diff options
author | Fernando Sahmkow <fsahmkow27@gmail.com> | 2023-04-23 03:58:16 +0200 |
---|---|---|
committer | Fernando Sahmkow <fsahmkow27@gmail.com> | 2023-04-29 00:46:31 +0200 |
commit | 80480fe3f70997d0520ef8bf38f5fe530c54f8e5 (patch) | |
tree | e8a44f90839f01fdee44434861b317785f1c3878 /src | |
parent | Implement Async downloads in normal and fix a few issues. (diff) | |
download | yuzu-80480fe3f70997d0520ef8bf38f5fe530c54f8e5.tar yuzu-80480fe3f70997d0520ef8bf38f5fe530c54f8e5.tar.gz yuzu-80480fe3f70997d0520ef8bf38f5fe530c54f8e5.tar.bz2 yuzu-80480fe3f70997d0520ef8bf38f5fe530c54f8e5.tar.lz yuzu-80480fe3f70997d0520ef8bf38f5fe530c54f8e5.tar.xz yuzu-80480fe3f70997d0520ef8bf38f5fe530c54f8e5.tar.zst yuzu-80480fe3f70997d0520ef8bf38f5fe530c54f8e5.zip |
Diffstat (limited to '')
-rw-r--r-- | src/video_core/buffer_cache/buffer_base.h | 26 | ||||
-rw-r--r-- | src/video_core/buffer_cache/buffer_cache.h | 4 | ||||
-rw-r--r-- | src/video_core/buffer_cache/buffer_cache_base.h | 32 | ||||
-rw-r--r-- | src/video_core/buffer_cache/memory_tracker_base.h | 82 | ||||
-rw-r--r-- | src/video_core/renderer_vulkan/vk_buffer_cache.h | 2 |
5 files changed, 78 insertions, 68 deletions
diff --git a/src/video_core/buffer_cache/buffer_base.h b/src/video_core/buffer_cache/buffer_base.h index 66d8bb43c..095f79387 100644 --- a/src/video_core/buffer_cache/buffer_base.h +++ b/src/video_core/buffer_cache/buffer_base.h @@ -55,53 +55,53 @@ public: [[nodiscard]] std::pair<u64, u64> ModifiedCpuRegion(VAddr query_cpu_addr, u64 query_size) const noexcept { const u64 offset = query_cpu_addr - cpu_addr; - return word_manager.ModifiedRegion<Type::CPU>(offset, query_size); + return word_manager.template ModifiedRegion<Type::CPU>(offset, query_size); } /// Returns the inclusive GPU modified range in a begin end pair [[nodiscard]] std::pair<u64, u64> ModifiedGpuRegion(VAddr query_cpu_addr, u64 query_size) const noexcept { const u64 offset = query_cpu_addr - cpu_addr; - return word_manager.ModifiedRegion<Type::GPU>(offset, query_size); + return word_manager.template ModifiedRegion<Type::GPU>(offset, query_size); } /// Returns true if a region has been modified from the CPU [[nodiscard]] bool IsRegionCpuModified(VAddr query_cpu_addr, u64 query_size) const noexcept { const u64 offset = query_cpu_addr - cpu_addr; - return word_manager.IsRegionModified<Type::CPU>(offset, query_size); + return word_manager.template IsRegionModified<Type::CPU>(offset, query_size); } /// Returns true if a region has been modified from the GPU [[nodiscard]] bool IsRegionGpuModified(VAddr query_cpu_addr, u64 query_size) const noexcept { const u64 offset = query_cpu_addr - cpu_addr; - return word_manager.IsRegionModified<Type::GPU>(offset, query_size); + return word_manager.template IsRegionModified<Type::GPU>(offset, query_size); } /// Mark region as CPU modified, notifying the rasterizer about this change void MarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 size) { - word_manager.ChangeRegionState<Type::CPU, true>(dirty_cpu_addr, size); + word_manager.template ChangeRegionState<Type::CPU, true>(dirty_cpu_addr, size); } /// Unmark region as CPU modified, notifying the rasterizer about this change void UnmarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 size) { - word_manager.ChangeRegionState<Type::CPU, false>(dirty_cpu_addr, size); + word_manager.template ChangeRegionState<Type::CPU, false>(dirty_cpu_addr, size); } /// Mark region as modified from the host GPU void MarkRegionAsGpuModified(VAddr dirty_cpu_addr, u64 size) noexcept { - word_manager.ChangeRegionState<Type::GPU, true>(dirty_cpu_addr, size); + word_manager.template ChangeRegionState<Type::GPU, true>(dirty_cpu_addr, size); } /// Unmark region as modified from the host GPU void UnmarkRegionAsGpuModified(VAddr dirty_cpu_addr, u64 size) noexcept { - word_manager.ChangeRegionState<Type::GPU, false>(dirty_cpu_addr, size); + word_manager.template ChangeRegionState<Type::GPU, false>(dirty_cpu_addr, size); } /// Mark region as modified from the CPU /// but don't mark it as modified until FlusHCachedWrites is called. void CachedCpuWrite(VAddr dirty_cpu_addr, u64 size) { flags |= BufferFlagBits::CachedWrites; - word_manager.ChangeRegionState<Type::CachedCPU, true>(dirty_cpu_addr, size); + word_manager.template