Added SSE and MMX optimization for SDL_FillRect()

--HG--
extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402611

src/video/SDL_blit.c

@@ -110,7 +110,8 @@ SDL_SoftBlit(SDL_Surface * src, SDL_Rect * srcrect,
 #ifdef __MACOSX__
 #include <sys/sysctl.h>
 
-static SDL_bool SDL_UseAltivecPrefetch()
+static SDL_bool
+SDL_UseAltivecPrefetch()
 {
     const char key[] = "hw.l3cachesize";
     u_int64_t result = 0;
@@ -123,14 +124,16 @@ static SDL_bool SDL_UseAltivecPrefetch()
     }
 }
 #else
-static SDL_bool SDL_UseAltivecPrefetch()
+static SDL_bool
+SDL_UseAltivecPrefetch()
 {
     /* Just guess G4 */
     return SDL_TRUE;
 }
 #endif /* __MACOSX__ */
 
-static SDL_loblit SDL_ChooseBlitFunc(SDL_BlitEntry *entries, int count)
+static SDL_loblit
+SDL_ChooseBlitFunc(SDL_BlitEntry * entries, int count)
 {
     int i;
     static Uint32 features = 0xffffffff;

src/video/SDL_blit.h

@@ -24,6 +24,13 @@
 #ifndef _SDL_blit_h
 #define _SDL_blit_h
 
+#ifdef __MMX__
+#include <mmintrin.h>
+#endif
+#ifdef __SSE__
+#include <xmmintrin.h>
+#endif
+
 #include "SDL_endian.h"
 
 /* The structure passed to the low level blit functions */
@@ -92,6 +99,14 @@ extern SDL_loblit SDL_CalculateAlphaBlit(SDL_Surface * surface, int complex);
  * Useful macros for blitting routines
  */
 
+#if defined(__GNUC__)
+#define DECLARE_ALIGNED(t,v,a)  t __attribute__((aligned(a))) v
+#elif defined(_MSC_VER)
+#define DECLARE_ALIGNED(t,v,a)  t __declspec(align(a)) v
+#else
+#define DECLARE_ALIGNED(t,v,a)  t v
+#endif
+
 #define FORMAT_EQUAL(A, B)						\
     ((A)->BitsPerPixel == (B)->BitsPerPixel				\
      && ((A)->Rmask == (B)->Rmask) && ((A)->Amask == (B)->Amask))

src/video/SDL_blit_copy.c

@@ -23,13 +23,8 @@
 
 #include "SDL_video.h"
 #include "SDL_blit.h"
+#include "SDL_blit_copy.h"
 
-#ifdef __MMX__
-#include <mmintrin.h>
-#endif
-#ifdef __SSE__
-#include <xmmintrin.h>
-#endif
 
 #ifdef __MMX__
 static __inline__ void