@@ -952,6 +952,7 @@ CONFIG_LIST="

     mdct

     memalign_hack

     mlib

+    mpegaudiodsp

     network

     nonfree

     pic

@@ -1235,6 +1236,7 @@ symver_if_any="symver_asm_label symver_gnu_asm"

 dct_select="rdft"

 mdct_select="fft"

 rdft_select="fft"

+mpegaudiodsp_select="dct"

 # decoders / encoders / hardware accelerators

 aac_decoder_select="mdct sinewin"

@@ -1286,11 +1288,16 @@ ljpeg_encoder_select="aandct"

 loco_decoder_select="golomb"

 mjpeg_encoder_select="aandct"

 mlp_decoder_select="mlp_parser"

-mp1float_decoder_select="dct"

-mp2float_decoder_select="dct"

-mp3adufloat_decoder_select="dct"

-mp3float_decoder_select="dct"

-mp3on4float_decoder_select="dct"

+mp1_decoder_select="mpegaudiodsp"

+mp2_decoder_select="mpegaudiodsp"

+mp3adu_decoder_select="mpegaudiodsp"

+mp3_decoder_select="mpegaudiodsp"

+mp3on4_decoder_select="mpegaudiodsp"

+mp1float_decoder_select="mpegaudiodsp"

+mp2float_decoder_select="mpegaudiodsp"

+mp3adufloat_decoder_select="mpegaudiodsp"

+mp3float_decoder_select="mpegaudiodsp"

+mp3on4float_decoder_select="mpegaudiodsp"

 mpeg1video_encoder_select="aandct"

 mpeg2video_encoder_select="aandct"

 mpeg4_decoder_select="h263_decoder mpeg4video_parser"

@@ -1315,7 +1322,7 @@ nellymoser_encoder_select="mdct sinewin"

 png_decoder_select="zlib"

 png_encoder_select="zlib"

 qcelp_decoder_select="lsp"

-qdm2_decoder_select="mdct rdft"

+qdm2_decoder_select="mdct rdft mpegaudiodsp"

 ra_144_encoder_select="lpc"

 rv10_decoder_select="h263_decoder"

 rv10_encoder_select="h263_encoder"

@@ -40,6 +40,9 @@ OBJS-$(CONFIG_HUFFMAN)                 += huffman.o

 OBJS-$(CONFIG_LPC)                     += lpc.o

 OBJS-$(CONFIG_LSP)                     += lsp.o

 OBJS-$(CONFIG_MDCT)                    += mdct_fixed.o mdct_float.o

+OBJS-$(CONFIG_MPEGAUDIODSP)            += mpegaudiodsp.o                \

+                                          mpegaudiodsp_fixed.o          \

+                                          mpegaudiodsp_float.o

 RDFT-OBJS-$(CONFIG_HARDCODED_TABLES)   += sin_tables.o

 OBJS-$(CONFIG_RDFT)                    += rdft.o $(RDFT-OBJS-yes)

 OBJS-$(CONFIG_SINEWIN)                 += sinewin.o

@@ -29,6 +29,7 @@

 #include "avcodec.h"

 #include "get_bits.h"

 #include "dsputil.h"

+#include "mpegaudiodsp.h"

 #include "mpegaudio.h"

 #include "mpc.h"

@@ -51,7 +52,8 @@ static void mpc_synth(MPCContext *c, int16_t *out, int channels)

     for(ch = 0;  ch < channels; ch++){

         samples_ptr = samples + ch;

         for(i = 0; i < SAMPLES_PER_BAND; i++) {

-            ff_mpa_synth_filter_fixed(c->synth_buf[ch], &(c->synth_buf_offset[ch]),

+            ff_mpa_synth_filter_fixed(&c->mpadsp,

+                                c->synth_buf[ch], &(c->synth_buf_offset[ch]),

                                 ff_mpa_synth_window_fixed, &dither_state,

                                 samples_ptr, channels,

                                 c->sb_samples[ch][i]);

@@ -52,6 +52,7 @@ typedef struct {

 typedef struct {

     DSPContext dsp;

+    MPADSPContext mpadsp;

     GetBitContext gb;

     int IS, MSS, gapless;

     int lastframelen;

