diff options
Diffstat (limited to 'src/core/memory.cpp')
| -rw-r--r-- | src/core/memory.cpp | 94 |
1 files changed, 83 insertions, 11 deletions
diff --git a/src/core/memory.cpp b/src/core/memory.cpp index 514ba0d66..805963178 100644 --- a/src/core/memory.cpp +++ b/src/core/memory.cpp @@ -3,6 +3,7 @@ #include <algorithm> #include <cstring> +#include <span> #include "common/assert.h" #include "common/atomic_ops.h" @@ -13,6 +14,7 @@ #include "common/swap.h" #include "core/core.h" #include "core/device_memory.h" +#include "core/gpu_dirty_memory_manager.h" #include "core/hardware_properties.h" #include "core/hle/kernel/k_page_table.h" #include "core/hle/kernel/k_process.h" @@ -264,6 +266,22 @@ struct Memory::Impl { ReadBlockImpl<true>(*system.ApplicationProcess(), src_addr, dest_buffer, size); } + const u8* GetSpan(const VAddr src_addr, const std::size_t size) const { + if (current_page_table->blocks[src_addr >> YUZU_PAGEBITS] == + current_page_table->blocks[(src_addr + size) >> YUZU_PAGEBITS]) { + return GetPointerSilent(src_addr); + } + return nullptr; + } + + u8* GetSpan(const VAddr src_addr, const std::size_t size) { + if (current_page_table->blocks[src_addr >> YUZU_PAGEBITS] == + current_page_table->blocks[(src_addr + size) >> YUZU_PAGEBITS]) { + return GetPointerSilent(src_addr); + } + return nullptr; + } + template <bool UNSAFE> void WriteBlockImpl(const Kernel::KProcess& process, const Common::ProcessAddress dest_addr, const void* src_buffer, const std::size_t size) { @@ -557,7 +575,7 @@ struct Memory::Impl { } } - const Common::ProcessAddress end = base + size; + const auto end = base + size; ASSERT_MSG(end <= page_table.pointers.size(), "out of range mapping at {:016X}", base + page_table.pointers.size()); @@ -568,14 +586,18 @@ struct Memory::Impl { while (base != end) { page_table.pointers[base].Store(nullptr, type); page_table.backing_addr[base] = 0; - + page_table.blocks[base] = 0; base += 1; } } else { + auto orig_base = base; while (base != end) { - page_table.pointers[base].Store( - system.DeviceMemory().GetPointer<u8>(target) - (base << YUZU_PAGEBITS), type); - page_table.backing_addr[base] = GetInteger(target) - (base << YUZU_PAGEBITS); + auto host_ptr = + system.DeviceMemory().GetPointer<u8>(target) - (base << YUZU_PAGEBITS); + auto backing = GetInteger(target) - (base << YUZU_PAGEBITS); + page_table.pointers[base].Store(host_ptr, type); + page_table.backing_addr[base] = backing; + page_table.blocks[base] = orig_base << YUZU_PAGEBITS; ASSERT_MSG(page_table.pointers[base].Pointer(), "memory mapping base yield a nullptr within the table"); @@ -678,7 +700,7 @@ struct Memory::Impl { LOG_ERROR(HW_Memory, "Unmapped Write{} @ 0x{:016X} = 0x{:016X}", sizeof(T) * 8, GetInteger(vaddr), static_cast<u64>(data)); }, - [&]() { system.GPU().InvalidateRegion(GetInteger(vaddr), sizeof(T)); }); + [&]() { HandleRasterizerWrite(GetInteger(vaddr), sizeof(T)); }); if (ptr) { std::memcpy(ptr, &data, sizeof(T)); } @@ -692,7 +714,7 @@ struct Memory::Impl { LOG_ERROR(HW_Memory, "Unmapped WriteExclusive{} @ 0x{:016X} = 0x{:016X}", sizeof(T) * 8, GetInteger(vaddr), static_cast<u64>(data)); }, - [&]() { system.GPU().InvalidateRegion(GetInteger(vaddr), sizeof(T)); }); + [&]() { HandleRasterizerWrite(GetInteger(vaddr), sizeof(T)); }); if (ptr) { const auto volatile_pointer = reinterpret_cast<volatile T*>(ptr); return Common::AtomicCompareAndSwap(volatile_pointer, data, expected); @@ -707,7 +729,7 @@ struct Memory::Impl { LOG_ERROR(HW_Memory, "Unmapped WriteExclusive128 @ 0x{:016X} = 