Mercurial > sdl-ios-xcode
comparison src/audio/SDL_mixer_MMX_VC.c @ 574:64fe373be3dc
Cth converted the MMX audio mixing routines to VC++ syntax
author | Sam Lantinga <slouken@libsdl.org> |
---|---|
date | Mon, 20 Jan 2003 16:01:20 +0000 |
parents | |
children | 72ef7ce609ef |
comparison
equal
deleted
inserted
replaced
573:6c3fa3b5e397 | 574:64fe373be3dc |
---|---|
1 // MMX assembler version of SDL_MixAudio for signed little endian 16 bit samples and signed 8 bit samples | |
2 // Copyright 2002 Stephane Marchesin (stephane.marchesin@wanadoo.fr) | |
3 // Converted to Intel ASM notation by Cth | |
4 // This code is licensed under the LGPL (see COPYING for details) | |
5 // | |
6 // Assumes buffer size in bytes is a multiple of 16 | |
7 // Assumes SDL_MIX_MAXVOLUME = 128 | |
8 | |
9 | |
10 //////////////////////////////////////////////// | |
11 // Mixing for 16 bit signed buffers | |
12 //////////////////////////////////////////////// | |
13 | |
14 #if defined(USE_ASM_MIXER_VC) | |
15 #include <windows.h> | |
16 #include <stdio.h> | |
17 | |
18 void SDL_MixAudio_MMX_S16_VC(char* dst,char* src,unsigned int nSize,int volume) | |
19 { | |
20 __asm | |
21 { | |
22 align 16 | |
23 | |
24 push edi | |
25 push esi | |
26 push ebx | |
27 | |
28 mov edi, dst // edi = dst | |
29 mov esi, src // esi = src | |
30 mov eax, volume // eax = volume | |
31 mov ebx, nSize // ebx = size | |
32 shr ebx, 4 // process 16 bytes per iteration = 8 samples | |
33 jz endS16 | |
34 | |
35 pxor mm0, mm0 | |
36 movd mm0, eax //%%eax,%%mm0 | |
37 movq mm1, mm0 //%%mm0,%%mm1 | |
38 psllq mm0, 16 //$16,%%mm0 | |
39 por mm0, mm1 //%%mm1,%%mm0 | |
40 psllq mm0, 16 //$16,%%mm0 | |
41 por mm0, mm1 //%%mm1,%%mm0 | |
42 psllq mm0, 16 //$16,%%mm0 | |
43 por mm0, mm1 //%%mm1,%%mm0 // mm0 = vol|vol|vol|vol | |
44 | |
45 mixloopS16: | |
46 movq mm1, [esi] //(%%esi),%%mm1\n" // mm1 = a|b|c|d | |
47 movq mm2, mm1 //%%mm1,%%mm2\n" // mm2 = a|b|c|d | |
48 movq mm4, [esi + 8] //8(%%esi),%%mm4\n" // mm4 = e|f|g|h | |
49 // pre charger le buffer dst dans mm7 | |
50 movq mm7, [edi] //(%%edi),%%mm7\n" // mm7 = dst[0]" | |
51 // multiplier par le volume | |
52 pmullw mm1, mm0 //%%mm0,%%mm1\n" // mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v) | |
53 pmulhw mm2, mm0 //%%mm0,%%mm2\n" // mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v) | |
54 movq mm5, mm4 //%%mm4,%%mm5\n" // mm5 = e|f|g|h | |
55 pmullw mm4, mm0 //%%mm0,%%mm4\n" // mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v) | |
56 pmulhw mm5, mm0 //%%mm0,%%mm5\n" // mm5 = h(e*v)|h(f*v)|h(g*v)|h(h*v) | |
57 movq mm3, mm1 //%%mm1,%%mm3\n" // mm3 = l(a*v)|l(b*v)|l(c*v)|l(d*v) | |
58 punpckhwd mm1, mm2 //%%mm2,%%mm1\n" // mm1 = a*v|b*v | |
59 movq mm6, mm4 //%%mm4,%%mm6\n" // mm6 = l(e*v)|l(f*v)|l(g*v)|l(h*v) | |
60 punpcklwd mm3, mm2 //%%mm2,%%mm3\n" // mm3 = c*v|d*v | |
61 punpckhwd mm4, mm5 //%%mm5,%%mm4\n" // mm4 = e*f|f*v | |
62 punpcklwd mm6, mm5 //%%mm5,%%mm6\n" // mm6 = g*v|h*v | |
63 // pre charger le buffer dst dans mm5 | |
64 movq mm5, [edi + 8] //8(%%edi),%%mm5\n" // mm5 = dst[1] | |
65 // diviser par 128 | |
66 psrad mm1, 7 //$7,%%mm1\n" // mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME | |
67 add esi, 16 //$16,%%esi\n" | |
68 psrad mm3, 7 //$7,%%mm3\n" // mm3 = c*v/128|d*v/128 | |
69 psrad mm4, 7 //$7,%%mm4\n" // mm4 = e*v/128|f*v/128 | |
70 // mm1 = le sample avec le volume modifie | |
71 packssdw mm3, mm1 //%%mm1,%%mm3\n" // mm3 = s(a*v|b*v|c*v|d*v) | |
72 psrad mm6, 7 //$7,%%mm6\n" // mm6= g*v/128|h*v/128 | |
73 paddsw mm3, mm7 //%%mm7,%%mm3\n" // mm3 = adjust_volume(src)+dst | |
74 // mm4 = le sample avec le volume modifie | |
75 packssdw mm6, mm4 //%%mm4,%%mm6\n" // mm6 = s(e*v|f*v|g*v|h*v) | |
76 movq [edi], mm3 //%%mm3,(%%edi)\n" | |