ChangeRegionState<Type::CachedCPU, true>(dirty_cpu_addr, size); } /// Flushes cached CPU writes, and notify the rasterizer about the deltas @@ -113,24 +113,24 @@ public: /// Call 'func' for each CPU modified range and unmark those pages as CPU modified template <typename Func> void ForEachUploadRange(VAddr query_cpu_range, u64 size, Func&& func) { - word_manager.ForEachModifiedRange<Type::CPU>(query_cpu_range, size, true, func); + word_manager.template ForEachModifiedRange<Type::CPU>(query_cpu_range, size, true, func); } /// Call 'func' for each GPU modified range and unmark those pages as GPU modified template <typename Func> void ForEachDownloadRange(VAddr query_cpu_range, u64 size, bool clear, Func&& func) { - word_manager.ForEachModifiedRange<Type::GPU>(query_cpu_range, size, clear, func); + word_manager.template ForEachModifiedRange<Type::GPU>(query_cpu_range, size, clear, func); } template <typename Func> void ForEachDownloadRangeAndClear(VAddr query_cpu_range, u64 size, Func&& func) { - word_manager.ForEachModifiedRange<Type::GPU>(query_cpu_range, size, true, func); + word_manager.template ForEachModifiedRange<Type::GPU>(query_cpu_range, size, true, func); } /// Call 'func' for each GPU modified range and unmark those pages as GPU modified template <typename Func> void ForEachDownloadRange(Func&& func) { - word_manager.ForEachModifiedRange<Type::GPU>(cpu_addr, SizeBytes(), true, func); + word_manager.template ForEachModifiedRange<Type::GPU>(cpu_addr, SizeBytes(), true, func); } /// Mark buffer as picked diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index faa48a678..8fed08dab 100644 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -1395,10 +1395,10 @@ bool BufferCache<P>::SynchronizeBufferNoModified(Buffer& buffer, VAddr cpu_addr, auto make_copies = [&] { for (auto& interval : found_sets) { const std::size_t sub_size = interval.upper() - interval.lower(); - const VAddr cpu_addr = interval.lower(); + const VAddr cpu_addr_ = interval.lower(); copies.push_back(BufferCopy{ .src_offset = total_size_bytes, - .dst_offset = cpu_addr - buffer.CpuAddr(), + .dst_offset = cpu_addr_ - buffer.CpuAddr(), .size = sub_size, }); total_size_bytes += sub_size; diff --git a/src/video_core/buffer_cache/buffer_cache_base.h b/src/video_core/buffer_cache/buffer_cache_base.h index d4914a8f5..acff22d4f 100644 --- a/src/video_core/buffer_cache/buffer_cache_base.h +++ b/src/video_core/buffer_cache/buffer_cache_base.h @@ -5,6 +5,7 @@ #include <algorithm> #include <array> +#include <functional> #include <memory> #include <mutex> #include <numeric> @@ -16,10 +17,13 @@ #define BOOST_NO_MT #include <boost/pool/detail/mutex.hpp> #undef BOOST_NO_MT +#include <boost/icl/interval.hpp> +#include <boost/icl/interval_base_set.hpp> #include <boost/icl/interval_set.hpp> #include <boost/icl/split_interval_map.hpp> #include <boost/pool/pool.hpp> #include <boost/pool/pool_alloc.hpp> +#include <boost/pool/poolfwd.hpp> #include "common/common_types.h" #include "common/div_ceil.h" @@ -42,7 +46,6 @@ #include "video_core/texture_cache/slot_vector.h" #include "video_core/texture_cache/types.h" - namespace boost { template <typename T> class fast_pool_allocator<T, default_user_allocator_new_delete, details::pool::null_mutex, 4096, 0>; @@ -116,11 +119,10 @@ class BufferCache : public VideoCommon::ChannelSetupCaches<VideoCommon::ChannelI using Async_Buffer = typename P::Async_Buffer; using MemoryTracker = typename P::MemoryTracker; - using IntervalCompare = ICL_COMPARE_INSTANCE(ICL_COMPARE_DEFAULT, VAddr); - using IntervalInstance = ICL_INTERVAL_INSTANCE(ICL_INTERVAL_DEFAULT, VAddr, IntervalCompare); - using IntervalAllocator = boost::fast_pool_allocator; - using IntervalSet = - boost::icl::interval_set<VAddr, IntervalCompare, IntervalInstance, IntervalAllocator>; + using IntervalCompare = std::less<VAddr>; + using IntervalInstance = boost::icl::interval_type_default<VAddr, std::less>; + using IntervalAllocator = boost::fast_pool_allocator<VAddr>; + using IntervalSet = boost::icl::interval_set<VAddr>; using IntervalType = typename IntervalSet::interval_type; template <typename Type> @@ -141,12 +143,9 @@ class BufferCache : public VideoCommon::ChannelSetupCaches<VideoCommon::ChannelI