[host] added multi-threaded memcopy for high resolutions

host/Capture/DXGI.cpp

@@ -21,7 +21,6 @@ Place, Suite 330, Boston, MA 02111-1307 USA
 using namespace Capture;
 
 #include "common/debug.h"
-#include "common/memcpySSE.h"
 
 DXGI::DXGI() :
   m_options(NULL),
@@ -33,11 +32,11 @@ DXGI::DXGI() :
   m_texture(),
   m_pointer(NULL)
 {
+
 }
 
 DXGI::~DXGI()
 {
-
 }
 
 bool DXGI::Initialize(CaptureOptions * options)
@@ -310,7 +309,7 @@ GrabStatus DXGI::GrabFrame(FrameInfo & frame)
         frame.stride = m_mapping.RowPitch / 4;
 
         unsigned int size = m_height * m_mapping.RowPitch;
-        memcpySSE(frame.buffer, m_mapping.pData, size < frame.bufferSize ? size : frame.bufferSize);
+        m_memcpy.Copy(frame.buffer, m_mapping.pData, size < frame.bufferSize ? size : frame.bufferSize);
         return GRAB_STATUS_OK;
       }
 
@@ -448,7 +447,7 @@ GrabStatus DXGI::GrabFrame(FrameInfo & frame)
   frame.stride  = m_mapping.RowPitch / 4;
   unsigned int size = m_height * m_mapping.RowPitch;
 
-  memcpySSE(frame.buffer, m_mapping.pData, size < frame.bufferSize ? size : frame.bufferSize);
+  m_memcpy.Copy(frame.buffer, m_mapping.pData, size < frame.bufferSize ? size : frame.bufferSize);
 
   return GRAB_STATUS_OK;
 }

host/Capture/DXGI.h

@@ -20,6 +20,7 @@ Place, Suite 330, Boston, MA 02111-1307 USA
 #pragma once
 
 #include "ICapture.h"
+#include "MultiMemcpy.h"
 
 #define W32_LEAN_AND_MEAN
 #include <windows.h>
@@ -75,6 +76,7 @@ namespace Capture
     unsigned int  m_width;
     unsigned int  m_height;
 
+    MultiMemcpy                     m_memcpy;
     IDXGIFactory1Ptr                m_dxgiFactory;
     ID3D11DevicePtr                 m_device;
     D3D_FEATURE_LEVEL               m_featureLevel;

host/MultiMemcpy.cpp

@@ -0,0 +1,70 @@
+/*
+Looking Glass - KVM FrameRelay (KVMFR) Client
+Copyright (C) 2017 Geoffrey McRae <geoff@hostfission.com>
+https://looking-glass.hostfission.com
+
+This program is free software; you can redistribute it and/or modify it under
+the terms of the GNU General Public License as published by the Free Software
+Foundation; either version 2 of the License, or (at your option) any later
+version.
+
+This program is distributed in the hope that it will be useful, but WITHOUT ANY
+WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
+PARTICULAR PURPOSE. See the GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along with
+this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+Place, Suite 330, Boston, MA 02111-1307 USA
+*/
+
+#include "MultiMemcpy.h"
+#include "Util.h"
+#include "common/memcpySSE.h"
+
+MultiMemcpy::MultiMemcpy()
+{
+  for (int i = 0; i < MULTIMEMCPY_THREADS; ++i)
+  {
+    m_workers[i].start  = CreateSemaphore(NULL, 0, 1, NULL);
+    m_workers[i].stop   = CreateSemaphore(NULL, 0, 1, NULL);
+    m_semaphores[i]     = m_workers[i].stop;
+
+    m_workers[i].thread = CreateThread(0, 0, WorkerFunction, &m_workers[i], 0, NULL);
+  }
+}
+
+MultiMemcpy::~MultiMemcpy()
+{
+  for(int i = 0; i < MULTIMEMCPY_THREADS; ++i)
+  {
+    TerminateThread(m_workers[i].thread, 0);
+    CloseHandle(m_workers[i].start);
+    CloseHandle(m_workers[i].stop );
+  }
+}
+
+void MultiMemcpy::Copy(void * dst, void * src, size_t size)
+{
+  const size_t block = size / MULTIMEMCPY_THREADS;
+  for (int i = 0; i < MULTIMEMCPY_THREADS; ++i)
+  {
+    m_workers[i].dst  = (uint8_t *)dst + i * block;
+    m_workers[i].src  = (uint8_t *)src + i * block;
+    m_workers[i].size = (i + 1) * block - i * block;
+    ReleaseSemaphore(m_workers[i].start, 1, NULL);
+  }
+  
+  WaitForMultipleObjects(MULTIMEMCPY_THREADS, m_semaphores, TRUE, INFINITE);
+}
+
+DWORD WINAPI MultiMemcpy::WorkerFunction(LPVOID param)
+{
+  struct Worker * w = (struct Worker *)param;
+  
+  for(;;)
+  {
+    WaitForSingleObject(w->start, INFINITE);
+    memcpySSE(w->dst, w->src, w->size);
+    ReleaseSemaphore(w->stop, 1, NULL);
+  }
+}

