Merge remote-tracking branch 'qatar/master'

* qatar/master:
  float_dsp: ppc: add a separate header for Altivec function prototypes
  ARM: fix float_dsp breakage from d5a7229
  Add a float DSP framework to libavutil
  PPC: Move types_altivec.h and util_altivec.h from libavcodec to libavutil
  ARM: Move asm.S from libavcodec to libavutil
  vc1dsp: mark put/avg_vc1_mspel_mc() always_inline

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

75 files changed, 125 insertions, 646 deletions

diff --git a/libavcodec/aac.h b/libavcodec/aac.h
index 9c5c6c044a..29ba2f1402 100644
--- a/libavcodec/aac.h
+++ b/libavcodec/aac.h
@@ -30,6 +30,7 @@
 #ifndef AVCODEC_AAC_H
 #define AVCODEC_AAC_H
 
+#include "libavutil/float_dsp.h"
 #include "avcodec.h"
 #include "dsputil.h"
 #include "fft.h"
@@ -292,6 +293,7 @@ typedef struct {
     FFTContext mdct_ltp;
     DSPContext dsp;
     FmtConvertContext fmt_conv;
+    AVFloatDSPContext fdsp;
     int random_state;
     /** @} */
 
diff --git a/libavcodec/aacdec.c b/libavcodec/aacdec.c
index 4d4291dc50..b638e3b68e 100644
--- a/libavcodec/aacdec.c
+++ b/libavcodec/aacdec.c
@@ -79,7 +79,7 @@
            Parametric Stereo.
  */
 
-
+#include "libavutil/float_dsp.h"
 #include "avcodec.h"
 #include "internal.h"
 #include "get_bits.h"
@@ -901,6 +901,7 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
 
     ff_dsputil_init(&ac->dsp, avctx);
     ff_fmt_convert_init(&ac->fmt_conv, avctx);
+    avpriv_float_dsp_init(&ac->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
 
     ac->random_state = 0x1f2e3d4c;
 
@@ -2069,10 +2070,10 @@ static void windowing_and_mdct_ltp(AACContext *ac, float *out,
     const float *swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
 
     if (ics->window_sequence[0] != LONG_STOP_SEQUENCE) {
-        ac->dsp.vector_fmul(in, in, lwindow_prev, 1024);
+        ac->fdsp.vector_fmul(in, in, lwindow_prev, 1024);
     } else {
         memset(in, 0, 448 * sizeof(float));
-        ac->dsp.vector_fmul(in + 448, in + 448, swindow_prev, 128);
+        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);
diff --git a/libavcodec/aacenc.c b/libavcodec/aacenc.c
index 234ebff39a..7696e96bdf 100644
--- a/libavcodec/aacenc.c
+++ b/libavcodec/aacenc.c
@@ -30,6 +30,7 @@
  * add temporal noise shaping
  ***********************************/
 
+#include "libavutil/float_dsp.h"
 #include "libavutil/opt.h"
 #include "avcodec.h"
 #include "put_bits.h"
@@ -182,7 +183,9 @@ static void put_audio_specific_config(AVCodecContext *avctx)
 }
 
 #define WINDOW_FUNC(type) \
-static void apply_ ##type ##_window(DSPContext *dsp, SingleChannelElement *sce, const float *audio)
+static void apply_ ##type ##_window(DSPContext *dsp, AVFloatDSPContext *fdsp, \
+                                    SingleChannelElement *sce, \
+                                    const float *audio)
 
 WINDOW_FUNC(only_long)
 {
@@ -190,7 +193,7 @@ 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;
 
-    dsp->vector_fmul        (out,        audio,        lwindow, 1024);
+    fdsp->vector_fmul       (out,        audio,        lwindow, 1024);
     dsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024);
 }
 
@@ -200,7 +203,7 @@ WINDOW_FUNC(long_start)
     const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
     float *out = sce->ret;
 
-    dsp->vector_fmul(out, audio, lwindow, 1024);
+    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);
     memset(out + 1024 + 576, 0, sizeof(out[0]) * 448);
@@ -213,7 +216,7 @@ WINDOW_FUNC(long_stop)
     float *out = sce->ret;
 
     memset(out, 0, sizeof(out[0]) * 448);
-    dsp->vector_fmul(out + 448, audio + 448, swindow, 128);
+    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);
 }
@@ -227,7 +230,7 @@ WINDOW_FUNC(eight_short)
     int w;
 
     for (w = 0; w < 8; w++) {
-        dsp->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);
@@ -235,7 +238,9 @@ WINDOW_FUNC(eight_short)
     }
 }
 
-static void (*const apply_window[4])(DSPContext *dsp, SingleChannelElement *sce, const float *audio) = {
+static void (*const apply_window[4])(DSPContext *dsp, AVFloatDSPContext *fdsp,
+                                     SingleChannelElement *sce,
+                                     const float *audio) = {
     [ONLY_LONG_SEQUENCE]   = apply_only_long_window,
     [LONG_START_SEQUENCE]  = apply_long_start_window,
     [EIGHT_SHORT_SEQUENCE] = apply_eight_short_window,
@@ -248,7 +253,7 @@ static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce,
     int i;
     float *output = sce->ret;
 
-    apply_window[sce->ics.window_sequence[0]](&s->dsp, sce, audio);
+    apply_window[sce->ics.window_sequence[0]](&s->dsp, &s->fdsp, sce, audio);
 
     if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE)
         s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
@@ -693,6 +698,7 @@ 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
     ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
diff --git a/libavcodec/aacenc.h b/libavcodec/aacenc.h
index 35a2565ef4..fbc3da8879 100644
--- a/libavcodec/aacenc.h
+++ b/libavcodec/aacenc.h
@@ -22,6 +22,7 @@
 #ifndef AVCODEC_AACENC_H
 #define AVCODEC_AACENC_H
 
+#include "libavutil/float_dsp.h"
 #include "avcodec.h"
 #include "put_bits.h"
 #include "dsputil.h"
@@ -61,6 +62,7 @@ typedef struct AACEncContext {
     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
 
     int samplerate_index;                        ///< MPEG-4 samplerate index
diff --git a/libavcodec/ac3enc.c b/libavcodec/ac3enc.c
index e04e70d8bc..05cdddafe6 100644
--- a/libavcodec/ac3enc.c
+++ b/libavcodec/ac3enc.c
@@ -2491,6 +2491,7 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx)
 #endif
 
     ff_dsputil_init(&s->dsp, avctx);
+    avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
     ff_ac3dsp_init(&s->ac3dsp, avctx->flags & CODEC_FLAG_BITEXACT);
 
     dprint_options(s);
diff --git a/libavcodec/ac3enc.h b/libavcodec/ac3enc.h
index e8415a2d69..be9dcf21c1 100644
--- a/libavcodec/ac3enc.h
+++ b/libavcodec/ac3enc.h
@@ -29,6 +29,8 @@
 #define AVCODEC_AC3ENC_H
 