@@ -29,7 +29,7 @@

 #include "avcodec.h"

 #include "get_bits.h"

 #include "dsputil.h"

-#include "mpegaudio.h"

+#include "mpegaudiodsp.h"

 #include "libavutil/audioconvert.h"

 #include "mpc.h"

@@ -68,6 +68,7 @@ static av_cold int mpc7_decode_init(AVCodecContext * avctx)

     memset(c->oldDSCF, 0, sizeof(c->oldDSCF));

     av_lfg_init(&c->rnd, 0xDEADBEEF);

     dsputil_init(&c->dsp, avctx);

+    ff_mpadsp_init(&c->mpadsp);

     c->dsp.bswap_buf((uint32_t*)buf, (const uint32_t*)avctx->extradata, 4);

     ff_mpc_init();

     init_get_bits(&gb, buf, 128);

@@ -29,7 +29,7 @@

 #include "avcodec.h"

 #include "get_bits.h"

 #include "dsputil.h"

-#include "mpegaudio.h"

+#include "mpegaudiodsp.h"

 #include "libavutil/audioconvert.h"

 #include "mpc.h"

@@ -120,6 +120,7 @@ static av_cold int mpc8_decode_init(AVCodecContext * avctx)

     memset(c->oldDSCF, 0, sizeof(c->oldDSCF));

     av_lfg_init(&c->rnd, 0xDEADBEEF);

     dsputil_init(&c->dsp, avctx);

+    ff_mpadsp_init(&c->mpadsp);

     ff_mpc_init();

@@ -33,7 +33,6 @@

 #include "avcodec.h"

 #include "get_bits.h"

 #include "dsputil.h"

-#include "dct.h"

 /* max frame size, in samples */

 #define MPA_FRAME_SIZE 1152

@@ -69,7 +68,6 @@

 typedef float OUT_INT;

 #else

 typedef int16_t OUT_INT;

-#define OUT_SHIFT (WFRAC_BITS + FRAC_BITS - 15)

 #endif

 #if CONFIG_FLOAT

@@ -142,11 +140,7 @@ typedef struct MPADecodeContext {

     int dither_state;

     int error_recognition;

     AVCodecContext* avctx;

-#if CONFIG_FLOAT

-    DCTContext dct;

-#endif

-    void (*apply_window_mp3)(MPA_INT *synth_buf, MPA_INT *window,

-                             int *dither_state, OUT_INT *samples, int incr);

+    MPADSPContext mpadsp;

 } MPADecodeContext;

 /* layer 3 huffman tables */

@@ -158,22 +152,6 @@ typedef struct HuffTable {

 int ff_mpa_l2_select_table(int bitrate, int nb_channels, int freq, int lsf);

 int ff_mpa_decode_header(AVCodecContext *avctx, uint32_t head, int *sample_rate, int *channels, int *frame_size, int *bitrate);

-extern MPA_INT ff_mpa_synth_window_fixed[];

-void ff_mpa_synth_init_fixed(MPA_INT *window);

-void ff_mpa_synth_filter_fixed(MPA_INT *synth_buf_ptr, int *synth_buf_offset,

-                         MPA_INT *window, int *dither_state,

-                         OUT_INT *samples, int incr,

-                         INTFLOAT sb_samples[SBLIMIT]);

-

-void ff_mpa_synth_init_float(MPA_INT *window);

-void ff_mpa_synth_filter_float(MPADecodeContext *s,

-                         MPA_INT *synth_buf_ptr, int *synth_buf_offset,

-                         MPA_INT *window, int *dither_state,

-                         OUT_INT *samples, int incr,

-                         INTFLOAT sb_samples[SBLIMIT]);

-

-void ff_mpegaudiodec_init_mmx(MPADecodeContext *s);

-void ff_mpegaudiodec_init_altivec(MPADecodeContext *s);

 /* fast header check for resync */

 static inline int ff_mpa_check_header(uint32_t header){

@@ -29,7 +29,7 @@

 #include "get_bits.h"

 #include "dsputil.h"

 #include "mathops.h"

-#include "dct32.h"

+#include "mpegaudiodsp.h"

 /*

  * TODO:

