early-access version 1950

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

@@ -5,10 +5,8 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
-#include "core/core_timing.h"
 #include "core/hle/service/nvdrv/devices/nvdisp_disp0.h"
 #include "core/hle/service/nvdrv/devices/nvmap.h"
-#include "core/perf_stats.h"
 #include "video_core/gpu.h"
 #include "video_core/renderer_base.h"
 
@@ -41,7 +39,7 @@ void nvdisp_disp0::OnClose(DeviceFD fd) {}
 
 void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height,
                         u32 stride, NVFlinger::BufferQueue::BufferTransformFlags transform,
-                        const Common::Rectangle<int>& crop_rect) {
+                        const Common::Rectangle<int>& crop_rect, const MultiFence& fences) {
     VAddr addr = nvmap_dev->GetObjectAddress(buffer_handle);
     LOG_TRACE(Service,
               "Drawing from address {:X} offset {:08X} Width {} Height {} Stride {} Format {}",
@@ -52,10 +50,7 @@ void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u3
         addr,      offset,   width, height, stride, static_cast<PixelFormat>(format),
         transform, crop_rect};
 
-    system.GetPerfStats().EndSystemFrame();
-    system.GPU().SwapBuffers(&framebuffer);
-    system.SpeedLimiter().DoSpeedLimiting(system.CoreTiming().GetGlobalTimeUs());
-    system.GetPerfStats().BeginSystemFrame();
+    system.GPU().QueueFrame(&framebuffer, fences);
 }
 
 } // namespace Service::Nvidia::Devices

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

@@ -33,7 +33,7 @@ public:
     /// Performs a screen flip, drawing the buffer pointed to by the handle.
     void flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height, u32 stride,
               NVFlinger::BufferQueue::BufferTransformFlags transform,
-              const Common::Rectangle<int>& crop_rect);
+              const Common::Rectangle<int>& crop_rect, const MultiFence& fence);
 
 private:
     std::shared_ptr<nvmap> nvmap_dev;

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

@@ -88,6 +88,10 @@ const IGBPBuffer& BufferQueue::RequestBuffer(u32 slot) const {
     return buffers[slot].igbp_buffer;
 }
 
+const BufferQueue::Buffer& BufferQueue::AccessBuffer(u32 slot) const {
+    return buffers[slot];
+}
+
 void BufferQueue::QueueBuffer(u32 slot, BufferTransformFlags transform,
                               const Common::Rectangle<int>& crop_rect, u32 swap_interval,
                               Service::Nvidia::MultiFence& multi_fence) {

src/core/hle/service/nvflinger/buffer_queue.h

@@ -107,6 +107,7 @@ public:
     void Connect();
     void Disconnect();
     u32 Query(QueryType type);
+    const Buffer& AccessBuffer(u32 slot) const;
 
     u32 GetId() const {
         return id;

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

@@ -274,8 +274,6 @@ void NVFlinger::Compose() {
             continue;
         }
 
-        const auto& igbp_buffer = buffer->get().igbp_buffer;
-
         if (!system.IsPoweredOn()) {
             return; // We are likely shutting down
         }
@@ -289,23 +287,31 @@ void NVFlinger::Compose() {
         }
         guard->lock();
 
+        system.GetPerfStats().EndSystemFrame();
         MicroProfileFlip();
-
-        // Now send the buffer to the GPU for drawing.
-        // TODO(Subv): Support more than just disp0. The display device selection is probably based
-        // on which display we're drawing (Default, Internal, External, etc)
-        auto nvdisp = nvdrv->GetDevice<Nvidia::Devices::nvdisp_disp0>("/dev/nvdisp_disp0");
-        ASSERT(nvdisp);
-
-        nvdisp->flip(igbp_buffer.gpu_buffer_id, igbp_buffer.offset, igbp_buffer.format,
-                     igbp_buffer.width, igbp_buffer.height, igbp_buffer.stride,
-                     buffer->get().transform, buffer->get().crop_rect);
+        system.SpeedLimiter().DoSpeedLimiting(system.CoreTiming().GetGlobalTimeUs());
+        system.GetPerfStats().BeginSystemFrame();
 
         swap_interval = buffer->get().swap_interval;
         buffer_queue.ReleaseBuffer(buffer->get().slot);
     }
 }
 
+void NVFlinger::PrequeeFrame(u32 buffer_queue_id, u32 slot) {
+    auto& buffer_queue = *FindBufferQueue(buffer_queue_id);
+    const auto& buffer = buffer_queue.AccessBuffer(slot);
+    const auto& igbp_buffer = buffer.igbp_buffer;
+
+    // Now send the buffer to the GPU for drawing.
+    // TODO(Subv): Support more than just disp0. The display device selection is probably based
+    // on which display we're drawing (Default, Internal, External, etc)
+    auto nvdisp = nvdrv->GetDevice<Nvidia::Devices::nvdisp_disp0>("/dev/nvdisp_disp0");
+    ASSERT(nvdisp);
+    nvdisp->flip(igbp_buffer.gpu_buffer_id, igbp_buffer.offset, igbp_buffer.format,
+                 igbp_buffer.width, igbp_buffer.height, igbp_buffer.stride, buffer.transform,
+                 buffer.crop_rect, buffer.multi_fence);
+}
+
 s64 NVFlinger::GetNextTicks() const {
     static constexpr s64 max_hertz = 120LL;
 

src/core/hle/service/nvflinger/nvflinger.h

@@ -77,6 +77,8 @@ public:
     /// Obtains a buffer queue identified by the ID.
     [[nodiscard]] BufferQueue* FindBufferQueue(u32 id);
 
+    void PrequeeFrame(u32 buffer_queue_id, u32 slot);
+
     /// Performs a composition request to the emulated nvidia GPU and triggers the vsync events when
     /// finished.
     void Compose();

src/core/hle/service/vi/vi.cpp

@@ -592,6 +592,7 @@ private:
             buffer_queue.QueueBuffer(request.data.slot, request.data.transform,
                                      request.data.GetCropRect(), request.data.swap_interval,
                                      request.data.multi_fence);
+            nv_flinger.PrequeeFrame(id, request.data.slot);
 
             IGBPQueueBufferResponseParcel response{1280, 720};
             ctx.WriteBuffer(response.Serialize());