 #include <stdint.h>
+
+#include "libavutil/float_dsp.h"
 #include "ac3.h"
 #include "ac3dsp.h"
 #include "avcodec.h"
@@ -158,6 +160,7 @@ typedef struct AC3EncodeContext {
     AVCodecContext *avctx;                  ///< parent AVCodecContext
     PutBitContext pb;                       ///< bitstream writer context
     DSPContext dsp;
+    AVFloatDSPContext fdsp;
     AC3DSPContext ac3dsp;                   ///< AC-3 optimized functions
     FFTContext mdct;                        ///< FFT context for MDCT calculation
     const SampleType *mdct_window;          ///< MDCT window function array
diff --git a/libavcodec/ac3enc_fixed.c b/libavcodec/ac3enc_fixed.c
index daf84cbd9a..689f822981 100644
--- a/libavcodec/ac3enc_fixed.c
+++ b/libavcodec/ac3enc_fixed.c
@@ -68,10 +68,11 @@ av_cold int AC3_NAME(mdct_init)(AC3EncodeContext *s)
 /*
  * Apply KBD window to input samples prior to MDCT.
  */
-static void apply_window(DSPContext *dsp, int16_t *output, const int16_t *input,
+static void apply_window(void *dsp, int16_t *output, const int16_t *input,
                          const int16_t *window, unsigned int len)
 {
-    dsp->apply_window_int16(output, input, window, len);
+    DSPContext *dsp0 = dsp;
+    dsp0->apply_window_int16(output, input, window, len);
 }
 
 
diff --git a/libavcodec/ac3enc_float.c b/libavcodec/ac3enc_float.c
index 1fc8c3625e..747fbeb536 100644
--- a/libavcodec/ac3enc_float.c
+++ b/libavcodec/ac3enc_float.c
@@ -86,10 +86,12 @@ av_cold int ff_ac3_float_mdct_init(AC3EncodeContext *s)
 /*
  * Apply KBD window to input samples prior to MDCT.
  */
-static void apply_window(DSPContext *dsp, float *output, const float *input,
-                         const float *window, unsigned int len)
+static void apply_window(void *dsp, float *output,
+                         const float *input, const float *window,
+                         unsigned int len)
 {
-    dsp->vector_fmul(output, input, window, len);
+    AVFloatDSPContext *fdsp = dsp;
+    fdsp->vector_fmul(output, input, window, len);
 }
 
 
diff --git a/libavcodec/ac3enc_template.c b/libavcodec/ac3enc_template.c
index 20ac301a43..e81bfce5a8 100644
--- a/libavcodec/ac3enc_template.c
+++ b/libavcodec/ac3enc_template.c
@@ -33,7 +33,7 @@
 
 static void scale_coefficients(AC3EncodeContext *s);
 
-static void apply_window(DSPContext *dsp, SampleType *output,
+static void apply_window(void *dsp, SampleType *output,
                          const SampleType *input, const SampleType *window,
                          unsigned int len);
 
@@ -110,8 +110,13 @@ static void apply_mdct(AC3EncodeContext *s)
             AC3Block *block = &s->blocks[blk];
             const SampleType *input_samples = &s->planar_samples[ch][blk * AC3_BLOCK_SIZE];
 
+#if CONFIG_AC3ENC_FLOAT
+            apply_window(&s->fdsp, s->windowed_samples, input_samples,
+                         s->mdct_window, AC3_WINDOW_SIZE);
+#else
             apply_window(&s->dsp, s->windowed_samples, input_samples,
                          s->mdct_window, AC3_WINDOW_SIZE);
+#endif
 
             if (s->fixed_point)
                 block->coeff_shift[ch+1] = normalize_samples(s);
diff --git a/libavcodec/arm/aacpsdsp_neon.S b/libavcodec/arm/aacpsdsp_neon.S
index e75760a2d6..fb00900a4d 100644
--- a/libavcodec/arm/aacpsdsp_neon.S
+++ b/libavcodec/arm/aacpsdsp_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 function ff_ps_add_squares_neon, export=1
         mov             r3,  r0
diff --git a/libavcodec/arm/ac3dsp_arm.S b/libavcodec/arm/ac3dsp_arm.S
index 9a7d20eb7b..ed8eb37845 100644
--- a/libavcodec/arm/ac3dsp_arm.S
+++ b/libavcodec/arm/ac3dsp_arm.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 function ff_ac3_update_bap_counts_arm, export=1
         push            {lr}
diff --git a/libavcodec/arm/ac3dsp_armv6.S b/libavcodec/arm/ac3dsp_armv6.S
index df8bfbaa03..7e2f40edf0 100644
--- a/libavcodec/arm/ac3dsp_armv6.S
+++ b/libavcodec/arm/ac3dsp_armv6.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 function ff_ac3_bit_alloc_calc_bap_armv6, export=1
         ldr             r12, [sp]
diff --git a/libavcodec/arm/ac3dsp_neon.S b/libavcodec/arm/ac3dsp_neon.S
index 152afbfd3c..42f35e3d0d 100644
--- a/libavcodec/arm/ac3dsp_neon.S
+++ b/libavcodec/arm/ac3dsp_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 function ff_ac3_max_msb_abs_int16_neon, export=1
         vmov.i16        q0,  #0
diff --git a/libavcodec/arm/asm.S b/libavcodec/arm/asm.S
deleted file mode 100644
index 28d52cfd53..0000000000
--- a/libavcodec/arm/asm.S
+++ /dev/null
@@ -1,231 +0,0 @@
-/*
- * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
- *
- * 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"
-
-#ifdef __ELF__
-#   define ELF
-#else
-#   define ELF @
-#endif
-
-#if CONFIG_THUMB
-#   define A @
-#   define T
-#else
-#   define A
-#   define T @
-#endif
-
-#if   HAVE_NEON
-        .arch           armv7-a
-#elif HAVE_ARMV6T2
-        .arch           armv6t2
-#elif HAVE_ARMV6
-        .arch           armv6
-#elif HAVE_ARMV5TE
-        .arch           armv5te
-#endif
-
-#if   HAVE_NEON
-        .fpu            neon
-#elif HAVE_ARMVFP
-        .fpu            vfp
-#endif
-
-        .syntax unified
-T       .thumb
-
-.macro  require8 val=1
-ELF     .eabi_attribute 24, \val
-.endm
-
-.macro  preserve8 val=1
-ELF     .eabi_attribute 25, \val
-.endm
-
-.macro  function name, export=0
-    .macro endfunc
-ELF     .size   \name, . - \name
-        .endfunc
-        .purgem endfunc
-    .endm
-        .text
-        .align          2
-    .if \export
-        .global EXTERN_ASM\name
-EXTERN_ASM\name:
-    .endif
-ELF     .type   \name, %function
-        .func   \name
-\name:
-.endm
-
-.macro  const   name, align=2
-    .macro endconst
-ELF     .size   \name, . - \name
-        .purgem endconst
-    .endm
-        .section        .rodata
-        .align          \align
-\name:
-.endm
-
-#if !HAVE_ARMV6T2
-.macro  movw    rd, val
-        mov     \rd, \val &  255
-        orr     \rd, \val & ~255
-.endm
-#endif
-
-.macro  mov32   rd, val
-#if HAVE_ARMV6T2
-        movw            \rd, #(\val) & 0xffff
-    .if (\val) >> 16
-        movt            \rd, #(\val) >> 16
-    .endif
-#else
-        ldr             \rd, =\val
-#endif
-.endm
-
-.macro  movrel rd, val
-#if HAVE_ARMV6T2 && !CONFIG_PIC && !defined(__APPLE__)
-        movw            \rd, #:lower16:\val
-        movt            \rd, #:upper16:\val
-#else
-        ldr             \rd, =\val
-#endif
-.endm
-
-.macro  ldr_pre         rt,  rn,  rm:vararg
-A       ldr             \rt, [\rn, \rm]!
-T       add             \rn, \rn, \rm
-T       ldr             \rt, [\rn]
-.endm
-
-.macro  ldr_dpre        rt,  rn,  rm:vararg
-A       ldr             \rt, [\rn, -\rm]!
-T       sub             \rn, \rn, \rm
-T       ldr             \rt, [\rn]
-.endm
-
-.macro  ldr_nreg        rt,  rn,  rm:vararg
-A       ldr             \rt, [\rn, -\rm]
-T       sub             \rt, \rn, \rm
-T       ldr             \rt, [\rt]
-.endm
-
-.macro  ldr_post        rt,  rn,  rm:vararg
-A       ldr             \rt, [\rn], \rm
-T       ldr             \rt, [\rn]
-T       add             \rn, \rn, \rm
-.endm
-
-.macro  ldrd_reg        rt,  rt2, rn,  rm
-A       ldrd            \rt, \rt2, [\rn, \rm]
-T       add             \rt, \rn, \rm
-T       ldrd            \rt, \rt2, [\rt]
-.endm
-
-.macro  ldrd_post       rt,  rt2, rn,  rm
-A       ldrd            \rt, \rt2, [\rn], \rm
-T       ldrd            \rt, \rt2, [\rn]
-T       add             \rn, \rn, \rm
-.endm
-
-.macro  ldrh_pre        rt,  rn,  rm
-A       ldrh            \rt, [\rn, \rm]!
-T       add             \rn, \rn, \rm
-T       ldrh            \rt, [\rn]
-.endm
-
-.macro  ldrh_dpre       rt,  rn,  rm
-A       ldrh            \rt, [\rn, -\rm]!
-T       sub             \rn, \rn, \rm
-T       ldrh            \rt, [\rn]
-.endm
-
-.macro  ldrh_post       rt,  rn,  rm
-A       ldrh            \rt, [\rn], \rm
-T       ldrh            \rt, [\rn]
-T       add             \rn, \rn, \rm
-.endm
-
-.macro  ldrb_post       rt,  rn,  rm
-A       ldrb            \rt, [\rn], \rm
-T       ldrb            \rt, [\rn]
-T       add             \rn, \rn, \rm
-.endm
-
-.macro  str_post       rt,  rn,  rm:vararg
-A       str             \rt, [\rn], \rm
-T       str             \rt, [\rn]
-T       add             \rn, \rn, \rm
-.endm
-
-.macro  strb_post       rt,  rn,  rm:vararg
-A       strb            \rt, [\rn], \rm
-T       strb            \rt, [\rn]
-T       add             \rn, \rn, \rm
-.endm
-
-.macro  strd_post       rt,  rt2, rn,  rm
-A       strd            \rt, \rt2, [\rn], \rm
-T       strd            \rt, \rt2, [\rn]
-T       add             \rn, \rn, \rm
-.endm
-
-.macro  strh_pre        rt,  rn,  rm
-A       strh            \rt, [\rn, \rm]!
-T       add             \rn, \rn, \rm
-T       strh            \rt, [\rn]
-.endm
-
-.macro  strh_dpre       rt,  rn,  rm
-A       strh            \rt, [\rn, -\rm]!
-T       sub             \rn, \rn, \rm
-T       strh            \rt, [\rn]
-.endm
-
-.macro  strh_post       rt,  rn,  rm
-A       strh            \rt, [\rn], \rm
-T       strh            \rt, [\rn]
-T       add             \rn, \rn, \rm
-.endm
-
-.macro  strh_dpost       rt,  rn,  rm
-A       strh            \rt, [\rn], -\rm
-T       strh            \rt, [\rn]
-T       sub             \rn, \rn, \rm
-.endm
-
-#if HAVE_VFP_ARGS
-        .eabi_attribute 28, 1
-#   define VFP
-#   define NOVFP @
-#else
-#   define VFP   @
-#   define NOVFP
-#endif
-
-#define GLUE(a, b) a ## b
-#define JOIN(a, b) GLUE(a, b)
-#define X(s) JOIN(EXTERN_ASM, s)
diff --git a/libavcodec/arm/dcadsp_neon.S b/libavcodec/arm/dcadsp_neon.S
index 852527a59e..6a6c77aca8 100644
--- a/libavcodec/arm/dcadsp_neon.S
+++ b/libavcodec/arm/dcadsp_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 function ff_dca_lfe_fir_neon, export=1
         push            {r4-r6,lr}
diff --git a/libavcodec/arm/dsputil_arm.S b/libavcodec/arm/dsputil_arm.S
index 1247b0fa84..3f0d07155f 100644
--- a/libavcodec/arm/dsputil_arm.S
+++ b/libavcodec/arm/dsputil_arm.S
@@ -20,7 +20,7 @@
 @
 
