SDL_mixer_MMX.c
1
2
/*
SDL - Simple DirectMedia Layer
3
Copyright (C) 1997-2009 Sam Lantinga
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
Sam Lantinga
slouken@libsdl.org
*/
#include "SDL_config.h"
24
25
26
27
28
29
30
31
32
33
34
35
36
/*
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)
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
***********************************************/
38
#if defined(__GNUC__) && defined(__i386__) && defined(SDL_ASSEMBLY_ROUTINES)
39
40
41
42
void SDL_MixAudio_MMX_S16(char* dst,char* src,unsigned int size,int volume)
{
__asm__ __volatile__ (
43
" movl %3,%%eax\n" /* eax = volume */
45
" movl %2,%%edx\n" /* edx = size */
47
" shrl $4,%%edx\n" /* process 16 bytes per iteration = 8 samples */
48
49
50
51
52
53
54
55
56
57
58
59
" 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"
60
" por %%mm1,%%mm0\n" /* mm0 = vol|vol|vol|vol */
62
".align 8\n"
63
64
" .mixloopS16:\n"
65
" movq (%1),%%mm1\n" /* mm1 = a|b|c|d */
67
" movq %%mm1,%%mm2\n" /* mm2 = a|b|c|d */
69
" movq 8(%1),%%mm4\n" /* mm4 = e|f|g|h */
71
72
/* pré charger le buffer dst dans mm7 */
" movq (%0),%%mm7\n" /* mm7 = dst[0] */
74
75
/* multiplier par le volume */
" pmullw %%mm0,%%mm1\n" /* mm1 = l(a*v)|l(b*v)|l(c*v)|l(d*v) */
77
78
" 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 */
80
" pmullw %%mm0,%%mm4\n" /* mm4 = l(e*v)|l(f*v)|l(g*v)|l(h*v) */
82
83
" 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) */
85
" punpckhwd %%mm2,%%mm1\n" /* mm1 = a*v|b*v */
87
88
" 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 */
90
" punpckhwd %%mm5,%%mm4\n" /* mm4 = e*f|f*v */
92
" punpcklwd %%mm5,%%mm6\n" /* mm6 = g*v|h*v */
94
95
/* pré charger le buffer dst dans mm5 */
" movq 8(%0),%%mm5\n" /* mm5 = dst[1] */
97
98
/* diviser par 128 */
" psrad $7,%%mm1\n" /* mm1 = a*v/128|b*v/128 , 128 = SDL_MIX_MAXVOLUME */
99
" add $16,%1\n"
101
" psrad $7,%%mm3\n" /* mm3 = c*v/128|d*v/128 */
103
" psrad $7,%%mm4\n" /* mm4 = e*v/128|f*v/128 */
105
106
/* mm1 = le sample avec le volume modifié */
" packssdw %%mm1,%%mm3\n" /* mm3 = s(a*v|b*v|c*v|d*v) */
108
109
" psrad $7,%%mm6\n" /* mm6= g*v/128|h*v/128 */
" paddsw %%mm7,%%mm3\n" /* mm3 = adjust_volume(src)+dst */
111
112
/* mm4 = le sample avec le volume modifié */
" packssdw %%mm4,%%mm6\n" /* mm6 = s(e*v|f*v|g*v|h*v) */
113
" movq %%mm3,(%0)\n"
115
" paddsw %%mm5,%%mm6\n" /* mm6 = adjust_volume(src)+dst */
117
" movq %%mm6,8(%0)\n"
119
" add $16,%0\n"
121
" dec %%edx\n"
122
123
124
125
126
127
128
" jnz .mixloopS16\n"
" emms\n"
".endS16:\n"
:
129
: "r" (dst), "r"(src),"m"(size),
130
"m"(volume)
131
: "eax","edx","memory"
132
133
134
135
136
);
}
137
138
139
/*////////////////////////////////////////////// */
/* Mixing for 8 bit signed buffers */
/*////////////////////////////////////////////// */
140
141
142
143
144
void SDL_MixAudio_MMX_S8(char* dst,char* src,unsigned int size,int volume)
{
__asm__ __volatile__ (
145
" movl %3,%%eax\n" /* eax = volume */
147
" movd %%eax,%%mm0\n"
148
149
150
151
152
153
154
155
" 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"
156
157
" movl %2,%%edx\n" /* edx = size */
" shr $3,%%edx\n" /* process 8 bytes per iteration = 8 samples */
159
" cmp $0,%%edx\n"
160
161
" je .endS8\n"
162
".align 8\n"
163
164
" .mixloopS8:\n"
165
166
" pxor %%mm2,%%mm2\n" /* mm2 = 0 */
" movq (%1),%%mm1\n" /* mm1 = a|b|c|d|e|f|g|h */
168
" movq %%mm1,%%mm3\n" /* mm3 = a|b|c|d|e|f|g|h */
170
171
/* 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.... */
173
" punpckhbw %%mm2,%%mm1\n" /* mm1 = 0|a|0|b|0|c|0|d */
175
176
" punpcklbw %%mm2,%%mm3\n" /* mm3 = 0|e|0|f|0|g|0|h */
" movq (%0),%%mm2\n" /* mm2 = destination */
178
" pmullw %%mm0,%%mm1\n" /* mm1 = v*a|v*b|v*c|v*d */
179
" add $8,%1\n"
181
182
" 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 */
184
" psraw $7,%%mm3\n" /* mm3 = v*e/128|v*f/128|v*g/128|v*h/128 */
186
" 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 */
188
" paddsb %%mm2,%%mm3\n" /* add to destination buffer */
190
" movq %%mm3,(%0)\n" /* store back to ram */
191
" add $8,%0\n"
193
" dec %%edx\n"
194
195
196
197
198
199
" jnz .mixloopS8\n"
".endS8:\n"
" emms\n"
:
200
: "r" (dst), "r"(src),"m"(size),
201
"m"(volume)
202
: "eax","edx","memory"
203
204
205
);
}
#endif