libsdl: added gcc4 patch

3 files changed, 475 insertions, 5 deletions

diff --git a/libsdl/.md5sum b/libsdl/.md5sum
index 589a19680..06316eb7a 100644
--- a/libsdl/.md5sum
+++ b/libsdl/.md5sum
@@ -1 +1,2 @@
 80919ef556425ff82a8555ff40a579a0  SDL-1.2.9.tar.gz
+ba148e589d8577925883564356a25bbc  libsdl-1.2.9-gcc4.patch
diff --git a/libsdl/Pkgfile b/libsdl/Pkgfile
index 5f7992618..841cbc9a4 100644
--- a/libsdl/Pkgfile
+++ b/libsdl/Pkgfile
@@ -1,18 +1,20 @@
 # Description: SDL, Simple DirectMedia Layer 
+# Description: a multimedia library which provides fast access to both graphics framebuffers and audio devices
 # URL: http://www.libsdl.org
-# Maintainer: Simone Rota, sip at varlock dot com
-# Packager: Daniel Mueller, daniel at danm dot de
+# Maintainer: Simone Rota, sip at crux dot nu
+# Packager: sten, nick dot steeves at shaw dot ca
 # Depends on: alsa-lib
 
 name=libsdl
 version=1.2.9
-release=1
-source=(http://www.libsdl.org/release/SDL-$version.tar.gz) 
+release=2
+source=(http://www.libsdl.org/release/SDL-$version.tar.gz $name-$version-gcc4.patch) 
 
 build () {
     cd SDL-$version
+    patch -p1 -i $SRC/$name-$version-gcc4.patch
     ./configure	--prefix=/usr \
-                --enable-alsa 
+        --enable-alsa 
     make 
     make DESTDIR=$PKG install
 }
diff --git a/libsdl/libsdl-1.2.9-gcc4.patch b/libsdl/libsdl-1.2.9-gcc4.patch
new file mode 100644
index 000000000..8d1d66f31
--- /dev/null
+++ b/libsdl/libsdl-1.2.9-gcc4.patch
@@ -0,0 +1,467 @@
+# The origins of this patch are the following two URLs
+# http://cvs.fedora.redhat.com/viewcvs/*checkout*/rpms/SDL/FC-5/SDL-1.2.9-yuv_mmx_gcc4.patch?rev=1.1
+# http://cvs.fedora.redhat.com/viewcvs/*checkout*/rpms/SDL/FC-5/SDL-1.2.8-gcc4.patch?rev=1.2
+
+diff -ru SDL-1.2.9.orig/src/hermes/HeadMMX.h SDL-1.2.9/src/hermes/HeadMMX.h
+--- SDL-1.2.9.orig/src/hermes/HeadMMX.h	2006-03-21 16:34:57.000000000 -0700
++++ SDL-1.2.9/src/hermes/HeadMMX.h	2006-03-21 16:36:07.000000000 -0700
+@@ -50,21 +50,19 @@
+   extern "C" {   
+   #endif
+ 
+-  void ConvertMMX(HermesConverterInterface *) __attribute__ ((alias ("_ConvertMMX")));
+-#if 0
+-  void ClearMMX_32(HermesClearInterface *) __attribute__ ((alias ("_ClearMMX_32")));
+-  void ClearMMX_24(HermesClearInterface *)  __attribute__ ((alias ("_ClearMMX_24")));
+-  void ClearMMX_16(HermesClearInterface *)  __attribute__ ((alias ("_ClearMMX_16")));
+-  void ClearMMX_8(HermesClearInterface *)  __attribute__ ((alias ("_ClearMMX_8")));
+-
+-  void ConvertMMXp32_16RGB555() __attribute__ ((alias ("_ConvertMMXp32_16RGB555")));
+-#endif
+-
+-  void ConvertMMXpII32_24RGB888() __attribute__ ((alias ("_ConvertMMXpII32_24RGB888")));
+-  void ConvertMMXpII32_16RGB565() __attribute__ ((alias ("_ConvertMMXpII32_16RGB565")));
+-  void ConvertMMXpII32_16BGR565() __attribute__ ((alias ("_ConvertMMXpII32_16BGR565")));
+-  void ConvertMMXpII32_16RGB555() __attribute__ ((alias ("_ConvertMMXpII32_16RGB555")));
+-  void ConvertMMXpII32_16BGR555() __attribute__ ((alias ("_ConvertMMXpII32_16BGR555")));