 #include "config.h"
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
 
diff --git a/libavcodec/arm/dsputil_armv6.S b/libavcodec/arm/dsputil_armv6.S
index 1adedf7b8f..fac431d393 100644
--- a/libavcodec/arm/dsputil_armv6.S
+++ b/libavcodec/arm/dsputil_armv6.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
 
diff --git a/libavcodec/arm/dsputil_init_neon.c b/libavcodec/arm/dsputil_init_neon.c
index 4c73276a95..46b22a412c 100644
--- a/libavcodec/arm/dsputil_init_neon.c
+++ b/libavcodec/arm/dsputil_init_neon.c
@@ -150,7 +150,6 @@ void ff_avg_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int);
 void ff_vp3_v_loop_filter_neon(uint8_t *, int, int *);
 void ff_vp3_h_loop_filter_neon(uint8_t *, int, int *);
 
-void ff_vector_fmul_neon(float *dst, const float *src0, const float *src1, int len);
 void ff_vector_fmul_window_neon(float *dst, const float *src0,
                                 const float *src1, const float *win, int len);
 void ff_vector_fmul_scalar_neon(float *dst, const float *src, float mul,
@@ -328,7 +327,6 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
         c->vp3_idct_dc_add   = ff_vp3_idct_dc_add_neon;
     }
 
-    c->vector_fmul                = ff_vector_fmul_neon;
     c->vector_fmul_window         = ff_vector_fmul_window_neon;
     c->vector_fmul_scalar         = ff_vector_fmul_scalar_neon;
     c->vector_fmac_scalar         = ff_vector_fmac_scalar_neon;
diff --git a/libavcodec/arm/dsputil_init_vfp.c b/libavcodec/arm/dsputil_init_vfp.c
index 91fb3d24e4..5713c71f80 100644
--- a/libavcodec/arm/dsputil_init_vfp.c
+++ b/libavcodec/arm/dsputil_init_vfp.c
@@ -18,20 +18,13 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "libavutil/arm/cpu.h"
 #include "libavcodec/dsputil.h"
 #include "dsputil_arm.h"
 
-void ff_vector_fmul_vfp(float *dst, const float *src0,
-                        const float *src1, int len);
 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)
 {
-    int cpu_flags = av_get_cpu_flags();
-
-    if (!have_vfpv3(cpu_flags))
-        c->vector_fmul = ff_vector_fmul_vfp;
     c->vector_fmul_reverse = ff_vector_fmul_reverse_vfp;
 }
diff --git a/libavcodec/arm/dsputil_neon.S b/libavcodec/arm/dsputil_neon.S
index 344f28bea9..7ddd76a153 100644
--- a/libavcodec/arm/dsputil_neon.S
+++ b/libavcodec/arm/dsputil_neon.S
@@ -20,7 +20,7 @@
  */
 
 #include "config.h"
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
 
@@ -534,45 +534,6 @@ function ff_add_pixels_clamped_neon, export=1
         bx              lr
 endfunc
 
-function ff_vector_fmul_neon, export=1
-        subs            r3,  r3,  #8
-        vld1.32         {d0-d3},  [r1,:128]!
-        vld1.32         {d4-d7},  [r2,:128]!
-        vmul.f32        q8,  q0,  q2
-        vmul.f32        q9,  q1,  q3
-        beq             3f
-        bics            ip,  r3,  #15
-        beq             2f
-1:      subs            ip,  ip,  #16
-        vld1.32         {d0-d1},  [r1,:128]!
-        vld1.32         {d4-d5},  [r2,:128]!
-        vmul.f32        q10, q0,  q2
-        vld1.32         {d2-d3},  [r1,:128]!
-        vld1.32         {d6-d7},  [r2,:128]!
-        vmul.f32        q11, q1,  q3
-        vst1.32         {d16-d19},[r0,:128]!
-        vld1.32         {d0-d1},  [r1,:128]!
-        vld1.32         {d4-d5},  [r2,:128]!
-        vmul.f32        q8,  q0,  q2
-        vld1.32         {d2-d3},  [r1,:128]!
-        vld1.32         {d6-d7},  [r2,:128]!
-        vmul.f32        q9,  q1,  q3
-        vst1.32         {d20-d23},[r0,:128]!
-        bne             1b
-        ands            r3,  r3,  #15
-        beq             3f
-2:      vld1.32         {d0-d1},  [r1,:128]!
-        vld1.32         {d4-d5},  [r2,:128]!
-        vst1.32         {d16-d17},[r0,:128]!
-        vmul.f32        q8,  q0,  q2
-        vld1.32         {d2-d3},  [r1,:128]!
-        vld1.32         {d6-d7},  [r2,:128]!
-        vst1.32         {d18-d19},[r0,:128]!
-        vmul.f32        q9,  q1,  q3
-3:      vst1.32         {d16-d19},[r0,:128]!
-        bx              lr
-endfunc
-
 function ff_vector_fmul_window_neon, export=1
         push            {r4,r5,lr}
         ldr             lr,  [sp, #12]
diff --git a/libavcodec/arm/dsputil_vfp.S b/libavcodec/arm/dsputil_vfp.S
index 108208174d..63ba766859 100644
--- a/libavcodec/arm/dsputil_vfp.S
+++ b/libavcodec/arm/dsputil_vfp.S
@@ -19,7 +19,7 @@
  */
 
