Mercurial > sdl-ios-xcode

comparison src/audio/SDL_mixer_MMX.c @ 887:b4b64bb88f2f

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
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
author Sam Lantinga <slouken@libsdl.org>
date Sun, 16 May 2004 17:19:48 +0000
parents a9e38f3b8e4d
children e3b3130f3af8
comparison
equal deleted inserted replaced
886:05c551e5bc64 887:b4b64bb88f2f
13 #if defined(i386) && defined(__GNUC__) && defined(USE_ASMBLIT) 13 #if defined(i386) && defined(__GNUC__) && defined(USE_ASMBLIT)
14 void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume) 14 void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume)
15 { 15 {
16 __asm__ __volatile__ ( 16 __asm__ __volatile__ (
17 17
18 " movl %0,%%edi\n" // edi = dst
19 " movl %1,%%esi\n" // esi = src
20 " movl %3,%%eax\n" // eax = volume 18 " movl %3,%%eax\n" // eax = volume
21 19
22 " movl %2,%%ebx\n" // ebx = size 20 " movl %2,%%edx\n" // edx = size
23 21
24 " shrl $4,%%ebx\n" // process 16 bytes per iteration = 8 samples 22 " shrl $4,%%edx\n" // process 16 bytes per iteration = 8 samples
25 23
26 " jz .endS16\n" 24 " jz .endS16\n"
27 25
28 " pxor %%mm0,%%mm0\n" 26 " pxor %%mm0,%%mm0\n"
29 27
37 " por %%mm1,%%mm0\n" // mm0 = vol|vol|vol|vol 35 " por %%mm1,%%mm0\n" // mm0 = vol|vol|vol|vol
38 36
39 ".align 16\n" 37 ".align 16\n"
40 " .mixloopS16:\n" 38 " .mixloopS16:\n"
41 39
42 " movq (%%esi),%%mm1\n" // mm1 = a|b|c|d 40 " movq (%1),%%mm1\n" // mm1 = a|b|c|d
43 41
44 " movq %%mm1,%%mm2\n" // mm2 = a|b|c|d 42 " movq %%mm1,%%mm2\n" // mm2 = a|b|c|d
45 43
46 " movq 8(%%esi),%%mm4\n" // mm4 = e|f|g|h 44 " movq 8(%1),%%mm4\n" // mm4 = e|f|g|h
47 45
48 // pré charger le buffer dst dans mm7 46 // pré charger le buffer dst dans mm7
49 " movq (%%edi),%%mm7\n" // mm7 = dst[0]" 47 " movq (%0),%%mm7\n" // mm7 = dst[0]"
50 48
51 // multiplier par le volume 49 // multiplier par le volume
52 " pmullw %%mm0,%%mm1\n" // mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v) 50 " pmullw %%mm0,%%mm1\n" // mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v)
53 51
54 " pmulhw %%mm0,%%mm2\n" // mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v) 52 " pmulhw %%mm0,%%mm2\n" // mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v)
67 " punpckhwd %%mm5,%%mm4\n" // mm4 = e*f|f*v 65 " punpckhwd %%mm5,%%mm4\n" // mm4 = e*f|f*v
68 66
69 " punpcklwd %%mm5,%%mm6\n" // mm6 = g*v|h*v 67 " punpcklwd %%mm5,%%mm6\n" // mm6 = g*v|h*v
70 68
71 // pré charger le buffer dst dans mm5 69 // pré charger le buffer dst dans mm5
72 " movq 8(%%edi),%%mm5\n" // mm5 = dst[1] 70 " movq 8(%0),%%mm5\n" // mm5 = dst[1]
73 71
74 // diviser par 128 72 // diviser par 128
75 " psrad $7,%%mm1\n" // mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME 73 " psrad $7,%%mm1\n" // mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME
76 " addl $16,%%esi\n" 74 " add $16,%1\n"
77 75
78 " psrad $7,%%mm3\n" // mm3 = c*v/128|d*v/128 76 " psrad $7,%%mm3\n" // mm3 = c*v/128|d*v/128
79 77
80 " psrad $7,%%mm4\n" // mm4 = e*v/128|f*v/128 78 " psrad $7,%%mm4\n" // mm4 = e*v/128|f*v/128
81 79
85 " psrad $7,%%mm6\n" // mm6= g*v/128|h*v/128 83 " psrad $7,%%mm6\n" // mm6= g*v/128|h*v/128
86 " paddsw %%mm7,%%mm3\n" // mm3 = adjust_volume(src)+dst 84 " paddsw %%mm7,%%mm3\n" // mm3 = adjust_volume(src)+dst
87 85
88 // mm4 = le sample avec le volume modifié 86 // mm4 = le sample avec le volume modifié
89 " packssdw %%mm4,%%mm6\n" // mm6 = s(e*v|f*v|g*v|h*v) 87 " packssdw %%mm4,%%mm6\n" // mm6 = s(e*v|f*v|g*v|h*v)
90 " movq %%mm3,(%%edi)\n" 88 " movq %%mm3,(%0)\n"
91 89
92 " paddsw %%mm5,%%mm6\n" // mm6 = adjust_volume(src)+dst 90 " paddsw %%mm5,%%mm6\n" // mm6 = adjust_volume(src)+dst
93 91
94 " movq %%mm6,8(%%edi)\n" 92 " movq %%mm6,8(%0)\n"
95 93
96 " addl $16,%%edi\n" 94 " add $16,%0\n"
97 95
98 " dec %%ebx\n" 96 " dec %%edx\n"