++  extern void _ConvertMMX(HermesConverterInterface *);
++  extern void _ConvertMMXpII32_24RGB888();
++  extern void _ConvertMMXpII32_16RGB565();
++  extern void _ConvertMMXpII32_16BGR565();
++  extern void _ConvertMMXpII32_16RGB555();
++  extern void _ConvertMMXpII32_16BGR555();
++
++  #define ConvertMMX _ConvertMMX
++  #define ConvertMMXpII32_24RGB888 _ConvertMMXpII32_24RGB888
++  #define ConvertMMXpII32_16RGB565 _ConvertMMXpII32_16RGB565
++  #define ConvertMMXpII32_16BGR565 _ConvertMMXpII32_16BGR565
++  #define ConvertMMXpII32_16RGB555 _ConvertMMXpII32_16RGB555
++  #define ConvertMMXpII32_16BGR555 _ConvertMMXpII32_16BGR555
+ 
+   #ifdef __cplusplus
+   }
+diff -ru SDL-1.2.9.orig/src/hermes/HeadX86.h SDL-1.2.9/src/hermes/HeadX86.h
+--- SDL-1.2.9.orig/src/hermes/HeadX86.h	2006-03-21 16:34:57.000000000 -0700
++++ SDL-1.2.9/src/hermes/HeadX86.h	2006-03-21 16:36:07.000000000 -0700
+@@ -79,56 +79,47 @@
+   extern "C" {
+   #endif
+ 
+-  int Hermes_X86_CPU() __attribute__ ((alias ("_Hermes_X86_CPU")));
++  extern int _Hermes_X86_CPU();
+ 
+-  void ConvertX86(HermesConverterInterface *) __attribute__ ((alias ("_ConvertX86")));
++  extern void _ConvertX86(HermesConverterInterface *);
++
++  extern void _ConvertX86p32_32BGR888();
++  extern void _ConvertX86p32_32RGBA888();
++  extern void _ConvertX86p32_32BGRA888();
++  extern void _ConvertX86p32_24RGB888();
++  extern void _ConvertX86p32_24BGR888();
++  extern void _ConvertX86p32_16RGB565();
++  extern void _ConvertX86p32_16BGR565();
++  extern void _ConvertX86p32_16RGB555();
++  extern void _ConvertX86p32_16BGR555();
++  extern void _ConvertX86p32_8RGB332();
++
++  extern void _ConvertX86p16_16BGR565();
++  extern void _ConvertX86p16_16RGB555();
++  extern void _ConvertX86p16_16BGR555();
++  extern void _ConvertX86p16_8RGB332();
++
++
++  #define Hermes_X86_CPU _Hermes_X86_CPU
++
++  #define ConvertX86 _ConvertX86
++
++  #define ConvertX86p32_32BGR888 _ConvertX86p32_32BGR888
++  #define ConvertX86p32_32RGBA888 _ConvertX86p32_32RGBA888
++  #define ConvertX86p32_32BGRA888 _ConvertX86p32_32BGRA888
++  #define ConvertX86p32_24RGB888 _ConvertX86p32_24RGB888
++  #define ConvertX86p32_24BGR888 _ConvertX86p32_24BGR888
++  #define ConvertX86p32_16RGB565 _ConvertX86p32_16RGB565
++  #define ConvertX86p32_16BGR565 _ConvertX86p32_16BGR565
++  #define ConvertX86p32_16RGB555 _ConvertX86p32_16RGB555
++  #define ConvertX86p32_16BGR555 _ConvertX86p32_16BGR555
++  #define ConvertX86p32_8RGB332 _ConvertX86p32_8RGB332
++
++  #define ConvertX86p16_16BGR565 _ConvertX86p16_16BGR565
++  #define ConvertX86p16_16RGB555 _ConvertX86p16_16RGB555
++  #define ConvertX86p16_16BGR555 _ConvertX86p16_16BGR555
++  #define ConvertX86p16_8RGB332 _ConvertX86p16_8RGB332
+ 
+-#if 0
+-  void ClearX86_32(HermesClearInterface *) __attribute__ ((alias ("_ClearX86_32")));
+-  void ClearX86_24(HermesClearInterface *)  __attribute__ ((alias ("_ClearX86_24")));
+-  void ClearX86_16(HermesClearInterface *)  __attribute__ ((alias ("_ClearX86_16")));
