Libav
mpegaudiodsp.c
Go to the documentation of this file.
1 /*
2  * SIMD-optimized MP3 decoding functions
3  * Copyright (c) 2010 Vitor Sessak
4  *
5  * This file is part of Libav.
6  *
7  * Libav 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  * Libav 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 Libav; 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/attributes.h"
23 #include "libavutil/cpu.h"
24 #include "libavutil/internal.h"
25 #include "libavutil/x86/asm.h"
26 #include "libavutil/x86/cpu.h"
28 
29 void ff_imdct36_float_sse(float *out, float *buf, float *in, float *win);
30 void ff_imdct36_float_sse2(float *out, float *buf, float *in, float *win);
31 void ff_imdct36_float_sse3(float *out, float *buf, float *in, float *win);
32 void ff_imdct36_float_ssse3(float *out, float *buf, float *in, float *win);
33 void ff_imdct36_float_avx(float *out, float *buf, float *in, float *win);
34 void ff_four_imdct36_float_sse(float *out, float *buf, float *in, float *win,
35  float *tmpbuf);
36 void ff_four_imdct36_float_avx(float *out, float *buf, float *in, float *win,
37  float *tmpbuf);
38 
39 DECLARE_ALIGNED(16, static float, mdct_win_sse)[2][4][4*40];
40 
41 #if HAVE_SSE2_INLINE
42 
43 #define MACS(rt, ra, rb) rt+=(ra)*(rb)
44 #define MLSS(rt, ra, rb) rt-=(ra)*(rb)
45 
46 #define SUM8(op, sum, w, p) \
47 { \
48  op(sum, (w)[0 * 64], (p)[0 * 64]); \
49  op(sum, (w)[1 * 64], (p)[1 * 64]); \
50  op(sum, (w)[2 * 64], (p)[2 * 64]); \
51  op(sum, (w)[3 * 64], (p)[3 * 64]); \
52  op(sum, (w)[4 * 64], (p)[4 * 64]); \
53  op(sum, (w)[5 * 64], (p)[5 * 64]); \
54  op(sum, (w)[6 * 64], (p)[6 * 64]); \
55  op(sum, (w)[7 * 64], (p)[7 * 64]); \
56 }
57 
58 static void apply_window(const float *buf, const float *win1,
59  const float *win2, float *sum1, float *sum2, int len)
60 {
61  x86_reg count = - 4*len;
62  const float *win1a = win1+len;
63  const float *win2a = win2+len;
64  const float *bufa = buf+len;
65  float *sum1a = sum1+len;
66  float *sum2a = sum2+len;
67 
68 
69 #define MULT(a, b) \
70  "movaps " #a "(%1,%0), %%xmm1 \n\t" \
71  "movaps " #a "(%3,%0), %%xmm2 \n\t" \
72  "mulps %%xmm2, %%xmm1 \n\t" \
73  "subps %%xmm1, %%xmm0 \n\t" \
74  "mulps " #b "(%2,%0), %%xmm2 \n\t" \
75  "subps %%xmm2, %%xmm4 \n\t" \
76 
77  __asm__ volatile(
78  "1: \n\t"
79  "xorps %%xmm0, %%xmm0 \n\t"
80  "xorps %%xmm4, %%xmm4 \n\t"
81 
82  MULT( 0, 0)
83  MULT( 256, 64)
84  MULT( 512, 128)
85  MULT( 768, 192)
86  MULT(1024, 256)
87  MULT(1280, 320)
88  MULT(1536, 384)
89  MULT(1792, 448)
90 
91  "movaps %%xmm0, (%4,%0) \n\t"
92  "movaps %%xmm4, (%5,%0) \n\t"
93  "add $16, %0 \n\t"
94  "jl 1b \n\t"
95  :"+&r"(count)
96  :"r"(win1a), "r"(win2a), "r"(bufa), "r"(sum1a), "r"(sum2a)
97  );
98 
99 #undef MULT
100 }
101 
102 static void apply_window_mp3(float *in, float *win, int *unused, float *out,
103  int incr)
104 {
105  LOCAL_ALIGNED_16(float, suma, [17]);
106  LOCAL_ALIGNED_16(float, sumb, [17]);
107  LOCAL_ALIGNED_16(float, sumc, [17]);
108  LOCAL_ALIGNED_16(float, sumd, [17]);
109 
110  float sum;
111 
112  /* copy to avoid wrap */
113  __asm__ volatile(
114  "movaps 0(%0), %%xmm0 \n\t" \
115  "movaps 16(%0), %%xmm1 \n\t" \
116  "movaps 32(%0), %%xmm2 \n\t" \
117  "movaps 48(%0), %%xmm3 \n\t" \
118  "movaps %%xmm0, 0(%1) \n\t" \
