto big_check = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) { + const VAddr cpu_addr_base = + (static_cast(big_page_table_cpu[page_index]) << cpu_page_bits) + offset; + if (old_page_addr && *old_page_addr != cpu_addr_base) { + return true; + } + range_so_far += copy_amount; + old_page_addr = {cpu_addr_base + copy_amount}; + return false; + }; + auto check_short_pages = [&](std::size_t page_index, std::size_t offset, + std::size_t copy_amount) { + GPUVAddr base = (page_index << big_page_bits) + offset; + MemoryOperation(base, copy_amount, short_check, fail, fail); + return result; + }; + MemoryOperation(gpu_addr, size, big_check, fail, check_short_pages); + return range_so_far; +} + +void MemoryManager::InvalidateRegion(GPUVAddr gpu_addr, size_t size) const { + auto do_nothing = [&]([[maybe_unused]] std::size_t page_index, + [[maybe_unused]] std::size_t offset, + [[maybe_unused]] std::size_t copy_amount) {}; + + auto mapped_normal = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) { + const VAddr cpu_addr_base = + (static_cast(page_table[page_index]) << cpu_page_bits) + offset; + rasterizer->InvalidateRegion(cpu_addr_base, copy_amount); + }; + auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) { + const VAddr cpu_addr_base = + (static_cast(big_page_table_cpu[page_index]) << cpu_page_bits) + offset; + rasterizer->InvalidateRegion(cpu_addr_base, copy_amount); + }; + auto invalidate_short_pages = [&](std::size_t page_index, std::size_t offset, + std::size_t copy_amount) { + GPUVAddr base = (page_index << big_page_bits) + offset; + MemoryOperation(base, copy_amount, mapped_normal, do_nothing, do_nothing); + }; + MemoryOperation(gpu_addr, size, mapped_big, do_nothing, invalidate_short_pages); +} + void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size) { std::vector tmp_buffer(size); ReadBlock(gpu_src_addr, tmp_buffer.data(), size); diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index 8f8877a92..9c08edc20 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -104,6 +104,12 @@ public: void FlushRegion(GPUVAddr gpu_addr, size_t size) const; + void InvalidateRegion(GPUVAddr gpu_addr, size_t size) const; + + bool IsMemoryDirty(GPUVAddr gpu_addr, size_t size) const; + + size_t MaxContinousRange(GPUVAddr gpu_addr, size_t size) const; + private: template inline void MemoryOperation(GPUVAddr gpu_src_addr, std::size_t size, FuncMapped&& func_mapped, diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h index cb07f3d38..d2d40884c 100644 --- a/src/video_core/rasterizer_interface.h +++ b/src/video_core/rasterizer_interface.h @@ -129,7 +129,7 @@ public: [[nodiscard]] virtual Tegra::Engines::AccelerateDMAInterface& AccessAccelerateDMA() = 0; virtual void AccelerateInlineToMemory(GPUVAddr address, size_t copy_size, - std::span memory) = 0; + std::span memory) = 0; /// Attempt to use a faster method to display the framebuffer to screen [[nodiscard]] virtual bool AccelerateDisplay(const Tegra::FramebufferConfig& config, diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 02bb17715..c2d80605d 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -476,7 +476,7 @@ Tegra::Engines::AccelerateDMAInterface& RasterizerOpenGL::AccessAccelerateDMA() } void RasterizerOpenGL::AccelerateInlineToMemory(GPUVAddr address, size_t copy_size, - std::span memory) { + std::span memory) { auto cpu_addr = gpu_memory->GpuToCpuAddress(address); if (!cpu_addr) [[unlikely]] { gpu_memory->WriteBlock(address, memory.data(), copy_size); diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index fe0ba979a..45131b785 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -99,7 +99,7 @@ public: const Tegra::Engines::Fermi2D::Config& copy_config) override; Tegra::Engines::AccelerateDMAInterface& AccessAccelerateDMA() override; void AccelerateInlineToMemory(GPUVAddr address, size_t copy_size, - std::span memory) override; + std::span memory) override; bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr, u32 pixel_stride) override; void LoadDiskResources(u64 title_id, std::stop_token stop_loading, diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.cpp b/src/video_core/renderer_vulkan/vk_compute_pass.cpp index f17a5ccd6..241d7573e 100644 --- a/src/video_core/renderer_vulkan/vk_compute_pass.cpp +++ b/src/video_core/renderer_vulkan/vk_compute_pass.cpp @@ -26,8 +26,6 @@ namespace Vulkan { -using Tegra::Texture::SWIZZLE_TABLE; - namespace { constexpr u32 ASTC_BINDING_INPUT_BUFFER = 0; diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index a35e41199..acfd5da7d 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -548,7 +548,7 @@ Tegra::Engines::AccelerateDMAInterface& RasterizerVulkan::AccessAccelerateDMA() } void RasterizerVulkan::AccelerateInlineToMemory(GPUVAddr address, size_t copy_size, - std::span memory) { + std::span memory) { auto cpu_addr = gpu_memory->GpuToCpuAddress(address); if (!cpu_addr) [[unlikely]] { gpu_memory->WriteBlock(address, memory.data(), copy_size); diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h index fb9e83e8f..4cde3c983 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.h +++ b/src/video_core/renderer_vulkan/vk_rasterizer.h @@ -95,7 +95,7 @@ public: const Tegra::Engines::Fermi2D::Config& copy_config) override; Tegra::Engines::AccelerateDMAInterface& AccessAccelerateDMA() override; void AccelerateInlineToMemory(GPUVAddr address, size_t copy_size, - std::span memory) override; + std::span memory) override; bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr, u32 pixel_stride) override; void LoadDiskResources(u64 title_id, std::stop_token stop_loading, diff --git a/src/video_core/texture_cache/util.cpp b/src/video_core/texture_cache/util.cpp