@@ -68,8 +68,6 @@

 #include "mpegaudiodectab.h"

 static void RENAME(compute_antialias)(MPADecodeContext *s, GranuleDef *g);

-static void apply_window_mp3_c(MPA_INT *synth_buf, MPA_INT *window,

-                               int *dither_state, OUT_INT *samples, int incr);

 /* vlc structure for decoding layer 3 huffman tables */

 static VLC huff_vlc[16];

@@ -119,8 +117,6 @@ static const int32_t scale_factor_mult2[3][3] = {

     SCALE_GEN(4.0 / 9.0), /* 9 steps */

 };

-DECLARE_ALIGNED(16, MPA_INT, RENAME(ff_mpa_synth_window))[512+256];

-

 /**

  * Convert region offsets to region sizes and truncate

  * size to big_values.

@@ -259,14 +255,8 @@ static av_cold int decode_init(AVCodecContext * avctx)

     int i, j, k;

     s->avctx = avctx;

-    s->apply_window_mp3 = apply_window_mp3_c;

-#if HAVE_MMX && CONFIG_FLOAT

-    ff_mpegaudiodec_init_mmx(s);

-#endif

-#if CONFIG_FLOAT

-    ff_dct_init(&s->dct, 5, DCT_II);

-#endif

-    if (HAVE_ALTIVEC && CONFIG_FLOAT) ff_mpegaudiodec_init_altivec(s);

+

+    ff_mpadsp_init(&s->mpadsp);

     avctx->sample_fmt= OUT_FMT;

     s->error_recognition= avctx->error_recognition;

@@ -461,183 +451,6 @@ static av_cold int decode_init(AVCodecContext * avctx)

     return 0;

 }

-

-#if CONFIG_FLOAT

-static inline float round_sample(float *sum)

-{

-    float sum1=*sum;

-    *sum = 0;

-    return sum1;

-}

-

-/* signed 16x16 -> 32 multiply add accumulate */

-#define MACS(rt, ra, rb) rt+=(ra)*(rb)

-

-/* signed 16x16 -> 32 multiply */

-#define MULS(ra, rb) ((ra)*(rb))

-

-#define MLSS(rt, ra, rb) rt-=(ra)*(rb)

-

-#else

-

-static inline int round_sample(int64_t *sum)

-{

-    int sum1;

-    sum1 = (int)((*sum) >> OUT_SHIFT);

-    *sum &= (1<<OUT_SHIFT)-1;

-    return av_clip_int16(sum1);

-}

-

-#   define MULS(ra, rb) MUL64(ra, rb)

-#   define MACS(rt, ra, rb) MAC64(rt, ra, rb)

-#   define MLSS(rt, ra, rb) MLS64(rt, ra, rb)