77 paddsw mm6, mm5 //%%mm5,%%mm6\n" // mm6 = adjust_volume(src)+dst | |
78 movq [edi + 8], mm6 //%%mm6,8(%%edi)\n" | |
79 add edi, 16 //$16,%%edi\n" | |
80 dec ebx //%%ebx\n" | |
81 jnz mixloopS16 | |
82 | |
83 ends16: | |
84 emms | |
85 | |
86 pop ebx | |
87 pop esi | |
88 pop edi | |
89 } | |
90 | |
91 } | |
92 | |
93 //////////////////////////////////////////////// | |
94 // Mixing for 8 bit signed buffers | |
95 //////////////////////////////////////////////// | |
96 | |
97 void SDL_MixAudio_MMX_S8_VC(char* dst,char* src,unsigned int nSize,int volume) | |
98 { | |
99 _asm | |
100 { | |
101 align 16 | |
102 | |
103 push edi | |
104 push esi | |
105 push ebx | |
106 | |
107 mov edi, dst //movl %0,%%edi // edi = dst | |
108 mov esi, src //%1,%%esi // esi = src | |
109 mov eax, volume //%3,%%eax // eax = volume | |
110 | |
111 movd mm0, ebx //%%ebx,%%mm0 | |
112 movq mm1, mm0 //%%mm0,%%mm1 | |
113 psllq mm0, 16 //$16,%%mm0 | |
114 por mm0, mm1 //%%mm1,%%mm0 | |
115 psllq mm0, 16 //$16,%%mm0 | |
116 por mm0, mm1 //%%mm1,%%mm0 | |
117 psllq mm0, 16 //$16,%%mm0 | |
118 por mm0, mm1 //%%mm1,%%mm0 | |
119 | |
120 mov ebx, nSize //%2,%%ebx // ebx = size | |
121 shr ebx, 3 //$3,%%ebx // process 8 bytes per iteration = 8 samples | |
122 cmp ebx, 0 //$0,%%ebx | |
123 je endS8 | |
124 | |
125 mixloopS8: | |
126 pxor mm2, mm2 //%%mm2,%%mm2 // mm2 = 0 | |
127 movq mm1, [esi] //(%%esi),%%mm1 // mm1 = a|b|c|d|e|f|g|h | |
128 movq mm3, mm1 //%%mm1,%%mm3 // mm3 = a|b|c|d|e|f|g|h | |
129 // on va faire le "sign extension" en faisant un cmp avec 0 qui retourne 1 si <0, 0 si >0 | |
130 pcmpgtb mm2, mm1 //%%mm1,%%mm2 // mm2 = 11111111|00000000|00000000.... | |
131 punpckhbw mm1, mm2 //%%mm2,%%mm1 // mm1 = 0|a|0|b|0|c|0|d | |
132 punpcklbw mm3, mm2 //%%mm2,%%mm3 // mm3 = 0|e|0|f|0|g|0|h | |
133 movq mm2, [edi] //(%%edi),%%mm2 // mm2 = destination | |
134 pmullw mm1, mm0 //%%mm0,%%mm1 // mm1 = v*a|v*b|v*c|v*d | |
135 add esi, 8 //$8,%%esi | |
136 pmullw mm3, mm0 //%%mm0,%%mm3 // mm3 = v*e|v*f|v*g|v*h | |
137 psraw mm1, 7 //$7,%%mm1 // mm1 = v*a/128|v*b/128|v*c/128|v*d/128 | |
138 psraw mm3, 7 //$7,%%mm3 // mm3 = v*e/128|v*f/128|v*g/128|v*h/128 | |
139 packsswb mm3, mm1 //%%mm1,%%mm3 // 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 | |
140 paddsb mm3, mm2 //%%mm2,%%mm3 // add to destination buffer | |
141 movq [edi], mm3 //%%mm3,(%%edi) // store back to ram | |
142 add edi, 8 //$8,%%edi | |
143 dec ebx //%%ebx | |
144 jnz mixloopS8 | |
145 | |
146 endS8: | |
147 emms | |
148 | |
149 pop ebx | |
150 pop esi | |
151 pop edi | |
152 } | |
153 } | |
154 | |
155 int _SDL_IsMMX_VC() | |
156 { | |
157 // This bit flag can get set on calling cpuid | |
158 // with register eax set to 1 | |
159 const int _MMX_FEATURE_BIT = 0x00800000; | |
160 DWORD dwFeature = 0; | |
161 __try { | |
162 _asm { | |
163 mov eax,1 | |
164 cpuid | |
165 mov dwFeature,edx | |
166 } | |
167 } __except ( EXCEPTION_EXECUTE_HANDLER) { | |
168 return 0; | |
169 } | |
170 if (dwFeature & _MMX_FEATURE_BIT) { | |
171 __try { | |
172 __asm { | |
173 pxor mm0, mm0 | |
174 emms | |
175 } | |
176 } __except (EXCEPTION_EXECUTE_HANDLER) { | |
177 return(0); | |
178 } | |
179 return 1; | |
180 } | |
181 | |
182 return 0; | |
183 } | |
184 | |
185 static int _notTwice = 2; | |
186 | |
187 int SDL_IsMMX_VC() | |
188 { | |
189 if (_notTwice > 1) | |
190 { | |
191 _notTwice = _SDL_IsMMX_VC(); | |
192 /* | |
193 #ifdef _DEBUG | |
194 if (_notTwice) | |
195 MessageBox( NULL, "Using MMX!!!", "Error", MB_OK | MB_ICONINFORMATION ); | |
196 else | |
197 MessageBox( NULL, "Not sing MMX!!!", "Error", MB_OK | MB_ICONINFORMATION ); | |
198 #endif | |
199 */ | |
200 } | |
201 return _notTwice; | |
202 } | |
203 | |
204 #endif | |
205 |