From abea6fa90c901d0b47487ed38d44511b18f0addf Mon Sep 17 00:00:00 2001 From: bunnei Date: Fri, 23 Nov 2018 23:20:56 -0500 Subject: gpu: Rewrite GPU command list processing with DmaPusher class. - More accurate impl., fixes Undertale (among other games). --- src/video_core/gpu.cpp | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 58 insertions(+) (limited to 'src/video_core/gpu.cpp') diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 51b3904f6..4a96530ae 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -3,6 +3,7 @@ // Refer to the license.txt file included. #include "common/assert.h" +#include "common/microprofile.h" #include "video_core/engines/fermi_2d.h" #include "video_core/engines/kepler_memory.h" #include "video_core/engines/maxwell_3d.h" @@ -26,6 +27,7 @@ u32 FramebufferConfig::BytesPerPixel(PixelFormat format) { GPU::GPU(VideoCore::RasterizerInterface& rasterizer) { memory_manager = std::make_unique(); + dma_pusher = std::make_unique(*this); maxwell_3d = std::make_unique(rasterizer, *memory_manager); fermi_2d = std::make_unique(rasterizer, *memory_manager); maxwell_compute = std::make_unique(); @@ -51,6 +53,14 @@ const MemoryManager& GPU::MemoryManager() const { return *memory_manager; } +DmaPusher& GPU::DmaPusher() { + return *dma_pusher; +} + +const DmaPusher& GPU::DmaPusher() const { + return *dma_pusher; +} + u32 RenderTargetBytesPerPixel(RenderTargetFormat format) { ASSERT(format != RenderTargetFormat::NONE); @@ -113,4 +123,52 @@ u32 DepthFormatBytesPerPixel(DepthFormat format) { } } +enum class BufferMethods { + BindObject = 0, + CountBufferMethods = 0x40, +}; + +MICROPROFILE_DEFINE(ProcessCommandLists, "GPU", "Execute command buffer", MP_RGB(128, 128, 192)); + +void GPU::CallMethod(const MethodCall& method_call) { + MICROPROFILE_SCOPE(ProcessCommandLists); + + LOG_TRACE(HW_GPU, + "Processing method {:08X} on subchannel {} value " + "{:08X} remaining params {}", + MethCall.method, MethCall.subchannel, value, remaining_params); + + ASSERT(method_call.subchannel < bound_engines.size()); + + if (method_call.method == static_cast(BufferMethods::BindObject)) { + // Bind the current subchannel to the desired engine id. + LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {}", method_call.subchannel, + method_call.argument); + bound_engines[method_call.subchannel] = static_cast(method_call.argument); + return; + } + + const EngineID engine = bound_engines[method_call.subchannel]; + + switch (engine) { + case EngineID::FERMI_TWOD_A: + fermi_2d->CallMethod(method_call); + break; + case EngineID::MAXWELL_B: + maxwell_3d->CallMethod(method_call); + break; + case EngineID::MAXWELL_COMPUTE_B: + maxwell_compute->CallMethod(method_call); + break; + case EngineID::MAXWELL_DMA_COPY_A: + maxwell_dma->CallMethod(method_call); + break; + case EngineID::KEPLER_INLINE_TO_MEMORY_B: + kepler_memory->CallMethod(method_call); + break; + default: + UNIMPLEMENTED_MSG("Unimplemented engine"); + } +} + } // namespace Tegra -- cgit v1.2.3