index bea1c27d0..1223df5a0 100644 --- a/src/video_core/texture_cache/util.cpp +++ b/src/video_core/texture_cache/util.cpp @@ -517,7 +517,6 @@ void SwizzleBlockLinearImage(Tegra::MemoryManager& gpu_memory, GPUVAddr gpu_addr const u32 host_bytes_per_layer = num_blocks_per_layer * bytes_per_block; UNIMPLEMENTED_IF(info.tile_width_spacing > 0); - UNIMPLEMENTED_IF(copy.image_offset.x != 0); UNIMPLEMENTED_IF(copy.image_offset.y != 0); UNIMPLEMENTED_IF(copy.image_offset.z != 0); diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp index 913f8ebcb..fcc636e0b 100644 --- a/src/video_core/textures/decoders.cpp +++ b/src/video_core/textures/decoders.cpp @@ -89,6 +89,69 @@ void SwizzleImpl(std::span output, std::span input, u32 width, u32 } } +template +void SwizzleSubrectImpl(std::span output, std::span input, u32 width, u32 height, + u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, u32 num_lines, + u32 block_height, u32 block_depth, u32 pitch_linear) { + // The origin of the transformation can be configured here, leave it as zero as the current API + // doesn't expose it. + static constexpr u32 origin_z = 0; + + // We can configure here a custom pitch + // As it's not exposed 'width * BYTES_PER_PIXEL' will be the expected pitch. + const u32 pitch = pitch_linear; + const u32 stride = Common::AlignUpLog2(width * BYTES_PER_PIXEL, GOB_SIZE_X_SHIFT); + + const u32 gobs_in_x = Common::DivCeilLog2(stride, GOB_SIZE_X_SHIFT); + const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth); + const u32 slice_size = + Common::DivCeilLog2(height, block_height + GOB_SIZE_Y_SHIFT) * block_size; + + const u32 block_height_mask = (1U << block_height) - 1; + const u32 block_depth_mask = (1U << block_depth) - 1; + const u32 x_shift = GOB_SIZE_SHIFT + block_height + block_depth; + + u32 unprocessed_lines = num_lines; + u32 extent_y = std::min(num_lines, height - origin_y); + + for (u32 slice = 0; slice < depth; ++slice) { + const u32 z = slice + origin_z; + const u32 offset_z = (z >> block_depth) * slice_size + + ((z & block_depth_mask) << (GOB_SIZE_SHIFT + block_height)); + const u32 lines_in_y = std::min(unprocessed_lines, extent_y); + for (u32 line = 0; line < lines_in_y; ++line) { + const u32 y = line + origin_y; + const u32 swizzled_y = pdep(y); + + const u32 block_y = y >> GOB_SIZE_Y_SHIFT; + const u32 offset_y = (block_y >> block_height) * block_size + + ((block_y & block_height_mask) << GOB_SIZE_SHIFT); + + u32 swizzled_x = pdep(origin_x * BYTES_PER_PIXEL); + for (u32 column = 0; column < extent_x; + ++column, incrpdep(swizzled_x)) { + const u32 x = (column + origin_x) * BYTES_PER_PIXEL; + const u32 offset_x = (x >> GOB_SIZE_X_SHIFT) << x_shift; + + const u32 base_swizzled_offset = offset_z + offset_y + offset_x; + const u32 swizzled_offset = base_swizzled_offset + (swizzled_x | swizzled_y); + + const u32 unswizzled_offset = + slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL; + + u8* const dst = &output[TO_LINEAR ? swizzled_offset : unswizzled_offset]; + const u8* const src = &input[TO_LINEAR ? unswizzled_offset : swizzled_offset]; + + std::memcpy(dst, src, BYTES_PER_PIXEL); + } + } + unprocessed_lines -= lines_in_y; + if (unprocessed_lines == 0) { + return; + } + } +} + template void Swizzle(std::span output, std::span input, u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth, u32 stride_alignment) { @@ -111,97 +174,6 @@ void Swizzle(std::span output, std::span input, u32 bytes_per_pixe } } -template -void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width, - u8* swizzled_data, const u8* unswizzled_data, u32 block_height_bit, - u32 offset_x, u32 offset_y) { - const u32 block_height = 1U << block_height_bit; - const u32 image_width_in_gobs = - (swizzled_width * BYTES_PER_PIXEL + (GOB_SIZE_X - 1)) / GOB_SIZE_X; - for (u32 line = 0; line < subrect_height; ++line) { - const u32 dst_y = line + offset_y; - const u32 gob_address_y = - (dst_y / (GOB_SIZE_Y * block_height)) * GOB_SIZE * block_height * image_width_in_gobs + - ((dst_y % (GOB_SIZE_Y * block_height)) / GOB_SIZE_Y) * GOB_SIZE; - - const u32 swizzled_y = pdep(dst_y); - u32 swizzled_x = pdep(offset_x * BYTES_PER_PIXEL); - for (u32 x = 0; x < subrect_width; - ++x, incrpdep(swizzled_x)) { - const u32 dst_x = x + offset_x; - const u32 gob_address = - gob_address_y + (dst_x * BYTES_PER_PIXEL / GOB_SIZE_X) * GOB_SIZE * block_height; - const u32 swizzled_offset = gob_address + (swizzled_x | swizzled_y); - const u32 unswizzled_offset = line * source_pitch + x * BYTES_PER_PIXEL; - - const u8* const source_line = unswizzled_data + unswizzled_offset; - u8* const dest_addr = swizzled_data + swizzled_offset; - std::memcpy(dest_addr, source_line, BYTES_PER_PIXEL); - } - } -} - -template -void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 block_height, - u32 origin_x, u32 origin_y, u8* output, const u8* input) { - const u32 stride = width * BYTES_PER_PIXEL; - const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) / GOB_SIZE_X; - const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height); - - const u32 block_height_mask = (1U << block_height) - 1; - const u32 x_shift = GOB_SIZE_SHIFT + block_height; - - for (u32 line = 0; line < line_count; ++line) { - const u32 src_y = line + origin_y; - const u32 swizzled_y = pdep(src_y); - - const u32 block_y = src_y >> GOB_SIZE_Y_SHIFT; - const u32 src_offset_y = (block_y >> block_height) * block_size + - ((block_y & block_height_mask) << GOB_SIZE_SHIFT); - - u32 swizzled_x = pdep(origin_x * BYTES_PER_PIXEL); - for (u32 column = 0; column < line_length_in; - ++column, incrpdep(swizzled_x)) { - const u32 src_x = (column + origin_x) * BYTES_PER_PIXEL; - const u32 src_offset_x = (src_x >> GOB_SIZE_X_SHIFT) << x_shift; - - const u32 swizzled_offset = src_offset_y + src_offset_x + (swizzled_x | swizzled_y); - const u32 unswizzled_offset = line * pitch + column * BYTES_PER_PIXEL; - - std::memcpy(output + unswizzled_offset, input + swizzled_offset, BYTES_PER_PIXEL); - } - } -} - -template -void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height, - u32 block_height, u32 block_depth, u32 origin_x, u32 origin_y, u8* output, - const u8* input) { - UNIMPLEMENTED_IF(origin_x > 0); - UNIMPLEMENTED_IF(origin_y > 0); - - const u32 stride = width * BYTES_PER_PIXEL; - const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) / GOB_SIZE_X; - const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth); - - const u32 block_height_mask = (1U << block_height) - 1; - const u32 x_shift = static_cast(GOB_SIZE_SHIFT) + block_height + block_depth; - - for (u32 line = 0; line < line_count; ++line) { - const u32 swizzled_y = pdep(line); - const u32 block_y = line / GOB_SIZE_Y; - const u32 dst_offset_y = - (block_y >> block_height) * block_size + (block_y & block_height_mask) * GOB_SIZE; - - u32 swizzled_x = 0; - for (u32 x = 0; x < line_length_in; ++x, incrpdep(swizzled_x)) { - const u32 dst_offset = - ((x / GOB_SIZE_X) << x_shift) + dst_offset_y + (swizzled_x | swizzled_y); - const u32 src_offset = x * BYTES_PER_PIXEL + line * pitch; - std::memcpy(output + dst_offset, input + src_offset, BYTES_PER_PIXEL); - } - } -} } // Anonymous namespace void UnswizzleTexture(std::span output, std::span input, u32 bytes_per_pixel, @@ -218,15 +190,15 @@ void SwizzleTexture(std::span output, std::span input, u32 bytes_p stride_alignment); } -void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width, - u32 bytes_per_pixel, u8* swizzled_data, const u8* unswizzled_data, - u32 block_height_bit, u32 offset_x, u32 offset_y) { +void SwizzleSubrect(std::span output, std::span input, u32 bytes_per_pixel, u32 width, + u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, u32 extent_y, + u32 block_height, u32 block_depth, u32 pitch_linear) { switch (bytes_per_pixel) { #define BPP_CASE(x) \ case x: \ - return SwizzleSubrect(subrect_width, subrect_height, source_pitch, swizzled_width, \ - swizzled_data, unswizzled_data, block_height_bit, offset_x, \ - offset_y); + return SwizzleSubrectImpl(output, input, width, height, depth, origin_x, \ + origin_y, extent_x, extent_y, block_height, \ + block_depth, pitch_linear); BPP_CASE(1) BPP_CASE(2) BPP_CASE(3) @@ -241,13 +213,15 @@ void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 } } -void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 bytes_per_pixel, - u32 block_height, u32 origin_x, u32 origin_y, u8* output, const u8* input) { +void UnswizzleSubrect(std::span output, std::span input, u32 bytes_per_pixel, + u32 width, u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, + u32 extent_y, u32 block_height, u32 block_depth, u32 pitch_linear) { switch (bytes_per_pixel) { #define BPP_CASE(x) \ case x: \ - return UnswizzleSubrect(line_length_in, line_count, pitch, width, block_height, \ - origin_x, origin_y, output, input); + return SwizzleSubrectImpl(output, input, width, height, depth, origin_x, \ + origin_y, extent_x, extent_y, block_height, \ + block_depth, pitch_linear); BPP_CASE(1) BPP_CASE(2) BPP_CASE(3) @@ -262,55 +236,6 @@ void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width, } } -void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height, - u32 bytes_per_pixel, u32 block_height, u32 block_depth, u32 origin_x, - u32 origin_y, u8* output, const u8* input) { - switch (bytes_per_pixel) { -#define BPP_CASE(x) \ - case x: \ - return SwizzleSliceToVoxel(line_length_in, line_count, pitch, width, height, \ - block_height, block_depth, origin_x, origin_y, output, \ - input); - BPP_CASE(1) - BPP_CASE(2) - BPP_CASE(3) - BPP_CASE(4) - BPP_CASE(6) - BPP_CASE(8) - BPP_CASE(12) - BPP_CASE(16) -#undef BPP_CASE - default: - ASSERT_MSG(false, "Invalid bytes_per_pixel={}", bytes_per_pixel); - } -} - -void SwizzleKepler(const u32 width, const u32 height, const u32 dst_x, const u32 dst_y, - const u32 block_height_bit, const std::size_t copy_size, const u8* source_data, - u8* swizzle_data) { - const u32 block_height = 1U << block_height_bit; - const u32 image_width_in_gobs{(width + GOB_SIZE_X - 1) / GOB_SIZE_X}; - std::size_t count = 0; - for (std::size_t y = dst_y; y < height && count < copy_size; ++y) { - const std::size_t gob_address_y = - (y / (GOB_SIZE_Y * block_height)) * GOB_SIZE * block_height * image_width_in_gobs + - ((y % (GOB_SIZE_Y * block_height)) / GOB_SIZE_Y) * GOB_SIZE; - const u32 swizzled_y = pdep(static_cast(y)); - u32 swizzled_x = pdep(dst_x); - for (std::size_t x = dst_x; x < width && count < copy_size; - ++x, incrpdep(swizzled_x)) { - const std::size_t gob_address = - gob_address_y + (x / GOB_SIZE_X) * GOB_SIZE * block_height; - const std::size_t swizzled_offset = gob_address + (swizzled_x | swizzled_y); - const u8* source_line = source_data + count; - u8* dest_addr = swizzle_data + swizzled_offset; - count++; - - *dest_addr = *source_line; - } - } -} - 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 31a11708f..e70407692 100644 --- a/src/video_core/textures/decoders.h +++ b/src/video_core/textures/decoders.h @@ -40,7 +40,6 @@ constexpr SwizzleTable MakeSwizzleTable() { } return table; } -constexpr SwizzleTable SWIZZLE_TABLE = MakeSwizzleTable(); /// Unswizzles a block linear texture into linear memory. void UnswizzleTexture(std::span output, std::span input, u32 bytes_per_pixel, @@ -57,34 +56,14 @@ std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height u32 block_height, u32 block_depth); /// Copies