-#endif

-

-#define SUM8(op, sum, w, p)               \

-{                                         \

-    op(sum, (w)[0 * 64], (p)[0 * 64]);    \

-    op(sum, (w)[1 * 64], (p)[1 * 64]);    \

-    op(sum, (w)[2 * 64], (p)[2 * 64]);    \

-    op(sum, (w)[3 * 64], (p)[3 * 64]);    \

-    op(sum, (w)[4 * 64], (p)[4 * 64]);    \

-    op(sum, (w)[5 * 64], (p)[5 * 64]);    \

-    op(sum, (w)[6 * 64], (p)[6 * 64]);    \

-    op(sum, (w)[7 * 64], (p)[7 * 64]);    \

-}

-

-#define SUM8P2(sum1, op1, sum2, op2, w1, w2, p) \

-{                                               \

-    INTFLOAT tmp;\

-    tmp = p[0 * 64];\

-    op1(sum1, (w1)[0 * 64], tmp);\

-    op2(sum2, (w2)[0 * 64], tmp);\

-    tmp = p[1 * 64];\

-    op1(sum1, (w1)[1 * 64], tmp);\

-    op2(sum2, (w2)[1 * 64], tmp);\

-    tmp = p[2 * 64];\

-    op1(sum1, (w1)[2 * 64], tmp);\

-    op2(sum2, (w2)[2 * 64], tmp);\

-    tmp = p[3 * 64];\

-    op1(sum1, (w1)[3 * 64], tmp);\

-    op2(sum2, (w2)[3 * 64], tmp);\

-    tmp = p[4 * 64];\

-    op1(sum1, (w1)[4 * 64], tmp);\

-    op2(sum2, (w2)[4 * 64], tmp);\

-    tmp = p[5 * 64];\

-    op1(sum1, (w1)[5 * 64], tmp);\

-    op2(sum2, (w2)[5 * 64], tmp);\

-    tmp = p[6 * 64];\

-    op1(sum1, (w1)[6 * 64], tmp);\

-    op2(sum2, (w2)[6 * 64], tmp);\

-    tmp = p[7 * 64];\

-    op1(sum1, (w1)[7 * 64], tmp);\

-    op2(sum2, (w2)[7 * 64], tmp);\

-}

-

-void av_cold RENAME(ff_mpa_synth_init)(MPA_INT *window)

-{

-    int i, j;

-

-    /* max = 18760, max sum over all 16 coefs : 44736 */

-    for(i=0;i<257;i++) {

-        INTFLOAT v;

-        v = ff_mpa_enwindow[i];

-#if CONFIG_FLOAT

-        v *= 1.0 / (1LL<<(16 + FRAC_BITS));

-#endif

-        window[i] = v;

-        if ((i & 63) != 0)

-            v = -v;

-        if (i != 0)

-            window[512 - i] = v;

-    }

-

-    // Needed for avoiding shuffles in ASM implementations

-    for(i=0; i < 8; i++)

-        for(j=0; j < 16; j++)

-            window[512+16*i+j] = window[64*i+32-j];

-

-    for(i=0; i < 8; i++)

-        for(j=0; j < 16; j++)

-            window[512+128+16*i+j] = window[64*i+48-j];

-}

-

-static void apply_window_mp3_c(MPA_INT *synth_buf, MPA_INT *window,

-                               int *dither_state, OUT_INT *samples, int incr)

-{

-    register const MPA_INT *w, *w2, *p;

-    int j;

-    OUT_INT *samples2;

-#if CONFIG_FLOAT

-    float sum, sum2;

-#else

-    int64_t sum, sum2;

-#endif

-

-    /* copy to avoid wrap */

-    memcpy(synth_buf + 512, synth_buf, 32 * sizeof(*synth_buf));

-

-    samples2 = samples + 31 * incr;

-    w = window;

-    w2 = window + 31;

-

-    sum = *dither_state;

-    p = synth_buf + 16;

-    SUM8(MACS, sum, w, p);

-    p = synth_buf + 48;

-    SUM8(MLSS, sum, w + 32, p);

-    *samples = round_sample(&sum);

-    samples += incr;

-    w++;

-

-    /* we calculate two samples at the same time to avoid one memory

-       access per two sample */

-    for(j=1;j<16;j++) {

-        sum2 = 0;

-        p = synth_buf + 16 + j;

-        SUM8P2(sum, MACS, sum2, MLSS, w, w2, p);

-        p = synth_buf + 48 - j;

-        SUM8P2(sum, MLSS, sum2, MLSS, w + 32, w2 + 32, p);

-

-        *samples = round_sample(&sum);

-        samples += incr;

-        sum += sum2;

-        *samples2 = round_sample(&sum);

-        samples2 -= incr;

-        w++;

-        w2--;

-    }

-

-    p = synth_buf + 32;

-    SUM8(MLSS, sum, w + 32, p);

-    *samples = round_sample(&sum);

-    *dither_state= sum;

-}

-

-

-/* 32 sub band synthesis filter. Input: 32 sub band samples, Output:

-   32 samples. */

-/* XXX: optimize by avoiding ring buffer usage */

-#if !CONFIG_FLOAT

-void ff_mpa_synth_filter_fixed(MPA_INT *synth_buf_ptr, int *synth_buf_offset,

-                         MPA_INT *window, int *dither_state,

-                         OUT_INT *samples, int incr,

-                         INTFLOAT sb_samples[SBLIMIT])

-{

-    register MPA_INT *synth_buf;

-    int offset;

-

-    offset = *synth_buf_offset;