+-  void ClearX86_8(HermesClearInterface *)  __attribute__ ((alias ("_ClearX86_8")));
+-#endif
+-
+-  void ConvertX86p32_32BGR888() __attribute__ ((alias ("_ConvertX86p32_32BGR888")));
+-  void ConvertX86p32_32RGBA888() __attribute__ ((alias ("_ConvertX86p32_32RGBA888")));
+-  void ConvertX86p32_32BGRA888() __attribute__ ((alias ("_ConvertX86p32_32BGRA888")));
+-  void ConvertX86p32_24RGB888() __attribute__ ((alias ("_ConvertX86p32_24RGB888")));
+-  void ConvertX86p32_24BGR888() __attribute__ ((alias ("_ConvertX86p32_24BGR888")));
+-  void ConvertX86p32_16RGB565() __attribute__ ((alias ("_ConvertX86p32_16RGB565")));
+-  void ConvertX86p32_16BGR565() __attribute__ ((alias ("_ConvertX86p32_16BGR565")));
+-  void ConvertX86p32_16RGB555() __attribute__ ((alias ("_ConvertX86p32_16RGB555")));
+-  void ConvertX86p32_16BGR555() __attribute__ ((alias ("_ConvertX86p32_16BGR555")));
+-  void ConvertX86p32_8RGB332() __attribute__ ((alias ("_ConvertX86p32_8RGB332")));
+-
+-#if 0
+-  void ConvertX86p16_32RGB888() __attribute__ ((alias ("_ConvertX86p16_32RGB888")));
+-  void ConvertX86p16_32BGR888() __attribute__ ((alias ("_ConvertX86p16_32BGR888")));
+-  void ConvertX86p16_32RGBA888() __attribute__ ((alias ("_ConvertX86p16_32RGBA888")));
+-  void ConvertX86p16_32BGRA888() __attribute__ ((alias ("_ConvertX86p16_32BGRA888")));
+-  void ConvertX86p16_24RGB888() __attribute__ ((alias ("_ConvertX86p16_24RGB888")));
+-  void ConvertX86p16_24BGR888() __attribute__ ((alias ("_ConvertX86p16_24BGR888")));
+-#endif
+-  void ConvertX86p16_16BGR565() __attribute__ ((alias ("_ConvertX86p16_16BGR565")));
+-  void ConvertX86p16_16RGB555() __attribute__ ((alias ("_ConvertX86p16_16RGB555")));
+-  void ConvertX86p16_16BGR555() __attribute__ ((alias ("_ConvertX86p16_16BGR555")));
+-  void ConvertX86p16_8RGB332() __attribute__ ((alias ("_ConvertX86p16_8RGB332")));
+-
+-#if 0
+-  void CopyX86p_4byte() __attribute__ ((alias ("_CopyX86p_4byte")));
+-  void CopyX86p_3byte() __attribute__ ((alias ("_CopyX86p_3byte")));
+-  void CopyX86p_2byte() __attribute__ ((alias ("_CopyX86p_2byte")));
+-  void CopyX86p_1byte() __attribute__ ((alias ("_CopyX86p_1byte")));
+-
+-  void ConvertX86pI8_32() __attribute__ ((alias ("_ConvertX86pI8_32")));
+-  void ConvertX86pI8_24() __attribute__ ((alias ("_ConvertX86pI8_24")));
+-  void ConvertX86pI8_16() __attribute__ ((alias ("_ConvertX86pI8_16")));
+-
+-  extern int ConvertX86p16_32RGB888_LUT_X86[512] __attribute__ ((alias ("_ConvertX86p16_32RGB888_LUT_X86")));
+-  extern int ConvertX86p16_32BGR888_LUT_X86[512] __attribute__ ((alias ("_ConvertX86p16_32BGR888_LUT_X86")));
+-  extern int ConvertX86p16_32RGBA888_LUT_X86[512] __attribute__ ((alias ("_ConvertX86p16_32RGBA888_LUT_X86")));
+-  extern int ConvertX86p16_32BGRA888_LUT_X86[512] __attribute__ ((alias ("_ConvertX86p16_32BGRA888_LUT_X86")));
+-#endif
+ 
+   #ifdef __cplusplus
+   }
+diff -ru SDL-1.2.9.orig/src/video/SDL_yuv_mmx.c SDL-1.2.9/src/video/SDL_yuv_mmx.c