host/MultiMemcpy.h

@@ -0,0 +1,48 @@
+/*
+Looking Glass - KVM FrameRelay (KVMFR) Client
+Copyright (C) 2017 Geoffrey McRae <geoff@hostfission.com>
+https://looking-glass.hostfission.com
+
+This program is free software; you can redistribute it and/or modify it under
+the terms of the GNU General Public License as published by the Free Software
+Foundation; either version 2 of the License, or (at your option) any later
+version.
+
+This program is distributed in the hope that it will be useful, but WITHOUT ANY
+WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
+PARTICULAR PURPOSE. See the GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along with
+this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+Place, Suite 330, Boston, MA 02111-1307 USA
+*/
+
+#define MULTIMEMCPY_THREADS 4
+
+#include <Windows.h>
+#include <stdint.h>
+
+#pragma once
+class MultiMemcpy
+{
+public:
+  MultiMemcpy();
+  ~MultiMemcpy();
+
+  void Copy(void * dst, void * src, size_t size);
+private:
+  struct Worker
+  {
+    HANDLE  start;
+    HANDLE  stop;
+    HANDLE  thread;
+    void  * dst;
+    void  * src;
+    size_t  size;
+  };
+
+  HANDLE m_semaphores[MULTIMEMCPY_THREADS];
+  struct Worker m_workers[MULTIMEMCPY_THREADS];
+  static DWORD WINAPI WorkerFunction(LPVOID param);
+};
+

host/Service.cpp

@@ -23,16 +23,9 @@ Place, Suite 330, Boston, MA 02111-1307 USA
 #include "common/debug.h"
 #include "common/KVMFR.h"
 
+#include "Util.h"
 #include "CaptureFactory.h"
 
-#if __MINGW32__
-#define INTERLOCKED_AND8 __sync_and_and_fetch
-#define INTERLOCKED_OR8 __sync_or_and_fetch
-#else
-#define INTERLOCKED_OR8 InterlockedOr8
-#define INTERLOCKED_AND8 InterlockedAnd8
-#endif
-
 Service * Service::m_instance = NULL;
 
 Service::Service() :

host/Util.h

@@ -29,6 +29,14 @@ Place, Suite 330, Boston, MA 02111-1307 USA
 #define min(a, b) ((a) < (b) ? (a) : (b))
 #endif
 