-    synth_buf = synth_buf_ptr + offset;

-

-    ff_dct32_fixed(synth_buf, sb_samples);

-    apply_window_mp3_c(synth_buf, window, dither_state, samples, incr);

-

-    offset = (offset - 32) & 511;

-    *synth_buf_offset = offset;

-}

-#endif

-

 #define C3 FIXHR(0.86602540378443864676/2)

 /* 0.5 / cos(pi*(2*i+1)/36) */

@@ -1915,9 +1728,7 @@ static int mp_decode_frame(MPADecodeContext *s,

         samples_ptr = samples + ch;

         for(i=0;i<nb_frames;i++) {

             RENAME(ff_mpa_synth_filter)(

-#if CONFIG_FLOAT

-                         s,

-#endif

+                         &s->mpadsp,

                          s->synth_buf[ch], &(s->synth_buf_offset[ch]),

                          RENAME(ff_mpa_synth_window), &s->dither_state,

                          samples_ptr, s->nb_channels,

@@ -22,25 +22,6 @@

 #define CONFIG_FLOAT 1

 #include "mpegaudiodec.c"

-void ff_mpa_synth_filter_float(MPADecodeContext *s, float *synth_buf_ptr,

-                               int *synth_buf_offset,

-                               float *window, int *dither_state,

-                               float *samples, int incr,

-                               float sb_samples[SBLIMIT])

-{

-    float *synth_buf;

-    int offset;

-

-    offset = *synth_buf_offset;

-    synth_buf = synth_buf_ptr + offset;

-

-    s->dct.dct32(synth_buf, sb_samples);

-    s->apply_window_mp3(synth_buf, window, dither_state, samples, incr);

-

-    offset = (offset - 32) & 511;

-    *synth_buf_offset = offset;

-}

-

 static void compute_antialias_float(MPADecodeContext *s,

                               GranuleDef *g)

 {

new file mode 100644

@@ -0,0 +1,40 @@

+/*

+ * Copyright (c) 2011 Mans Rullgard

+ *

+ * This file is part of Libav.

+ *

+ * Libav 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.

+ *

+ * Libav 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 Libav; if not, write to the Free Software

+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

+ */

+

+#include "config.h"

+#include "mpegaudiodsp.h"

+#include "dct.h"

+#include "dct32.h"

+

+void ff_mpadsp_init(MPADSPContext *s)

+{

+    DCTContext dct;

+

+    ff_dct_init(&dct, 5, DCT_II);

+

+    s->apply_window_float = ff_mpadsp_apply_window_float;

+    s->apply_window_fixed = ff_mpadsp_apply_window_fixed;

+

+    s->dct32_float = dct.dct32;

+    s->dct32_fixed = ff_dct32_fixed;

+

+    if (HAVE_MMX)     ff_mpadsp_init_mmx(s);

+    if (HAVE_ALTIVEC) ff_mpadsp_init_altivec(s);

+}

new file mode 100644

@@ -0,0 +1,63 @@

+/*

+ * This file is part of Libav.

+ *

+ * Libav 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.

+ *

+ * Libav 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 Libav; if not, write to the Free Software

+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

+ */

+

+#ifndef AVCODEC_MPEGAUDIODSP_H

+#define AVCODEC_MPEGAUDIODSP_H

+

+#include <stdint.h>

+

+typedef struct MPADSPContext {

+    void (*apply_window_float)(float *synth_buf, float *window,

+                               int *dither_state, float *samples, int incr);

+    void (*apply_window_fixed)(int32_t *synth_buf, int32_t *window,

+                               int *dither_state, int16_t *samples, int incr);

+    void (*dct32_float)(float *dst, const float *src);

+    void (*dct32_fixed)(int *dst, const int *src);

+} MPADSPContext;

+

+void ff_mpadsp_init(MPADSPContext *s);

+

+extern int32_t ff_mpa_synth_window_fixed[];

+extern float   ff_mpa_synth_window_float[];

+

+void ff_mpa_synth_filter_fixed(MPADSPContext *s,

+                               int32_t *synth_buf_ptr, int *synth_buf_offset,

+                               int32_t *window, int *dither_state,

+                               int16_t *samples, int incr,

+                               int *sb_samples);