+--- SDL-1.2.9.orig/src/video/SDL_yuv_mmx.c	2006-03-21 16:34:57.000000000 -0700
++++ SDL-1.2.9/src/video/SDL_yuv_mmx.c	2006-03-21 16:36:37.000000000 -0700
+@@ -30,42 +30,31 @@
+ 
+ #include "SDL_types.h"
+ 
+-#if __GNUC__ > 2
+-#    undef GCC2_HACK
+-#else
+-#    define GCC2_HACK
+-#endif
+-
+- 
+-#if defined(GCC2_HACK) && defined (__ELF__)
+-#define ASM_VAR(X) _##X
+-#else
+-#define ASM_VAR(X) X
+-#endif
++#define ASM_ARRAY(x) x[] __asm__("_" #x) __attribute__((used))
+  
+-static volatile unsigned int  ASM_VAR(MMX_0080w)[]    = {0x00800080, 0x00800080};
+-static volatile unsigned int  ASM_VAR(MMX_00FFw)[]    = {0x00ff00ff, 0x00ff00ff}; 
+-static volatile unsigned int  ASM_VAR(MMX_FF00w)[]    = {0xff00ff00, 0xff00ff00}; 
+-
+-static volatile unsigned short ASM_VAR(MMX_Ycoeff)[]  = {0x004a, 0x004a, 0x004a, 0x004a}; 
+-
+-static volatile unsigned short ASM_VAR(MMX_UbluRGB)[] = {0x0072, 0x0072, 0x0072, 0x0072};    
+-static volatile unsigned short ASM_VAR(MMX_VredRGB)[] = {0x0059, 0x0059, 0x0059, 0x0059};  
+-static volatile unsigned short ASM_VAR(MMX_UgrnRGB)[] = {0xffea, 0xffea, 0xffea, 0xffea}; 
+-static volatile unsigned short ASM_VAR(MMX_VgrnRGB)[] = {0xffd2, 0xffd2, 0xffd2, 0xffd2};  
+-
+-static volatile unsigned short ASM_VAR(MMX_Ublu5x5)[] = {0x0081, 0x0081, 0x0081, 0x0081};
+-static volatile unsigned short ASM_VAR(MMX_Vred5x5)[] = {0x0066, 0x0066, 0x0066, 0x0066};
+-static volatile unsigned short ASM_VAR(MMX_Ugrn555)[] = {0xffe7, 0xffe7, 0xffe7, 0xffe7};
+-static volatile unsigned short ASM_VAR(MMX_Vgrn555)[] = {0xffcc, 0xffcc, 0xffcc, 0xffcc};
+-static volatile unsigned short ASM_VAR(MMX_Ugrn565)[] = {0xffe8, 0xffe8, 0xffe8, 0xffe8};
+-static volatile unsigned short ASM_VAR(MMX_Vgrn565)[] = {0xffcd, 0xffcd, 0xffcd, 0xffcd};
+-
+-static volatile unsigned short ASM_VAR(MMX_red555)[]  = {0x7c00, 0x7c00, 0x7c00, 0x7c00};
+-static volatile unsigned short ASM_VAR(MMX_red565)[]  = {0xf800, 0xf800, 0xf800, 0xf800};
+-static volatile unsigned short ASM_VAR(MMX_grn555)[]  = {0x03e0, 0x03e0, 0x03e0, 0x03e0};
+-static volatile unsigned short ASM_VAR(MMX_grn565)[]  = {0x07e0, 0x07e0, 0x07e0, 0x07e0};
+-static volatile unsigned short ASM_VAR(MMX_blu5x5)[]  = {0x001f, 0x001f, 0x001f, 0x001f};
++static unsigned int  ASM_ARRAY(MMX_0080w)    = {0x00800080, 0x00800080};
++static unsigned int  ASM_ARRAY(MMX_00FFw)    = {0x00ff00ff, 0x00ff00ff};
++static unsigned int  ASM_ARRAY(MMX_FF00w)    = {0xff00ff00, 0xff00ff00};
++
++static unsigned short ASM_ARRAY(MMX_Ycoeff)  = {0x004a, 0x004a, 0x004a, 0x004a};
++
++static unsigned short ASM_ARRAY(MMX_UbluRGB) = {0x0072, 0x0072, 0x0072, 0x0072};
++static unsigned short ASM_ARRAY(MMX_VredRGB) = {0x0059, 0x0059, 0x0059, 0x0059};
++static unsigned short ASM_ARRAY(MMX_UgrnRGB) = {0xffea, 0xffea, 0xffea, 0xffea};
++static unsigned short ASM_ARRAY(MMX_VgrnRGB) = {0xffd2, 0xffd2, 0xffd2, 0xffd2};
++
++static unsigned short ASM_ARRAY(MMX_Ublu5x5) = {0x0081, 0x0081, 0x0081, 0x0081};
++static unsigned short ASM_ARRAY(MMX_Vred5x5) = {0x0066, 0x0066, 0x0066, 0x0066};
