Mercurial > sdl-ios-xcode
diff src/audio/SDL_mixer_MMX.c @ 1895:c121d94672cb
SDL 1.2 is moving to a branch, and SDL 1.3 is becoming the head.
| author | Sam Lantinga <slouken@libsdl.org> |
|---|---|
| date | Mon, 10 Jul 2006 21:04:37 +0000 |
| parents | 97d0966f4bf7 |
| children | 99210400e8b9 |
line wrap: on
line diff
--- a/src/audio/SDL_mixer_MMX.c Thu Jul 06 18:01:37 2006 +0000 +++ b/src/audio/SDL_mixer_MMX.c Mon Jul 10 21:04:37 2006 +0000 @@ -36,100 +36,50 @@ ***********************************************/ #if defined(__GNUC__) && defined(__i386__) && defined(SDL_ASSEMBLY_ROUTINES) -void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume) +void +SDL_MixAudio_MMX_S16(char *dst, char *src, unsigned int size, int volume) { - __asm__ __volatile__ ( - -" movl %3,%%eax\n" /* eax = volume */ - -" movl %2,%%edx\n" /* edx = size */ - -" shrl $4,%%edx\n" /* process 16 bytes per iteration = 8 samples */ - -" jz .endS16\n" - -" pxor %%mm0,%%mm0\n" - -" movd %%eax,%%mm0\n" -" movq %%mm0,%%mm1\n" -" psllq $16,%%mm0\n" -" por %%mm1,%%mm0\n" -" psllq $16,%%mm0\n" -" por %%mm1,%%mm0\n" -" psllq $16,%%mm0\n" -" por %%mm1,%%mm0\n" /* mm0 = vol|vol|vol|vol */ - -".align 8\n" -" .mixloopS16:\n" - -" movq (%1),%%mm1\n" /* mm1 = a|b|c|d */ - -" movq %%mm1,%%mm2\n" /* mm2 = a|b|c|d */ - -" movq 8(%1),%%mm4\n" /* mm4 = e|f|g|h */ - - /* pré charger le buffer dst dans mm7 */ -" 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) */ - -" pmulhw %%mm0,%%mm2\n" /* mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v) */ -" movq %%mm4,%%mm5\n" /* mm5 = e|f|g|h */ - -" pmullw %%mm0,%%mm4\n" /* mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */ - -" pmulhw %%mm0,%%mm5\n" /* mm5 = h(e*v)|h(f*v)|h(g*v)|h(h*v) */ -" movq %%mm1,%%mm3\n" /* mm3 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */ - -" punpckhwd %%mm2,%%mm1\n" /* mm1 = a*v|b*v */ - -" movq %%mm4,%%mm6\n" /* mm6 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */ -" punpcklwd %%mm2,%%mm3\n" /* mm3 = c*v|d*v */ - -" punpckhwd %%mm5,%%mm4\n" /* mm4 = e*f|f*v */ - -" punpcklwd %%mm5,%%mm6\n" /* mm6 = g*v|h*v */ - - /* pré charger le buffer dst dans mm5 */ -" 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 */ -" add $16,%1\n" - -" psrad $7,%%mm3\n" /* mm3 = c*v/128|d*v/128 */ - -" psrad $7,%%mm4\n" /* mm4 = e*v/128|f*v/128 */ - - /* mm1 = le sample avec le volume modifié */ -" packssdw %%mm1,%%mm3\n" /* mm3 = s(a*v|b*v|c*v|d*v) */ - -" psrad $7,%%mm6\n" /* mm6= g*v/128|h*v/128 */ -" paddsw %%mm7,%%mm3\n" /* mm3 = adjust_volume(src)+dst */ - - /* mm4 = le sample avec le volume modifié */ -" packssdw %%mm4,%%mm6\n" /* mm6 = s(e*v|f*v|g*v|h*v) */ -" movq %%mm3,(%0)\n" - -" paddsw %%mm5,%%mm6\n" /* mm6 = adjust_volume(src)+dst */ - -" movq %%mm6,8(%0)\n" - -" add $16,%0\n" - -" dec %%edx\n" - -" jnz .mixloopS16\n" - -" emms\n" - -".endS16:\n" - : - : "r" (dst), "r"(src),"m"(size), - "m"(volume) - : "eax","edx","memory" - ); + __asm__ __volatile__(" movl %3,%%eax\n" /* eax = volume */ + " movl %2,%%edx\n" /* edx = size */ + " shrl $4,%%edx\n" /* process 16 bytes per iteration = 8 samples */ + " jz .endS16\n" " pxor %%mm0,%%mm0\n" " movd %%eax,%%mm0\n" " movq %%mm0,%%mm1\n" " psllq $16,%%mm0\n" " por %%mm1,%%mm0\n" " psllq $16,%%mm0\n" " por %%mm1,%%mm0\n" " psllq $16,%%mm0\n" " por %%mm1,%%mm0\n" /* mm0 = vol|vol|vol|vol */ + ".align 8\n" " .mixloopS16:\n" " movq (%1),%%mm1\n" /* mm1 = a|b|c|d */ + " movq %%mm1,%%mm2\n" /* mm2 = a|b|c|d */ + " movq 8(%1),%%mm4\n" /* mm4 = e|f|g|h */ + /* pré charger le buffer dst dans mm7 */ + " 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) */ + " pmulhw %%mm0,%%mm2\n" /* mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v) */ + " movq %%mm4,%%mm5\n" /* mm5 = e|f|g|h */ + " pmullw %%mm0,%%mm4\n" /* mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */ + " pmulhw %%mm0,%%mm5\n" /* mm5 = h(e*v)|h(f*v)|h(g*v)|h(h*v) */ + " movq %%mm1,%%mm3\n" /* mm3 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */ + " punpckhwd %%mm2,%%mm1\n" /* mm1 = a*v|b*v */ + " movq %%mm4,%%mm6\n" /* mm6 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */ + " punpcklwd %%mm2,%%mm3\n" /* mm3 = c*v|d*v */ + " punpckhwd %%mm5,%%mm4\n" /* mm4 = e*f|f*v */ + " punpcklwd %%mm5,%%mm6\n" /* mm6 = g*v|h*v */ + /* pré charger le buffer dst dans mm5 */ + " 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 */ + " add $16,%1\n" " psrad $7,%%mm3\n" /* mm3 = c*v/128|d*v/128 */ + " psrad $7,%%mm4\n" /* mm4 = e*v/128|f*v/128 */ + /* mm1 = le sample avec le volume modifié */ + " packssdw %%mm1,%%mm3\n" /* mm3 = s(a*v|b*v|c*v|d*v) */ + " psrad $7,%%mm6\n" /* mm6= g*v/128|h*v/128 */ + " paddsw %%mm7,%%mm3\n" /* mm3 = adjust_volume(src)+dst */ + /* mm4 = le sample avec le volume modifié */ + " packssdw %%mm4,%%mm6\n" /* mm6 = s(e*v|f*v|g*v|h*v) */ + " movq %%mm3,(%0)\n" " paddsw %%mm5,%%mm6\n" /* mm6 = adjust_volume(src)+dst */ + " movq %%mm6,8(%0)\n" + " add $16,%0\n" + " dec %%edx\n" + " jnz .mixloopS16\n" + " emms\n" + ".endS16:\n"::"r"(dst), "r"(src), + "m"(size), "m"(volume):"eax", "edx", "memory"); } @@ -138,69 +88,33 @@ /* Mixing for 8 bit signed buffers */ /*////////////////////////////////////////////// */ -void SDL_MixAudio_MMX_S8(char* dst,char* src,unsigned int size,int volume) +void +SDL_MixAudio_MMX_S8(char *dst, char *src, unsigned int size, int volume) { - __asm__ __volatile__ ( - -" movl %3,%%eax\n" /* eax = volume */ - -" movd %%eax,%%mm0\n" -" movq %%mm0,%%mm1\n" -" psllq $16,%%mm0\n" -" por %%mm1,%%mm0\n" -" psllq $16,%%mm0\n" -" por %%mm1,%%mm0\n" -" psllq $16,%%mm0\n" -" por %%mm1,%%mm0\n" - -" movl %2,%%edx\n" /* edx = size */ -" shr $3,%%edx\n" /* process 8 bytes per iteration = 8 samples */ - -" cmp $0,%%edx\n" -" je .endS8\n" - -".align 8\n" -" .mixloopS8:\n" - -" pxor %%mm2,%%mm2\n" /* mm2 = 0 */ -" 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 */ - - /* on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0 */ -" pcmpgtb %%mm1,%%mm2\n" /* mm2 = 11111111|00000000|00000000.... */ - -" 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 (%0),%%mm2\n" /* mm2 = destination */ - -" pmullw %%mm0,%%mm1\n" /* mm1 = v*a|v*b|v*c|v*d */ -" 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 */ - -" psraw $7,%%mm3\n" /* mm3 = v*e/128|v*f/128|v*g/128|v*h/128 */ - -" packsswb %%mm1,%%mm3\n" /* mm1 = v*a/128|v*b/128|v*c/128|v*d/128|v*e/128|v*f/128|v*g/128|v*h/128 */ - -" paddsb %%mm2,%%mm3\n" /* add to destination buffer */ - -" movq %%mm3,(%0)\n" /* store back to ram */ -" add $8,%0\n" - -" dec %%edx\n" - -" jnz .mixloopS8\n" - -".endS8:\n" -" emms\n" - : - : "r" (dst), "r"(src),"m"(size), - "m"(volume) - : "eax","edx","memory" - ); + __asm__ __volatile__(" movl %3,%%eax\n" /* eax = volume */ + " movd %%eax,%%mm0\n" " movq %%mm0,%%mm1\n" " psllq $16,%%mm0\n" " por %%mm1,%%mm0\n" " psllq $16,%%mm0\n" " por %%mm1,%%mm0\n" " psllq $16,%%mm0\n" " por %%mm1,%%mm0\n" " movl %2,%%edx\n" /* edx = size */ + " shr $3,%%edx\n" /* process 8 bytes per iteration = 8 samples */ + " cmp $0,%%edx\n" " je .endS8\n" ".align 8\n" " .mixloopS8:\n" " pxor %%mm2,%%mm2\n" /* mm2 = 0 */ + " 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 */ + /* on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0 */ + " pcmpgtb %%mm1,%%mm2\n" /* mm2 = 11111111|00000000|00000000.... */ + " 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 (%0),%%mm2\n" /* mm2 = destination */ + " pmullw %%mm0,%%mm1\n" /* mm1 = v*a|v*b|v*c|v*d */ + " 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 */ + " psraw $7,%%mm3\n" /* mm3 = v*e/128|v*f/128|v*g/128|v*h/128 */ + " packsswb %%mm1,%%mm3\n" /* mm1 = v*a/128|v*b/128|v*c/128|v*d/128|v*e/128|v*f/128|v*g/128|v*h/128 */ + " paddsb %%mm2,%%mm3\n" /* add to destination buffer */ + " movq %%mm3,(%0)\n" /* store back to ram */ + " add $8,%0\n" + " dec %%edx\n" + " jnz .mixloopS8\n" + ".endS8:\n" + " emms\n"::"r"(dst), "r"(src), "m"(size), + "m"(volume):"eax", "edx", "memory"); } #endif - +/* vi: set ts=4 sw=4 expandtab: */