ADDED asm-filters thread-safe patches by chrono.

ADDED maxCpuCores option to ini file and autodetect best value at first run.

doc/ReadMe.txt

@@ -63,9 +63,6 @@ Important:
 
 - Improve automatic 64k/128k flash save detection
 
-- HQ3x/4x ASM implementation produces wrong interpolation on the image's border
- - This has already been fixed in the C version; look at hq_base.h / line 343 - 372. The ASM version most likely only has something like skipLine instead of skipLinePlus and skipLineMinus, which is however necessary in order to work correctly.
-
 - Fix OpenGL issues
 
 - Remove 16 bit hack for filters
@@ -107,7 +104,7 @@ Performance:
 
 - Apply pixel filter only to changed parts of the image
 
-- Make use of multi-core CPUs
+- Make even more use of multi-core CPUs
 
 - Make use of 64 bit CPUs
 
@@ -150,4 +147,6 @@ Jonas Quinn
 Spacy
 
 Special Thanks go to:
-chrono - ASM genius, fixed a bug in the HQ3x/4x filters and made them thread-safe
+chrono:
+ASM genius, fixed a bug in the HQ3x/4x filters.
+Made HQ3x/4x and Bilinear filters (ASM versions) thread-safe

src/bilinear.cpp

@@ -24,12 +24,6 @@
 
 #include "System.h"
 
-static u8 row_cur[3*322];
-static u8 row_next[3*322];
-
-static u8 *rgb_row_cur = row_cur;
-static u8 *rgb_row_next = row_next;
-
 #define RGB(r,g,b) ((r)>>3) << systemRedShift |\
   ((g) >> 3) << systemGreenShift |\
   ((b) >> 3) << systemBlueShift\
@@ -83,6 +77,11 @@ static void fill_rgb_row_32(u32 *from, int src_width, u8 *row, int width)
 void Bilinear(u8 *srcPtr, u32 srcPitch, u8 * /* deltaPtr */,
               u8 *dstPtr, u32 dstPitch, int width, int height)
 {
+  u8 row_cur[3*322];
+  u8 row_next[3*322];
+  u8 *rgb_row_cur = row_cur;
+  u8 *rgb_row_next = row_next;
+
   u16 *to = (u16 *)dstPtr;
   u16 *to_odd = (u16 *)(dstPtr + dstPitch);
 
@@ -161,6 +160,11 @@ void Bilinear(u8 *srcPtr, u32 srcPitch, u8 * /* deltaPtr */,
 void BilinearPlus(u8 *srcPtr, u32 srcPitch, u8 * /* deltaPtr */,
                   u8 *dstPtr, u32 dstPitch, int width, int height)
 {
+  u8 row_cur[3*322];
+  u8 row_next[3*322];
+  u8 *rgb_row_cur = row_cur;
+  u8 *rgb_row_next = row_next;
+
   u16 *to = (u16 *)dstPtr;
   u16 *to_odd = (u16 *)(dstPtr + dstPitch);
 
@@ -250,6 +254,11 @@ void BilinearPlus(u8 *srcPtr, u32 srcPitch, u8 * /* deltaPtr */,
 void Bilinear32(u8 *srcPtr, u32 srcPitch, u8 * /* deltaPtr */,
                 u8 *dstPtr, u32 dstPitch, int width, int height)
 {
+  u8 row_cur[3*322];
+  u8 row_next[3*322];
+  u8 *rgb_row_cur = row_cur;
+  u8 *rgb_row_next = row_next;
+
   u32 *to = (u32 *)dstPtr;
   u32 *to_odd = (u32 *)(dstPtr + dstPitch);
 
@@ -330,6 +339,11 @@ void Bilinear32(u8 *srcPtr, u32 srcPitch, u8 * /* deltaPtr */,
 void BilinearPlus32(u8 *srcPtr, u32 srcPitch, u8 * /* deltaPtr */,
                     u8 *dstPtr, u32 dstPitch, int width, int height)
 {
+  u8 row_cur[3*322];
+  u8 row_next[3*322];
+  u8 *rgb_row_cur = row_cur;
+  u8 *rgb_row_next = row_next;
+
   u32 *to = (u32 *)dstPtr;
   u32 *to_odd = (u32 *)(dstPtr + dstPitch);
 

src/win32/Direct3D.cpp

@@ -21,7 +21,7 @@
 #ifndef NO_D3D
 
 // The number of pixel-filter threads to be created
-#define NTHREADS 4
+#define NTHREADS ( theApp.maxCpuCores )
 
 #pragma comment( lib, "d3d9" )
 #pragma comment( lib, "d3dx9" )
@@ -454,7 +454,7 @@ void Direct3DDisplay::render()
 			u32 dst_bytes_per_thread = lr.Pitch * dst_height_per_thread;
 