++static unsigned short ASM_ARRAY(MMX_Ugrn555) = {0xffe7, 0xffe7, 0xffe7, 0xffe7};
++static unsigned short ASM_ARRAY(MMX_Vgrn555) = {0xffcc, 0xffcc, 0xffcc, 0xffcc};
++static unsigned short ASM_ARRAY(MMX_Ugrn565) = {0xffe8, 0xffe8, 0xffe8, 0xffe8};
++static unsigned short ASM_ARRAY(MMX_Vgrn565) = {0xffcd, 0xffcd, 0xffcd, 0xffcd};
++
++static unsigned short ASM_ARRAY(MMX_red555)  = {0x7c00, 0x7c00, 0x7c00, 0x7c00};
++static unsigned short ASM_ARRAY(MMX_red565)  = {0xf800, 0xf800, 0xf800, 0xf800};
++static unsigned short ASM_ARRAY(MMX_grn555)  = {0x03e0, 0x03e0, 0x03e0, 0x03e0};
++static unsigned short ASM_ARRAY(MMX_grn565)  = {0x07e0, 0x07e0, 0x07e0, 0x07e0};
++static unsigned short ASM_ARRAY(MMX_blu5x5)  = {0x001f, 0x001f, 0x001f, 0x001f};
+ 
+ /**
+    This MMX assembler is my first assembler/MMX program ever.
+@@ -127,21 +116,12 @@
+ 		 "movd (%2), %%mm2\n"           //    0  0  0  0 l3 l2 l1 l0
+ 		 "punpcklbw %%mm7,%%mm1\n" //         0  v3 0  v2 00 v1 00 v0
+ 		 "punpckldq %%mm1,%%mm1\n" //         00 v1 00 v0 00 v1 00 v0
+-#ifdef GCC2_HACK
+ 		 "psubw _MMX_0080w,%%mm1\n"  // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 
+-#else
+-		 "psubw %[_MMX_0080w],%%mm1\n"  // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 
+-#endif
+ 
+ 		 // create Cr_g (result in mm0)
+ 		 "movq %%mm1,%%mm0\n"           // r1 r1 r0 r0 r1 r1 r0 r0
+-#ifdef GCC2_HACK
+ 		 "pmullw _MMX_VgrnRGB,%%mm0\n"// red*-46dec=0.7136*64
+ 		 "pmullw _MMX_VredRGB,%%mm1\n"// red*89dec=1.4013*64
+-#else
+-		 "pmullw %[_MMX_VgrnRGB],%%mm0\n"// red*-46dec=0.7136*64
+-		 "pmullw %[_MMX_VredRGB],%%mm1\n"// red*89dec=1.4013*64
+-#endif
+ 		 "psraw  $6, %%mm0\n"           // red=red/64
+ 		 "psraw  $6, %%mm1\n"           // red=red/64
+ 		 
+@@ -150,13 +130,8 @@
+ 		 "movq (%2,%4),%%mm3\n"         //    0  0  0  0 L3 L2 L1 L0
+ 		 "punpckldq %%mm3,%%mm2\n"      //   L3 L2 L1 L0 l3 l2 l1 l0
+ 		 "movq %%mm2,%%mm4\n"           //   L3 L2 L1 L0 l3 l2 l1 l0
+-#ifdef GCC2_HACK
+ 		 "pand _MMX_FF00w,%%mm2\n"      //   L3 0  L1  0 l3  0 l1  0
+ 		 "pand _MMX_00FFw,%%mm4\n"      //   0  L2  0 L0  0 l2  0 l0
+-#else
+-		 "pand %[_MMX_FF00w],%%mm2\n"      //   L3 0  L1  0 l3  0 l1  0
+-		 "pand %[_MMX_00FFw],%%mm4\n"      //   0  L2  0 L0  0 l2  0 l0
+-#endif
+ 		 "psrlw $8,%%mm2\n"             //   0  L3  0 L1  0 l3  0 l1
+ 
+ 		 // create R (result in mm6)
+@@ -173,20 +148,11 @@
+ 		 "movd (%1), %%mm1\n"      //         0  0  0  0  u3 u2 u1 u0
+ 		 "punpcklbw %%mm7,%%mm1\n" //         0  u3 0  u2 00 u1 00 u0
+ 		 "punpckldq %%mm1,%%mm1\n" //         00 u1 00 u0 00 u1 00 u0
+-#ifdef GCC2_HACK
+ 		 "psubw _MMX_0080w,%%mm1\n"  // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 
+-#else
+-		 "psubw %[_MMX_0080w],%%mm1\n"  // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 
+-#endif
+ 		 // create Cb_g (result in mm5)
+ 		 "movq %%mm1,%%mm5\n"            // u1 u1 u0 u0 u1 u1 u0 u0
+-#ifdef GCC2_HACK
+ 		 "pmullw _MMX_UgrnRGB,%%mm5\n"    // blue*-109dec=1.7129*64