0x{:016X}{:016X}", GetInteger(vaddr), static_cast<u64>(data[1]), static_cast<u64>(data[0])); }, - [&]() { system.GPU().InvalidateRegion(GetInteger(vaddr), sizeof(u128)); }); + [&]() { HandleRasterizerWrite(GetInteger(vaddr), sizeof(u128)); }); if (ptr) { const auto volatile_pointer = reinterpret_cast<volatile u64*>(ptr); return Common::AtomicCompareAndSwap(volatile_pointer, data, expected); @@ -717,7 +739,7 @@ struct Memory::Impl { void HandleRasterizerDownload(VAddr address, size_t size) { const size_t core = system.GetCurrentHostThreadID(); - auto& current_area = rasterizer_areas[core]; + auto& current_area = rasterizer_read_areas[core]; const VAddr end_address = address + size; if (current_area.start_address <= address && end_address <= current_area.end_address) [[likely]] { @@ -726,9 +748,39 @@ struct Memory::Impl { current_area = system.GPU().OnCPURead(address, size); } - Common::PageTable* current_page_table = nullptr; - std::array<VideoCore::RasterizerDownloadArea, Core::Hardware::NUM_CPU_CORES> rasterizer_areas{}; + void HandleRasterizerWrite(VAddr address, size_t size) { + const size_t core = system.GetCurrentHostThreadID(); + auto& current_area = rasterizer_write_areas[core]; + VAddr subaddress = address >> YUZU_PAGEBITS; + bool do_collection = current_area.last_address == subaddress; + if (!do_collection) [[unlikely]] { + do_collection = system.GPU().OnCPUWrite(address, size); + if (!do_collection) { + return; + } + current_area.last_address = subaddress; + } + gpu_dirty_managers[core].Collect(address, size); + } + + struct GPUDirtyState { + VAddr last_address; + }; + + void InvalidateRegion(Common::ProcessAddress dest_addr, size_t size) { + system.GPU().InvalidateRegion(GetInteger(dest_addr), size); + } + + void FlushRegion(Common::ProcessAddress dest_addr, size_t size) { + system.GPU().FlushRegion(GetInteger(dest_addr), size); + } + Core::System& system; + Common::PageTable* current_page_table = nullptr; + std::array<VideoCore::RasterizerDownloadArea, Core::Hardware::NUM_CPU_CORES> + rasterizer_read_areas{}; + std::array<GPUDirtyState, Core::Hardware::NUM_CPU_CORES> rasterizer_write_areas{}; + std::span<Core::GPUDirtyMemoryManager> gpu_dirty_managers; }; Memory::Memory(Core::System& system_) : system{system_} { @@ -857,6 +909,14 @@ void Memory::ReadBlockUnsafe(const Common::ProcessAddress src_addr, void* dest_b impl->ReadBlockUnsafe(src_addr, dest_buffer, size); } +const u8* Memory::GetSpan(const VAddr src_addr, const std::size_t size) const { + return impl->GetSpan(src_addr, size); +} + +u8* Memory::GetSpan(const VAddr src_addr, const std::size_t size) { + return impl->GetSpan(src_addr, size); +} + void Memory::WriteBlock(const Common::ProcessAddress dest_addr, const void* src_buffer, const std::size_t size) { impl->WriteBlock(dest_addr, src_buffer, size); @@ -876,6 +936,10 @@ void Memory::ZeroBlock(Common::ProcessAddress dest_addr, const std::size_t size) impl->ZeroBlock(*system.ApplicationProcess(), dest_addr, size); } +void Memory::SetGPUDirtyManagers(std::span<Core::GPUDirtyMemoryManager> managers) { + impl->gpu_dirty_managers = managers; +} + Result Memory::InvalidateDataCache(Common::ProcessAddress dest_addr, const std::size_t size) { return impl->InvalidateDataCache(*system.ApplicationProcess(), dest_addr, size); } @@ -896,4 +960,12 @@ void Memory::MarkRegionDebug(Common::ProcessAddress vaddr, u64 size, bool debug) impl->MarkRegionDebug(GetInteger(vaddr), size, debug); } +void Memory::InvalidateRegion(Common::ProcessAddress dest_addr, size_t size) { + impl->InvalidateRegion(dest_addr, size); +} + +void Memory::FlushRegion(Common::ProcessAddress dest_addr, size_t size) { + impl->FlushRegion(dest_addr, size); +} + } // namespace Core::Memory |