From f3c65c4548d6a2b921a5fbf1f37c346fd90e86a3 Mon Sep 17 00:00:00 2001 From: Sam Lantinga Date: Sun, 16 May 2004 17:19:48 +0000 Subject: [PATCH] Date: Mon, 10 May 2004 10:17:46 -0400 From: Mike Frysinger Subject: Re: [SDL] gcc-3.4.0 / PIC fix here's a combined patch (yours and the one i mentioned earlier) that i tested with gcc-3.4.0 and gcc-3.3.3 --- src/audio/SDL_mixer_MMX.c | 54 ++++++++++++++++++--------------------- src/cpuinfo/SDL_cpuinfo.c | 4 +-- src/video/SDL_yuv_mmx.c | 38 ++++++++++++++++++--------- 3 files changed, 53 insertions(+), 43 deletions(-) diff --git a/src/audio/SDL_mixer_MMX.c b/src/audio/SDL_mixer_MMX.c index 179aa92be..51a3b9d4e 100644 --- a/src/audio/SDL_mixer_MMX.c +++ b/src/audio/SDL_mixer_MMX.c @@ -15,13 +15,11 @@ void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume) { __asm__ __volatile__ ( -" movl %0,%%edi\n" // edi = dst -" movl %1,%%esi\n" // esi = src " movl %3,%%eax\n" // eax = volume -" movl %2,%%ebx\n" // ebx = size +" movl %2,%%edx\n" // edx = size -" shrl $4,%%ebx\n" // process 16 bytes per iteration = 8 samples +" shrl $4,%%edx\n" // process 16 bytes per iteration = 8 samples " jz .endS16\n" @@ -39,14 +37,14 @@ void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume) ".align 16\n" " .mixloopS16:\n" -" movq (%%esi),%%mm1\n" // mm1 = a|b|c|d +" movq (%1),%%mm1\n" // mm1 = a|b|c|d " movq %%mm1,%%mm2\n" // mm2 = a|b|c|d -" movq 8(%%esi),%%mm4\n" // mm4 = e|f|g|h +" movq 8(%1),%%mm4\n" // mm4 = e|f|g|h // pré charger le buffer dst dans mm7 -" movq (%%edi),%%mm7\n" // mm7 = dst[0]" +" movq (%0),%%mm7\n" // mm7 = dst[0]" // multiplier par le volume " pmullw %%mm0,%%mm1\n" // mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v) @@ -69,11 +67,11 @@ void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume) " punpcklwd %%mm5,%%mm6\n" // mm6 = g*v|h*v // pré charger le buffer dst dans mm5 -" movq 8(%%edi),%%mm5\n" // mm5 = dst[1] +" movq 8(%0),%%mm5\n" // mm5 = dst[1] // diviser par 128 " psrad $7,%%mm1\n" // mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME -" addl $16,%%esi\n" +" add $16,%1\n" " psrad $7,%%mm3\n" // mm3 = c*v/128|d*v/128 @@ -87,15 +85,15 @@ void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume) // mm4 = le sample avec le volume modifié " packssdw %%mm4,%%mm6\n" // mm6 = s(e*v|f*v|g*v|h*v) -" movq %%mm3,(%%edi)\n" +" movq %%mm3,(%0)\n" " paddsw %%mm5,%%mm6\n" // mm6 = adjust_volume(src)+dst -" movq %%mm6,8(%%edi)\n" +" movq %%mm6,8(%0)\n" -" addl $16,%%edi\n" +" add $16,%0\n" -" dec %%ebx\n" +" dec %%edx\n" " jnz .mixloopS16\n" @@ -103,9 +101,9 @@ void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume) ".endS16:\n" : - : "m" (dst), "m"(src),"m"(size), + : "r" (dst), "r"(src),"m"(size), "m"(volume) - : "eax","ebx", "esi", "edi","memory" + : "eax","edx","memory" ); } @@ -119,11 +117,9 @@ void SDL_MixAudio_MMX_S8(char* dst,char* src,unsigned int size,int volume) { __asm__ __volatile__ ( -" movl %0,%%edi\n" // edi = dst -" movl %1,%%esi\n" // esi = src " movl %3,%%eax\n" // eax = volume -" movd %%ebx,%%mm0\n" +" movd %%edx,%%mm0\n" " movq %%mm0,%%mm1\n" " psllq $16,%%mm0\n" " por %%mm1,%%mm0\n" @@ -132,17 +128,17 @@ void SDL_MixAudio_MMX_S8(char* dst,char* src,unsigned int size,int volume) " psllq $16,%%mm0\n" " por %%mm1,%%mm0\n" -" movl %2,%%ebx\n" // ebx = size -" shr $3,%%ebx\n" // process 8 bytes per iteration = 8 samples +" movl %2,%%edx\n" // edx = size +" shr $3,%%edx\n" // process 8 bytes per iteration = 8 samples -" cmp $0,%%ebx\n" +" cmp $0,%%edx\n" " je .endS8\n" ".align 16\n" " .mixloopS8:\n" " pxor %%mm2,%%mm2\n" // mm2 = 0 -" movq (%%esi),%%mm1\n" // mm1 = a|b|c|d|e|f|g|h +" movq (%1),%%mm1\n" // mm1 = a|b|c|d|e|f|g|h " movq %%mm1,%%mm3\n" // mm3 = a|b|c|d|e|f|g|h @@ -152,10 +148,10 @@ void SDL_MixAudio_MMX_S8(char* dst,char* src,unsigned int size,int volume) " punpckhbw %%mm2,%%mm1\n" // mm1 = 0|a|0|b|0|c|0|d " punpcklbw %%mm2,%%mm3\n" // mm3 = 0|e|0|f|0|g|0|h -" movq (%%edi),%%mm2\n" // mm2 = destination +" movq (%0),%%mm2\n" // mm2 = destination " pmullw %%mm0,%%mm1\n" // mm1 = v*a|v*b|v*c|v*d -" addl $8,%%esi\n" +" add $8,%1\n" " pmullw %%mm0,%%mm3\n" // mm3 = v*e|v*f|v*g|v*h " psraw $7,%%mm1\n" // mm1 = v*a/128|v*b/128|v*c/128|v*d/128 @@ -166,19 +162,19 @@ void SDL_MixAudio_MMX_S8(char* dst,char* src,unsigned int size,int volume) " paddsb %%mm2,%%mm3\n" // add to destination buffer -" movq %%mm3,(%%edi)\n" // store back to ram -" addl $8,%%edi\n" +" movq %%mm3,(%0)\n" // store back to ram +" add $8,%0\n" -" dec %%ebx\n" +" dec %%edx\n" " jnz .mixloopS8\n" ".endS8:\n" " emms\n" : - : "m" (dst), "m"(src),"m"(size), + : "r" (dst), "r"(src),"m"(size), "m"(volume) - : "eax","ebx", "esi", "edi","memory" + : "eax","edx","memory" ); } #endif diff --git a/src/cpuinfo/SDL_cpuinfo.c b/src/cpuinfo/SDL_cpuinfo.c index 7c63522e7..a8fa97203 100644 --- a/src/cpuinfo/SDL_cpuinfo.c +++ b/src/cpuinfo/SDL_cpuinfo.c @@ -138,7 +138,7 @@ static __inline__ int CPU_getCPUIDFeatures() " movl %%edi,%%ebx\n" : "=m" (features) : - : "%eax", "%ebx", "%ecx", "%edx", "%edi" + : "%eax", "%ecx", "%edx", "%edi" ); #elif defined(_MSC_VER) __asm { @@ -173,7 +173,7 @@ static __inline__ int CPU_getCPUIDFeaturesExt() " movl %%edi,%%ebx\n" : "=m" (features) : - : "%eax", "%ebx", "%ecx", "%edx", "%edi" + : "%eax", "%ecx", "%edx", "%edi" ); #elif defined(_MSC_VER) __asm { diff --git a/src/video/SDL_yuv_mmx.c b/src/video/SDL_yuv_mmx.c index 070711c1c..0cb8e8d58 100644 --- a/src/video/SDL_yuv_mmx.c +++ b/src/video/SDL_yuv_mmx.c @@ -120,12 +120,12 @@ void ColorRGBDitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, "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 %[_MMX_0080w],%%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 %[_MMX_VgrnRGB],%%mm0\n"// red*-46dec=0.7136*64 + "pmullw %[_MMX_VredRGB],%%mm1\n"// red*89dec=1.4013*64 "psraw $6, %%mm0\n" // red=red/64 "psraw $6, %%mm1\n" // red=red/64 @@ -134,8 +134,8 @@ void ColorRGBDitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, "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 %[_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 "psrlw $8,%%mm2\n" // 0 L3 0 L1 0 l3 0 l1 // create R (result in mm6) @@ -152,11 +152,11 @@ void ColorRGBDitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, "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 %[_MMX_0080w],%%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 %[_MMX_UgrnRGB],%%mm5\n" // blue*-109dec=1.7129*64 + "pmullw %[_MMX_UbluRGB],%%mm1\n" // blue*114dec=1.78125*64 "psraw $6, %%mm5\n" // blue=red/64 "psraw $6, %%mm1\n" // blue=blue/64 @@ -238,8 +238,14 @@ void ColorRGBDitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, "popl %%ebx\n" : : "m" (cr), "r"(cb),"r"(lum), - "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod) - : "%ebx" + "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod), + [_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) ); } @@ -413,8 +419,16 @@ void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, "popl %%ebx\n" : :"m" (cr), "r"(cb),"r"(lum), - "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod) - : "%ebx" + "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod), + [_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) ); }