gpu: Refactor command and swap buffers interface for asynch.

src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp

@@ -36,7 +36,7 @@ void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u3
 
     auto& instance = Core::System::GetInstance();
     instance.GetPerfStats().EndGameFrame();
-    instance.Renderer().SwapBuffers(framebuffer);
+    instance.GPU().SwapBuffers(framebuffer);
 }
 
 } // namespace Service::Nvidia::Devices

src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp

@@ -136,16 +136,6 @@ u32 nvhost_gpu::AllocateObjectContext(const std::vector<u8>& input, std::vector<
     return 0;
 }
 
-static void PushGPUEntries(Tegra::CommandList&& entries) {
-    if (entries.empty()) {
-        return;
-    }
-
-    auto& dma_pusher{Core::System::GetInstance().GPU().DmaPusher()};
-    dma_pusher.Push(std::move(entries));
-    dma_pusher.DispatchCalls();
-}
-
 u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& output) {
     if (input.size() < sizeof(IoctlSubmitGpfifo)) {
         UNIMPLEMENTED();
@@ -163,7 +153,7 @@ u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& outp
     std::memcpy(entries.data(), &input[sizeof(IoctlSubmitGpfifo)],
                 params.num_entries * sizeof(Tegra::CommandListHeader));
 
-    PushGPUEntries(std::move(entries));
+    Core::System::GetInstance().GPU().PushGPUEntries(std::move(entries));
 
     params.fence_out.id = 0;
     params.fence_out.value = 0;
@@ -184,7 +174,7 @@ u32 nvhost_gpu::KickoffPB(const std::vector<u8>& input, std::vector<u8>& output)
     Memory::ReadBlock(params.address, entries.data(),
                       params.num_entries * sizeof(Tegra::CommandListHeader));
 
-    PushGPUEntries(std::move(entries));
+    Core::System::GetInstance().GPU().PushGPUEntries(std::move(entries));
 
     params.fence_out.id = 0;
     params.fence_out.value = 0;

src/core/hle/service/nvflinger/nvflinger.cpp

@@ -141,7 +141,7 @@ void NVFlinger::Compose() {
 
             // There was no queued buffer to draw, render previous frame
             system_instance.GetPerfStats().EndGameFrame();
-            system_instance.Renderer().SwapBuffers({});
+            system_instance.GPU().SwapBuffers({});
             continue;
         }
 

src/video_core/dma_pusher.cpp

@@ -17,6 +17,13 @@ DmaPusher::~DmaPusher() = default;
 
 MICROPROFILE_DEFINE(DispatchCalls, "GPU", "Execute command buffer", MP_RGB(128, 128, 192));
 
+void DmaPusher::QueuePendingCalls() {
+    for (auto& entry : dma_writebuffer) {
+        dma_readbuffer.push(std::move(entry));
+    }
+    dma_writebuffer.clear();
+}
+
 void DmaPusher::DispatchCalls() {
     MICROPROFILE_SCOPE(DispatchCalls);
 
@@ -89,9 +96,9 @@ bool DmaPusher::Step() {
                 break;
             }
         }
-    } else if (ib_enable && !dma_pushbuffer.empty()) {
+    } else if (ib_enable && !dma_readbuffer.empty()) {
         // Current pushbuffer empty, but we have more IB entries to read
-        const CommandList& command_list{dma_pushbuffer.front()};
+        const CommandList& command_list{dma_readbuffer.front()};
         const CommandListHeader& command_list_header{command_list[dma_pushbuffer_subindex++]};
         dma_get = command_list_header.addr;
         dma_put = dma_get + command_list_header.size * sizeof(u32);
@@ -99,7 +106,7 @@ bool DmaPusher::Step() {
 
         if (dma_pushbuffer_subindex >= command_list.size()) {
             // We've gone through the current list, remove it from the queue
-            dma_pushbuffer.pop();
+            dma_readbuffer.pop();
             dma_pushbuffer_subindex = 0;
         }
     } else {

src/video_core/dma_pusher.h

@@ -61,9 +61,10 @@ public:
     ~DmaPusher();
 
     void Push(CommandList&& entries) {
-        dma_pushbuffer.push(std::move(entries));
+        dma_writebuffer.push_back(std::move(entries));
     }
 
+    void QueuePendingCalls();
     void DispatchCalls();
 
 private:
@@ -75,8 +76,9 @@ private:
 
     GPU& gpu;
 
-    std::queue<CommandList> dma_pushbuffer; ///< Queue of command lists to be processed
-    std::size_t dma_pushbuffer_subindex{};  ///< Index within a command list within the pushbuffer
+    std::vector<CommandList> dma_writebuffer;
+    std::queue<CommandList> dma_readbuffer;
+    std::size_t dma_pushbuffer_subindex{}; ///< Index within a command list within the pushbuffer
 
     struct DmaState {
         u32 method;            ///< Current method

src/video_core/gpu.cpp

@@ -61,6 +61,17 @@ const DmaPusher& GPU::DmaPusher() const {
     return *dma_pusher;
 }
 
+void GPU::PushGPUEntries(Tegra::CommandList&& entries) {
+    dma_pusher->Push(std::move(entries));
+    dma_pusher->QueuePendingCalls();
+    dma_pusher->DispatchCalls();
+}
+
+void GPU::SwapBuffers(
+    std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
+    renderer.SwapBuffers(std::move(framebuffer));
+}
+
 u32 RenderTargetBytesPerPixel(RenderTargetFormat format) {
     ASSERT(format != RenderTargetFormat::NONE);
 

src/video_core/gpu.h

@@ -156,6 +156,13 @@ public:
     /// Returns a const reference to the GPU DMA pusher.
     const Tegra::DmaPusher& DmaPusher() const;
 
+    /// Push GPU command entries to be processed
+    void PushGPUEntries(Tegra::CommandList&& entries);
+
+    /// Swap buffers (render frame)
+    void SwapBuffers(
+        std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer);
+
 private:
     std::unique_ptr<Tegra::DmaPusher> dma_pusher;
     std::unique_ptr<Tegra::MemoryManager> memory_manager;