+ 		 "pmullw _MMX_UbluRGB,%%mm1\n"    // blue*114dec=1.78125*64
+-#else
+-		 "pmullw %[_MMX_UgrnRGB],%%mm5\n"    // blue*-109dec=1.7129*64
+-		 "pmullw %[_MMX_UbluRGB],%%mm1\n"    // blue*114dec=1.78125*64
+-#endif
+ 		 "psraw  $6, %%mm5\n"            // blue=red/64
+ 		 "psraw  $6, %%mm1\n"            // blue=blue/64
+ 
+@@ -268,17 +234,7 @@
+ 		 "popl %%ebx\n"
+ 		 :
+ 		 : "m" (cr), "r"(cb),"r"(lum),
+-		 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)
+-#ifndef GCC2_HACK
+-		 ,[_MMX_0080w] "m" (*MMX_0080w),
+-		 [_MMX_00FFw] "m" (*MMX_00FFw),
+-		 [_MMX_FF00w] "m" (*MMX_FF00w),
+-		 [_MMX_VgrnRGB] "m" (*MMX_VgrnRGB),
+-		 [_MMX_VredRGB] "m" (*MMX_VredRGB),
+-		 [_MMX_UgrnRGB] "m" (*MMX_UgrnRGB),
+-		 [_MMX_UbluRGB] "m" (*MMX_UbluRGB)
+-#endif
+-			 );
++		 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod));
+ }
+ 
+ void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix,
+@@ -307,48 +263,21 @@
+          "movd           (%%ebx),                %%mm1\n" // 4 Cr                0  0  0  0 v3 v2 v1 v0
+          "punpcklbw      %%mm7,                  %%mm0\n" // 4 W cb   0 u3  0 u2  0 u1  0 u0
+          "punpcklbw      %%mm7,                  %%mm1\n" // 4 W cr   0 v3  0 v2  0 v1  0 v0
+-#ifdef GCC2_HACK
+          "psubw          _MMX_0080w,             %%mm0\n"
+          "psubw          _MMX_0080w,             %%mm1\n"
+-#else
+-         "psubw          %[_MMX_0080w],             %%mm0\n"
+-         "psubw          %[_MMX_0080w],             %%mm1\n"
+-#endif
+          "movq           %%mm0,                  %%mm2\n" // Cb                   0 u3  0 u2  0 u1  0 u0
+          "movq           %%mm1,                  %%mm3\n" // Cr
+-#ifdef GCC2_HACK
+          "pmullw         _MMX_Ugrn565,           %%mm2\n" // Cb2green 0 R3  0 R2  0 R1  0 R0
+-#else
+-         "pmullw         %[_MMX_Ugrn565],           %%mm2\n" // Cb2green 0 R3  0 R2  0 R1  0 R0
+-#endif
+          "movq           (%2),                   %%mm6\n" // L1      l7 L6 L5 L4 L3 L2 L1 L0
+-#ifdef GCC2_HACK
+          "pmullw         _MMX_Ublu5x5,           %%mm0\n" // Cb2blue
+          "pand           _MMX_00FFw,             %%mm6\n" // L1      00 L6 00 L4 00 L2 00 L0
+          "pmullw         _MMX_Vgrn565,           %%mm3\n" // Cr2green
+-#else
+-         "pmullw         %[_MMX_Ublu5x5],           %%mm0\n" // Cb2blue
+-         "pand           %[_MMX_00FFw],             %%mm6\n" // L1      00 L6 00 L4 00 L2 00 L0
+-         "pmullw         %[_MMX_Vgrn565],           %%mm3\n" // Cr2green
+-#endif
+          "movq           (%2),                   %%mm7\n" // L2
+-#ifdef GCC2_HACK
+          "pmullw         _MMX_Vred5x5,           %%mm1\n" // Cr2red
+-#else
+-         "pmullw         %[_MMX_Vred5x5],           %%mm1\n" // Cr2red
+-#endif
+          "psrlw          $8,                     %%mm7\n"        // L2           00 L7 00 L5 00 L3 00 L1
+-#ifdef GCC2_HACK
+          "pmullw         _MMX_Ycoeff,            %%mm6\n" // lum1
+-#else
