Mercurial > sdl-ios-xcode

comparison src/audio/SDL_mixer_MMX.c @ 1668:4da1ee79c9af SDL-1.3

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
more tweaking indent options
author Sam Lantinga <slouken@libsdl.org>
date Mon, 29 May 2006 04:04:35 +0000
parents 782fd950bd46
children
comparison
equal deleted inserted replaced
1667:1fddae038bc8 1668:4da1ee79c9af
35 * Mixing for 16 bit signed buffers 35 * Mixing for 16 bit signed buffers
36 ***********************************************/ 36 ***********************************************/
37 37
38 #if defined(__GNUC__) && defined(__i386__) && defined(SDL_ASSEMBLY_ROUTINES) 38 #if defined(__GNUC__) && defined(__i386__) && defined(SDL_ASSEMBLY_ROUTINES)
39 void 39 void
40 SDL_MixAudio_MMX_S16 (char *dst, char *src, unsigned int size, int volume) 40 SDL_MixAudio_MMX_S16(char *dst, char *src, unsigned int size, int volume)
41 { 41 {
42 __asm__ __volatile__ (" movl %3,%%eax\n" /* eax = volume */ 42 __asm__ __volatile__(" movl %3,%%eax\n" /* eax = volume */
43 " movl %2,%%edx\n" /* edx = size */ 43 " movl %2,%%edx\n" /* edx = size */
44 " shrl $4,%%edx\n" /* process 16 bytes per iteration = 8 samples */ 44 " shrl $4,%%edx\n" /* process 16 bytes per iteration = 8 samples */
45 " 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 */ 45 " 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 */
46 ".align 8\n" " .mixloopS16:\n" " movq (%1),%%mm1\n" /* mm1 = a|b|c|d */ 46 ".align 8\n" " .mixloopS16:\n" " movq (%1),%%mm1\n" /* mm1 = a|b|c|d */
47 " movq %%mm1,%%mm2\n" /* mm2 = a|b|c|d */ 47 " movq %%mm1,%%mm2\n" /* mm2 = a|b|c|d */
48 " movq 8(%1),%%mm4\n" /* mm4 = e|f|g|h */ 48 " movq 8(%1),%%mm4\n" /* mm4 = e|f|g|h */
49 /* pré charger le buffer dst dans mm7 */ 49 /* pré charger le buffer dst dans mm7 */
50 " movq (%0),%%mm7\n" /* mm7 = dst[0] */ 50 " movq (%0),%%mm7\n" /* mm7 = dst[0] */
51 /* multiplier par le volume */ 51 /* multiplier par le volume */
52 " pmullw %%mm0,%%mm1\n" /* mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */ 52 " pmullw %%mm0,%%mm1\n" /* mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */
53 " pmulhw %%mm0,%%mm2\n" /* mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v) */ 53 " pmulhw %%mm0,%%mm2\n" /* mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v) */
54 " movq %%mm4,%%mm5\n" /* mm5 = e|f|g|h */ 54 " movq %%mm4,%%mm5\n" /* mm5 = e|f|g|h */
55 " pmullw %%mm0,%%mm4\n" /* mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */ 55 " pmullw %%mm0,%%mm4\n" /* mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */
56 " pmulhw %%mm0,%%mm5\n" /* mm5 = h(e*v)|h(f*v)|h(g*v)|h(h*v) */ 56 " pmulhw %%mm0,%%mm5\n" /* mm5 = h(e*v)|h(f*v)|h(g*v)|h(h*v) */
57 " movq %%mm1,%%mm3\n" /* mm3 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */ 57 " movq %%mm1,%%mm3\n" /* mm3 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */
