early-access version 1870

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -69,7 +69,7 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra
       buffer_cache(*this, maxwell3d, kepler_compute, gpu_memory, cpu_memory_, buffer_cache_runtime),
       shader_cache(*this, emu_window_, maxwell3d, kepler_compute, gpu_memory, device, texture_cache,
                    buffer_cache, program_manager, state_tracker, gpu.ShaderNotify()),
-      query_cache(*this, maxwell3d, gpu_memory),
+      query_cache(*this, maxwell3d, gpu_memory), accelerate_dma(buffer_cache),
       fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache) {}
 
 RasterizerOpenGL::~RasterizerOpenGL() = default;
@@ -502,6 +502,10 @@ bool RasterizerOpenGL::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surf
     return true;
 }
 
+Tegra::Engines::AccelerateDMAInterface& RasterizerOpenGL::AccessAccelerateDMA() {
+    return accelerate_dma;
+}
+
 bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
                                          VAddr framebuffer_addr, u32 pixel_stride) {
     if (framebuffer_addr == 0) {
@@ -1040,4 +1044,11 @@ void RasterizerOpenGL::EndTransformFeedback() {
     }
 }
 
+AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_) : buffer_cache{buffer_cache_} {}
+
+bool AccelerateDMA::BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) {
+    std::scoped_lock lock{buffer_cache.mutex};
+    return buffer_cache.DMACopy(src_address, dest_address, amount);
+}
+
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -19,6 +19,7 @@
 #include "common/common_types.h"
 #include "video_core/engines/const_buffer_info.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/engines/maxwell_dma.h"
 #include "video_core/rasterizer_accelerated.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_buffer_cache.h"
@@ -56,6 +57,16 @@ struct BindlessSSBO {
 };
 static_assert(sizeof(BindlessSSBO) * CHAR_BIT == 128);
 
+class AccelerateDMA : public Tegra::Engines::AccelerateDMAInterface {
+public:
+    explicit AccelerateDMA(BufferCache& buffer_cache);
+
+    bool BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) override;
+
+private:
+    BufferCache& buffer_cache;
+};
+
 class RasterizerOpenGL : public VideoCore::RasterizerAccelerated {
 public:
     explicit RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
@@ -94,6 +105,7 @@ public:
     bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surface& src,
                                const Tegra::Engines::Fermi2D::Surface& dst,
                                const Tegra::Engines::Fermi2D::Config& copy_config) override;
+    Tegra::Engines::AccelerateDMAInterface& AccessAccelerateDMA() override;
     bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
                            u32 pixel_stride) override;
     void LoadDiskResources(u64 title_id, std::stop_token stop_loading,
@@ -203,6 +215,7 @@ private:
     BufferCache buffer_cache;
     ShaderCache shader_cache;
     QueryCache query_cache;
+    AccelerateDMA accelerate_dma;
     FenceManagerOpenGL fence_manager;
 
     boost::container::static_vector<u32, MAX_IMAGE_VIEWS> image_view_indices;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -211,6 +211,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .has_gl_component_indexing_bug = device.HasComponentIndexingBug(),
           .has_gl_precise_bug = device.HasPreciseBug(),
           .ignore_nan_fp_comparisons = true,
+          .gl_max_compute_smem_size = device.GetMaxComputeSharedMemorySize(),
       },
       host_info{
           .support_float16 = false,
@@ -318,7 +319,7 @@ GraphicsPipeline* ShaderCache::CurrentGraphicsPipeline() {
         SetXfbState(graphics_key.xfb_state, regs);
     }
     if (current_pipeline && graphics_key == current_pipeline->Key()) {
-        return current_pipeline->IsBuilt() ? current_pipeline : nullptr;
+        return BuiltPipeline(current_pipeline);
     }
     return CurrentGraphicsPipelineSlowPath();
 }