+

+void ff_mpa_synth_filter_float(MPADSPContext *s,

+                               float *synth_buf_ptr, int *synth_buf_offset,

+                               float *window, int *dither_state,

+                               float *samples, int incr,

+                               float *sb_samples);

+

+void ff_mpadsp_init_mmx(MPADSPContext *s);

+void ff_mpadsp_init_altivec(MPADSPContext *s);

+

+void ff_mpa_synth_init_float(float *window);

+void ff_mpa_synth_init_fixed(int32_t *window);

+

+void ff_mpadsp_apply_window_float(float *synth_buf, float *window,

+                                  int *dither_state, float *samples,

+                                  int incr);

+void ff_mpadsp_apply_window_fixed(int32_t *synth_buf, int32_t *window,

+                                  int *dither_state, int16_t *samples,

+                                  int incr);

+

+#endif

new file mode 100644

@@ -0,0 +1,20 @@

+/*

+ * This file is part of Libav.

+ *

+ * Libav 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.

+ *

+ * Libav 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 Libav; if not, write to the Free Software

+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

+ */

+

+#define CONFIG_FLOAT 0

+#include "mpegaudiodsp_template.c"

new file mode 100644

@@ -0,0 +1,20 @@

+/*

+ * This file is part of Libav.

+ *

+ * Libav 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.

+ *

+ * Libav 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 Libav; if not, write to the Free Software

+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

+ */

+

+#define CONFIG_FLOAT 1

+#include "mpegaudiodsp_template.c"

new file mode 100644

@@ -0,0 +1,205 @@

+/*

+ * Copyright (c) 2001, 2002 Fabrice Bellard

+ *

+ * This file is part of Libav.

+ *

+ * Libav 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.

+ *

+ * Libav 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 Libav; if not, write to the Free Software

+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

+ */

+

+#include <stdint.h>

+

+#include "libavutil/mem.h"

+#include "dct32.h"

+#include "mathops.h"

+#include "mpegaudiodsp.h"

+#include "mpegaudio.h"

+#include "mpegaudiodata.h"

+

+#if CONFIG_FLOAT

+#define RENAME(n) n##_float

+

+static inline float round_sample(float *sum)

+{

+    float sum1=*sum;

+    *sum = 0;

+    return sum1;

+}

+

+#define MACS(rt, ra, rb) rt+=(ra)*(rb)

+#define MULS(ra, rb) ((ra)*(rb))

+#define MLSS(rt, ra, rb) rt-=(ra)*(rb)

+

+#else

+

+#define RENAME(n) n##_fixed

+#define OUT_SHIFT (WFRAC_BITS + FRAC_BITS - 15)

+

+static inline int round_sample(int64_t *sum)

+{

+    int sum1;

+    sum1 = (int)((*sum) >> OUT_SHIFT);

+    *sum &= (1<<OUT_SHIFT)-1;

+    return av_clip_int16(sum1);

+}

+

+#   define MULS(ra, rb) MUL64(ra, rb)

+#   define MACS(rt, ra, rb) MAC64(rt, ra, rb)

+#   define MLSS(rt, ra, rb) MLS64(rt, ra, rb)

+#endif

+

+DECLARE_ALIGNED(16, MPA_INT, RENAME(ff_mpa_synth_window))[512+256];