an untiled subrectangle into a tiled surface. -void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width, - u32 bytes_per_pixel, u8* swizzled_data, const u8* unswizzled_data, - u32 block_height_bit, u32 offset_x, u32 offset_y); +void SwizzleSubrect(std::span output, std::span input, u32 bytes_per_pixel, u32 width, + u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, u32 extent_y, + u32 block_height, u32 block_depth, u32 pitch_linear); /// Copies a tiled subrectangle into a linear surface. -void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 bytes_per_pixel, - u32 block_height, u32 origin_x, u32 origin_y, u8* output, const u8* input); - -/// @brief Swizzles a 2D array of pixels into a 3D texture -/// @param line_length_in Number of pixels per line -/// @param line_count Number of lines -/// @param pitch Number of bytes per line -/// @param width Width of the swizzled texture -/// @param height Height of the swizzled texture -/// @param bytes_per_pixel Number of bytes used per pixel -/// @param block_height Block height shift -/// @param block_depth Block depth shift -/// @param origin_x Column offset in pixels of the swizzled texture -/// @param origin_y Row offset in pixels of the swizzled texture -/// @param output Pointer to the pixels of the swizzled texture -/// @param input Pointer to the 2D array of pixels used as input -/// @pre input and output points to an array large enough to hold the number of bytes used -void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height, - u32 bytes_per_pixel, u32 block_height, u32 block_depth, u32 origin_x, - u32 origin_y, u8* output, const u8* input); - -void SwizzleKepler(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height, - std::size_t copy_size, const u8* source_data, u8* swizzle_data); +void UnswizzleSubrect(std::span output, std::span input, u32 bytes_per_pixel, + u32 width, u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, + u32 extent_y, u32 block_height, u32 block_depth, u32 pitch_linear); /// Obtains the offset of the gob for positions 'dst_x' & 'dst_y' u64 GetGOBOffset(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height, -- cgit v1.2.3 From fd7afda1e802fffe843ac28811470432949fe7ee Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Wed, 13 Apr 2022 20:19:06 +0200 Subject: VideoCore: Implement formats needed for N64 emulation. --- src/video_core/renderer_opengl/maxwell_to_gl.h | 2 +- src/video_core/renderer_vulkan/maxwell_to_vk.cpp | 2 +- src/video_core/renderer_vulkan/vk_texture_cache.cpp | 2 +- src/video_core/surface.h | 8 ++++---- src/video_core/texture_cache/format_lookup_table.cpp | 2 +- src/video_core/texture_cache/formatter.h | 4 ++-- 6 files changed, 10 insertions(+), 10 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h index dfe7f26ca..004421236 100644 --- a/src/video_core/renderer_opengl/maxwell_to_gl.h +++ b/src/video_core/renderer_opengl/maxwell_to_gl.h @@ -87,7 +87,7 @@ constexpr std::array FORMAT_TAB {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_R8, GL_RED, GL_UNSIGNED_BYTE}, // R4G4_UNORM + {GL_R8, GL_RED, GL_UNSIGNED_BYTE}, // G4R4_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 diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp index 6703b8e68..e7104d377 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp @@ -184,7 +184,7 @@ struct FormatTuple { {VK_FORMAT_BC3_SRGB_BLOCK}, // BC3_SRGB {VK_FORMAT_BC7_SRGB_BLOCK}, // BC7_SRGB {VK_FORMAT_R4G4B4A4_UNORM_PACK16, Attachable}, // A4B4G4R4_UNORM - {VK_FORMAT_R4G4_UNORM_PACK8}, // R4G4_UNORM + {VK_FORMAT_R4G4_UNORM_PACK8}, // G4R4_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 diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index caca79d79..8d2728cd7 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -592,7 +592,7 @@ void TryTransformSwizzleIfNeeded(PixelFormat format, std::array BLOCK_WIDTH_TABLE = {{ 4, // BC3_SRGB 4, // BC7_SRGB 1, // A4B4G4R4_UNORM - 1, // R4G4_UNORM + 1, // G4R4_UNORM 4, // ASTC_2D_4X4_SRGB 8, // ASTC_2D_8X8_SRGB 8, // ASTC_2D_8X5_SRGB @@ -323,7 +323,7 @@ constexpr std::array BLOCK_HEIGHT_TABLE = {{ 4, // BC3_SRGB 4, // BC7_SRGB 1, // A4B4G4R4_UNORM - 1, // R4G4_UNORM + 1, // G4R4_UNORM 4, // ASTC_2D_4X4_SRGB 8, // ASTC_2D_8X8_SRGB 5, // ASTC_2D_8X5_SRGB @@ -428,7 +428,7 @@ constexpr std::array BITS_PER_BLOCK_TABLE = {{ 128, // BC3_SRGB 128, // BC7_UNORM 16, // A4B4G4R4_UNORM - 8, // R4G4_UNORM + 8, // G4R4_UNORM 128, // ASTC_2D_4X4_SRGB 128, // ASTC_2D_8X8_SRGB 128, // ASTC_2D_8X5_SRGB diff --git a/src/video_core/texture_cache/format_lookup_table.cpp b/src/video_core/texture_cache/format_lookup_table.cpp index c71694d2a..ad935d386 100644 --- a/src/video_core/texture_cache/format_lookup_table.cpp +++ b/src/video_core/texture_cache/format_lookup_table.cpp @@ -63,7 +63,7 @@ PixelFormat PixelFormatFromTextureInfo(TextureFormat format, ComponentType red, case Hash(TextureFormat::A4B4G4R4, UNORM): return PixelFormat::A4B4G4R4_UNORM; case Hash(TextureFormat::G4R4, UNORM): - return PixelFormat::R4G4_UNORM; + return PixelFormat::G4R4_UNORM; case Hash(TextureFormat::A5B5G5R1, UNORM): return PixelFormat::A5B5G5R1_UNORM; case Hash(TextureFormat::R8, UNORM): diff --git a/src/video_core/texture_cache/formatter.h b/src/video_core/texture_cache/formatter.h index 6881e4c90..acc854715 100644 --- a/src/video_core/texture_cache/formatter.h +++ b/src/video_core/texture_cache/formatter.h @@ -153,8 +153,8 @@ struct fmt::formatter : fmt::formatter Date: Fri, 15 Apr 2022 13:43:27 +0200 Subject: ImageBase: Basic fixes. --- src/video_core/texture_cache/image_base.cpp | 13 +++++-------- 1 file changed, 5 insertions(+), 8 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/video_core/texture_cache/image_base.cpp