58 " punpckhwd %%mm2,%%mm1\n" /* mm1 = a*v|b*v */ 58 " punpckhwd %%mm2,%%mm1\n" /* mm1 = a*v|b*v */
59 " movq %%mm4,%%mm6\n" /* mm6 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */ 59 " movq %%mm4,%%mm6\n" /* mm6 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */
60 " punpcklwd %%mm2,%%mm3\n" /* mm3 = c*v|d*v */ 60 " punpcklwd %%mm2,%%mm3\n" /* mm3 = c*v|d*v */
61 " punpckhwd %%mm5,%%mm4\n" /* mm4 = e*f|f*v */ 61 " punpckhwd %%mm5,%%mm4\n" /* mm4 = e*f|f*v */
62 " punpcklwd %%mm5,%%mm6\n" /* mm6 = g*v|h*v */ 62 " punpcklwd %%mm5,%%mm6\n" /* mm6 = g*v|h*v */
63 /* pré charger le buffer dst dans mm5 */ 63 /* pré charger le buffer dst dans mm5 */
64 " movq 8(%0),%%mm5\n" /* mm5 = dst[1] */ 64 " movq 8(%0),%%mm5\n" /* mm5 = dst[1] */
65 /* diviser par 128 */ 65 /* diviser par 128 */
66 " psrad $7,%%mm1\n" /* mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME */ 66 " psrad $7,%%mm1\n" /* mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME */
67 " add $16,%1\n" " psrad $7,%%mm3\n" /* mm3 = c*v/128|d*v/128 */ 67 " add $16,%1\n" " psrad $7,%%mm3\n" /* mm3 = c*v/128|d*v/128 */
68 " psrad $7,%%mm4\n" /* mm4 = e*v/128|f*v/128 */ 68 " psrad $7,%%mm4\n" /* mm4 = e*v/128|f*v/128 */
69 /* mm1 = le sample avec le volume modifié */ 69 /* mm1 = le sample avec le volume modifié */
70 " packssdw %%mm1,%%mm3\n" /* mm3 = s(a*v|b*v|c*v|d*v) */ 70 " packssdw %%mm1,%%mm3\n" /* mm3 = s(a*v|b*v|c*v|d*v) */
71 " psrad $7,%%mm6\n" /* mm6= g*v/128|h*v/128 */ 71 " psrad $7,%%mm6\n" /* mm6= g*v/128|h*v/128 */
72 " paddsw %%mm7,%%mm3\n" /* mm3 = adjust_volume(src)+dst */ 72 " paddsw %%mm7,%%mm3\n" /* mm3 = adjust_volume(src)+dst */
73 /* mm4 = le sample avec le volume modifié */ 73 /* mm4 = le sample avec le volume modifié */
74 " packssdw %%mm4,%%mm6\n" /* mm6 = s(e*v|f*v|g*v|h*v) */ 74 " packssdw %%mm4,%%mm6\n" /* mm6 = s(e*v|f*v|g*v|h*v) */
75 " movq %%mm3,(%0)\n" " paddsw %%mm5,%%mm6\n" /* mm6 = adjust_volume(src)+dst */ 75 " movq %%mm3,(%0)\n" " paddsw %%mm5,%%mm6\n" /* mm6 = adjust_volume(src)+dst */
76 " movq %%mm6,8(%0)\n" 76 " movq %%mm6,8(%0)\n"
77 " add $16,%0\n" 77 " add $16,%0\n"
78 " dec %%edx\n" 78 " dec %%edx\n"
79 " jnz .mixloopS16\n" 79 " jnz .mixloopS16\n"
80 " emms\n" 80 " emms\n"
81 ".endS16:\n"::"r" (dst), "r" (src), 81 ".endS16:\n"::"r"(dst), "r"(src),
82 "m" (size), "m" (volume):"eax", "edx", "memory"); 82 "m"(size), "m"(volume):"eax", "edx", "memory");
83 } 83 }
84 84
85 85
86 86
87 /*////////////////////////////////////////////// */ 87 /*////////////////////////////////////////////// */
88 /* Mixing for 8 bit signed buffers */ 88 /* Mixing for 8 bit signed buffers */
89 /*////////////////////////////////////////////// */ 89 /*////////////////////////////////////////////// */
90 90
91 void 91 void