 #include "config.h"
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 /*
  * VFP is a floating point coprocessor used in some ARM cores. VFP11 has 1 cycle
@@ -37,53 +37,6 @@
  */
 
 /**
- * ARM VFP optimized implementation of 'vector_fmul_c' function.
- * Assume that len is a positive number and is multiple of 8
- */
-@ void ff_vector_fmul_vfp(float *dst, const float *src0, const float *src1, int len)
-function ff_vector_fmul_vfp, export=1
-        vpush           {d8-d15}
-        fmrx            r12, fpscr
-        orr             r12, r12, #(3 << 16) /* set vector size to 4 */
-        fmxr            fpscr, r12
-
-        vldmia          r1!, {s0-s3}
-        vldmia          r2!, {s8-s11}
-        vldmia          r1!, {s4-s7}
-        vldmia          r2!, {s12-s15}
-        vmul.f32        s8,  s0,  s8
-1:
-        subs            r3,  r3,  #16
-        vmul.f32        s12, s4,  s12
-        itttt           ge
-        vldmiage        r1!, {s16-s19}
-        vldmiage        r2!, {s24-s27}
-        vldmiage        r1!, {s20-s23}
-        vldmiage        r2!, {s28-s31}
-        it              ge
-        vmulge.f32      s24, s16, s24
-        vstmia          r0!, {s8-s11}
-        vstmia          r0!, {s12-s15}
-        it              ge
-        vmulge.f32      s28, s20, s28
-        itttt           gt
-        vldmiagt        r1!, {s0-s3}
-        vldmiagt        r2!, {s8-s11}
-        vldmiagt        r1!, {s4-s7}
-        vldmiagt        r2!, {s12-s15}
-        ittt            ge
-        vmulge.f32      s8,  s0,  s8
-        vstmiage        r0!, {s24-s27}
-        vstmiage        r0!, {s28-s31}
-        bgt             1b
-
-        bic             r12, r12, #(7 << 16) /* set vector size back to 1 */
-        fmxr            fpscr, r12
-        vpop            {d8-d15}
-        bx              lr
-endfunc
-
-/**
  * ARM VFP optimized implementation of 'vector_fmul_reverse_c' function.
  * Assume that len is a positive number and is multiple of 8
  */
diff --git a/libavcodec/arm/fft_fixed_neon.S b/libavcodec/arm/fft_fixed_neon.S
index 0316b80bce..2430ada0ff 100644
--- a/libavcodec/arm/fft_fixed_neon.S
+++ b/libavcodec/arm/fft_fixed_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 .macro  bflies          d0,  d1,  r0,  r1
         vrev64.32       \r0, \d1                @ t5, t6, t1, t2
diff --git a/libavcodec/arm/fft_neon.S b/libavcodec/arm/fft_neon.S
index ef8e4d4210..fe69f58cda 100644
--- a/libavcodec/arm/fft_neon.S
+++ b/libavcodec/arm/fft_neon.S
@@ -24,7 +24,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 #define M_SQRT1_2 0.70710678118654752440
 
diff --git a/libavcodec/arm/fmtconvert_neon.S b/libavcodec/arm/fmtconvert_neon.S
index 17af9f524a..83fc124a29 100644
--- a/libavcodec/arm/fmtconvert_neon.S
+++ b/libavcodec/arm/fmtconvert_neon.S
@@ -20,7 +20,7 @@
  */
 
 #include "config.h"
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
 
diff --git a/libavcodec/arm/fmtconvert_vfp.S b/libavcodec/arm/fmtconvert_vfp.S
index 7e2eb83620..4bdbe82e52 100644
--- a/libavcodec/arm/fmtconvert_vfp.S
+++ b/libavcodec/arm/fmtconvert_vfp.S
@@ -19,7 +19,7 @@
  */
 
 #include "config.h"
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 /**
  * ARM VFP optimized float to int16 conversion.
diff --git a/libavcodec/arm/h264cmc_neon.S b/libavcodec/arm/h264cmc_neon.S
index a6feadd189..e82394d899 100644
--- a/libavcodec/arm/h264cmc_neon.S
+++ b/libavcodec/arm/h264cmc_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 /* chroma_mc8(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y) */
 .macro  h264_chroma_mc8 type, codec=h264
diff --git a/libavcodec/arm/h264dsp_neon.S b/libavcodec/arm/h264dsp_neon.S
index fc6b3b35d5..341863ef40 100644
--- a/libavcodec/arm/h264dsp_neon.S
+++ b/libavcodec/arm/h264dsp_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 #include "neon.S"
 
         /* H.264 loop filter */
diff --git a/libavcodec/arm/h264idct_neon.S b/libavcodec/arm/h264idct_neon.S
index 49b301247f..612aec36d9 100644
--- a/libavcodec/arm/h264idct_neon.S
+++ b/libavcodec/arm/h264idct_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
 
diff --git a/libavcodec/arm/h264pred_neon.S b/libavcodec/arm/h264pred_neon.S
index 0dac20b4a0..4dc47ba8f1 100644
--- a/libavcodec/arm/h264pred_neon.S
+++ b/libavcodec/arm/h264pred_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         .macro ldcol.8  rd,  rs,  rt,  n=8,  hi=0
 .if \n == 8 || \hi == 0
diff --git a/libavcodec/arm/int_neon.S b/libavcodec/arm/int_neon.S
index 311c99d95e..bce84516c6 100644
--- a/libavcodec/arm/int_neon.S
+++ b/libavcodec/arm/int_neon.S
@@ -19,7 +19,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
         .fpu neon
diff --git a/libavcodec/arm/jrevdct_arm.S b/libavcodec/arm/jrevdct_arm.S
index 93cbbbe8eb..331ad85328 100644
--- a/libavcodec/arm/jrevdct_arm.S
+++ b/libavcodec/arm/jrevdct_arm.S
@@ -25,7 +25,7 @@
 
 */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 #define FIX_0_298631336 2446
 #define FIX_0_541196100 4433
diff --git a/libavcodec/arm/mdct_fixed_neon.S b/libavcodec/arm/mdct_fixed_neon.S
index d219216a20..08a388777a 100644
--- a/libavcodec/arm/mdct_fixed_neon.S
+++ b/libavcodec/arm/mdct_fixed_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
 
diff --git a/libavcodec/arm/mdct_neon.S b/libavcodec/arm/mdct_neon.S
index 9b7dc6ce83..a8f376d760 100644
--- a/libavcodec/arm/mdct_neon.S
+++ b/libavcodec/arm/mdct_neon.S
@@ -19,7 +19,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
 
diff --git a/libavcodec/arm/mpegaudiodsp_fixed_armv6.S b/libavcodec/arm/mpegaudiodsp_fixed_armv6.S
index b517b973e7..49bd0bcaf2 100644
--- a/libavcodec/arm/mpegaudiodsp_fixed_armv6.S
+++ b/libavcodec/arm/mpegaudiodsp_fixed_armv6.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 .macro  skip            args:vararg
 .endm
diff --git a/libavcodec/arm/mpegvideo_armv5te_s.S b/libavcodec/arm/mpegvideo_armv5te_s.S
index 3db9c734e9..8f9dd42b22 100644
--- a/libavcodec/arm/mpegvideo_armv5te_s.S
+++ b/libavcodec/arm/mpegvideo_armv5te_s.S
@@ -20,7 +20,7 @@
  */
 
