Merge commit '42d324694883cdf1fff1612ac70fa403692a1ad4'

* commit '42d324694883cdf1fff1612ac70fa403692a1ad4':
  floatdsp: move vector_fmul_reverse from dsputil to avfloatdsp.

Conflicts:
	libavcodec/arm/dsputil_init_vfp.c
	libavcodec/arm/dsputil_vfp.S
	libavcodec/dsputil.c
	libavcodec/ppc/float_altivec.c
	libavcodec/x86/dsputil.asm
	libavutil/x86/float_dsp.asm

Merged-by: Michael Niedermayer <michaelni@gmx.at>

libavcodec/aacdec.c

@@ -2131,9 +2131,9 @@ static void windowing_and_mdct_ltp(AACContext *ac, float *out,
         ac->fdsp.vector_fmul(in + 448, in + 448, swindow_prev, 128);
     }
     if (ics->window_sequence[0] != LONG_START_SEQUENCE) {
-        ac->dsp.vector_fmul_reverse(in + 1024, in + 1024, lwindow, 1024);
+        ac->fdsp.vector_fmul_reverse(in + 1024, in + 1024, lwindow, 1024);
     } else {
-        ac->dsp.vector_fmul_reverse(in + 1024 + 448, in + 1024 + 448, swindow, 128);
+        ac->fdsp.vector_fmul_reverse(in + 1024 + 448, in + 1024 + 448, swindow, 128);
         memset(in + 1024 + 576, 0, 448 * sizeof(float));
     }
     ac->mdct_ltp.mdct_calc(&ac->mdct_ltp, out, in);
@@ -2186,17 +2186,17 @@ static void update_ltp(AACContext *ac, SingleChannelElement *sce)
     if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
         memcpy(saved_ltp,       saved, 512 * sizeof(float));
         memset(saved_ltp + 576, 0,     448 * sizeof(float));
-        ac->dsp.vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960,     &swindow[64],      64);
+        ac->fdsp.vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960,     &swindow[64],      64);
         for (i = 0; i < 64; i++)
             saved_ltp[i + 512] = ac->buf_mdct[1023 - i] * swindow[63 - i];
     } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
         memcpy(saved_ltp,       ac->buf_mdct + 512, 448 * sizeof(float));
         memset(saved_ltp + 576, 0,                  448 * sizeof(float));
-        ac->dsp.vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960,     &swindow[64],      64);
+        ac->fdsp.vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960,     &swindow[64],      64);
         for (i = 0; i < 64; i++)
             saved_ltp[i + 512] = ac->buf_mdct[1023 - i] * swindow[63 - i];
     } else { // LONG_STOP or ONLY_LONG
-        ac->dsp.vector_fmul_reverse(saved_ltp,       ac->buf_mdct + 512,     &lwindow[512],     512);
+        ac->fdsp.vector_fmul_reverse(saved_ltp,       ac->buf_mdct + 512,     &lwindow[512],     512);
         for (i = 0; i < 512; i++)
             saved_ltp[i + 512] = ac->buf_mdct[1023 - i] * lwindow[511 - i];
     }

libavcodec/aacenc.c

@@ -183,7 +183,7 @@ static void put_audio_specific_config(AVCodecContext *avctx)
 }
 
 #define WINDOW_FUNC(type) \
-static void apply_ ##type ##_window(DSPContext *dsp, AVFloatDSPContext *fdsp, \
+static void apply_ ##type ##_window(AVFloatDSPContext *fdsp, \
                                     SingleChannelElement *sce, \
                                     const float *audio)
 
@@ -193,8 +193,8 @@ WINDOW_FUNC(only_long)
     const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
     float *out = sce->ret_buf;
 
-    fdsp->vector_fmul       (out,        audio,        lwindow, 1024);
-    dsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024);
+    fdsp->vector_fmul        (out,        audio,        lwindow, 1024);
+    fdsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024);
 }
 
