diff options
Diffstat (limited to 'src/video_core/macro.cpp')
| -rw-r--r-- | src/video_core/macro.cpp | 50 |
1 files changed, 45 insertions, 5 deletions
diff --git a/src/video_core/macro.cpp b/src/video_core/macro.cpp index 3422cb16fa..485b8eaf10 100644 --- a/src/video_core/macro.cpp +++ b/src/video_core/macro.cpp @@ -17,6 +17,7 @@ #ifdef ARCHITECTURE_x86_64 #include "common/x64/xbyak.h" +#include "video_core/macro_turing_jit.h" #endif #include "common/assert.h" @@ -1404,7 +1405,10 @@ void MacroEngine::Execute(Core::System& system, Engines::Maxwell3D& maxwell3d, u modes[0], modes[1], modes[2], modes[3]); } } - maxwell3d.CallMethod(system, address, value, true); + { + VideoCommon::Nx2Profile::Scope method_scope{VideoCommon::Nx2Profile::Zone::Method}; + maxwell3d.CallMethod(system, address, value, true); + } // Turing MME memory reads (methods 0x550 SET_MME_MEM_ADDRESS_A/B, 0x558 data RAM // address, 0x55c MME_DMA_READ into data RAM, 0x560 MME_DMA_READ_FIFOED into // the load FIFO, which the macro then reads after a load barrier). @@ -1420,6 +1424,14 @@ void MacroEngine::Execute(Core::System& system, Engines::Maxwell3D& maxwell3d, u // from a culling dispatch) and still be only in host buffers: flush them first, // or the macro draws with an older frame's arguments (objects pop in and out). const size_t read_bytes = size_t{std::min<u32>(value, 0x1000)} * 4; + if (VideoCommon::Nx2Profile::Enabled()) { + VideoCommon::Nx2Profile::Add(VideoCommon::Nx2Profile::Counter::DmaReadBytes, read_bytes); + for (GPUVAddr page = source & ~GPUVAddr{0xfff}; page < source + read_bytes; page += 0x1000) { + if (const auto device = memory.GpuToCpuAddress(page)) { + VideoCommon::Nx2Profile::DmaReadPages().insert(*device >> 12); + } + } + } if (memory.IsMemoryDirty(source, read_bytes, VideoCommon::CacheType::BufferCache)) { memory.FlushRegion(source, read_bytes, VideoCommon::CacheType::BufferCache); static u64 flushes{}; @@ -1451,6 +1463,33 @@ void MacroEngine::Execute(Core::System& system, Engines::Maxwell3D& maxwell3d, u } return true; }; +#ifdef ARCHITECTURE_x86_64 + // Macro code runs compiled to x86-64 (macro_turing_jit.h) instead of through the memo + // and the interpreter; EDEN_NX2_MME_JIT=0 goes back to those. + static const bool use_jit = [] { + const char* value = std::getenv("EDEN_NX2_MME_JIT"); + return value == nullptr || value[0] != '0'; + }(); + if (use_jit) { + if (!native_jit) { + native_jit = std::make_shared<TuringMacroJit>(); + } + native_jit->Prepare(native_code, native_code_generation); + fifo.assign(parameters.begin(), parameters.end()); + const auto result = native_jit->Execute( + method, fifo, std::span<u32>{*native_interpreter.active_dram}, + [&](u32 address) { + return address < Maxwell3D::Regs::NUM_REGS ? maxwell3d.GetRegisterValue(address) : 0U; + }, + emit); + if (!result.success) { + native_execution_failed = true; + LOG_ERROR(HW_GPU, "Native MME (JIT) failed: start={:#x}, consumed={}/{}, error={}", + method, result.consumed, parameters.size(), result.error); + } + return; + } +#endif // EDEN_NX2_MME_MEMO=0: interpret every call (see macro_memo.h). static const bool use_memo = [] { const char* value = std::getenv("EDEN_NX2_MME_MEMO"); @@ -1462,10 +1501,11 @@ void MacroEngine::Execute(Core::System& system, Engines::Maxwell3D& maxwell3d, u memo_stats && (++native_calls & 0xffffff) == 0) { const auto& stats = native_memo.GetStatistics(); LOG_INFO(HW_GPU, - "Macro memo: {} hits, {} misses, {} guard failures, {} stores ({} in " - "place), {} clears; {} entries, {} KiB", - stats.hits, stats.misses, stats.guard_failures, stats.stores, - stats.overwrites, stats.clears, native_memo.Size(), + "Macro memo: {} hits ({} on the last call's slot), {} misses, {} guard " + "failures, {} stores ({} in place), {} compactions, {} clears; {} entries, " + "{} KiB", + stats.hits, stats.slot_hits, stats.misses, stats.guard_failures, stats.stores, + stats.overwrites, stats.compactions, stats.clears, native_memo.Size(), native_memo.ArenaWords() * 4 / 1024); } native_memo.Synchronize(native_code_generation); |