SDL_mixer_MMX_VC.c
1
2
3
#include "SDL_mixer_MMX_VC.h"
4
#if defined(USE_ASM_MIXER_VC)
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
// MMX assembler version of SDL_MixAudio for signed little endian 16 bit samples and signed 8 bit samples
// Copyright 2002 Stephane Marchesin (stephane.marchesin@wanadoo.fr)
// Converted to Intel ASM notation by Cth
// This code is licensed under the LGPL (see COPYING for details)
//
// Assumes buffer size in bytes is a multiple of 16
// Assumes SDL_MIX_MAXVOLUME = 128
////////////////////////////////////////////////
// Mixing for 16 bit signed buffers
////////////////////////////////////////////////
void SDL_MixAudio_MMX_S16_VC(char* dst,char* src,unsigned int nSize,int volume)
{
__asm
{
push edi
push esi
push ebx
mov edi, dst // edi = dst
mov esi, src // esi = src
mov eax, volume // eax = volume
mov ebx, nSize // ebx = size
shr ebx, 4 // process 16 bytes per iteration = 8 samples
jz endS16
pxor mm0, mm0
movd mm0, eax //%%eax,%%mm0
movq mm1, mm0 //%%mm0,%%mm1
psllq mm0, 16 //$16,%%mm0
por mm0, mm1 //%%mm1,%%mm0
psllq mm0, 16 //$16,%%mm0
por mm0, mm1 //%%mm1,%%mm0
psllq mm0, 16 //$16,%%mm0
por mm0, mm1 //%%mm1,%%mm0 // mm0 = vol|vol|vol|vol
44
#ifndef __WATCOMC__
46
#endif
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
mixloopS16:
movq mm1, [esi] //(%%esi),%%mm1\n" // mm1 = a|b|c|d
movq mm2, mm1 //%%mm1,%%mm2\n" // mm2 = a|b|c|d
movq mm4, [esi + 8] //8(%%esi),%%mm4\n" // mm4 = e|f|g|h
// pre charger le buffer dst dans mm7
movq mm7, [edi] //(%%edi),%%mm7\n" // mm7 = dst[0]"
// multiplier par le volume
pmullw mm1, mm0 //%%mm0,%%mm1\n" // mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v)
pmulhw mm2, mm0 //%%mm0,%%mm2\n" // mm2 = h(a*v)|h(b*v)|h(c*v)|h(d*v)
movq mm5, mm4 //%%mm4,%%mm5\n" // mm5 = e|f|g|h
pmullw mm4, mm0 //%%mm0,%%mm4\n" // mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v)
pmulhw mm5, mm0 //%%mm0,%%mm5\n" // mm5 = h(e*v)|h(f*v)|h(g*v)|h(h*v)
movq mm3, mm1 //%%mm1,%%mm3\n" // mm3 = l(a*v)|l(b*v)|l(c*v)|l(d*v)
punpckhwd mm1, mm2 //%%mm2,%%mm1\n" // mm1 = a*v|b*v
movq mm6, mm4 //%%mm4,%%mm6\n" // mm6 = l(e*v)|l(f*v)|l(g*v)|l(h*v)
punpcklwd mm3, mm2 //%%mm2,%%mm3\n" // mm3 = c*v|d*v
punpckhwd mm4, mm5 //%%mm5,%%mm4\n" // mm4 = e*f|f*v
punpcklwd mm6, mm5 //%%mm5,%%mm6\n" // mm6 = g*v|h*v
// pre charger le buffer dst dans mm5
movq mm5, [edi + 8] //8(%%edi),%%mm5\n" // mm5 = dst[1]
// diviser par 128
psrad mm1, 7 //$7,%%mm1\n" // mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME
add esi, 16 //$16,%%esi\n"
