22 #include "libavutil/attributes.h"
23 #include "libavutil/cpu.h"
43 #define MACS(rt, ra, rb) rt+=(ra)*(rb)
44 #define MLSS(rt, ra, rb) rt-=(ra)*(rb)
46 #define SUM8(op, sum, w, p) \
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]); \
58 static void apply_window(
const float *buf,
const float *win1,
59 const float *win2,
float *sum1,
float *sum2,
int 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;
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" \
79 "xorps %%xmm0, %%xmm0 \n\t"
80 "xorps %%xmm4, %%xmm4 \n\t"
91 "movaps %%xmm0, (%4,%0) \n\t"
92 "movaps %%xmm4, (%5,%0) \n\t"
96 :
"r"(win1a),
"r"(win2a),
"r"(bufa),
"r"(sum1a),
"r"(sum2a)
102 static void apply_window_mp3(
float *
in,
float *win,
int *unused,
float *
out,
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)
135 apply_window(in + 32, win + 48, win + 640, sumb, sumd, 16);
137 SUM8(
MACS, suma[0], win + 32, in + 48);
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" \
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"
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)
162 :"
r"(&suma[0]), "
r"(&sumb[0]), "
r"(&sumc[0]), "
r"(&sumd[0])
168 float *out2 = out + 32 * incr;
173 *out = -suma[ j] + sumd[16-j];
174 *out2 = sumb[16-j] + sumc[ j];
181 SUM8(
MLSS, sum, win + 16 + 32, in + 32);
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) \
192 int align_end = count - (count & 3); \
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]; \
200 ff_four_imdct36_float_ ## CPU2(out, buf, in, win, tmpbuf); \
205 for (; j < count; j++) { \
209 int win_idx = (switch_point && j < 2) ? 0 : block_type; \
210 float *win = ff_mdct_win_float[win_idx + (4 & -(j & 1))]; \
212 ff_imdct36_float_ ## CPU1(out, buf, in, win); \
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)
232 for (j = 0; j < 4; j++) {
233 for (i = 0; i < 40; i ++) {
#define INLINE_SSE2(flags)
#define EXTERNAL_SSE(flags)
#define DECLARE_ALIGNED(n, t, v)
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)
float ff_mdct_win_float[8][MDCT_BUF_SIZE]
#define EXTERNAL_SSE3(flags)
#define EXTERNAL_SSSE3(flags)
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
void ff_imdct36_float_avx(float *out, float *buf, float *in, float *win)
#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
#define EXTERNAL_AVX(flags)
void(* imdct36_blocks_float)(float *out, float *buf, float *in, int count, int switch_point, int block_type)
static float mdct_win_sse[2][4][4 *40]
static void(*const apply_window[4])(AVFloatDSPContext *fdsp, SingleChannelElement *sce, const float *audio)
av_cold void ff_mpadsp_init_x86(MPADSPContext *s)
void ff_imdct36_float_sse3(float *out, float *buf, float *in, float *win)
void ff_four_imdct36_float_avx(float *out, float *buf, float *in, float *win, float *tmpbuf)
#define EXTERNAL_SSE2(flags)
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)
#define LOCAL_ALIGNED_16(t, v,...)
void ff_imdct36_float_sse2(float *out, float *buf, float *in, float *win)