Index: src/hermes/mmx_main.asm =================================================================== --- src/hermes/mmx_main.asm (revision 3073) +++ src/hermes/mmx_main.asm (working copy) @@ -12,7 +12,6 @@ %include "common.inc" SDL_FUNC _ConvertMMX -SDL_FUNC _mmxreturn SECTION .text @@ -51,9 +50,8 @@ y_loop: mov ecx,[ebp+4] - jmp [ebp+32] + call [ebp+32] -_mmxreturn: add esi,[ebp+12] add edi,[ebp+28] Index: src/hermes/mmxp2_32.asm =================================================================== --- src/hermes/mmxp2_32.asm (revision 3073) +++ src/hermes/mmxp2_32.asm (working copy) @@ -28,8 +28,6 @@ SDL_FUNC _ConvertMMXpII32_16RGB555 SDL_FUNC _ConvertMMXpII32_16BGR555 -EXTERN _mmxreturn - ;; Macros for conversion routines %macro _push_immq_mask 1 @@ -121,7 +119,7 @@ dec ecx jnz .L3 .L4: - jmp _mmxreturn + return @@ -190,7 +188,7 @@ jnz .L3 .L4: - jmp _mmxreturn + retn _ConvertMMXpII32_16BGR565: @@ -260,7 +258,7 @@ jnz .L3 .L4: - jmp _mmxreturn + retn _ConvertMMXpII32_16BGR555: @@ -400,7 +398,7 @@ jnz .L3 .L4: - jmp _mmxreturn + retn %ifidn __OUTPUT_FORMAT__,elf section .note.GNU-stack noalloc noexec nowrite progbits Index: src/hermes/x86_main.asm =================================================================== --- src/hermes/x86_main.asm (revision 3073) +++ src/hermes/x86_main.asm (working copy) @@ -14,9 +14,6 @@ %include "common.inc" SDL_FUNC _ConvertX86 -SDL_FUNC _x86return - -SDL_FUNC _Hermes_X86_CPU SECTION .text @@ -55,9 +52,8 @@ y_loop: mov ecx,[ebp+4] - jmp [ebp+32] + call [ebp+32] -_x86return: add esi,[ebp+12] add edi,[ebp+28] @@ -74,49 +70,6 @@ -;; Hermes_X86_CPU returns the CPUID flags in eax - -_Hermes_X86_CPU: - pushfd - pop eax - - mov ecx,eax - - xor eax,040000h - push eax - - popfd - pushfd - - pop eax - xor eax,ecx - jz .L1 ; Processor is 386 - - push ecx - popfd - - mov eax,ecx - xor eax,200000h - - push eax - popfd - pushfd - - pop eax - xor eax,ecx - je .L1 - - push ebx - - mov eax,1 - cpuid - mov eax,edx - - pop ebx - -.L1: - ret - %ifidn __OUTPUT_FORMAT__,elf section .note.GNU-stack noalloc noexec nowrite progbits %endif Index: src/hermes/x86p_32.asm =================================================================== --- src/hermes/x86p_32.asm (revision 3073) +++ src/hermes/x86p_32.asm (working copy) @@ -24,8 +24,6 @@ SDL_FUNC _ConvertX86p32_16BGR555 SDL_FUNC _ConvertX86p32_8RGB332 -EXTERN _x86return - SECTION .text ;; _Convert_* @@ -53,7 +51,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; save ebp push ebp @@ -113,7 +111,7 @@ jnz .L5 .L6 pop ebp - jmp _x86return + retn @@ -133,7 +131,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; save ebp push ebp @@ -184,7 +182,7 @@ jnz .L5 .L6 pop ebp - jmp _x86return + retn @@ -204,7 +202,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; save ebp push ebp @@ -257,7 +255,7 @@ jnz .L5 .L6 pop ebp - jmp _x86return + retn @@ -282,7 +280,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; head mov edx,edi @@ -354,7 +352,7 @@ jnz .L6 .L7 pop ebp - jmp _x86return + retn @@ -380,7 +378,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; head mov edx,edi @@ -455,7 +453,7 @@ .L7 pop ebp - jmp _x86return + retn @@ -484,7 +482,7 @@ jnz .L1 .L2: ; End of short loop - jmp _x86return + retn .L3 ; head @@ -570,7 +568,7 @@ add edi,BYTE 2 .L7: - jmp _x86return + retn @@ -599,7 +597,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; head mov ebx,edi @@ -684,7 +682,7 @@ add edi,BYTE 2 .L7 - jmp _x86return + retn @@ -713,7 +711,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; head mov ebx,edi @@ -795,7 +793,7 @@ add edi,BYTE 2 .L7 - jmp _x86return + retn @@ -825,7 +823,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; head mov ebx,edi @@ -910,7 +908,7 @@ add edi,BYTE 2 .L7 - jmp _x86return + retn @@ -1040,7 +1038,7 @@ jnz .L3 .L4: - jmp _x86return + retn %ifidn __OUTPUT_FORMAT__,elf section .note.GNU-stack noalloc noexec nowrite progbits Index: src/hermes/x86p_16.asm =================================================================== --- src/hermes/x86p_16.asm (revision 3073) +++ src/hermes/x86p_16.asm (working copy) @@ -20,7 +20,6 @@ SDL_FUNC _ConvertX86p16_8RGB332 EXTERN _ConvertX86 -EXTERN _x86return SECTION .text @@ -49,7 +48,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; head mov eax,edi @@ -127,7 +126,7 @@ add edi,BYTE 2 .L7 - jmp _x86return + retn @@ -156,7 +155,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; head mov eax,edi @@ -235,7 +234,7 @@ jmp SHORT .L6 .L7 pop ebp - jmp _x86return + retn @@ -269,7 +268,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 ; head mov eax,edi @@ -353,7 +352,7 @@ add edi,BYTE 2 .L7 - jmp _x86return + retn @@ -386,7 +385,7 @@ dec ecx jnz .L1 .L2 - jmp _x86return + retn .L3 mov eax,edi and eax,BYTE 11b @@ -484,7 +483,7 @@ jnz .L6 .L7 pop ebp - jmp _x86return + retn %ifidn __OUTPUT_FORMAT__,elf section .note.GNU-stack noalloc noexec nowrite progbits Index: src/video/SDL_yuv_sw.c =================================================================== --- src/video/SDL_yuv_sw.c (revision 3073) +++ src/video/SDL_yuv_sw.c (working copy) @@ -121,7 +121,7 @@ /* The colorspace conversion functions */ -#if 0 /*defined(__GNUC__) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES*/ +#if (__GNUC__ > 2) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES extern void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, unsigned char *lum, unsigned char *cr, unsigned char *cb, unsigned char *out, @@ -1061,7 +1061,7 @@ case SDL_YV12_OVERLAY: case SDL_IYUV_OVERLAY: if ( display->format->BytesPerPixel == 2 ) { -#if 0 /*defined(__GNUC__) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES*/ +#if (__GNUC__ > 2) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES /* inline assembly functions */ if ( SDL_HasMMX() && (Rmask == 0xF800) && (Gmask == 0x07E0) && @@ -1083,7 +1083,7 @@ swdata->Display2X = Color24DitherYV12Mod2X; } if ( display->format->BytesPerPixel == 4 ) { -#if 0 /*defined(__GNUC__) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES*/ +#if (__GNUC__ > 2) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES /* inline assembly functions */ if ( SDL_HasMMX() && (Rmask == 0x00FF0000) && (Gmask == 0x0000FF00) && Index: src/video/SDL_yuv_mmx.c =================================================================== --- src/video/SDL_yuv_mmx.c (revision 3073) +++ src/video/SDL_yuv_mmx.c (working copy) @@ -21,37 +21,36 @@ */ #include "SDL_config.h" -#if 0 /* FIXME: This code needs to be rewritten to reference the static data using relocatable addresses (e.g. http://www.gentoo.org/proj/en/hardened/pic-fix-guide.xml or http://nasm.sourceforge.net/doc/html/nasmdoc8.html#section-8.2) This code currently breaks on systems with readonly text segments (hardened Linux / Intel Mac) */ -#if defined(__GNUC__) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES +#if (__GNUC__ > 2) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES #include "SDL_stdinc.h" -#define ASM_ARRAY(x) x[] __asm__("_" #x) __attribute__((used)) - -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}; +#include "mmx.h" -static unsigned short ASM_ARRAY(MMX_Ycoeff) = {0x004a, 0x004a, 0x004a, 0x004a}; +static mmx_t MMX_0080w = { .ud = {0x00800080, 0x00800080} }; +static mmx_t MMX_00FFw = { .ud = {0x00ff00ff, 0x00ff00ff} }; +static mmx_t MMX_FF00w = { .ud = {0xff00ff00, 0xff00ff00} }; -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 mmx_t MMX_Ycoeff = { .uw = {0x004a, 0x004a, 0x004a, 0x004a} }; -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 mmx_t MMX_UbluRGB = { .uw = {0x0072, 0x0072, 0x0072, 0x0072} }; +static mmx_t MMX_VredRGB = { .uw = {0x0059, 0x0059, 0x0059, 0x0059} }; +static mmx_t MMX_UgrnRGB = { .uw = {0xffea, 0xffea, 0xffea, 0xffea} }; +static mmx_t MMX_VgrnRGB = { .uw = {0xffd2, 0xffd2, 0xffd2, 0xffd2} }; -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}; +static mmx_t MMX_Ublu5x5 = { .uw = {0x0081, 0x0081, 0x0081, 0x0081} }; +static mmx_t MMX_Vred5x5 = { .uw = {0x0066, 0x0066, 0x0066, 0x0066} }; +static mmx_t MMX_Ugrn555 = { .uw = {0xffe7, 0xffe7, 0xffe7, 0xffe7} }; +static mmx_t MMX_Vgrn555 = { .uw = {0xffcc, 0xffcc, 0xffcc, 0xffcc} }; +static mmx_t MMX_Ugrn565 = { .uw = {0xffe8, 0xffe8, 0xffe8, 0xffe8} }; +static mmx_t MMX_Vgrn565 = { .uw = {0xffcd, 0xffcd, 0xffcd, 0xffcd} }; +static mmx_t MMX_red555 = { .uw = {0x7c00, 0x7c00, 0x7c00, 0x7c00} }; +static mmx_t MMX_red565 = { .uw = {0xf800, 0xf800, 0xf800, 0xf800} }; +static mmx_t MMX_grn555 = { .uw = {0x03e0, 0x03e0, 0x03e0, 0x03e0} }; +static mmx_t MMX_grn565 = { .uw = {0x07e0, 0x07e0, 0x07e0, 0x07e0} }; +static mmx_t MMX_blu5x5 = { .uw = {0x001f, 0x001f, 0x001f, 0x001f} }; + /** This MMX assembler is my first assembler/MMX program ever. Thus it maybe buggy. @@ -96,10 +95,6 @@ mod = (mod+cols+mod)*4; // increment for row1 in byte __asm__ __volatile__ ( -/* We don't really care about PIC - the code should be rewritten to use - relative addressing for the static tables, so right now we take the - COW hit on the pages this code resides. Big deal. - This spill is just to reduce register pressure in the PIC case. */ "pushl %%ebx\n" "movl %0, %%ebx\n" @@ -112,12 +107,12 @@ "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 - "psubw _MMX_0080w,%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 + "psubw %9,%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 // create Cr_g (result in mm0) "movq %%mm1,%%mm0\n" // r1 r1 r0 r0 r1 r1 r0 r0 - "pmullw _MMX_VgrnRGB,%%mm0\n"// red*-46dec=0.7136*64 - "pmullw _MMX_VredRGB,%%mm1\n"// red*89dec=1.4013*64 + "pmullw %10,%%mm0\n" // red*-46dec=0.7136*64 + "pmullw %11,%%mm1\n" // red*89dec=1.4013*64 "psraw $6, %%mm0\n" // red=red/64 "psraw $6, %%mm1\n" // red=red/64 @@ -126,8 +121,8 @@ "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 - "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 + "pand %12,%%mm2\n" // L3 0 L1 0 l3 0 l1 0 + "pand %13,%%mm4\n" // 0 L2 0 L0 0 l2 0 l0 "psrlw $8,%%mm2\n" // 0 L3 0 L1 0 l3 0 l1 // create R (result in mm6) @@ -144,11 +139,11 @@ "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 - "psubw _MMX_0080w,%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 + "psubw %9,%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0 // create Cb_g (result in mm5) "movq %%mm1,%%mm5\n" // u1 u1 u0 u0 u1 u1 u0 u0 - "pmullw _MMX_UgrnRGB,%%mm5\n" // blue*-109dec=1.7129*64 - "pmullw _MMX_UbluRGB,%%mm1\n" // blue*114dec=1.78125*64 + "pmullw %14,%%mm5\n" // blue*-109dec=1.7129*64 + "pmullw %15,%%mm1\n" // blue*114dec=1.78125*64 "psraw $6, %%mm5\n" // blue=red/64 "psraw $6, %%mm1\n" // blue=blue/64 @@ -230,7 +225,10 @@ "popl %%ebx\n" : : "m" (cr), "r"(cb),"r"(lum), - "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)); + "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod), + "m"(MMX_0080w),"m"(MMX_VgrnRGB),"m"(MMX_VredRGB), + "m"(MMX_FF00w),"m"(MMX_00FFw),"m"(MMX_UgrnRGB), + "m"(MMX_UbluRGB)); } void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, @@ -259,21 +257,21 @@ "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 - "psubw _MMX_0080w, %%mm0\n" - "psubw _MMX_0080w, %%mm1\n" + "psubw %9, %%mm0\n" + "psubw %9, %%mm1\n" "movq %%mm0, %%mm2\n" // Cb 0 u3 0 u2 0 u1 0 u0 "movq %%mm1, %%mm3\n" // Cr - "pmullw _MMX_Ugrn565, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0 + "pmullw %10, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0 "movq (%2), %%mm6\n" // L1 l7 L6 L5 L4 L3 L2 L1 L0 - "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 + "pmullw %11, %%mm0\n" // Cb2blue + "pand %12, %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0 + "pmullw %13, %%mm3\n" // Cr2green "movq (%2), %%mm7\n" // L2 - "pmullw _MMX_Vred5x5, %%mm1\n" // Cr2red + "pmullw %14, %%mm1\n" // Cr2red "psrlw $8, %%mm7\n" // L2 00 L7 00 L5 00 L3 00 L1 - "pmullw _MMX_Ycoeff, %%mm6\n" // lum1 + "pmullw %15, %%mm6\n" // lum1 "paddw %%mm3, %%mm2\n" // Cb2green + Cr2green == green - "pmullw _MMX_Ycoeff, %%mm7\n" // lum2 + "pmullw %15, %%mm7\n" // lum2 "movq %%mm6, %%mm4\n" // lum1 "paddw %%mm0, %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0 @@ -291,11 +289,11 @@ "punpcklbw %%mm4, %%mm4\n" "punpcklbw %%mm5, %%mm5\n" - "pand _MMX_red565, %%mm4\n" + "pand %16, %%mm4\n" "psllw $3, %%mm5\n" // GREEN 1 "punpcklbw %%mm6, %%mm6\n" - "pand _MMX_grn565, %%mm5\n" - "pand _MMX_red565, %%mm6\n" + "pand %17, %%mm5\n" + "pand %16, %%mm6\n" "por %%mm5, %%mm4\n" // "psrlw $11, %%mm6\n" // BLUE 1 "movq %%mm3, %%mm5\n" // lum2 @@ -309,23 +307,23 @@ "packuswb %%mm3, %%mm3\n" "packuswb %%mm5, %%mm5\n" "packuswb %%mm7, %%mm7\n" - "pand _MMX_00FFw, %%mm6\n" // L3 + "pand %12, %%mm6\n" // L3 "punpcklbw %%mm3, %%mm3\n" "punpcklbw %%mm5, %%mm5\n" - "pmullw _MMX_Ycoeff, %%mm6\n" // lum3 + "pmullw %15, %%mm6\n" // lum3 "punpcklbw %%mm7, %%mm7\n" "psllw $3, %%mm5\n" // GREEN 2 - "pand _MMX_red565, %%mm7\n" - "pand _MMX_red565, %%mm3\n" + "pand %16, %%mm7\n" + "pand %16, %%mm3\n" "psrlw $11, %%mm7\n" // BLUE 2 - "pand _MMX_grn565, %%mm5\n" + "pand %17, %%mm5\n" "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" - "pmullw _MMX_Ycoeff, %%mm7\n" // lum4 + "pmullw %15, %%mm7\n" // lum4 "punpckhwd %%mm3, %%mm5\n" "movq %%mm4, (%3)\n" // write row1 @@ -352,11 +350,11 @@ "punpcklbw %%mm5, %%mm5\n" "punpcklbw %%mm6, %%mm6\n" "psllw $3, %%mm5\n" // GREEN 3 - "pand _MMX_red565, %%mm4\n" + "pand %16, %%mm4\n" "psraw $6, %%mm3\n" // psr 6 "psraw $6, %%mm0\n" - "pand _MMX_red565, %%mm6\n" // BLUE - "pand _MMX_grn565, %%mm5\n" + "pand %16, %%mm6\n" // BLUE + "pand %17, %%mm5\n" "psrlw $11, %%mm6\n" // BLUE 3 "por %%mm5, %%mm4\n" "psraw $6, %%mm7\n" @@ -367,11 +365,11 @@ "punpcklbw %%mm3, %%mm3\n" "punpcklbw %%mm0, %%mm0\n" "punpcklbw %%mm7, %%mm7\n" - "pand _MMX_red565, %%mm3\n" - "pand _MMX_red565, %%mm7\n" // BLUE + "pand %16, %%mm3\n" + "pand %16, %%mm7\n" // BLUE "psllw $3, %%mm0\n" // GREEN 4 "psrlw $11, %%mm7\n" - "pand _MMX_grn565, %%mm0\n" + "pand %17, %%mm0\n" "por %%mm7, %%mm3\n" "por %%mm0, %%mm3\n" @@ -403,8 +401,11 @@ "popl %%ebx\n" : :"m" (cr), "r"(cb),"r"(lum), - "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)); + "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod), + "m"(MMX_0080w),"m"(MMX_Ugrn565),"m"(MMX_Ublu5x5), + "m"(MMX_00FFw),"m"(MMX_Vgrn565),"m"(MMX_Vred5x5), + "m"(MMX_Ycoeff),"m"(MMX_red565),"m"(MMX_grn565)); } -#endif /* GCC i386 inline assembly */ -#endif /* 0 */ +#endif /* GCC3 i386 inline assembly */ +