92 SDL_MixAudio_MMX_S8 (char *dst, char *src, unsigned int size, int volume) 92 SDL_MixAudio_MMX_S8(char *dst, char *src, unsigned int size, int volume)
93 { 93 {
94 __asm__ __volatile__ (" movl %3,%%eax\n" /* eax = volume */ 94 __asm__ __volatile__(" movl %3,%%eax\n" /* eax = volume */
95 " 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 */ 95 " 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 */
96 " shr $3,%%edx\n" /* process 8 bytes per iteration = 8 samples */ 96 " shr $3,%%edx\n" /* process 8 bytes per iteration = 8 samples */
97 " cmp $0,%%edx\n" " je .endS8\n" ".align 8\n" " .mixloopS8:\n" " pxor %%mm2,%%mm2\n" /* mm2 = 0 */ 97 " cmp $0,%%edx\n" " je .endS8\n" ".align 8\n" " .mixloopS8:\n" " pxor %%mm2,%%mm2\n" /* mm2 = 0 */
98 " movq (%1),%%mm1\n" /* mm1 = a|b|c|d|e|f|g|h */ 98 " movq (%1),%%mm1\n" /* mm1 = a|b|c|d|e|f|g|h */
99 " movq %%mm1,%%mm3\n" /* mm3 = a|b|c|d|e|f|g|h */ 99 " movq %%mm1,%%mm3\n" /* mm3 = a|b|c|d|e|f|g|h */
100 /* on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0 */ 100 /* on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0 */
101 " pcmpgtb %%mm1,%%mm2\n" /* mm2 = 11111111|00000000|00000000.... */ 101 " pcmpgtb %%mm1,%%mm2\n" /* mm2 = 11111111|00000000|00000000.... */
102 " punpckhbw %%mm2,%%mm1\n" /* mm1 = 0|a|0|b|0|c|0|d */ 102 " punpckhbw %%mm2,%%mm1\n" /* mm1 = 0|a|0|b|0|c|0|d */
103 " punpcklbw %%mm2,%%mm3\n" /* mm3 = 0|e|0|f|0|g|0|h */ 103 " punpcklbw %%mm2,%%mm3\n" /* mm3 = 0|e|0|f|0|g|0|h */
104 " movq (%0),%%mm2\n" /* mm2 = destination */ 104 " movq (%0),%%mm2\n" /* mm2 = destination */
105 " pmullw %%mm0,%%mm1\n" /* mm1 = v*a|v*b|v*c|v*d */ 105 " pmullw %%mm0,%%mm1\n" /* mm1 = v*a|v*b|v*c|v*d */
106 " add $8,%1\n" " pmullw %%mm0,%%mm3\n" /* mm3 = v*e|v*f|v*g|v*h */ 106 " add $8,%1\n" " pmullw %%mm0,%%mm3\n" /* mm3 = v*e|v*f|v*g|v*h */
107 " psraw $7,%%mm1\n" /* mm1 = v*a/128|v*b/128|v*c/128|v*d/128 */ 107 " psraw $7,%%mm1\n" /* mm1 = v*a/128|v*b/128|v*c/128|v*d/128 */
108 " psraw $7,%%mm3\n" /* mm3 = v*e/128|v*f/128|v*g/128|v*h/128 */ 108 " psraw $7,%%mm3\n" /* mm3 = v*e/128|v*f/128|v*g/128|v*h/128 */
109 " 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 */ 109 " 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 */
110 " paddsb %%mm2,%%mm3\n" /* add to destination buffer */ 110 " paddsb %%mm2,%%mm3\n" /* add to destination buffer */
111 " movq %%mm3,(%0)\n" /* store back to ram */ 111 " movq %%mm3,(%0)\n" /* store back to ram */
112 " add $8,%0\n" 112 " add $8,%0\n"
113 " dec %%edx\n" 113 " dec %%edx\n"
114 " jnz .mixloopS8\n" 114 " jnz .mixloopS8\n"
115 ".endS8:\n" 115 ".endS8:\n"
116 " emms\n"::"r" (dst), "r" (src), "m" (size), 116 " emms\n"::"r"(dst), "r"(src), "m"(size),
117 "m" (volume):"eax", "edx", "memory"); 117 "m"(volume):"eax", "edx", "memory");
118 } 118 }
119 #endif 119 #endif
120 /* vi: set ts=4 sw=4 expandtab: */ 120 /* vi: set ts=4 sw=4 expandtab: */