Mercurial > libavcodec.hg
comparison x86/mpegaudiodec_mmx.c @ 11939:ef338bd70180 libavcodec
SSE-optimized MP3 floating point windowing functions
author | vitor |
---|---|
date | Thu, 24 Jun 2010 07:44:50 +0000 |
parents | |
children | 9e12316c508a |
comparison
equal
deleted
inserted
replaced
11938:a5c32cb046eb | 11939:ef338bd70180 |
---|---|
1 /* | |
2 * MMX optimized MP3 decoding functions | |
3 * Copyright (c) 2010 Vitor Sessak | |
4 * | |
5 * This file is part of FFmpeg. | |
6 * | |
7 * FFmpeg is free software; you can redistribute it and/or | |
8 * modify it under the terms of the GNU Lesser General Public | |
9 * License as published by the Free Software Foundation; either | |
10 * version 2.1 of the License, or (at your option) any later version. | |
11 * | |
12 * FFmpeg is distributed in the hope that it will be useful, | |
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
15 * Lesser General Public License for more details. | |
16 * | |
17 * You should have received a copy of the GNU Lesser General Public | |
18 * License along with FFmpeg; if not, write to the Free Software | |
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
20 */ | |
21 | |
22 #include "libavutil/x86_cpu.h" | |
23 | |
24 #define CONFIG_FLOAT 1 | |
25 #include "libavcodec/mpegaudio.h" | |
26 | |
27 #define MACS(rt, ra, rb) rt+=(ra)*(rb) | |
28 #define MLSS(rt, ra, rb) rt-=(ra)*(rb) | |
29 | |
30 #define SUM8(op, sum, w, p) \ | |
31 { \ | |
32 op(sum, (w)[0 * 64], (p)[0 * 64]); \ | |
33 op(sum, (w)[1 * 64], (p)[1 * 64]); \ | |
34 op(sum, (w)[2 * 64], (p)[2 * 64]); \ | |
35 op(sum, (w)[3 * 64], (p)[3 * 64]); \ | |
36 op(sum, (w)[4 * 64], (p)[4 * 64]); \ | |
37 op(sum, (w)[5 * 64], (p)[5 * 64]); \ | |
38 op(sum, (w)[6 * 64], (p)[6 * 64]); \ | |
39 op(sum, (w)[7 * 64], (p)[7 * 64]); \ | |
40 } | |
41 | |
42 static void apply_window(const float *buf, const float *win1, | |
43 const float *win2, float *sum1, float *sum2, int len) | |
44 { | |
45 x86_reg count = - 4*len; | |
46 const float *win1a = win1+len; | |
47 const float *win2a = win2+len; | |
48 const float *bufa = buf+len; | |
49 float *sum1a = sum1+len; | |
50 float *sum2a = sum2+len; | |
51 | |
52 | |
53 #define MULT(a, b) \ | |
54 "movaps " #a "(%0,%5), %%xmm1 \n\t" \ | |
55 "movaps " #a "(%2,%5), %%xmm2 \n\t" \ | |
56 "mulps %%xmm2, %%xmm1 \n\t" \ | |
57 "subps %%xmm1, %%xmm0 \n\t" \ | |
58 "mulps " #b "(%1,%5), %%xmm2 \n\t" \ | |
59 "subps %%xmm2, %%xmm4 \n\t" \ | |
60 | |
61 __asm__ volatile( | |
62 "1: \n\t" | |
63 "xorps %%xmm0, %%xmm0 \n\t" | |
64 "xorps %%xmm4, %%xmm4 \n\t" | |
65 | |
66 MULT( 0, 0) | |
67 MULT( 256, 64) | |
68 MULT( 512, 128) | |
69 MULT( 768, 192) | |
70 MULT(1024, 256) | |
71 MULT(1280, 320) | |
72 MULT(1536, 384) | |
73 MULT(1792, 448) | |
74 | |
75 "movaps %%xmm0, (%3,%5) \n\t" | |
76 "movaps %%xmm4, (%4,%5) \n\t" | |
77 "addl $16, %5 \n\t" | |
78 "jl 1b \n\t" | |
79 :"+&r"(win1a), "+&r"(win2a), "+&r"(bufa), "+&r"(sum1a), "+&r"(sum2a), "+&r"(count) | |
80 ); | |
81 | |
82 #undef MULT | |
83 } | |
84 | |
85 static void apply_window_mp3(float *in, float *win, int *unused, float *out, | |
86 int incr) | |
87 { | |
88 LOCAL_ALIGNED_16(float, suma, [17]); | |
89 LOCAL_ALIGNED_16(float, sumb, [17]); | |
90 LOCAL_ALIGNED_16(float, sumc, [17]); | |
91 LOCAL_ALIGNED_16(float, sumd, [17]); | |
92 | |
93 float sum; | |
94 | |
95 /* copy to avoid wrap */ | |
96 memcpy(in + 512, in, 32 * sizeof(*in)); | |
97 | |
98 apply_window(in + 16, win , win + 512, suma, sumc, 16); | |
99 apply_window(in + 32, win + 48, win + 640, sumb, sumd, 16); | |
100 | |
101 SUM8(MACS, suma[0], win + 32, in + 48); | |
102 | |
103 sumc[ 0] = 0; | |
104 sumb[16] = 0; | |
105 sumd[16] = 0; | |
106 | |
107 #define SUMS(suma, sumb, sumc, sumd, out1, out2) \ | |
108 "movups " #sumd "(%4), %%xmm0 \n\t" \ | |
109 "shufps $0x1b, %%xmm0, %%xmm0 \n\t" \ | |
110 "subps " #suma "(%1), %%xmm0 \n\t" \ | |
111 "movaps %%xmm0," #out1 "(%0) \n\t" \ | |
112 \ | |
113 "movups " #sumc "(%3), %%xmm0 \n\t" \ | |
114 "shufps $0x1b, %%xmm0, %%xmm0 \n\t" \ | |
115 "addps " #sumb "(%2), %%xmm0 \n\t" \ | |
116 "movaps %%xmm0," #out2 "(%0) \n\t" | |
117 | |
118 if (incr == 1) { | |
119 __asm__ volatile( | |
120 SUMS( 0, 48, 4, 52, 0, 112) | |
121 SUMS(16, 32, 20, 36, 16, 96) | |
122 SUMS(32, 16, 36, 20, 32, 80) | |
123 SUMS(48, 0, 52, 4, 48, 64) | |
124 | |
125 :"+&r"(out) | |
126 :"r"(&suma[0]), "r"(&sumb[0]), "r"(&sumc[0]), "r"(&sumd[0]) | |
127 :"memory" | |
128 ); | |
129 out += 16*incr; | |
130 } else { | |
131 int j; | |
132 float *out2 = out + 32 * incr; | |
133 out[0 ] = -suma[ 0]; | |
134 out += incr; | |
135 out2 -= incr; | |
136 for(j=1;j<16;j++) { | |
137 *out = -suma[ j] + sumd[16-j]; | |
138 *out2 = sumb[16-j] + sumc[ j]; | |
139 out += incr; | |
140 out2 -= incr; | |
141 } | |
142 } | |
143 | |
144 sum = 0; | |
145 SUM8(MLSS, sum, win + 16 + 32, in + 32); | |
146 *out = sum; | |
147 } | |
148 | |
149 void ff_mpegaudiodec_init_mmx(MPADecodeContext *s) | |
150 { | |
151 mm_flags = mm_support(); | |
152 | |
153 if (mm_flags & FF_MM_SSE2) { | |
154 s->apply_window_mp3 = apply_window_mp3; | |
155 } | |
156 } |