 WINDOW_FUNC(long_start)
@@ -205,7 +205,7 @@ WINDOW_FUNC(long_start)
 
     fdsp->vector_fmul(out, audio, lwindow, 1024);
     memcpy(out + 1024, audio + 1024, sizeof(out[0]) * 448);
-    dsp->vector_fmul_reverse(out + 1024 + 448, audio + 1024 + 448, swindow, 128);
+    fdsp->vector_fmul_reverse(out + 1024 + 448, audio + 1024 + 448, swindow, 128);
     memset(out + 1024 + 576, 0, sizeof(out[0]) * 448);
 }
 
@@ -218,7 +218,7 @@ WINDOW_FUNC(long_stop)
     memset(out, 0, sizeof(out[0]) * 448);
     fdsp->vector_fmul(out + 448, audio + 448, swindow, 128);
     memcpy(out + 576, audio + 576, sizeof(out[0]) * 448);
-    dsp->vector_fmul_reverse(out + 1024, audio + 1024, lwindow, 1024);
+    fdsp->vector_fmul_reverse(out + 1024, audio + 1024, lwindow, 1024);
 }
 
 WINDOW_FUNC(eight_short)
@@ -230,15 +230,15 @@ WINDOW_FUNC(eight_short)
     int w;
 
     for (w = 0; w < 8; w++) {
-        fdsp->vector_fmul       (out, in, w ? pwindow : swindow, 128);
+        fdsp->vector_fmul        (out, in, w ? pwindow : swindow, 128);
         out += 128;
         in  += 128;
-        dsp->vector_fmul_reverse(out, in, swindow, 128);
+        fdsp->vector_fmul_reverse(out, in, swindow, 128);
         out += 128;
     }
 }
 
-static void (*const apply_window[4])(DSPContext *dsp, AVFloatDSPContext *fdsp,
+static void (*const apply_window[4])(AVFloatDSPContext *fdsp,
                                      SingleChannelElement *sce,
                                      const float *audio) = {
     [ONLY_LONG_SEQUENCE]   = apply_only_long_window,
@@ -253,7 +253,7 @@ static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce,
     int i;
     float *output = sce->ret_buf;
 
-    apply_window[sce->ics.window_sequence[0]](&s->dsp, &s->fdsp, sce, audio);
+    apply_window[sce->ics.window_sequence[0]](&s->fdsp, sce, audio);
 
     if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE)
         s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
@@ -692,7 +692,6 @@ static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
 {
     int ret = 0;
 
-    ff_dsputil_init(&s->dsp, avctx);
     avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
 
     // window init

libavcodec/aacenc.h

@@ -61,7 +61,6 @@ typedef struct AACEncContext {
     PutBitContext pb;
     FFTContext mdct1024;                         ///< long (1024 samples) frame transform context
     FFTContext mdct128;                          ///< short (128 samples) frame transform context
-    DSPContext  dsp;
     AVFloatDSPContext fdsp;
     float *planar_samples[6];                    ///< saved preprocessed input
 

libavcodec/aacsbr.c

@@ -1156,7 +1156,7 @@ static void sbr_dequant(SpectralBandReplication *sbr, int id_aac)
  * @param   x       pointer to the beginning of the first sample window
  * @param   W       array of complex-valued samples split into subbands
  */
-static void sbr_qmf_analysis(DSPContext *dsp, FFTContext *mdct,
+static void sbr_qmf_analysis(AVFloatDSPContext *dsp, FFTContext *mdct,
                              SBRDSPContext *sbrdsp, const float *in, float *x,
                              float z[320], float W[2][32][32][2], int buf_idx)
 {
@@ -1668,7 +1668,7 @@ void ff_sbr_apply(AACContext *ac, SpectralBandReplication *sbr, int id_aac,
     }
     for (ch = 0; ch < nch; ch++) {
         /* decode channel */
-        sbr_qmf_analysis(&ac->dsp, &sbr->mdct_ana, &sbr->dsp, ch ? R : L, sbr->data[ch].analysis_filterbank_samples,
+        sbr_qmf_analysis(&ac->fdsp, &sbr->mdct_ana, &sbr->dsp, ch ? R : L, sbr->data[ch].analysis_filterbank_samples,
                          (float*)sbr->qmf_filter_scratch,
                          sbr->data[ch].W, sbr->data[ch].Ypos);
         sbr_lf_gen(ac, sbr, sbr->X_low, sbr->data[ch].W, sbr->data[ch].Ypos);

libavcodec/arm/Makefile

@@ -56,9 +56,6 @@ ARMV6-OBJS                             += arm/dsputil_init_armv6.o      \
 