+-         "pmullw         %[_MMX_Ycoeff],            %%mm6\n" // lum1
+-#endif
+          "paddw          %%mm3,                  %%mm2\n" // Cb2green + Cr2green == green
+-#ifdef GCC2_HACK
+          "pmullw         _MMX_Ycoeff,            %%mm7\n" // lum2
+-#else
+-         "pmullw         %[_MMX_Ycoeff],            %%mm7\n" // lum2
+-#endif
+ 
+          "movq           %%mm6,                  %%mm4\n" // lum1
+          "paddw          %%mm0,                  %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0
+@@ -366,20 +295,11 @@
+          "punpcklbw      %%mm4,                  %%mm4\n"
+          "punpcklbw      %%mm5,                  %%mm5\n"
+ 
+-#ifdef GCC2_HACK
+          "pand           _MMX_red565,            %%mm4\n"
+-#else
+-         "pand           %[_MMX_red565],            %%mm4\n"
+-#endif
+          "psllw          $3,                     %%mm5\n" // GREEN       1
+          "punpcklbw      %%mm6,                  %%mm6\n"
+-#ifdef GCC2_HACK
+          "pand           _MMX_grn565,            %%mm5\n"
+          "pand           _MMX_red565,            %%mm6\n"
+-#else
+-         "pand           %[_MMX_grn565],            %%mm5\n"
+-         "pand           %[_MMX_red565],            %%mm6\n"
+-#endif
+          "por            %%mm5,                  %%mm4\n" //
+          "psrlw          $11,                    %%mm6\n" // BLUE        1
+          "movq           %%mm3,                  %%mm5\n" // lum2
+@@ -393,44 +313,23 @@
+          "packuswb       %%mm3,                  %%mm3\n"
+          "packuswb       %%mm5,                  %%mm5\n"
+          "packuswb       %%mm7,                  %%mm7\n"
+-#ifdef GCC2_HACK
+          "pand           _MMX_00FFw,             %%mm6\n" // L3
+-#else
+-         "pand           %[_MMX_00FFw],             %%mm6\n" // L3
+-#endif
+          "punpcklbw      %%mm3,                  %%mm3\n"
+          "punpcklbw      %%mm5,                  %%mm5\n"
+-#ifdef GCC2_HACK
+          "pmullw         _MMX_Ycoeff,            %%mm6\n" // lum3
+-#else
+-         "pmullw         %[_MMX_Ycoeff],            %%mm6\n" // lum3
+-#endif
+          "punpcklbw      %%mm7,                  %%mm7\n"
+          "psllw          $3,                     %%mm5\n" // GREEN 2
+-#ifdef GCC2_HACK
+          "pand           _MMX_red565,            %%mm7\n"
+          "pand           _MMX_red565,            %%mm3\n"
+-#else
+-         "pand           %[_MMX_red565],            %%mm7\n"
+-         "pand           %[_MMX_red565],            %%mm3\n"
+-#endif
+          "psrlw          $11,                    %%mm7\n" // BLUE  2
+-#ifdef GCC2_HACK
+          "pand           _MMX_grn565,            %%mm5\n"
+-#else
+-         "pand           %[_MMX_grn565],            %%mm5\n"
+-#endif
+          "por            %%mm7,                  %%mm3\n"
+          "movq           (%2,%4),                %%mm7\n" // L4 load lum2
+          "por            %%mm5,                  %%mm3\n" //
+          "psrlw          $8,                     %%mm7\n" // L4
