2 * MMX optimized MP3 decoding functions
3 * Copyright (c) 2010 Vitor Sessak
5 * This file is part of FFmpeg.
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.
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.
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
22 #include "libavutil/cpu.h"
23 #include "libavutil/x86_cpu.h"
24 #include "libavcodec/dsputil.h"
25 #include "libavcodec/mpegaudiodsp.h"
27 void ff_imdct36_float_sse(float *out, float *buf, float *in, float *win);
28 void ff_imdct36_float_sse2(float *out, float *buf, float *in, float *win);
29 void ff_imdct36_float_sse3(float *out, float *buf, float *in, float *win);
30 void ff_imdct36_float_ssse3(float *out, float *buf, float *in, float *win);
31 void ff_imdct36_float_avx(float *out, float *buf, float *in, float *win);
33 #define MACS(rt, ra, rb) rt+=(ra)*(rb)
34 #define MLSS(rt, ra, rb) rt-=(ra)*(rb)
36 #define SUM8(op, sum, w, p) \
38 op(sum, (w)[0 * 64], (p)[0 * 64]); \
39 op(sum, (w)[1 * 64], (p)[1 * 64]); \
40 op(sum, (w)[2 * 64], (p)[2 * 64]); \
41 op(sum, (w)[3 * 64], (p)[3 * 64]); \
42 op(sum, (w)[4 * 64], (p)[4 * 64]); \
43 op(sum, (w)[5 * 64], (p)[5 * 64]); \
44 op(sum, (w)[6 * 64], (p)[6 * 64]); \
45 op(sum, (w)[7 * 64], (p)[7 * 64]); \
48 static void apply_window(const float *buf, const float *win1,
49 const float *win2, float *sum1, float *sum2, int len)
51 x86_reg count = - 4*len;
52 const float *win1a = win1+len;
53 const float *win2a = win2+len;
54 const float *bufa = buf+len;
55 float *sum1a = sum1+len;
56 float *sum2a = sum2+len;
60 "movaps " #a "(%1,%0), %%xmm1 \n\t" \
61 "movaps " #a "(%3,%0), %%xmm2 \n\t" \
62 "mulps %%xmm2, %%xmm1 \n\t" \
63 "subps %%xmm1, %%xmm0 \n\t" \
64 "mulps " #b "(%2,%0), %%xmm2 \n\t" \
65 "subps %%xmm2, %%xmm4 \n\t" \
69 "xorps %%xmm0, %%xmm0 \n\t"
70 "xorps %%xmm4, %%xmm4 \n\t"
81 "movaps %%xmm0, (%4,%0) \n\t"
82 "movaps %%xmm4, (%5,%0) \n\t"
86 :"r"(win1a), "r"(win2a), "r"(bufa), "r"(sum1a), "r"(sum2a)
92 static void apply_window_mp3(float *in, float *win, int *unused, float *out,
95 LOCAL_ALIGNED_16(float, suma, [17]);
96 LOCAL_ALIGNED_16(float, sumb, [17]);
97 LOCAL_ALIGNED_16(float, sumc, [17]);
98 LOCAL_ALIGNED_16(float, sumd, [17]);
102 /* copy to avoid wrap */
103 memcpy(in + 512, in, 32 * sizeof(*in));
105 apply_window(in + 16, win , win + 512, suma, sumc, 16);
106 apply_window(in + 32, win + 48, win + 640, sumb, sumd, 16);
108 SUM8(MACS, suma[0], win + 32, in + 48);
114 #define SUMS(suma, sumb, sumc, sumd, out1, out2) \
115 "movups " #sumd "(%4), %%xmm0 \n\t" \
116 "shufps $0x1b, %%xmm0, %%xmm0 \n\t" \
117 "subps " #suma "(%1), %%xmm0 \n\t" \
118 "movaps %%xmm0," #out1 "(%0) \n\t" \
120 "movups " #sumc "(%3), %%xmm0 \n\t" \
121 "shufps $0x1b, %%xmm0, %%xmm0 \n\t" \
122 "addps " #sumb "(%2), %%xmm0 \n\t" \
123 "movaps %%xmm0," #out2 "(%0) \n\t"
127 SUMS( 0, 48, 4, 52, 0, 112)
128 SUMS(16, 32, 20, 36, 16, 96)
129 SUMS(32, 16, 36, 20, 32, 80)
130 SUMS(48, 0, 52, 4, 48, 64)
133 :"r"(&suma[0]), "r"(&sumb[0]), "r"(&sumc[0]), "r"(&sumd[0])
139 float *out2 = out + 32 * incr;
144 *out = -suma[ j] + sumd[16-j];
145 *out2 = sumb[16-j] + sumc[ j];
152 SUM8(MLSS, sum, win + 16 + 32, in + 32);
156 void ff_mpadsp_init_mmx(MPADSPContext *s)
158 int mm_flags = av_get_cpu_flags();
160 if (mm_flags & AV_CPU_FLAG_SSE2) {
161 s->apply_window_float = apply_window_mp3;
163 if (HAVE_YASM && mm_flags & AV_CPU_FLAG_AVX && HAVE_AVX) {
164 s->imdct36_float = ff_imdct36_float_avx;
166 else if (HAVE_YASM && mm_flags & AV_CPU_FLAG_SSSE3 && HAVE_SSE) {
167 s->imdct36_float = ff_imdct36_float_ssse3;
169 else if (HAVE_YASM && mm_flags & AV_CPU_FLAG_SSE3 && HAVE_SSE) {
170 s->imdct36_float = ff_imdct36_float_sse3;
172 else if (HAVE_YASM && mm_flags & AV_CPU_FLAG_SSE2 && HAVE_SSE) {
173 s->imdct36_float = ff_imdct36_float_sse2;
175 else if (HAVE_YASM && mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE) {
176 s->imdct36_float = ff_imdct36_float_sse;