psrad mm3, 7 //$7,%%mm3\n" // mm3 = c*v/128|d*v/128
psrad mm4, 7 //$7,%%mm4\n" // mm4 = e*v/128|f*v/128
// mm1 = le sample avec le volume modifie
packssdw mm3, mm1 //%%mm1,%%mm3\n" // mm3 = s(a*v|b*v|c*v|d*v)
psrad mm6, 7 //$7,%%mm6\n" // mm6= g*v/128|h*v/128
paddsw mm3, mm7 //%%mm7,%%mm3\n" // mm3 = adjust_volume(src)+dst
// mm4 = le sample avec le volume modifie
packssdw mm6, mm4 //%%mm4,%%mm6\n" // mm6 = s(e*v|f*v|g*v|h*v)
movq [edi], mm3 //%%mm3,(%%edi)\n"
paddsw mm6, mm5 //%%mm5,%%mm6\n" // mm6 = adjust_volume(src)+dst
movq [edi + 8], mm6 //%%mm6,8(%%edi)\n"
add edi, 16 //$16,%%edi\n"
dec ebx //%%ebx\n"
jnz mixloopS16
ends16:
emms
pop ebx
pop esi
pop edi
}
}
////////////////////////////////////////////////
// Mixing for 8 bit signed buffers
////////////////////////////////////////////////
void SDL_MixAudio_MMX_S8_VC(char* dst,char* src,unsigned int nSize,int volume)
{
_asm
{
push edi
push esi
push ebx
mov edi, dst //movl %0,%%edi // edi = dst
mov esi, src //%1,%%esi // esi = src
mov eax, volume //%3,%%eax // eax = volume
112
movd mm0, eax //%%eax,%%mm0
113
114
115
116
117
118
119
120
121
122
123
124
125
movq mm1, mm0 //%%mm0,%%mm1
psllq mm0, 16 //$16,%%mm0
por mm0, mm1 //%%mm1,%%mm0
psllq mm0, 16 //$16,%%mm0
por mm0, mm1 //%%mm1,%%mm0
psllq mm0, 16 //$16,%%mm0
por mm0, mm1 //%%mm1,%%mm0
mov ebx, nSize //%2,%%ebx // ebx = size
shr ebx, 3 //$3,%%ebx // process 8 bytes per iteration = 8 samples
cmp ebx, 0 //$0,%%ebx
je endS8
126
#ifndef __WATCOMC__
128
#endif
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
mixloopS8:
pxor mm2, mm2 //%%mm2,%%mm2 // mm2 = 0
movq mm1, [esi] //(%%esi),%%mm1 // mm1 = a|b|c|d|e|f|g|h
movq mm3, mm1 //%%mm1,%%mm3 // 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 mm2, mm1 //%%mm1,%%mm2 // mm2 = 11111111|00000000|00000000....
punpckhbw mm1, mm2 //%%mm2,%%mm1 // mm1 = 0|a|0|b|0|c|0|d
punpcklbw mm3, mm2 //%%mm2,%%mm3 // mm3 = 0|e|0|f|0|g|0|h
movq mm2, [edi] //(%%edi),%%mm2 // mm2 = destination
pmullw mm1, mm0 //%%mm0,%%mm1 // mm1 = v*a|v*b|v*c|v*d
add esi, 8 //$8,%%esi
pmullw mm3, mm0 //%%mm0,%%mm3 // mm3 = v*e|v*f|v*g|v*h
psraw mm1, 7 //$7,%%mm1 // mm1 = v*a/128|v*b/128|v*c/128|v*d/128
psraw mm3, 7 //$7,%%mm3 // mm3 = v*e/128|v*f/128|v*g/128|v*h/128
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
paddsb mm3, mm2 //%%mm2,%%mm3 // add to destination buffer
movq [edi], mm3 //%%mm3,(%%edi) // store back to ram
add edi, 8 //$8,%%edi
dec ebx //%%ebx
jnz mixloopS8
endS8:
emms
pop ebx
pop esi
pop edi
}
}
159
#endif /* USE_ASM_MIXER_VC */