 VFP-OBJS-$(HAVE_ARMV6)                 += arm/fmtconvert_vfp.o
 
-VFP-OBJS                               += arm/dsputil_vfp.o             \
-                                          arm/dsputil_init_vfp.o        \
-
 NEON-OBJS-$(CONFIG_FFT)                += arm/fft_neon.o                \
                                           arm/fft_fixed_neon.o          \
 

libavcodec/arm/dsputil_init_arm.c

@@ -121,6 +121,5 @@ void ff_dsputil_init_arm(DSPContext* c, AVCodecContext *avctx)
 
     if (have_armv5te(cpu_flags)) ff_dsputil_init_armv5te(c, avctx);
     if (have_armv6(cpu_flags))   ff_dsputil_init_armv6(c, avctx);
-    if (have_vfp(cpu_flags))     ff_dsputil_init_vfp(c, avctx);
     if (have_neon(cpu_flags))    ff_dsputil_init_neon(c, avctx);
 }

libavcodec/arm/dsputil_init_neon.c

@@ -144,8 +144,6 @@ void ff_avg_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int);
 
 void ff_butterflies_float_neon(float *v1, float *v2, int len);
 float ff_scalarproduct_float_neon(const float *v1, const float *v2, int len);
-void ff_vector_fmul_reverse_neon(float *dst, const float *src0,
-                                 const float *src1, int len);
 
 void ff_vector_clipf_neon(float *dst, const float *src, float min, float max,
                           int len);
@@ -298,7 +296,6 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
 
     c->butterflies_float          = ff_butterflies_float_neon;
     c->scalarproduct_float        = ff_scalarproduct_float_neon;
-    c->vector_fmul_reverse        = ff_vector_fmul_reverse_neon;
     c->vector_clipf               = ff_vector_clipf_neon;
     c->vector_clip_int32          = ff_vector_clip_int32_neon;
 

libavcodec/arm/dsputil_init_vfp.c

@@ -1,30 +0,0 @@
-/*
- * Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg 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.
- *
- * FFmpeg 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 FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#include "libavcodec/dsputil.h"
-#include "dsputil_arm.h"
-
-void ff_vector_fmul_reverse_vfp(float *dst, const float *src0,
-                                const float *src1, int len);
-
-void ff_dsputil_init_vfp(DSPContext* c, AVCodecContext *avctx)
-{
-    c->vector_fmul_reverse = ff_vector_fmul_reverse_vfp;
-}

libavcodec/arm/dsputil_neon.S

@@ -556,30 +556,6 @@ NOVFP   vmov.32         r0,  d0[0]
         bx              lr
 endfunc
 
-function ff_vector_fmul_reverse_neon, export=1
-        add             r2,  r2,  r3,  lsl #2
-        sub             r2,  r2,  #32
-        mov             r12, #-32
-        vld1.32         {q0-q1},  [r1,:128]!
-        vld1.32         {q2-q3},  [r2,:128], r12
-1:      pld             [r1, #32]
-        vrev64.32       q3,  q3
-        vmul.f32        d16, d0,  d7
-        vmul.f32        d17, d1,  d6
-        pld             [r2, #-32]
-        vrev64.32       q2,  q2
-        vmul.f32        d18, d2,  d5
-        vmul.f32        d19, d3,  d4
-        subs            r3,  r3,  #8
-        beq             2f
-        vld1.32         {q0-q1},  [r1,:128]!
-        vld1.32         {q2-q3},  [r2,:128], r12
-        vst1.32         {q8-q9},  [r0,:128]!
-        b               1b
-2:      vst1.32         {q8-q9},  [r0,:128]!
-        bx              lr
-endfunc
-
 function ff_vector_clipf_neon, export=1
 VFP     vdup.32         q1,  d0[1]
 VFP     vdup.32         q0,  d0[0]