 			// Use Multi Threading
-			assert( NTHREADS > 0 );
+			assert( ( NTHREADS > 0 ) && ( NTHREADS < MAXIMUM_PROCESSORS ) );
 			for( int i = ( NTHREADS - 1 ) ; i > -1 ; i-- ) {
 				// create last thread first because it could have more work than the others (for eg. if NTHREADS = 3)
 				// (last thread has to process the remaining lines if (height / NTHREADS) is not an integer)
@@ -819,7 +819,7 @@ void Direct3DDisplay::calculateDestRect()
 		rectangleFillsScreen = true; // no clear() necessary
 		destRect.left = 0;
 		destRect.top = 0;
-		destRect.right = dpp.BackBufferWidth;   // for some reason there'l be a black
+		destRect.right = dpp.BackBufferWidth;   // for some reason there will be a black
 		destRect.bottom = dpp.BackBufferHeight; // border line when using -1 at the end
 	} else {
 		// use aspect ratio

src/win32/VBA.cpp

@@ -24,6 +24,7 @@
 #endif
 
 #include "stdafx.h"
+#include <intrin.h>
 
 #include "AVIWrite.h"
 #include "LangSelect.h"
@@ -207,16 +208,6 @@ void directXMessage(const char *msg)
 /////////////////////////////////////////////////////////////////////////////
 // VBA
 
-BEGIN_MESSAGE_MAP(VBA, CWinApp)
-  //{{AFX_MSG_MAP(VBA)
-  // NOTE - the ClassWizard will add and remove mapping macros here.
-  //    DO NOT EDIT what you see in these blocks of generated code!
-  //}}AFX_MSG_MAP
-  END_MESSAGE_MAP()
-
-  /////////////////////////////////////////////////////////////////////////////
-// VBA construction
-
 VBA::VBA()
 {
   // COINIT_MULTITHREADED is not supported by SHBrowseForFolder with BIF_USENEWUI
@@ -229,6 +220,7 @@ VBA::VBA()
   mode800Available = false;
   mode1024Available = false;
   mode1280Available = false;
+  maxCpuCores = 1;
   windowPositionX = 0;
   windowPositionY = 0;
   filterFunction = NULL;
@@ -1486,6 +1478,14 @@ void VBA::loadSettings()
   if(windowPositionY < 0)
     windowPositionY = 0;
 
+  maxCpuCores = regQueryDwordValue("maxCpuCores", 0);
+  if(maxCpuCores < 0) {
+	  maxCpuCores = 0;
+  }
+  if(maxCpuCores == 0) {
+	  maxCpuCores = detectCpuCores();
+  }
+
   useBiosFileGBA = ( regQueryDwordValue("useBiosGBA", 0) == 1 ) ? true : false;
 
   useBiosFileGB = ( regQueryDwordValue("useBiosGB", 0) == 1 ) ? true : false;
@@ -2474,6 +2474,8 @@ void VBA::saveSettings()
   regSetDwordValue("windowX", windowPositionX);
   regSetDwordValue("windowY", windowPositionY);
 
+  regSetDwordValue("maxCpuCores", maxCpuCores);
+
   regSetDwordValue("useBiosGBA", useBiosFileGBA);
 
   regSetDwordValue("useBiosGB", useBiosFileGB);
@@ -2587,6 +2589,17 @@ void VBA::saveSettings()
 #endif
 }
 
+int VBA::detectCpuCores()
+{
+	int CPUInfo[4];
+
+	__cpuid( CPUInfo, 1 );
+
+	int processor_count = ( CPUInfo[1] & 0x00FF0000 ) >> 16;
+
+	return processor_count;
+}
+
 void winSignal(int, int)
 {
 }

src/win32/VBA.h

@@ -79,6 +79,7 @@ class VBA : public CWinApp
   bool mode800Available;
   bool mode1024Available;
   bool mode1280Available;
+  int maxCpuCores; // maximum number of CPU cores VBA should use, 0 means auto-detect
   int windowPositionX;
   int windowPositionY;
   void (*filterFunction)(u8*,u32,u8*,u8*,u32,int,int);
@@ -262,13 +263,10 @@ class VBA : public CWinApp
   void updateFrameSkip();
   void loadSettings();
   void addRecentFile(CString file);
-  //{{AFX_MSG(VBA)
+
-  afx_msg void OnAppAbout();
+  private:
-  // NOTE - the ClassWizard will add and remove member functions here.
+  int detectCpuCores();
-  //    DO NOT EDIT what you see in these blocks of generated code !
+};
-  //}}AFX_MSG
-  DECLARE_MESSAGE_MAP()
-    };
 
     extern VBA theApp;
 	extern int emulating;