 #include "config.h"
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 /*
  * Special optimized version of dct_unquantize_h263_helper_c, it
diff --git a/libavcodec/arm/mpegvideo_neon.S b/libavcodec/arm/mpegvideo_neon.S
index 849047e13c..e05df8e711 100644
--- a/libavcodec/arm/mpegvideo_neon.S
+++ b/libavcodec/arm/mpegvideo_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 #include "asm-offsets.h"
 
 function ff_dct_unquantize_h263_inter_neon, export=1
diff --git a/libavcodec/arm/rdft_neon.S b/libavcodec/arm/rdft_neon.S
index 19886e6d0b..dcd3eee0b4 100644
--- a/libavcodec/arm/rdft_neon.S
+++ b/libavcodec/arm/rdft_neon.S
@@ -19,7 +19,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
 
diff --git a/libavcodec/arm/rv34dsp_neon.S b/libavcodec/arm/rv34dsp_neon.S
index 15a015deef..522e387582 100644
--- a/libavcodec/arm/rv34dsp_neon.S
+++ b/libavcodec/arm/rv34dsp_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 #include "neon.S"
 
 .macro rv34_inv_transform    r0
diff --git a/libavcodec/arm/rv40dsp_neon.S b/libavcodec/arm/rv40dsp_neon.S
index f68f38234a..6bd45eb5ad 100644
--- a/libavcodec/arm/rv40dsp_neon.S
+++ b/libavcodec/arm/rv40dsp_neon.S
@@ -19,7 +19,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 #include "neon.S"
 
 .macro  qpel_lowpass    r0,  r1,  rc1, rc2, shift
diff --git a/libavcodec/arm/sbrdsp_neon.S b/libavcodec/arm/sbrdsp_neon.S
index 835c32caee..4b681bfe6a 100644
--- a/libavcodec/arm/sbrdsp_neon.S
+++ b/libavcodec/arm/sbrdsp_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 function ff_sbr_sum64x5_neon, export=1
         push            {lr}
diff --git a/libavcodec/arm/simple_idct_arm.S b/libavcodec/arm/simple_idct_arm.S
index 1d3e9175cc..42741a07a1 100644
--- a/libavcodec/arm/simple_idct_arm.S
+++ b/libavcodec/arm/simple_idct_arm.S
@@ -23,7 +23,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 /* useful constants for the algorithm, they are save in __constant_ptr__ at */
 /* the end of the source code.*/
diff --git a/libavcodec/arm/simple_idct_armv5te.S b/libavcodec/arm/simple_idct_armv5te.S
index 71727ceccc..20a2305b65 100644
--- a/libavcodec/arm/simple_idct_armv5te.S
+++ b/libavcodec/arm/simple_idct_armv5te.S
@@ -21,7 +21,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 #define W1  22725   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
 #define W2  21407   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
diff --git a/libavcodec/arm/simple_idct_armv6.S b/libavcodec/arm/simple_idct_armv6.S
index a176b3a7b4..25393bfd27 100644
--- a/libavcodec/arm/simple_idct_armv6.S
+++ b/libavcodec/arm/simple_idct_armv6.S
@@ -21,7 +21,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 #define W1  22725   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
 #define W2  21407   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
diff --git a/libavcodec/arm/simple_idct_neon.S b/libavcodec/arm/simple_idct_neon.S
index 5df8f6e4fc..3c048b0d56 100644
--- a/libavcodec/arm/simple_idct_neon.S
+++ b/libavcodec/arm/simple_idct_neon.S
@@ -23,7 +23,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 #define W1  22725  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
 #define W2  21407  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
diff --git a/libavcodec/arm/synth_filter_neon.S b/libavcodec/arm/synth_filter_neon.S
index 3f91d67506..0e38ffcf3a 100644
--- a/libavcodec/arm/synth_filter_neon.S
+++ b/libavcodec/arm/synth_filter_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
         preserve8
 
diff --git a/libavcodec/arm/vp3dsp_neon.S b/libavcodec/arm/vp3dsp_neon.S
index ae3e40201a..bcdd4ecc1d 100644
--- a/libavcodec/arm/vp3dsp_neon.S
+++ b/libavcodec/arm/vp3dsp_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 const   vp3_idct_constants, align=4
 .short 64277, 60547, 54491, 46341, 36410, 25080, 12785
diff --git a/libavcodec/arm/vp56dsp_neon.S b/libavcodec/arm/vp56dsp_neon.S
index 0353661009..03dd28d1cb 100644
--- a/libavcodec/arm/vp56dsp_neon.S
+++ b/libavcodec/arm/vp56dsp_neon.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 .macro  vp6_edge_filter
         vdup.16         q3,  r2                 @ t
diff --git a/libavcodec/arm/vp8_armv6.S b/libavcodec/arm/vp8_armv6.S
index 9399bb28d7..1412f43ad3 100644
--- a/libavcodec/arm/vp8_armv6.S
+++ b/libavcodec/arm/vp8_armv6.S
@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 .macro rac_get_prob     h, bs, buf, cw, pr, t0, t1
         adds            \bs, \bs, \t0
diff --git a/libavcodec/arm/vp8dsp_armv6.S b/libavcodec/arm/vp8dsp_armv6.S
index 08054ff262..a26a2a9813 100644
--- a/libavcodec/arm/vp8dsp_armv6.S
+++ b/libavcodec/arm/vp8dsp_armv6.S
@@ -52,7 +52,7 @@
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 
 @ idct
 
diff --git a/libavcodec/arm/vp8dsp_neon.S b/libavcodec/arm/vp8dsp_neon.S
index a9f698dfed..04e7c5ca66 100644
--- a/libavcodec/arm/vp8dsp_neon.S
+++ b/libavcodec/arm/vp8dsp_neon.S
@@ -21,7 +21,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "asm.S"
+#include "libavutil/arm/asm.S"
 #include "neon.S"
 
 function ff_vp8_luma_dc_wht_neon, export=1
diff --git a/libavcodec/atrac3.c b/libavcodec/atrac3.c
index f23fcfbf7f..585b9972bc 100644
--- a/libavcodec/atrac3.c
+++ b/libavcodec/atrac3.c
@@ -36,9 +36,9 @@
 #include <stddef.h>
 #include <stdio.h>
 
+#include "libavutil/float_dsp.h"
 #include "avcodec.h"
 #include "get_bits.h"
-#include "dsputil.h"
 #include "bytestream.h"
 #include "fft.h"
 #include "fmtconvert.h"
@@ -125,13 +125,13 @@ typedef struct {
 
     FFTContext          mdct_ctx;
     FmtConvertContext   fmt_conv;
+    AVFloatDSPContext   fdsp;
 } ATRAC3Context;
 
 static DECLARE_ALIGNED(32, float, mdct_window)[MDCT_SIZE];
 static VLC              spectral_coeff_tab[7];
 static float            gain_tab1[16];
 static float            gain_tab2[31];
-static DSPContext       dsp;
 
 
 /**
@@ -164,7 +164,7 @@ static void IMLT(ATRAC3Context *q, float *pInput, float *pOutput, int odd_band)
     q->mdct_ctx.imdct_calc(&q->mdct_ctx,pOutput,pInput);
 
     /* Perform windowing on the output. */
-    dsp.vector_fmul(pOutput, pOutput, mdct_window, MDCT_SIZE);
+    q->fdsp.vector_fmul(pOutput, pOutput, mdct_window, MDCT_SIZE);
 
 }
 
@@ -1039,7 +1039,7 @@ static av_cold int atrac3_decode_init(AVCodecContext *avctx)
         q->matrix_coeff_index_next[i] = 3;
     }
 