+          "movq           %%mm4,                  %%mm5\n"
+          "punpcklwd      %%mm3,                  %%mm4\n"
+-#ifdef GCC2_HACK
+          "pmullw         _MMX_Ycoeff,            %%mm7\n" // lum4
+-#else
+-         "pmullw         %[_MMX_Ycoeff],            %%mm7\n" // lum4
+-#endif
+          "punpckhwd      %%mm3,                  %%mm5\n"
+ 
+          "movq           %%mm4,                  (%3)\n"  // write row1
+@@ -457,20 +356,11 @@
+          "punpcklbw      %%mm5,                  %%mm5\n"
+          "punpcklbw      %%mm6,                  %%mm6\n"
+          "psllw          $3,                     %%mm5\n" // GREEN 3
+-#ifdef GCC2_HACK
+          "pand           _MMX_red565,            %%mm4\n"
+-#else
+-         "pand           %[_MMX_red565],            %%mm4\n"
+-#endif
+          "psraw          $6,                     %%mm3\n" // psr 6
+          "psraw          $6,                     %%mm0\n"
+-#ifdef GCC2_HACK
+          "pand           _MMX_red565,            %%mm6\n" // BLUE
+          "pand           _MMX_grn565,            %%mm5\n"
+-#else
+-         "pand           %[_MMX_red565],            %%mm6\n" // BLUE
+-         "pand           %[_MMX_grn565],            %%mm5\n"
+-#endif
+          "psrlw          $11,                    %%mm6\n" // BLUE  3
+          "por            %%mm5,                  %%mm4\n"
+          "psraw          $6,                     %%mm7\n"
+@@ -481,20 +371,11 @@
+          "punpcklbw      %%mm3,                  %%mm3\n"
+          "punpcklbw      %%mm0,                  %%mm0\n"
+          "punpcklbw      %%mm7,                  %%mm7\n"
+-#ifdef GCC2_HACK
+          "pand           _MMX_red565,            %%mm3\n"
+          "pand           _MMX_red565,            %%mm7\n" // BLUE
+-#else
+-         "pand           %[_MMX_red565],            %%mm3\n"
+-         "pand           %[_MMX_red565],            %%mm7\n" // BLUE
+-#endif
+          "psllw          $3,                     %%mm0\n" // GREEN 4
+          "psrlw          $11,                    %%mm7\n"
+-#ifdef GCC2_HACK
+          "pand           _MMX_grn565,            %%mm0\n"
+-#else
+-         "pand           %[_MMX_grn565],            %%mm0\n"
+-#endif
+          "por            %%mm7,                  %%mm3\n"
+          "por            %%mm0,                  %%mm3\n"
+ 
+@@ -526,21 +407,7 @@
+ 	 "popl %%ebx\n"
+          :
+          :"m" (cr), "r"(cb),"r"(lum),
+-	 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)
+-#ifndef GCC2_HACK
+-	 ,[_MMX_0080w] "m" (*MMX_0080w),
+-	 [_MMX_Ugrn565] "m" (*MMX_Ugrn565),
+-	 [_MMX_Ublu5x5] "m" (*MMX_Ublu5x5),
+-	 [_MMX_00FFw] "m" (*MMX_00FFw),
+-	 [_MMX_Vgrn565] "m" (*MMX_Vgrn565),
+-	 [_MMX_Vred5x5] "m" (*MMX_Vred5x5),
+-	 [_MMX_Ycoeff] "m" (*MMX_Ycoeff),
+-	 [_MMX_red565] "m" (*MMX_red565),
+-	 [_MMX_grn565] "m" (*MMX_grn565)
+-#endif
+-		 );
++	 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod));
+ }
+ 
+-#undef GCC2_HACK
+-
+ #endif /* GCC i386 inline assembly */