+

+#define SUM8(op, sum, w, p)               \

+{                                         \

+    op(sum, (w)[0 * 64], (p)[0 * 64]);    \

+    op(sum, (w)[1 * 64], (p)[1 * 64]);    \

+    op(sum, (w)[2 * 64], (p)[2 * 64]);    \

+    op(sum, (w)[3 * 64], (p)[3 * 64]);    \

+    op(sum, (w)[4 * 64], (p)[4 * 64]);    \

+    op(sum, (w)[5 * 64], (p)[5 * 64]);    \

+    op(sum, (w)[6 * 64], (p)[6 * 64]);    \

+    op(sum, (w)[7 * 64], (p)[7 * 64]);    \

+}

+

+#define SUM8P2(sum1, op1, sum2, op2, w1, w2, p) \

+{                                               \

+    INTFLOAT tmp;\

+    tmp = p[0 * 64];\

+    op1(sum1, (w1)[0 * 64], tmp);\

+    op2(sum2, (w2)[0 * 64], tmp);\

+    tmp = p[1 * 64];\

+    op1(sum1, (w1)[1 * 64], tmp);\

+    op2(sum2, (w2)[1 * 64], tmp);\

+    tmp = p[2 * 64];\

+    op1(sum1, (w1)[2 * 64], tmp);\

+    op2(sum2, (w2)[2 * 64], tmp);\

+    tmp = p[3 * 64];\

+    op1(sum1, (w1)[3 * 64], tmp);\

+    op2(sum2, (w2)[3 * 64], tmp);\

+    tmp = p[4 * 64];\

+    op1(sum1, (w1)[4 * 64], tmp);\

+    op2(sum2, (w2)[4 * 64], tmp);\

+    tmp = p[5 * 64];\

+    op1(sum1, (w1)[5 * 64], tmp);\

+    op2(sum2, (w2)[5 * 64], tmp);\

+    tmp = p[6 * 64];\

+    op1(sum1, (w1)[6 * 64], tmp);\

+    op2(sum2, (w2)[6 * 64], tmp);\

+    tmp = p[7 * 64];\

+    op1(sum1, (w1)[7 * 64], tmp);\

+    op2(sum2, (w2)[7 * 64], tmp);\

+}

+

+void RENAME(ff_mpadsp_apply_window)(MPA_INT *synth_buf, MPA_INT *window,

+                                  int *dither_state, OUT_INT *samples,

+                                  int incr)

+{

+    register const MPA_INT *w, *w2, *p;

+    int j;

+    OUT_INT *samples2;

+#if CONFIG_FLOAT

+    float sum, sum2;

+#else

+    int64_t sum, sum2;

+#endif

+

+    /* copy to avoid wrap */

+    memcpy(synth_buf + 512, synth_buf, 32 * sizeof(*synth_buf));

+

+    samples2 = samples + 31 * incr;

+    w = window;

+    w2 = window + 31;

+

+    sum = *dither_state;

+    p = synth_buf + 16;

+    SUM8(MACS, sum, w, p);

+    p = synth_buf + 48;

+    SUM8(MLSS, sum, w + 32, p);

+    *samples = round_sample(&sum);

+    samples += incr;

+    w++;

+

+    /* we calculate two samples at the same time to avoid one memory

+       access per two sample */

+    for(j=1;j<16;j++) {

+        sum2 = 0;

+        p = synth_buf + 16 + j;

+        SUM8P2(sum, MACS, sum2, MLSS, w, w2, p);

+        p = synth_buf + 48 - j;

+        SUM8P2(sum, MLSS, sum2, MLSS, w + 32, w2 + 32, p);

+

+        *samples = round_sample(&sum);

+        samples += incr;

+        sum += sum2;

+        *samples2 = round_sample(&sum);

+        samples2 -= incr;

+        w++;

+        w2--;

+    }

+

+    p = synth_buf + 32;

+    SUM8(MLSS, sum, w + 32, p);

+    *samples = round_sample(&sum);

+    *dither_state= sum;

+}

+

+/* 32 sub band synthesis filter. Input: 32 sub band samples, Output:

+   32 samples. */

+void RENAME(ff_mpa_synth_filter)(MPADSPContext *s, MPA_INT *synth_buf_ptr,

+                                 int *synth_buf_offset,

+                                 MPA_INT *window, int *dither_state,

+                                 OUT_INT *samples, int incr,

+                                 MPA_INT *sb_samples)

+{

+    MPA_INT *synth_buf;

+    int offset;

+

+    offset = *synth_buf_offset;

+    synth_buf = synth_buf_ptr + offset;

+

+    s->RENAME(dct32)(synth_buf, sb_samples);

+    s->RENAME(apply_window)(synth_buf, window, dither_state, samples, incr);

+

+    offset = (offset - 32) & 511;

+    *synth_buf_offset = offset;

+}

+

+void av_cold RENAME(ff_mpa_synth_init)(MPA_INT *window)

+{

+    int i, j;