-    ff_dsputil_init(&dsp, avctx);
+    avpriv_float_dsp_init(&q->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
     ff_fmt_convert_init(&q->fmt_conv, avctx);
 
     q->pUnits = av_mallocz(sizeof(channel_unit)*q->channels);
diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index bbcf761552..b2d5d357d5 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -2471,12 +2471,6 @@ WRAPPER8_16_SQ(quant_psnr8x8_c, quant_psnr16_c)
 WRAPPER8_16_SQ(rd8x8_c, rd16_c)
 WRAPPER8_16_SQ(bit8x8_c, bit16_c)
 
-static void vector_fmul_c(float *dst, const float *src0, const float *src1, int len){
-    int i;
-    for(i=0; i<len; i++)
-        dst[i] = src0[i] * src1[i];
-}
-
 static void vector_fmul_reverse_c(float *dst, const float *src0, const float *src1, int len){
     int i;
     src1 += len-1;
@@ -3054,7 +3048,6 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)
 #if CONFIG_AC3_DECODER
     c->ac3_downmix = ff_ac3_downmix_c;
 #endif
-    c->vector_fmul = vector_fmul_c;
     c->vector_fmul_reverse = vector_fmul_reverse_c;
     c->vector_fmul_add = vector_fmul_add_c;
     c->vector_fmul_window = vector_fmul_window_c;
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index 6c814407c1..f934c59245 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -403,7 +403,6 @@ typedef struct DSPContext {
     void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
     void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
     /* assume len is a multiple of 16, and arrays are 32-byte aligned */
-    void (*vector_fmul)(float *dst, const float *src0, const float *src1, int len);
     void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
     /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
     void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
diff --git a/libavcodec/nellymoserenc.c b/libavcodec/nellymoserenc.c
index b40e70866b..e3bae11706 100644
--- a/libavcodec/nellymoserenc.c
+++ b/libavcodec/nellymoserenc.c
@@ -35,6 +35,7 @@
  * http://wiki.multimedia.cx/index.php?title=Nellymoser
  */
 
+#include "libavutil/float_dsp.h"
 #include "libavutil/mathematics.h"
 #include "nellymoser.h"
 #include "avcodec.h"
@@ -55,6 +56,7 @@ typedef struct NellyMoserEncodeContext {
     AVCodecContext  *avctx;
     int             last_frame;
     DSPContext      dsp;
+    AVFloatDSPContext fdsp;
     FFTContext      mdct_ctx;
     AudioFrameQueue afq;
     DECLARE_ALIGNED(32, float, mdct_out)[NELLY_SAMPLES];
@@ -120,11 +122,11 @@ static void apply_mdct(NellyMoserEncodeContext *s)
     float *in1 = s->buf + NELLY_BUF_LEN;
     float *in2 = s->buf + 2 * NELLY_BUF_LEN;
 
-    s->dsp.vector_fmul        (s->in_buff,                 in0, ff_sine_128, NELLY_BUF_LEN);
+    s->fdsp.vector_fmul       (s->in_buff,                 in0, ff_sine_128, NELLY_BUF_LEN);
     s->dsp.vector_fmul_reverse(s->in_buff + NELLY_BUF_LEN, in1, ff_sine_128, NELLY_BUF_LEN);
     s->mdct_ctx.mdct_calc(&s->mdct_ctx, s->mdct_out, s->in_buff);
 
-    s->dsp.vector_fmul        (s->in_buff,                 in1, ff_sine_128, NELLY_BUF_LEN);
+    s->fdsp.vector_fmul       (s->in_buff,                 in1, ff_sine_128, NELLY_BUF_LEN);
     s->dsp.vector_fmul_reverse(s->in_buff + NELLY_BUF_LEN, in2, ff_sine_128, NELLY_BUF_LEN);
     s->mdct_ctx.mdct_calc(&s->mdct_ctx, s->mdct_out + NELLY_BUF_LEN, s->in_buff);
 }
@@ -172,6 +174,7 @@ static av_cold int encode_init(AVCodecContext *avctx)
     if ((ret = ff_mdct_init(&s->mdct_ctx, 8, 0, 32768.0)) < 0)
         goto error;
     ff_dsputil_init(&s->dsp, avctx);
+    avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
 
     /* Generate overlap window */
     ff_init_ff_sine_windows(7);
diff --git a/libavcodec/ppc/dsputil_altivec.c b/libavcodec/ppc/dsputil_altivec.c
index de8b85891b..5f14f1dea3 100644
--- a/libavcodec/ppc/dsputil_altivec.c
+++ b/libavcodec/ppc/dsputil_altivec.c
@@ -24,9 +24,9 @@
 #if HAVE_ALTIVEC_H
 #include <altivec.h>
 #endif
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/dsputil.h"
-#include "util_altivec.h"
-#include "types_altivec.h"
 #include "dsputil_altivec.h"
 
 static int sad16_x2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
diff --git a/libavcodec/ppc/fft_altivec.c b/libavcodec/ppc/fft_altivec.c
index e171665b37..651ee2680f 100644
--- a/libavcodec/ppc/fft_altivec.c
+++ b/libavcodec/ppc/fft_altivec.c
@@ -19,9 +19,10 @@
  * License along with FFmpeg; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
+
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/fft.h"
-#include "util_altivec.h"
-#include "types_altivec.h"
 
 /**
  * Do a complex FFT with the parameters defined in ff_fft_init(). The
diff --git a/libavcodec/ppc/float_altivec.c b/libavcodec/ppc/float_altivec.c
index a724a46d5c..e7b04ece6a 100644
--- a/libavcodec/ppc/float_altivec.c
+++ b/libavcodec/ppc/float_altivec.c
@@ -18,25 +18,10 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/dsputil.h"
 
 #include "dsputil_altivec.h"
-#include "util_altivec.h"
-
-static void vector_fmul_altivec(float *dst, const float *src0, const float *src1, int len)
-{
-    int i;
-    vector float d0, d1, s, zero = (vector float)vec_splat_u32(0);
-    for(i=0; i<len-7; i+=8) {
-        d0 = vec_ld(0, src0+i);
-        s = vec_ld(0, src1+i);
-        d1 = vec_ld(16, src0+i);
-        d0 = vec_madd(d0, s, zero);
-        d1 = vec_madd(d1, vec_ld(16,src1+i), zero);
-        vec_st(d0, 0, dst+i);
-        vec_st(d1, 16, dst+i);
-    }
-}
 
 static void vector_fmul_reverse_altivec(float *dst, const float *src0,
                                         const float *src1, int len)
@@ -124,7 +109,6 @@ static void vector_fmul_window_altivec(float *dst, const float *src0, const floa
 
 void ff_float_init_altivec(DSPContext* c, AVCodecContext *avctx)
 {
-    c->vector_fmul = vector_fmul_altivec;
     c->vector_fmul_reverse = vector_fmul_reverse_altivec;
     c->vector_fmul_add = vector_fmul_add_altivec;
     if(!(avctx->flags & CODEC_FLAG_BITEXACT)) {
diff --git a/libavcodec/ppc/fmtconvert_altivec.c b/libavcodec/ppc/fmtconvert_altivec.c
index 30de0e637f..6f24199023 100644
--- a/libavcodec/ppc/fmtconvert_altivec.c
+++ b/libavcodec/ppc/fmtconvert_altivec.c
@@ -20,8 +20,8 @@
 
 #include "libavcodec/fmtconvert.h"
 
+#include "libavutil/ppc/util_altivec.h"
 #include "dsputil_altivec.h"
-#include "util_altivec.h"
 
 static void int32_to_float_fmul_scalar_altivec(float *dst, const int *src, float mul, int len)
 {
diff --git a/libavcodec/ppc/gmc_altivec.c b/libavcodec/ppc/gmc_altivec.c
index 1341594ee5..fadf498f09 100644
--- a/libavcodec/ppc/gmc_altivec.c
+++ b/libavcodec/ppc/gmc_altivec.c
@@ -20,9 +20,9 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/dsputil.h"
-#include "util_altivec.h"
-#include "types_altivec.h"
 #include "dsputil_altivec.h"
 