static void version(Type&){}; }; - using OverlapCombine = ICL_COMBINE_INSTANCE(counter_add_functor, int); - using OverlapSection = ICL_SECTION_INSTANCE(boost::icl::inter_section, int); - using OverlapCounter = - boost::icl::split_interval_map<VAddr, int, boost::icl::partial_absorber, IntervalCompare, - OverlapCombine, OverlapSection, IntervalInstance, - IntervalAllocator>; + using OverlapCombine = counter_add_functor<int>; + using OverlapSection = boost::icl::inter_section<int>; + using OverlapCounter = boost::icl::split_interval_map<VAddr, int>; struct Empty {}; @@ -262,7 +261,8 @@ public: /// Return true when a CPU region is modified from the CPU [[nodiscard]] bool IsRegionCpuModified(VAddr addr, size_t size); - void SetDrawIndirect(const Tegra::Engines::DrawManager::IndirectParams* current_draw_indirect_) { + void SetDrawIndirect( + const Tegra::Engines::DrawManager::IndirectParams* current_draw_indirect_) { current_draw_indirect = current_draw_indirect_; } @@ -349,7 +349,8 @@ private: } } - void RemoveEachInOverlapCounter(OverlapCounter& current_range, const IntervalType search_interval, int subtract_value) { + void RemoveEachInOverlapCounter(OverlapCounter& current_range, + const IntervalType search_interval, int subtract_value) { bool any_removals = false; current_range.add(std::make_pair(search_interval, subtract_value)); do { @@ -469,7 +470,8 @@ private: void NotifyBufferDeletion(); - [[nodiscard]] Binding StorageBufferBinding(GPUVAddr ssbo_addr, u32 cbuf_index, bool is_written) const; + [[nodiscard]] Binding StorageBufferBinding(GPUVAddr ssbo_addr, u32 cbuf_index, + bool is_written) const; [[nodiscard]] TextureBufferBinding GetTextureBufferBinding(GPUVAddr gpu_addr, u32 size, PixelFormat format); diff --git a/src/video_core/buffer_cache/memory_tracker_base.h b/src/video_core/buffer_cache/memory_tracker_base.h index 93bd779c9..016d8430f 100644 --- a/src/video_core/buffer_cache/memory_tracker_base.h +++ b/src/video_core/buffer_cache/memory_tracker_base.h @@ -35,67 +35,71 @@ public: /// Returns the inclusive CPU modified range in a begin end pair [[nodiscard]] std::pair<u64, u64> ModifiedCpuRegion(VAddr query_cpu_addr, u64 query_size) noexcept { - return IteratePairs<true>(query_cpu_addr, query_size, - [](Manager* manager, u64 offset, size_t size) { - return manager->ModifiedRegion<Type::CPU>(offset, size); - }); + return IteratePairs<true>( + query_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) { + return manager->template ModifiedRegion<Type::CPU>(offset, size); + }); } /// Returns the inclusive GPU modified range in a begin end pair [[nodiscard]] std::pair<u64, u64> ModifiedGpuRegion(VAddr query_cpu_addr, u64 query_size) noexcept { - return IteratePairs<false>(query_cpu_addr, query_size, - [](Manager* manager, u64 offset, size_t size) { - return manager->ModifiedRegion<Type::GPU>(offset, size); - }); + return IteratePairs<false>( + query_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) { + return manager->template ModifiedRegion<Type::GPU>(offset, size); + }); } /// Returns true if a region has been modified from the CPU [[nodiscard]] bool IsRegionCpuModified(VAddr query_cpu_addr, u64 query_size) noexcept { - return IteratePages<true>(query_cpu_addr, query_size, - [](Manager* manager, u64 offset, size_t size) { - return manager->IsRegionModified<Type::CPU>(offset, size); - }); + return IteratePages<true>( + query_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) { + return manager->template IsRegionModified<Type::CPU>(offset, size); + }); } /// Returns true if a region has been modified from the GPU [[nodiscard]] bool IsRegionGpuModified(VAddr query_cpu_addr, u64 query_size) noexcept { - return IteratePages<false>(query_cpu_addr, query_size, - [](Manager* manager, u64 offset, size_t size) { - return manager->IsRegionModified<Type::GPU>(offset, size); - }); + return IteratePages<false>( + query_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) { + return manager->template IsRegionModified<Type::GPU>(offset, size); + }); } /// Mark region as CPU modified, notifying the rasterizer about this change void MarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 