b/src/video_core/texture_cache/image_base.cpp index f61e09ac7..91512022f 100644 --- a/src/video_core/texture_cache/image_base.cpp +++ b/src/video_core/texture_cache/image_base.cpp @@ -7,6 +7,7 @@ #include #include "common/common_types.h" +#include "common/div_ceil.h" #include "video_core/surface.h" #include "video_core/texture_cache/formatter.h" #include "video_core/texture_cache/image_base.h" @@ -182,10 +183,6 @@ void AddImageAlias(ImageBase& lhs, ImageBase& rhs, ImageId lhs_id, ImageId rhs_i }; const bool is_lhs_compressed = lhs_block.width > 1 || lhs_block.height > 1; const bool is_rhs_compressed = rhs_block.width > 1 || rhs_block.height > 1; - if (is_lhs_compressed && is_rhs_compressed) { - LOG_ERROR(HW_GPU, "Compressed to compressed image aliasing is not implemented"); - return; - } const s32 lhs_mips = lhs.info.resources.levels; const s32 rhs_mips = rhs.info.resources.levels; const s32 num_mips = std::min(lhs_mips - base->level, rhs_mips); @@ -199,12 +196,12 @@ void AddImageAlias(ImageBase& lhs, ImageBase& rhs, ImageId lhs_id, ImageId rhs_i Extent3D lhs_size = MipSize(lhs.info.size, base->level + mip_level); Extent3D rhs_size = MipSize(rhs.info.size, mip_level); if (is_lhs_compressed) { - lhs_size.width /= lhs_block.width; - lhs_size.height /= lhs_block.height; + lhs_size.width = Common::DivCeil(lhs_size.width, lhs_block.width); + lhs_size.height = Common::DivCeil(lhs_size.height, lhs_block.height); } if (is_rhs_compressed) { - rhs_size.width /= rhs_block.width; - rhs_size.height /= rhs_block.height; + rhs_size.width = Common::DivCeil(rhs_size.width, rhs_block.width); + rhs_size.height = Common::DivCeil(rhs_size.height, rhs_block.height); } const Extent3D copy_size{ .width = std::min(lhs_size.width, rhs_size.width), -- cgit v1.2.3 From ada09778d97d39d83353ca54d0d6c9abd5eefc60 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Fri, 15 Apr 2022 12:29:49 +0200 Subject: Vulkan Texture Cache: Limit render area to the max width/height of the targets. --- .../renderer_vulkan/vk_texture_cache.cpp | 30 +++++++++++++++++----- src/video_core/renderer_vulkan/vk_texture_cache.h | 5 ++-- src/video_core/texture_cache/render_targets.h | 1 + src/video_core/texture_cache/texture_cache.h | 2 ++ 4 files changed, 29 insertions(+), 9 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index 8d2728cd7..305ad8aee 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -1474,13 +1474,14 @@ bool Image::BlitScaleHelper(bool scale_up) { }; const VkExtent2D extent{ .width = std::max(scaled_width, info.size.width), - .height = std::max(scaled_height, info.size.width), + .height = std::max(scaled_height, info.size.height), }; auto* view_ptr = blit_view.get(); if (aspect_mask == VK_IMAGE_ASPECT_COLOR_BIT) { if (!blit_framebuffer) { - blit_framebuffer = std::make_unique(*runtime, view_ptr, nullptr, extent); + blit_framebuffer = + std::make_unique(*runtime, view_ptr, nullptr, extent, scale_up); } const auto color_view = blit_view->Handle(Shader::TextureType::Color2D); @@ -1488,7 +1489,8 @@ bool Image::BlitScaleHelper(bool scale_up) { src_region, operation, BLIT_OPERATION); } else if (aspect_mask == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) { if (!blit_framebuffer) { - blit_framebuffer = std::make_unique(*runtime, nullptr, view_ptr, extent); + blit_framebuffer = + std::make_unique(*runtime, nullptr, view_ptr, extent, scale_up); } runtime->blit_image_helper.BlitDepthStencil(blit_framebuffer.get(), blit_view->DepthView(), blit_view->StencilView(), dst_region, @@ -1756,34 +1758,42 @@ Framebuffer::Framebuffer(TextureCacheRuntime& runtime, std::span color_buffers{color_buffer}; - CreateFramebuffer(runtime, color_buffers, depth_buffer); + CreateFramebuffer(runtime, color_buffers, depth_buffer, is_rescaled); } Framebuffer::~Framebuffer() = default; void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime, std::span color_buffers, - ImageView* depth_buffer) { + ImageView* depth_buffer, bool is_rescaled) { std::vector attachments; RenderPassKey renderpass_key{}; s32 num_layers = 1; + const auto& resolution = runtime.resolution; + + u32 width = 0; + u32 height = 0; for (size_t index = 0; index < NUM_RT; ++index) { const ImageView* const color_buffer = color_buffers[index]; if (!color_buffer) { renderpass_key.color_formats[index] = PixelFormat::Invalid; continue; } + width = std::max(width, is_rescaled ? resolution.ScaleUp(color_buffer->size.width) + : color_buffer->size.width); + height = std::max(height, is_rescaled ? resolution.ScaleUp(color_buffer->size.height) + : color_buffer->size.height); attachments.push_back(color_buffer->RenderTarget()); renderpass_key.color_formats[index] = color_buffer->format; num_layers = std::max(num_layers, color_buffer->range.extent.layers); @@ -1794,6 +1804,10 @@ void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime, } const size_t num_colors = attachments.size(); if (depth_buffer) { + width = std::max(width, is_rescaled ? resolution.ScaleUp(depth_buffer->size.width) + : depth_buffer->size.width); + height = std::max(height, is_rescaled ? resolution.ScaleUp(depth_buffer->size.height) + : depth_buffer->size.height); attachments.push_back(depth_buffer->RenderTarget()); renderpass_key.depth_format = depth_buffer->format; num_layers = std::max(num_layers, depth_buffer->range.extent.layers); @@ -1810,6 +1824,8 @@ void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime, renderpass_key.samples = samples; renderpass = runtime.render_pass_cache.Get(renderpass_key); + render_area.width = std::min(render_area.width, width); + render_area.height = std::min(render_area.height, height); num_color_buffers = static_cast(num_colors); framebuffer = runtime.device.GetLogical().CreateFramebuffer({ diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h index 69f06ee7b..0b7ac0df1 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.h +++ b/src/video_core/renderer_vulkan/vk_texture_cache.h @@ -268,7 +268,7 @@ public: ImageView* depth_buffer, const VideoCommon::RenderTargets& key); explicit Framebuffer(TextureCacheRuntime& runtime, ImageView* color_buffer, - ImageView* depth_buffer, VkExtent2D extent); + ImageView* depth_buffer, VkExtent2D extent, bool is_rescaled); ~Framebuffer(); @@ -279,7 +279,8 @@ public: Framebuffer& operator=(Framebuffer&&) = default; void CreateFramebuffer(TextureCacheRuntime& runtime, - std::span color_buffers, ImageView* depth_buffer); + std::span color_buffers, ImageView* depth_buffer, + bool is_rescaled = false); [[nodiscard]] VkFramebuffer Handle() const noexcept { return *framebuffer; diff --git a/src/video_core/texture_cache/render_targets.h b/src/video_core/texture_cache/render_targets.h index da8ffe9ec..1efbd6507 100644 --- a/src/video_core/texture_cache/render_targets.h +++ b/src/video_core/texture_cache/render_targets.h @@ -26,6 +26,7 @@ struct RenderTargets { ImageViewId depth_buffer_id{}; std::array draw_buffers{}; Extent2D size{}; + bool is_rescaled{}; }; } // namespace VideoCommon diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 66de41f04..9a835cefc 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -357,6 +357,7 @@ void TextureCache

::UpdateRenderTargets(bool is_clear) { (maxwell3d->regs.render_area.width * up_scale) >> down_shift, (maxwell3d->regs.render_area.height * up_scale) >> down_shift, }; + render_targets.is_rescaled = is_rescaling; flags[Dirty::DepthBiasGlobal] = true; } @@ -1962,6 +1963,7 @@ std::pair TextureCache

::RenderTargetFromImage( .color_buffer_ids = {color_view_id}, .depth_buffer_id = depth_view_id, .size = {extent.width >> samples_x, extent.height >> samples_y}, + .is_rescaled = is_rescaled, }); return {framebuffer_id, view_id}; } -- cgit v1.2.3 From fedd983f96bcbcc0c39f651db1cca0503d582fd9 Mon Sep 17 00:00:00 2001 From: Morph <39850852+Morph1984@users.noreply.github.com> Date: Wed, 29 Jun 2022 19:27:49 -0400 Subject: general: Format licenses as per SPDX guidelines --- src/common/address_space.cpp | 5 ++--- src/common/address_space.h | 5 ++--- src/common/address_space.inc | 4 ++-- src/common/multi_level_page_table.cpp | 3 +++ src/common/multi_level_page_table.h | 5 ++--- src/common/multi_level_page_table.inc | 5 ++--- src/core/hle/service/nvdrv/core/container.cpp | 7 +++---- src/core/hle/service/nvdrv/core/container.h | 7 +++---- src/core/hle/service/nvdrv/core/nvmap.cpp | 7 +++---- src/core/hle/service/nvdrv/core/nvmap.h | 7 +++---- src/core/hle/service/nvdrv/core/syncpoint_manager.cpp | 7 +++---- src/core/hle/service/nvdrv/core/syncpoint_manager.h | 7 +++---- src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp | 7 +++---- src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h | 7 +++---- src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp | 7 +++---- src/core/hle/service/nvdrv/devices/nvhost_ctrl.h | 7 +++---- src/core/hle/service/nvdrv/nvdata.h | 7 +++---- src/core/hle/service/nvdrv/nvdrv.cpp | 7 +++---- src/core/hle/service/nvdrv/nvdrv.h | 7 +++---- src/core/hle/service/nvdrv/nvdrv_interface.cpp | 7 +++---- src/video_core/control/channel_state.cpp | 5 ++--- src/video_core/control/channel_state.h | 5 ++--- src/video_core/control/channel_state_cache.cpp | 3 +++ src/video_core/control/channel_state_cache.h | 5 ++--- src/video_core/control/channel_state_cache.inc | 2 ++ src/video_core/control/scheduler.cpp | 5 ++--- src/video_core/control/scheduler.h | 5 ++--- src/video_core/engines/puller.cpp | 5 ++--- src/video_core/engines/puller.h | 5 ++--- src/video_core/host1x/control.cpp | 5 ++--- src/video_core/host1x/control.h | 7 +++---- src/video_core/host1x/host1x.cpp | 5 ++--- src/video_core/host1x/host1x.h | 5 ++--- src/video_core/host1x/syncpoint_manager.cpp | 5 ++--- src/video_core/host1x/syncpoint_manager.h | 5 ++--- src/video_core/texture_cache/texture_cache.cpp | 5 ++--- src/video_core/texture_cache/texture_cache.h | 6 ++---- src/video_core/texture_cache/texture_cache_base.h | 6 ++---- 38 files changed, 93 insertions(+), 121 deletions(-) (limited to 'src/video_core/texture_cache') diff --git a/src/common/address_space.cpp b/src/common/address_space.cpp index 6db85be87..866e78dbe 100644 --- a/src/common/address_space.cpp +++ b/src/common/address_space.cpp @@ -1,6 +1,5 @@ -// Copyright © 2021 Skyline Team and Contributors (https://github.com/skyline-emu/) -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #include "common/address_space.inc" diff --git a/src/common/address_space.h b/src/common/address_space.h index 8e13935af..5b3832f07 100644 --- a/src/common/address_space.h +++ b/src/common/address_space.h @@ -1,6 +1,5 @@ -// Copyright © 2021 Skyline Team and Contributors (https://github.com/skyline-emu/) -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/common/address_space.inc b/src/common/address_space.inc index 7cfbb150b..a063782b3 100644 --- a/src/common/address_space.inc +++ b/src/common/address_space.inc @@ -1,5 +1,5 @@ -// SPDX-License-Identifier: GPLv3 or later -// Copyright © 2021 Skyline Team and Contributors (https://github.com/skyline-emu/) +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #include "common/address_space.h" #include "common/assert.h" diff --git a/src/common/multi_level_page_table.cpp b/src/common/multi_level_page_table.cpp index 3a7a75aa7..46e362f3b 100644 --- a/src/common/multi_level_page_table.cpp +++ b/src/common/multi_level_page_table.cpp @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + #include "common/multi_level_page_table.inc" namespace Common { diff --git a/src/common/multi_level_page_table.h b/src/common/multi_level_page_table.h index dde1cc962..08092c89a 100644 --- a/src/common/multi_level_page_table.h +++ b/src/common/multi_level_page_table.h @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later #pragma once diff --git a/src/common/multi_level_page_table.inc b/src/common/multi_level_page_table.inc index 4def6dba8..8ac506fa0 100644 --- a/src/common/multi_level_page_table.inc +++ b/src/common/multi_level_page_table.inc @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later #ifdef _WIN32 #include diff --git a/src/core/hle/service/nvdrv/core/container.cpp b/src/core/hle/service/nvdrv/core/container.cpp index 4175d3d9c..d2a632646 100644 --- a/src/core/hle/service/nvdrv/core/container.cpp +++ b/src/core/hle/service/nvdrv/core/container.cpp @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2022 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-FileCopyrightText: 2022 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #include "core/hle/service/nvdrv/core/container.h" #include "core/hle/service/nvdrv/core/nvmap.h" diff --git a/src/core/hle/service/nvdrv/core/container.h b/src/core/hle/service/nvdrv/core/container.h index e069ade4e..5c8b95803 100644 --- a/src/core/hle/service/nvdrv/core/container.h +++ b/src/core/hle/service/nvdrv/core/container.h @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2022 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-FileCopyrightText: 2022 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/core/hle/service/nvdrv/core/nvmap.cpp b/src/core/hle/service/nvdrv/core/nvmap.cpp index 9b21da6b1..e63ec7717 100644 --- a/src/core/hle/service/nvdrv/core/nvmap.cpp +++ b/src/core/hle/service/nvdrv/core/nvmap.cpp @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2022 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-FileCopyrightText: 2022 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #include "common/alignment.h" #include "common/assert.h" diff --git a/src/core/hle/service/nvdrv/core/nvmap.h b/src/core/hle/service/nvdrv/core/nvmap.h index ef2df3ad7..6d6dac023 100644 --- a/src/core/hle/service/nvdrv/core/nvmap.h +++ b/src/core/hle/service/nvdrv/core/nvmap.h @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2022 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-FileCopyrightText: 2022 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/core/hle/service/nvdrv/core/syncpoint_manager.cpp b/src/core/hle/service/nvdrv/core/syncpoint_manager.cpp index fc4ff3c2f..0bb2aec97 100644 --- a/src/core/hle/service/nvdrv/core/syncpoint_manager.cpp +++ b/src/core/hle/service/nvdrv/core/syncpoint_manager.cpp @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2022 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-FileCopyrightText: 2022 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #include "common/assert.h" #include "core/hle/service/nvdrv/core/syncpoint_manager.h" diff --git a/src/core/hle/service/nvdrv/core/syncpoint_manager.h b/src/core/hle/service/nvdrv/core/syncpoint_manager.h index da456f206..6b71cd33d 100644 --- a/src/core/hle/service/nvdrv/core/syncpoint_manager.h +++ b/src/core/hle/service/nvdrv/core/syncpoint_manager.h @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2022 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-FileCopyrightText: 2022 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp index d1beefba6..b48f7fcaf 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team, Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #include #include diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h index 12e881f0d..86fe71c75 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h +++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team, Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp index 7fffb8e48..5bee4a3d3 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team, Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #include #include diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h index f511c0296..4aa738b41 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h +++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team, Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/core/hle/service/nvdrv/nvdata.h b/src/core/hle/service/nvdrv/nvdata.h index 2ee91f9c4..0e2f47075 100644 --- a/src/core/hle/service/nvdrv/nvdata.h +++ b/src/core/hle/service/nvdrv/nvdata.h @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/core/hle/service/nvdrv/nvdrv.cpp b/src/core/hle/service/nvdrv/nvdrv.cpp index 20bf24ec8..7929443d2 100644 --- a/src/core/hle/service/nvdrv/nvdrv.cpp +++ b/src/core/hle/service/nvdrv/nvdrv.cpp @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #include diff --git a/src/core/hle/service/nvdrv/nvdrv.h b/src/core/hle/service/nvdrv/nvdrv.h index 22836529d..a2aeb80b4 100644 --- a/src/core/hle/service/nvdrv/nvdrv.h +++ b/src/core/hle/service/nvdrv/nvdrv.h @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/core/hle/service/nvdrv/nvdrv_interface.cpp b/src/core/hle/service/nvdrv/nvdrv_interface.cpp index 5e50a04e8..edbdfee43 100644 --- a/src/core/hle/service/nvdrv/nvdrv_interface.cpp +++ b/src/core/hle/service/nvdrv/nvdrv_interface.cpp @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #include #include "common/logging/log.h" diff --git a/src/video_core/control/channel_state.cpp b/src/video_core/control/channel_state.cpp index 3613c4992..b04922ac0 100644 --- a/src/video_core/control/channel_state.cpp +++ b/src/video_core/control/channel_state.cpp @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #include "common/assert.h" #include "video_core/control/channel_state.h" diff --git a/src/video_core/control/channel_state.h b/src/video_core/control/channel_state.h index 08a7591e1..305b21cba 100644 --- a/src/video_core/control/channel_state.h +++ b/src/video_core/control/channel_state.h @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/video_core/control/channel_state_cache.cpp b/src/video_core/control/channel_state_cache.cpp index ec7ba907c..4ebeb6356 100644 --- a/src/video_core/control/channel_state_cache.cpp +++ b/src/video_core/control/channel_state_cache.cpp @@ -1,3 +1,6 @@ +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + #include "video_core/control/channel_state_cache.inc" namespace VideoCommon { diff --git a/src/video_core/control/channel_state_cache.h b/src/video_core/control/channel_state_cache.h index 102947adb..5246192a8 100644 --- a/src/video_core/control/channel_state_cache.h +++ b/src/video_core/control/channel_state_cache.h @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/video_core/control/channel_state_cache.inc b/src/video_core/control/channel_state_cache.inc index d3ae758b2..460313893 100644 --- a/src/video_core/control/channel_state_cache.inc +++ b/src/video_core/control/channel_state_cache.inc @@ -1,3 +1,5 @@ +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #include diff --git a/src/video_core/control/scheduler.cpp b/src/video_core/control/scheduler.cpp index a9bb00aa7..733042690 100644 --- a/src/video_core/control/scheduler.cpp +++ b/src/video_core/control/scheduler.cpp @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #include diff --git a/src/video_core/control/scheduler.h b/src/video_core/control/scheduler.h index c1a773946..305a01e0a 100644 --- a/src/video_core/control/scheduler.h +++ b/src/video_core/control/scheduler.h @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/video_core/engines/puller.cpp b/src/video_core/engines/puller.cpp index c3ed11c13..cca890792 100644 --- a/src/video_core/engines/puller.cpp +++ b/src/video_core/engines/puller.cpp @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #include "common/assert.h" #include "common/logging/log.h" diff --git a/src/video_core/engines/puller.h b/src/video_core/engines/puller.h index b4619e9a8..d4175ee94 100644 --- a/src/video_core/engines/puller.h +++ b/src/video_core/engines/puller.h @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2022 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/video_core/host1x/control.cpp b/src/video_core/host1x/control.cpp index a81c635ae..dceefdb7f 100644 --- a/src/video_core/host1x/control.cpp +++ b/src/video_core/host1x/control.cpp @@ -1,6 +1,5 @@ -// Copyright 2022 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #include "common/assert.h" #include "video_core/host1x/control.h" diff --git a/src/video_core/host1x/control.h b/src/video_core/host1x/control.h index 18a9b56c0..e117888a3 100644 --- a/src/video_core/host1x/control.h +++ b/src/video_core/host1x/control.h @@ -1,7 +1,6 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team and Skyline Team and Contributors -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/video_core/host1x/host1x.cpp b/src/video_core/host1x/host1x.cpp index eb00f4855..7c317a85d 100644 --- a/src/video_core/host1x/host1x.cpp +++ b/src/video_core/host1x/host1x.cpp @@ -1,6 +1,5 @@ -// Copyright 2022 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #include "core/core.h" #include "video_core/host1x/host1x.h" diff --git a/src/video_core/host1x/host1x.h b/src/video_core/host1x/host1x.h index e4b69d75a..7ecf853d9 100644 --- a/src/video_core/host1x/host1x.h +++ b/src/video_core/host1x/host1x.h @@ -1,6 +1,5 @@ -// Copyright 2022 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/video_core/host1x/syncpoint_manager.cpp b/src/video_core/host1x/syncpoint_manager.cpp index 825bd551e..4471bacae 100644 --- a/src/video_core/host1x/syncpoint_manager.cpp +++ b/src/video_core/host1x/syncpoint_manager.cpp @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #include "common/microprofile.h" #include "video_core/host1x/syncpoint_manager.h" diff --git a/src/video_core/host1x/syncpoint_manager.h b/src/video_core/host1x/syncpoint_manager.h index 440b1508a..72220a09a 100644 --- a/src/video_core/host1x/syncpoint_manager.h +++ b/src/video_core/host1x/syncpoint_manager.h @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/video_core/texture_cache/texture_cache.cpp b/src/video_core/texture_cache/texture_cache.cpp index bc905a1a4..8a9a32f44 100644 --- a/src/video_core/texture_cache/texture_cache.cpp +++ b/src/video_core/texture_cache/texture_cache.cpp @@ -1,6 +1,5 @@ -// Copyright 2021 yuzu Emulator Project -// Licensed under GPLv3 or any later version -// Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #include "video_core/control/channel_state_cache.inc" #include "video_core/texture_cache/texture_cache_base.h" diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index 9a835cefc..eaf4a1c95 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -1,7 +1,5 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once diff --git a/src/video_core/texture_cache/texture_cache_base.h b/src/video_core/texture_cache/texture_cache_base.h index 2f4db5047..2fa8445eb 100644 --- a/src/video_core/texture_cache/texture_cache_base.h +++ b/src/video_core/texture_cache/texture_cache_base.h @@ -1,7 +1,5 @@ -// SPDX-FileCopyrightText: 2021 yuzu emulator team -// (https://github.com/skyline-emu/) -// SPDX-License-Identifier: GPL-3.0-or-later Licensed under GPLv3 -// or any later version Refer to the license.txt file included. +// SPDX-FileCopyrightText: 2021 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later #pragma once -- cgit v1.2.3