 /*
diff --git a/libavcodec/ppc/h264_altivec.c b/libavcodec/ppc/h264_altivec.c
index 4589c4f96d..3e7edc74df 100644
--- a/libavcodec/ppc/h264_altivec.c
+++ b/libavcodec/ppc/h264_altivec.c
@@ -19,13 +19,13 @@
  */
 
 #include "libavutil/cpu.h"
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/dsputil.h"
 #include "libavcodec/h264data.h"
 #include "libavcodec/h264dsp.h"
 
 #include "dsputil_altivec.h"
-#include "util_altivec.h"
-#include "types_altivec.h"
 
 #define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
 #define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)
diff --git a/libavcodec/ppc/idct_altivec.c b/libavcodec/ppc/idct_altivec.c
index dbf234536c..9a8b91740e 100644
--- a/libavcodec/ppc/idct_altivec.c
+++ b/libavcodec/ppc/idct_altivec.c
@@ -41,8 +41,8 @@
 #if HAVE_ALTIVEC_H
 #include <altivec.h>
 #endif
+#include "libavutil/ppc/types_altivec.h"
 #include "libavcodec/dsputil.h"
-#include "types_altivec.h"
 #include "dsputil_altivec.h"
 
 #define IDCT_HALF                                       \
diff --git a/libavcodec/ppc/int_altivec.c b/libavcodec/ppc/int_altivec.c
index 57e63da9f0..5a73876ad3 100644
--- a/libavcodec/ppc/int_altivec.c
+++ b/libavcodec/ppc/int_altivec.c
@@ -28,12 +28,11 @@
 #include <altivec.h>
 #endif
 
+#include "libavutil/ppc/types_altivec.h"
 #include "libavcodec/dsputil.h"
 
 #include "dsputil_altivec.h"
 
-#include "types_altivec.h"
-
 static int ssd_int8_vs_int16_altivec(const int8_t *pix1, const int16_t *pix2,
                                      int size) {
     int i, size16;
diff --git a/libavcodec/ppc/mpegaudiodec_altivec.c b/libavcodec/ppc/mpegaudiodec_altivec.c
index 2de5dd133a..7d829b86f4 100644
--- a/libavcodec/ppc/mpegaudiodec_altivec.c
+++ b/libavcodec/ppc/mpegaudiodec_altivec.c
@@ -20,7 +20,7 @@
  */
 
 #include "dsputil_altivec.h"
-#include "util_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/dsputil.h"
 #include "libavcodec/mpegaudiodsp.h"
 
diff --git a/libavcodec/ppc/mpegvideo_altivec.c b/libavcodec/ppc/mpegvideo_altivec.c
index 62cd8e0adc..bbc102bc21 100644
--- a/libavcodec/ppc/mpegvideo_altivec.c
+++ b/libavcodec/ppc/mpegvideo_altivec.c
@@ -23,12 +23,13 @@
 
 #include <stdlib.h>
 #include <stdio.h>
+
 #include "libavutil/cpu.h"
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/dsputil.h"
 #include "libavcodec/mpegvideo.h"
 
-#include "util_altivec.h"
-#include "types_altivec.h"
 #include "dsputil_altivec.h"
 
 /* AltiVec version of dct_unquantize_h263
diff --git a/libavcodec/ppc/types_altivec.h b/libavcodec/ppc/types_altivec.h
deleted file mode 100644
index 36b6e1f388..0000000000
--- a/libavcodec/ppc/types_altivec.h
+++ /dev/null
@@ -1,47 +0,0 @@
-/*
- * Copyright (c) 2006 Guillaume Poirier <gpoirier@mplayerhq.hu>
- *
- * 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
- */
-
-#ifndef AVCODEC_PPC_TYPES_ALTIVEC_H
-#define AVCODEC_PPC_TYPES_ALTIVEC_H
-
-/***********************************************************************
- * Vector types
- **********************************************************************/
-#define vec_u8  vector unsigned char
-#define vec_s8  vector signed char
-#define vec_u16 vector unsigned short
-#define vec_s16 vector signed short
-#define vec_u32 vector unsigned int
-#define vec_s32 vector signed int
-#define vec_f   vector float
-
-/***********************************************************************
- * Null vector
- **********************************************************************/
-#define LOAD_ZERO const vec_u8 zerov = vec_splat_u8( 0 )
-
-#define zero_u8v  (vec_u8)  zerov
-#define zero_s8v  (vec_s8)  zerov
-#define zero_u16v (vec_u16) zerov
-#define zero_s16v (vec_s16) zerov
-#define zero_u32v (vec_u32) zerov
-#define zero_s32v (vec_s32) zerov
-
-#endif /* AVCODEC_PPC_TYPES_ALTIVEC_H */
diff --git a/libavcodec/ppc/util_altivec.h b/libavcodec/ppc/util_altivec.h
deleted file mode 100644
index 6800348570..0000000000
--- a/libavcodec/ppc/util_altivec.h
+++ /dev/null
@@ -1,118 +0,0 @@
-/*
- * 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
- */
-
-/**
- * @file
- * Contains misc utility macros and inline functions
- */
-
-#ifndef AVCODEC_PPC_UTIL_ALTIVEC_H
-#define AVCODEC_PPC_UTIL_ALTIVEC_H
-
-#include <stdint.h>
-
-#include "config.h"
-
-#if HAVE_ALTIVEC_H
-#include <altivec.h>
-#endif
-
-#include "types_altivec.h"
-
-// used to build registers permutation vectors (vcprm)
-// the 's' are for words in the _s_econd vector
-#define WORD_0 0x00,0x01,0x02,0x03
-#define WORD_1 0x04,0x05,0x06,0x07
-#define WORD_2 0x08,0x09,0x0a,0x0b
-#define WORD_3 0x0c,0x0d,0x0e,0x0f
-#define WORD_s0 0x10,0x11,0x12,0x13
-#define WORD_s1 0x14,0x15,0x16,0x17
-#define WORD_s2 0x18,0x19,0x1a,0x1b
-#define WORD_s3 0x1c,0x1d,0x1e,0x1f
-
-#define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
-#define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d}
-
-// vcprmle is used to keep the same index as in the SSE version.
-// it's the same as vcprm, with the index inversed
-// ('le' is Little Endian)
-#define vcprmle(a,b,c,d) vcprm(d,c,b,a)
-
-// used to build inverse/identity vectors (vcii)
-// n is _n_egative, p is _p_ositive
-#define FLOAT_n -1.
-#define FLOAT_p 1.
-
-
-// Transpose 8x8 matrix of 16-bit elements (in-place)
-#define TRANSPOSE8(a,b,c,d,e,f,g,h) \
-do { \
-    vector signed short A1, B1, C1, D1, E1, F1, G1, H1; \
-    vector signed short A2, B2, C2, D2, E2, F2, G2, H2; \
- \
-    A1 = vec_mergeh (a, e); \
-    B1 = vec_mergel (a, e); \
-    C1 = vec_mergeh (b, f); \
-    D1 = vec_mergel (b, f); \
-    E1 = vec_mergeh (c, g); \
-    F1 = vec_mergel (c, g); \
-    G1 = vec_mergeh (d, h); \
-    H1 = vec_mergel (d, h); \
- \
-    A2 = vec_mergeh (A1, E1); \
-    B2 = vec_mergel (A1, E1); \
-    C2 = vec_mergeh (B1, F1); \
-    D2 = vec_mergel (B1, F1); \
-    E2 = vec_mergeh (C1, G1); \
-    F2 = vec_mergel (C1, G1); \
-    G2 = vec_mergeh (D1, H1); \
-    H2 = vec_mergel (D1, H1); \
- \
-    a = vec_mergeh (A2, E2); \
-    b = vec_mergel (A2, E2); \
-    c = vec_mergeh (B2, F2); \
-    d = vec_mergel (B2, F2); \
-    e = vec_mergeh (C2, G2); \
-    f = vec_mergel (C2, G2); \
-    g = vec_mergeh (D2, H2); \
-    h = vec_mergel (D2, H2); \
-} while (0)
-
-
-/** @brief loads unaligned vector @a *src with offset @a offset
-    and returns it */
-static inline vector unsigned char unaligned_load(int offset, uint8_t *src)
-{
-    register vector unsigned char first = vec_ld(offset, src);
-    register vector unsigned char second = vec_ld(offset+15, src);
-    register vector unsigned char mask = vec_lvsl(offset, src);
-    return vec_perm(first, second, mask);
-}
-
-/**
- * loads vector known misalignment
- * @param perm_vec the align permute vector to combine the two loads from lvsl
- */
-static inline vec_u8 load_with_perm_vec(int offset, uint8_t *src, vec_u8 perm_vec)
-{
-    vec_u8 a = vec_ld(offset, src);
-    vec_u8 b = vec_ld(offset+15, src);
-    return vec_perm(a, b, perm_vec);
-}
-
-#endif /* AVCODEC_PPC_UTIL_ALTIVEC_H */
diff --git a/libavcodec/ppc/vc1dsp_altivec.c b/libavcodec/ppc/vc1dsp_altivec.c
index 69670619da..bf76adb359 100644
--- a/libavcodec/ppc/vc1dsp_altivec.c
+++ b/libavcodec/ppc/vc1dsp_altivec.c
@@ -19,12 +19,11 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/dsputil.h"
 #include "libavcodec/vc1dsp.h"
 