99 97
100 " jnz .mixloopS16\n" 98 " jnz .mixloopS16\n"
101 99
102 " emms\n" 100 " emms\n"
103 101
104 ".endS16:\n" 102 ".endS16:\n"
105 : 103 :
106 : "m" (dst), "m"(src),"m"(size), 104 : "r" (dst), "r"(src),"m"(size),
107 "m"(volume) 105 "m"(volume)
108 : "eax","ebx", "esi", "edi","memory" 106 : "eax","edx","memory"
109 ); 107 );
110 } 108 }
111 109
112 110
113 111
117 115
118 void SDL_MixAudio_MMX_S8(char* dst,char* src,unsigned int size,int volume) 116 void SDL_MixAudio_MMX_S8(char* dst,char* src,unsigned int size,int volume)
119 { 117 {
120 __asm__ __volatile__ ( 118 __asm__ __volatile__ (
121 119
122 " movl %0,%%edi\n" // edi = dst
123 " movl %1,%%esi\n" // esi = src
124 " movl %3,%%eax\n" // eax = volume 120 " movl %3,%%eax\n" // eax = volume
125 121
126 " movd %%ebx,%%mm0\n" 122 " movd %%edx,%%mm0\n"
127 " movq %%mm0,%%mm1\n" 123 " movq %%mm0,%%mm1\n"
128 " psllq $16,%%mm0\n" 124 " psllq $16,%%mm0\n"
129 " por %%mm1,%%mm0\n" 125 " por %%mm1,%%mm0\n"
130 " psllq $16,%%mm0\n" 126 " psllq $16,%%mm0\n"
131 " por %%mm1,%%mm0\n" 127 " por %%mm1,%%mm0\n"
132 " psllq $16,%%mm0\n" 128 " psllq $16,%%mm0\n"
133 " por %%mm1,%%mm0\n" 129 " por %%mm1,%%mm0\n"
134 130
135 " movl %2,%%ebx\n" // ebx = size 131 " movl %2,%%edx\n" // edx = size
136 " shr $3,%%ebx\n" // process 8 bytes per iteration = 8 samples 132 " shr $3,%%edx\n" // process 8 bytes per iteration = 8 samples
137 133
138 " cmp $0,%%ebx\n" 134 " cmp $0,%%edx\n"
139 " je .endS8\n" 135 " je .endS8\n"
140 136
141 ".align 16\n" 137 ".align 16\n"
142 " .mixloopS8:\n" 138 " .mixloopS8:\n"
143 139
144 " pxor %%mm2,%%mm2\n" // mm2 = 0 140 " pxor %%mm2,%%mm2\n" // mm2 = 0
145 " movq (%%esi),%%mm1\n" // mm1 = a|b|c|d|e|f|g|h 141 " movq (%1),%%mm1\n" // mm1 = a|b|c|d|e|f|g|h
146 142
147 " movq %%mm1,%%mm3\n" // mm3 = a|b|c|d|e|f|g|h 143 " movq %%mm1,%%mm3\n" // mm3 = a|b|c|d|e|f|g|h
148 144
149 // on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0 145 // on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0
150 " pcmpgtb %%mm1,%%mm2\n" // mm2 = 11111111|00000000|00000000.... 146 " pcmpgtb %%mm1,%%mm2\n" // mm2 = 11111111|00000000|00000000....
151 147
152 " punpckhbw %%mm2,%%mm1\n" // mm1 = 0|a|0|b|0|c|0|d 148 " punpckhbw %%mm2,%%mm1\n" // mm1 = 0|a|0|b|0|c|0|d
153 149
154 " punpcklbw %%mm2,%%mm3\n" // mm3 = 0|e|0|f|0|g|0|h 150 " punpcklbw %%mm2,%%mm3\n" // mm3 = 0|e|0|f|0|g|0|h
155 " movq (%%edi),%%mm2\n" // mm2 = destination 151 " movq (%0),%%mm2\n" // mm2 = destination
156 152
157 " pmullw %%mm0,%%mm1\n" // mm1 = v*a|v*b|v*c|v*d 153 " pmullw %%mm0,%%mm1\n" // mm1 = v*a|v*b|v*c|v*d
158 " addl $8,%%esi\n" 154 " add $8,%1\n"
159 155
160 " pmullw %%mm0,%%mm3\n" // mm3 = v*e|v*f|v*g|v*h 156 " pmullw %%mm0,%%mm3\n" // mm3 = v*e|v*f|v*g|v*h
161 " psraw $7,%%mm1\n" // mm1 = v*a/128|v*b/128|v*c/128|v*d/128 157 " psraw $7,%%mm1\n" // mm1 = v*a/128|v*b/128|v*c/128|v*d/128
162 158
163 " psraw $7,%%mm3\n" // mm3 = v*e/128|v*f/128|v*g/128|v*h/128 159 " psraw $7,%%mm3\n" // mm3 = v*e/128|v*f/128|v*g/128|v*h/128
164 160
165 " 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/128 161 " 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
166 162
167 " paddsb %%mm2,%%mm3\n" // add to destination buffer 163 " paddsb %%mm2,%%mm3\n" // add to destination buffer
168 164
169 " movq %%mm3,(%%edi)\n" // store back to ram 165 " movq %%mm3,(%0)\n" // store back to ram
170 " addl $8,%%edi\n" 166 " add $8,%0\n"
171 167
172 " dec %%ebx\n" 168 " dec %%edx\n"
173 169
174 " jnz .mixloopS8\n" 170 " jnz .mixloopS8\n"
175 171
176 ".endS8:\n" 172 ".endS8:\n"
177 " emms\n" 173 " emms\n"
178 : 174 :
179 : "m" (dst), "m"(src),"m"(size), 175 : "r" (dst), "r"(src),"m"(size),
180 "m"(volume) 176 "m"(volume)
181 : "eax","ebx", "esi", "edi","memory" 177 : "eax","edx","memory"
182 ); 178 );
183 } 179 }
184 #endif 180 #endif
185 181