query_size) { - IteratePages<true>( - dirty_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) { - manager->ChangeRegionState<Type::CPU, true>(manager->GetCpuAddr() + offset, size); - }); + IteratePages<true>(dirty_cpu_addr, query_size, + [](Manager* manager, u64 offset, size_t size) { + manager->template ChangeRegionState<Type::CPU, true>( + manager->GetCpuAddr() + offset, size); + }); } /// Unmark region as CPU modified, notifying the rasterizer about this change void UnmarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 query_size) { - IteratePages<true>( - dirty_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) { - manager->ChangeRegionState<Type::CPU, false>(manager->GetCpuAddr() + offset, size); - }); + IteratePages<true>(dirty_cpu_addr, query_size, + [](Manager* manager, u64 offset, size_t size) { + manager->template ChangeRegionState<Type::CPU, false>( + manager->GetCpuAddr() + offset, size); + }); } /// Mark region as modified from the host GPU void MarkRegionAsGpuModified(VAddr dirty_cpu_addr, u64 query_size) noexcept { - IteratePages<true>( - dirty_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) { - manager->ChangeRegionState<Type::GPU, true>(manager->GetCpuAddr() + offset, size); - }); + IteratePages<true>(dirty_cpu_addr, query_size, + [](Manager* manager, u64 offset, size_t size) { + manager->template ChangeRegionState<Type::GPU, true>( + manager->GetCpuAddr() + offset, size); + }); } /// Unmark region as modified from the host GPU void UnmarkRegionAsGpuModified(VAddr dirty_cpu_addr, u64 query_size) noexcept { - IteratePages<true>( - dirty_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) { - manager->ChangeRegionState<Type::GPU, false>(manager->GetCpuAddr() + offset, size); - }); + IteratePages<true>(dirty_cpu_addr, query_size, + [](Manager* manager, u64 offset, size_t size) { + manager->template ChangeRegionState<Type::GPU, false>( + manager->GetCpuAddr() + offset, size); + }); } /// Mark region as modified from the CPU @@ -104,7 +108,7 @@ public: IteratePages<true>( dirty_cpu_addr, query_size, [this](Manager* manager, u64 offset, size_t size) { const VAddr cpu_address = manager->GetCpuAddr() + offset; - manager->ChangeRegionState<Type::CachedCPU, true>(cpu_address, size); + manager->template ChangeRegionState<Type::CachedCPU, true>(cpu_address, size); cached_pages.insert(static_cast<u32>(cpu_address >> HIGHER_PAGE_BITS)); }); } @@ -128,7 +132,7 @@ public: void ForEachUploadRange(VAddr query_cpu_range, u64 query_size, Func&& func) { IteratePages<true>(query_cpu_range, query_size, [&func](Manager* manager, u64 offset, size_t size) { - manager->ForEachModifiedRange<Type::CPU>( + manager->template ForEachModifiedRange<Type::CPU>( manager->GetCpuAddr() + offset, size, true, func); }); } @@ -138,7 +142,7 @@ public: void ForEachDownloadRange(VAddr query_cpu_range, u64 query_size, bool clear, Func&& func) { IteratePages<false>(query_cpu_range, query_size, [&func, clear](Manager* manager, u64 offset, size_t size) { - manager->ForEachModifiedRange<Type::GPU>( + manager->template ForEachModifiedRange<Type::GPU>( manager->GetCpuAddr() + offset, size, clear, func); }); } @@ -147,7 +151,7 @@ public: void ForEachDownloadRangeAndClear(VAddr query_cpu_range, u64 query_size, Func&& func) { IteratePages<false>(query_cpu_range, query_size, [&func](Manager* manager, u64 offset, size_t size) { - manager->ForEachModifiedRange<Type::GPU>( + manager->template ForEachModifiedRange<Type::GPU>( manager->GetCpuAddr() + offset, size, true, func); }); } @@ -218,7 +222,11 @@ private: page_offset = 0; remaining_size -= copy_amount; } - return begin < end ? std::make_pair(begin, end) : std::make_pair(0ULL, 0ULL); + if (begin < end) { + return std::make_pair(begin, end); + } else { + return std::make_pair(0ULL, 0ULL); + } } void CreateRegion(std::size_t page_index) { diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h index 05968e6a6..879f1ed94 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.h +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h @@ -3,7 +3,7 @@ #pragma once -#include "video_core/buffer_cache/buffer_cache.h" +#include "video_core/buffer_cache/buffer_cache_base.h" #include "video_core/buffer_cache/memory_tracker_base.h" #include "video_core/engines/maxwell_3d.h" #include "video_core/renderer_vulkan/vk_compute_pass.h" |