libavcodec/arm/dsputil_vfp.S

@@ -1,106 +0,0 @@
-/*
- * Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg 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.
- *
- * FFmpeg 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 FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#include "config.h"
-#include "libavutil/arm/asm.S"
-
-/*
- * VFP is a floating point coprocessor used in some ARM cores. VFP11 has 1 cycle
- * throughput for almost all the instructions (except for double precision
- * arithmetics), but rather high latency. Latency is 4 cycles for loads and 8 cycles
- * for arithmetic operations. Scheduling code to avoid pipeline stalls is very
- * important for performance. One more interesting feature is that VFP has
- * independent load/store and arithmetics pipelines, so it is possible to make
- * them work simultaneously and get more than 1 operation per cycle. Load/store
- * pipeline can process 2 single precision floating point values per cycle and
- * supports bulk loads and stores for large sets of registers. Arithmetic operations
- * can be done on vectors, which allows to keep the arithmetics pipeline busy,
- * while the processor may issue and execute other instructions. Detailed
- * optimization manuals can be found at http://www.arm.com
- */
-
-/**
- * ARM VFP optimized implementation of 'vector_fmul_reverse_c' function.
- * Assume that len is a positive number and is multiple of 8
- */
-@ void ff_vector_fmul_reverse_vfp(float *dst, const float *src0,
-@                                 const float *src1, int len)
-function ff_vector_fmul_reverse_vfp, export=1
-        vpush           {d8-d15}
-        add             r2,  r2,  r3, lsl #2
-        vldmdb          r2!, {s0-s3}
-        vldmia          r1!, {s8-s11}
-        vldmdb          r2!, {s4-s7}
-        vldmia          r1!, {s12-s15}
-        vmul.f32        s8,  s3,  s8
-        vmul.f32        s9,  s2,  s9
-        vmul.f32        s10, s1,  s10
-        vmul.f32        s11, s0,  s11
-1:
-        subs            r3,  r3,  #16
-        it              ge
-        vldmdbge        r2!, {s16-s19}
-        vmul.f32        s12, s7,  s12
-        it              ge
-        vldmiage        r1!, {s24-s27}
-        vmul.f32        s13, s6,  s13
-        it              ge
-        vldmdbge        r2!, {s20-s23}
-        vmul.f32        s14, s5,  s14
-        it              ge
-        vldmiage        r1!, {s28-s31}
-        vmul.f32        s15, s4,  s15
-        it              ge
-        vmulge.f32      s24, s19, s24
-        it              gt
-        vldmdbgt        r2!, {s0-s3}
-        it              ge
-        vmulge.f32      s25, s18, s25
-        vstmia          r0!, {s8-s13}
-        it              ge
-        vmulge.f32      s26, s17, s26
-        it              gt
-        vldmiagt        r1!, {s8-s11}
-        itt             ge
-        vmulge.f32      s27, s16, s27
-        vmulge.f32      s28, s23, s28
-        it              gt
-        vldmdbgt        r2!, {s4-s7}
-        it              ge
-        vmulge.f32      s29, s22, s29
-        vstmia          r0!, {s14-s15}
-        ittt            ge
-        vmulge.f32      s30, s21, s30
-        vmulge.f32      s31, s20, s31
-        vmulge.f32      s8,  s3,  s8
-        it              gt
-        vldmiagt        r1!, {s12-s15}
-        itttt           ge
-        vmulge.f32      s9,  s2,  s9
-        vmulge.f32      s10, s1,  s10
-        vstmiage        r0!, {s24-s27}
-        vmulge.f32      s11, s0,  s11
-        it              ge
-        vstmiage        r0!, {s28-s31}
-        bgt             1b
-
-        vpop            {d8-d15}
-        bx              lr
-endfunc