+

+    /* max = 18760, max sum over all 16 coefs : 44736 */

+    for(i=0;i<257;i++) {

+        INTFLOAT v;

+        v = ff_mpa_enwindow[i];

+#if CONFIG_FLOAT

+        v *= 1.0 / (1LL<<(16 + FRAC_BITS));

+#endif

+        window[i] = v;

+        if ((i & 63) != 0)

+            v = -v;

+        if (i != 0)

+            window[512 - i] = v;

+    }

+

+    // Needed for avoiding shuffles in ASM implementations

+    for(i=0; i < 8; i++)

+        for(j=0; j < 16; j++)

+            window[512+16*i+j] = window[64*i+32-j];

+

+    for(i=0; i < 8; i++)

+        for(j=0; j < 16; j++)

+            window[512+128+16*i+j] = window[64*i+48-j];

+}

@@ -21,9 +21,8 @@

 #include "dsputil_altivec.h"

 #include "util_altivec.h"

-

-#define CONFIG_FLOAT 1

-#include "libavcodec/mpegaudio.h"

+#include "libavcodec/dsputil.h"

+#include "libavcodec/mpegaudiodsp.h"

 #define MACS(rt, ra, rb) rt+=(ra)*(rb)

 #define MLSS(rt, ra, rb) rt-=(ra)*(rb)

@@ -124,7 +123,7 @@ static void apply_window_mp3(float *in, float *win, int *unused, float *out,

     *out = sum;

 }

-void ff_mpegaudiodec_init_altivec(MPADecodeContext *s)

+void ff_mpadsp_init_altivec(MPADSPContext *s)

 {

-    s->apply_window_mp3 = apply_window_mp3;

+    s->apply_window_float = apply_window_mp3;

 }

@@ -39,6 +39,7 @@

 #include "get_bits.h"

 #include "dsputil.h"

 #include "rdft.h"

+#include "mpegaudiodsp.h"

 #include "mpegaudio.h"

 #include "qdm2data.h"

@@ -170,6 +171,7 @@ typedef struct {

     float output_buffer[1024];

     /// Synthesis filter

+    MPADSPContext mpadsp;

     DECLARE_ALIGNED(16, MPA_INT, synth_buf)[MPA_MAX_CHANNELS][512*2];

     int synth_buf_offset[MPA_MAX_CHANNELS];

     DECLARE_ALIGNED(16, int32_t, sb_samples)[MPA_MAX_CHANNELS][128][SBLIMIT];

@@ -1616,7 +1618,8 @@ static void qdm2_synthesis_filter (QDM2Context *q, int index)

         OUT_INT *samples_ptr = samples + ch;

         for (i = 0; i < 8; i++) {

-            ff_mpa_synth_filter_fixed(q->synth_buf[ch], &(q->synth_buf_offset[ch]),

+            ff_mpa_synth_filter_fixed(&q->mpadsp,

+                q->synth_buf[ch], &(q->synth_buf_offset[ch]),

                 ff_mpa_synth_window_fixed, &dither_state,

                 samples_ptr, q->nb_channels,

                 q->sb_samples[ch][(8 * index) + i]);

@@ -1863,6 +1866,7 @@ static av_cold int qdm2_decode_init(AVCodecContext *avctx)

     }

     ff_rdft_init(&s->rdft_ctx, s->fft_order, IDFT_C2R);

+    ff_mpadsp_init(&s->mpadsp);

     qdm2_init(s);

@@ -21,9 +21,8 @@

 #include "libavutil/cpu.h"

 #include "libavutil/x86_cpu.h"

-

-#define CONFIG_FLOAT 1

-#include "libavcodec/mpegaudio.h"

+#include "libavcodec/dsputil.h"

+#include "libavcodec/mpegaudiodsp.h"

 #define MACS(rt, ra, rb) rt+=(ra)*(rb)

 #define MLSS(rt, ra, rb) rt-=(ra)*(rb)

@@ -148,11 +147,11 @@ static void apply_window_mp3(float *in, float *win, int *unused, float *out,

     *out = sum;

 }

-void ff_mpegaudiodec_init_mmx(MPADecodeContext *s)

+void ff_mpadsp_init_mmx(MPADSPContext *s)