119  "movaps %%xmm1, 16(%1) \n\t" \
120  "movaps %%xmm2, 32(%1) \n\t" \
121  "movaps %%xmm3, 48(%1) \n\t" \
122  "movaps 64(%0), %%xmm0 \n\t" \
123  "movaps 80(%0), %%xmm1 \n\t" \
124  "movaps 96(%0), %%xmm2 \n\t" \
125  "movaps 112(%0), %%xmm3 \n\t" \
126  "movaps %%xmm0, 64(%1) \n\t" \
127  "movaps %%xmm1, 80(%1) \n\t" \
128  "movaps %%xmm2, 96(%1) \n\t" \
129  "movaps %%xmm3, 112(%1) \n\t"
130  ::"r"(in), "r"(in+512)
131  :"memory"
132  );
133 
134  apply_window(in + 16, win , win + 512, suma, sumc, 16);
135  apply_window(in + 32, win + 48, win + 640, sumb, sumd, 16);
136 
137  SUM8(MACS, suma[0], win + 32, in + 48);
138 
139  sumc[ 0] = 0;
140  sumb[16] = 0;
141  sumd[16] = 0;
142 
143 #define SUMS(suma, sumb, sumc, sumd, out1, out2) \
144  "movups " #sumd "(%4), %%xmm0 \n\t" \
145  "shufps $0x1b, %%xmm0, %%xmm0 \n\t" \
146  "subps " #suma "(%1), %%xmm0 \n\t" \
147  "movaps %%xmm0," #out1 "(%0) \n\t" \
148 \
149  "movups " #sumc "(%3), %%xmm0 \n\t" \
150  "shufps $0x1b, %%xmm0, %%xmm0 \n\t" \
151  "addps " #sumb "(%2), %%xmm0 \n\t" \
152  "movaps %%xmm0," #out2 "(%0) \n\t"
153 
154  if (incr == 1) {
155  __asm__ volatile(
156  SUMS( 0, 48, 4, 52, 0, 112)
157  SUMS(16, 32, 20, 36, 16, 96)
158  SUMS(32, 16, 36, 20, 32, 80)
159  SUMS(48, 0, 52, 4, 48, 64)
160 
161  :"+&r"(out)
162  :"r"(&suma[0]), "r"(&sumb[0]), "r"(&sumc[0]), "r"(&sumd[0])
163  :"memory"
164  );
165  out += 16*incr;
166  } else {
167  int j;
168  float *out2 = out + 32 * incr;
169  out[0 ] = -suma[ 0];
170  out += incr;
171  out2 -= incr;
172  for(j=1;j<16;j++) {
173  *out = -suma[ j] + sumd[16-j];
174  *out2 = sumb[16-j] + sumc[ j];
175  out += incr;
176  out2 -= incr;
177  }
178  }
179 
180  sum = 0;
181  SUM8(MLSS, sum, win + 16 + 32, in + 32);
182  *out = sum;
183 }
184 
185 #endif /* HAVE_SSE2_INLINE */
186 
187 #if HAVE_YASM
188 #define DECL_IMDCT_BLOCKS(CPU1, CPU2) \
189 static void imdct36_blocks_ ## CPU1(float *out, float *buf, float *in, \
190  int count, int switch_point, int block_type) \
191 { \
192  int align_end = count - (count & 3); \
193  int j; \
194  for (j = 0; j < align_end; j+= 4) { \
195  LOCAL_ALIGNED_16(float, tmpbuf, [1024]); \
196  float *win = mdct_win_sse[switch_point && j < 4][block_type]; \
197  /* apply window & overlap with previous buffer */ \
198  \
199  /* select window */ \
200  ff_four_imdct36_float_ ## CPU2(out, buf, in, win, tmpbuf); \
201  in += 4*18; \
202  buf += 4*18; \
203  out += 4; \
204  } \
205  for (; j < count; j++) { \
206  /* apply window & overlap with previous buffer */ \
207  \
208  /* select window */ \
209  int win_idx = (switch_point && j < 2) ? 0 : block_type; \
210  float *win = ff_mdct_win_float[win_idx + (4 & -(j & 1))]; \
211  \
212  ff_imdct36_float_ ## CPU1(out, buf, in, win); \
213  \
214  in += 18; \
215  buf++; \
216  out++; \
217  } \
218 }
219 
220 DECL_IMDCT_BLOCKS(sse,sse)
221 DECL_IMDCT_BLOCKS(sse2,sse)
222 DECL_IMDCT_BLOCKS(sse3,sse)
223 DECL_IMDCT_BLOCKS(ssse3,sse)
224 DECL_IMDCT_BLOCKS(avx,avx)
225 #endif /* HAVE_YASM */
226 
228 {
229  int cpu_flags = av_get_cpu_flags();
230 
231  int i, j;
232  for (j = 0; j < 4; j++) {
233  for (i = 0; i < 40; i ++) {
234  mdct_win_sse[0][j][4*i ] = ff_mdct_win_float[j ][i];
235  mdct_win_sse[0][j][4*i + 1] = ff_mdct_win_float[j + 4][i];
236  mdct_win_sse[0][j][4*i + 2] = ff_mdct_win_float[j ][i];
