[d3d11] Synchronize with CS thread before present

Fixes flickering by synchronizing with the presenter. This has
to do for now, a more efficient solution can be added later.

src/d3d11/d3d11_context_imm.cpp

@@ -15,8 +15,8 @@ namespace dxvk {
   
   D3D11ImmediateContext::~D3D11ImmediateContext() {
     Flush();
-    SynchronizeCs();
-    Synchronize();
+    SynchronizeCsThread();
+    SynchronizeDevice();
   }
   
   
@@ -114,15 +114,14 @@ namespace dxvk {
       } else if (MapType != D3D11_MAP_WRITE_NO_OVERWRITE) {
         // Synchronize with CS thread so that we know whether
         // the buffer is currently in use by the GPU or not
-        SynchronizeCs();
+        SynchronizeCsThread();
         
         if (buffer->isInUse()) {
           if (MapFlags & D3D11_MAP_FLAG_DO_NOT_WAIT)
             return DXGI_ERROR_WAS_STILL_DRAWING;
           
           Flush();
-          SynchronizeCs();
-          Synchronize();
+          SynchronizeDevice();
         }
       }
       
@@ -200,8 +199,8 @@ namespace dxvk {
         });
         
         Flush();
-        SynchronizeCs();
-        Synchronize();
+        SynchronizeCsThread();
+        SynchronizeDevice();
         
         physicalSlice = textureInfo->imageBuffer->slice();
       }
@@ -249,7 +248,16 @@ namespace dxvk {
   }
   
   
-  void D3D11ImmediateContext::Synchronize() {
+  void D3D11ImmediateContext::SynchronizeCsThread() {
+    // Dispatch current chunk so that all commands
+    // recorded prior to this function will be run
+    EmitCsChunk();
+    
+    m_csThread.synchronize();
+  }
+  
+  
+  void D3D11ImmediateContext::SynchronizeDevice() {
     // FIXME waiting until the device finished executing *all*
     // pending commands is too pessimistic. Instead we should
     // wait for individual command submissions to complete.
@@ -258,13 +266,6 @@ namespace dxvk {
   }
   
   
-  void D3D11ImmediateContext::SynchronizeCs() {
-    EmitCsChunk();
-    
-    m_csThread.synchronize();
-  }
-  
-  
   void D3D11ImmediateContext::EmitCsChunk() {
     if (m_csChunk->commandCount() > 0)
       m_csChunk = m_csThread.dispatchChunk(std::move(m_csChunk));

src/d3d11/d3d11_context_imm.h

@@ -42,13 +42,13 @@ namespace dxvk {
             ID3D11Resource*             pResource,
             UINT                        Subresource) final;
     
+    void SynchronizeCsThread();
     
   private:
     
     DxvkCsThread m_csThread;
     
-    void Synchronize();
-    void SynchronizeCs();
+    void SynchronizeDevice();
     
     void EmitCsChunk();
     

src/d3d11/d3d11_present.cpp

@@ -1,4 +1,5 @@
 #include "d3d11_device.h"
+#include "d3d11_context_imm.h"
 #include "d3d11_present.h"
 
 namespace dxvk {
@@ -57,7 +58,11 @@ namespace dxvk {
     Com<ID3D11DeviceContext> deviceContext = nullptr;
     m_device->GetImmediateContext(&deviceContext);
     
-    deviceContext->Flush();
+    // The presentation code is run from the main rendering thread
+    // rather than the command stream thread, so we synchronize.
+    auto immediateContext = static_cast<D3D11ImmediateContext*>(deviceContext.ptr());
+    immediateContext->Flush();
+    immediateContext->SynchronizeCsThread();
     return S_OK;
   }