+#if __MINGW32__
+#define INTERLOCKED_AND8 __sync_and_and_fetch
+#define INTERLOCKED_OR8 __sync_or_and_fetch
+#else
+#define INTERLOCKED_OR8 InterlockedOr8
+#define INTERLOCKED_AND8 InterlockedAnd8
+#endif
+
 class Util
 {
 public:
@@ -120,68 +128,4 @@ public:
       _mm_stream_si128((__m128i *)&dest[32], v2);
     }
   }
-
-  static void DrawCursor(
-    const enum CursorType   type,
-    const uint8_t         * cursorData,
-    const POINT             cursorRect,
-    const unsigned int      cursorPitch,
-    const POINT             cursorPos,
-    FrameInfo             & frame
-  )
-  {
-    const int maxHeight = min(cursorRect.y, (int)frame.height - cursorPos.y);
-    const int maxWidth  = min(cursorRect.x, (int)frame.width  - cursorPos.x);
-
-    switch (type)
-    {
-      case CURSOR_TYPE_COLOR:
-      {
-        const unsigned int destPitch = frame.stride * 4;
-        for (int y = abs(min(0, cursorPos.y)); y < maxHeight; ++y)
-          for (int x = abs(min(0, cursorPos.x)); x < maxWidth; ++x)
-          {
-            uint8_t *src = (uint8_t *)cursorData + (cursorPitch * y) + (x * 4);
-            uint8_t *dst = (uint8_t *)frame.buffer + (destPitch * (y + cursorPos.y)) + ((x + cursorPos.x) * 4);
-
-            const unsigned int alpha = src[3] + 1;
-            const unsigned int inv = 256 - alpha;
-            dst[0] = (uint8_t)((alpha * src[0] + inv * dst[0]) >> 8);
-            dst[1] = (uint8_t)((alpha * src[1] + inv * dst[1]) >> 8);
-            dst[2] = (uint8_t)((alpha * src[2] + inv * dst[2]) >> 8);
-          }
-        break;
-      }
-
-      case CURSOR_TYPE_MASKED_COLOR:
-      {
-        for (int y = abs(min(0, cursorPos.y)); y < maxHeight; ++y)
-          for (int x = abs(min(0, cursorPos.x)); x < maxWidth; ++x)
-          {
-            uint32_t *src = (uint32_t *)cursorData + ((cursorPitch / 4) * y) + x;
-            uint32_t *dst = (uint32_t *)frame.buffer + (frame.stride * (y + cursorPos.y)) + (x + cursorPos.x);
-            if (*src & 0xff000000)
-              *dst = 0xff000000 | (*dst ^ *src);
-            else *dst = 0xff000000 | *src;
-          }
-        break;
-      }
-
-      case CURSOR_TYPE_MONOCHROME:
-      {
-        for (int y = abs(min(0, cursorPos.y)); y < maxHeight / 2; ++y)
-          for (int x = abs(min(0, cursorPos.x)); x < maxWidth; ++x)
-          {
-            uint8_t  *srcAnd = (uint8_t  *)cursorData + (cursorPitch * y) + (x / 8);
-            uint8_t  *srcXor = srcAnd + cursorPitch * (cursorRect.y / 2);
-            uint32_t *dst = (uint32_t *)frame.buffer + (frame.stride * (y + cursorPos.y)) + (x + cursorPos.x);
-            const uint8_t mask = 0x80 >> (x % 8);
-            const uint32_t andMask = (*srcAnd & mask) ? 0xFFFFFFFF : 0xFF000000;
-            const uint32_t xorMask = (*srcXor & mask) ? 0x00FFFFFF : 0x00000000;
-            *dst = (*dst & andMask) ^ xorMask;
-          }
-        break;
-      }
-    }
-  }
 };

host/looking-glass-host.vcxproj

@@ -335,6 +335,7 @@
     <ClCompile Include="CrashHandler.cpp" />
     <ClCompile Include="ivshmem.cpp" />
     <ClCompile Include="main.cpp" />
+    <ClCompile Include="MultiMemcpy.cpp" />
     <ClCompile Include="Service.cpp" />
   </ItemGroup>
   <ItemGroup>
@@ -344,6 +345,7 @@
     <ClInclude Include="CrashHandler.h" />
     <ClInclude Include="ICapture.h" />
     <ClInclude Include="ivshmem.h" />
+    <ClInclude Include="MultiMemcpy.h" />
     <ClInclude Include="Service.h" />
     <ClInclude Include="Util.h" />
   </ItemGroup>

host/looking-glass-host.vcxproj.filters

@@ -42,6 +42,9 @@
     <ClCompile Include="CrashHandler.cpp">
       <Filter>Source Files</Filter>
     </ClCompile>
+    <ClCompile Include="MultiMemcpy.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="ivshmem.h">
@@ -68,5 +71,8 @@
     <ClInclude Include="CrashHandler.h">
       <Filter>Header Files</Filter>
     </ClInclude>
+    <ClInclude Include="MultiMemcpy.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
   </ItemGroup>
 </Project>