237  mdct_win_sse[0][j][4*i + 3] = ff_mdct_win_float[j + 4][i];
238  mdct_win_sse[1][j][4*i ] = ff_mdct_win_float[0 ][i];
239  mdct_win_sse[1][j][4*i + 1] = ff_mdct_win_float[4 ][i];
240  mdct_win_sse[1][j][4*i + 2] = ff_mdct_win_float[j ][i];
241  mdct_win_sse[1][j][4*i + 3] = ff_mdct_win_float[j + 4][i];
242  }
243  }
244 
245 #if HAVE_SSE2_INLINE
246  if (INLINE_SSE2(cpu_flags)) {
247  s->apply_window_float = apply_window_mp3;
248  }
249 #endif /* HAVE_SSE2_INLINE */
250 
251 #if HAVE_YASM
252  if (EXTERNAL_SSE(cpu_flags)) {
253  s->imdct36_blocks_float = imdct36_blocks_sse;
254  }
255  if (EXTERNAL_SSE2(cpu_flags)) {
256  s->imdct36_blocks_float = imdct36_blocks_sse2;
257  }
258  if (EXTERNAL_SSE3(cpu_flags)) {
259  s->imdct36_blocks_float = imdct36_blocks_sse3;
260  }
261  if (EXTERNAL_SSSE3(cpu_flags)) {
262  s->imdct36_blocks_float = imdct36_blocks_ssse3;
263  }
264  if (EXTERNAL_AVX(cpu_flags)) {
265  s->imdct36_blocks_float = imdct36_blocks_avx;
266  }
267 #endif /* HAVE_YASM */
268 }
#define DECLARE_ALIGNED(n, t, v)
Definition: mem.h:58
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_dlog(ac->avr,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> in
static int sse(MpegEncContext *s, uint8_t *src1, uint8_t *src2, int w, int h, int stride)
#define INLINE_SSE2(flags)
Definition: cpu.h:66
#define EXTERNAL_SSE(flags)
Definition: cpu.h:49
float ff_mdct_win_float[8][MDCT_BUF_SIZE]
Macro definitions for various function/variable attributes.
#define MACS(rt, ra, rb)
#define av_cold
Definition: attributes.h:66
int x86_reg
Definition: asm.h:70
#define EXTERNAL_SSE3(flags)
Definition: cpu.h:51
#define EXTERNAL_SSE2(flags)
Definition: cpu.h:50
void ff_imdct36_float_avx(float *out, float *buf, float *in, float *win)
#define r
Definition: input.c:51
#define SUM8(op, sum, w, p)
void ff_imdct36_float_ssse3(float *out, float *buf, float *in, float *win)
void ff_imdct36_float_sse(float *out, float *buf, float *in, float *win)
common internal API header
void(* imdct36_blocks_float)(float *out, float *buf, float *in, int count, int switch_point, int block_type)
Definition: mpegaudiodsp.h:32
static float mdct_win_sse[2][4][4 *40]
Definition: mpegaudiodsp.c:39
static void(*const apply_window[4])(AVFloatDSPContext *fdsp, SingleChannelElement *sce, const float *audio)
Definition: aacenc.c:240
#define MLSS(rt, ra, rb)
av_cold void ff_mpadsp_init_x86(MPADSPContext *s)
Definition: mpegaudiodsp.c:227
void ff_imdct36_float_sse3(float *out, float *buf, float *in, float *win)
#define EXTERNAL_SSSE3(flags)
Definition: cpu.h:52
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition: cpu.c:47
void ff_four_imdct36_float_avx(float *out, float *buf, float *in, float *win, float *tmpbuf)
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_dlog(ac->avr,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> out
void ff_four_imdct36_float_sse(float *out, float *buf, float *in, float *win, float *tmpbuf)
void(* apply_window_float)(float *synth_buf, float *window, int *dither_state, float *samples, int incr)
Definition: mpegaudiodsp.h:26
int len
#define LOCAL_ALIGNED_16(t, v,...)
Definition: internal.h:114
#define EXTERNAL_AVX(flags)
Definition: cpu.h:55
void ff_imdct36_float_sse2(float *out, float *buf, float *in, float *win)