-#include "util_altivec.h"
-#include "dsputil_altivec.h"
-
 // main steps of 8x8 transform
 #define STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_rnd) \
 do { \
diff --git a/libavcodec/ppc/vp3dsp_altivec.c b/libavcodec/ppc/vp3dsp_altivec.c
index b0509d8b64..950e5c7419 100644
--- a/libavcodec/ppc/vp3dsp_altivec.c
+++ b/libavcodec/ppc/vp3dsp_altivec.c
@@ -18,9 +18,9 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/dsputil.h"
-#include "util_altivec.h"
-#include "types_altivec.h"
 #include "dsputil_altivec.h"
 
 static const vec_s16 constants =
diff --git a/libavcodec/ppc/vp8dsp_altivec.c b/libavcodec/ppc/vp8dsp_altivec.c
index 7f7d3bd35c..4c86ca5f74 100644
--- a/libavcodec/ppc/vp8dsp_altivec.c
+++ b/libavcodec/ppc/vp8dsp_altivec.c
@@ -21,10 +21,10 @@
  */
 
 #include "libavutil/cpu.h"
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
 #include "libavcodec/vp8dsp.h"
 #include "dsputil_altivec.h"
-#include "types_altivec.h"
-#include "util_altivec.h"
 
 #define REPT4(...) { __VA_ARGS__, __VA_ARGS__, __VA_ARGS__, __VA_ARGS__ }
 
diff --git a/libavcodec/ra288.c b/libavcodec/ra288.c
index 2d1208d8fd..ca10565a2e 100644
--- a/libavcodec/ra288.c
+++ b/libavcodec/ra288.c
@@ -19,6 +19,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavutil/float_dsp.h"
 #include "avcodec.h"
 #define BITSTREAM_READER_LE
 #include "get_bits.h"
@@ -26,7 +27,6 @@
 #include "lpc.h"
 #include "celp_math.h"
 #include "celp_filters.h"
-#include "dsputil.h"
 
 #define MAX_BACKWARD_FILTER_ORDER  36
 #define MAX_BACKWARD_FILTER_LEN    40
@@ -38,6 +38,7 @@
 typedef struct {
     AVFrame frame;
     DSPContext dsp;
+    AVFloatDSPContext fdsp;
     DECLARE_ALIGNED(32, float,   sp_lpc)[FFALIGN(36, 16)];   ///< LPC coefficients for speech data (spec: A)
     DECLARE_ALIGNED(32, float, gain_lpc)[FFALIGN(10, 16)];   ///< LPC coefficients for gain        (spec: GB)
 
@@ -62,7 +63,7 @@ static av_cold int ra288_decode_init(AVCodecContext *avctx)
 {
     RA288Context *ractx = avctx->priv_data;
     avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
-    ff_dsputil_init(&ractx->dsp, avctx);
+    avpriv_float_dsp_init(&ractx->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
 
     avcodec_get_frame_defaults(&ractx->frame);
     avctx->coded_frame = &ractx->frame;
@@ -137,7 +138,7 @@ static void do_hybrid_window(RA288Context *ractx,
                                             MAX_BACKWARD_FILTER_LEN   +
                                             MAX_BACKWARD_FILTER_NONREC, 16)]);
 
-    ractx->dsp.vector_fmul(work, window, hist, FFALIGN(order + n + non_rec, 16));
+    ractx->fdsp.vector_fmul(work, window, hist, FFALIGN(order + n + non_rec, 16));
 
     convolve(buffer1, work + order    , n      , order);
     convolve(buffer2, work + order + n, non_rec, order);
@@ -164,7 +165,7 @@ static void backward_filter(RA288Context *ractx,
     do_hybrid_window(ractx, order, n, non_rec, temp, hist, rec, window);
 
     if (!compute_lpc_coefs(temp, order, lpc, 0, 1, 1))
-        ractx->dsp.vector_fmul(lpc, lpc, tab, FFALIGN(order, 16));
+        ractx->fdsp.vector_fmul(lpc, lpc, tab, FFALIGN(order, 16));
 
     memmove(hist, hist + n, move_size*sizeof(*hist));
 }
diff --git a/libavcodec/twinvq.c b/libavcodec/twinvq.c
index d1ee695bce..071017da2a 100644
--- a/libavcodec/twinvq.c
+++ b/libavcodec/twinvq.c
@@ -19,6 +19,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavutil/float_dsp.h"
 #include "avcodec.h"
 #include "get_bits.h"
 #include "dsputil.h"
@@ -176,6 +177,7 @@ typedef struct TwinContext {
     AVCodecContext *avctx;
     AVFrame frame;
     DSPContext      dsp;
+    AVFloatDSPContext fdsp;
     FFTContext mdct_ctx[3];
 
     const ModeTab *mtab;
@@ -787,8 +789,8 @@ static void read_and_decode_spectrum(TwinContext *tctx, GetBitContext *gb,
             dec_bark_env(tctx, bark1[i][j], bark_use_hist[i][j], i,
                          tctx->tmp_buf, gain[sub*i+j], ftype);
 
-            tctx->dsp.vector_fmul(chunk + block_size*j, chunk + block_size*j, tctx->tmp_buf,
-                                  block_size);
+            tctx->fdsp.vector_fmul(chunk + block_size*j, chunk + block_size*j,
+                                   tctx->tmp_buf, block_size);
 
         }
 
@@ -809,7 +811,7 @@ static void read_and_decode_spectrum(TwinContext *tctx, GetBitContext *gb,
         dec_lpc_spectrum_inv(tctx, lsp, ftype, tctx->tmp_buf);
 
         for (j = 0; j < mtab->fmode[ftype].sub; j++) {
-            tctx->dsp.vector_fmul(chunk, chunk, tctx->tmp_buf, block_size);
+            tctx->fdsp.vector_fmul(chunk, chunk, tctx->tmp_buf, block_size);
             chunk += block_size;
         }
     }
@@ -1156,6 +1158,7 @@ static av_cold int twin_decode_init(AVCodecContext *avctx)
     }
 
     ff_dsputil_init(&tctx->dsp, avctx);
+    avpriv_float_dsp_init(&tctx->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
     if ((ret = init_mdct_win(tctx))) {
         av_log(avctx, AV_LOG_ERROR, "Error initializing MDCT\n");
         twin_decode_close(avctx);
diff --git a/libavcodec/vc1dsp.c b/libavcodec/vc1dsp.c
index da7593feca..b97fd5c533 100644
--- a/libavcodec/vc1dsp.c
+++ b/libavcodec/vc1dsp.c
@@ -561,7 +561,7 @@ static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int
 /** Function used to do motion compensation with bicubic interpolation
  */
 #define VC1_MSPEL_MC(OP, OPNAME)\
-static void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)\
+static av_always_inline void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)\
 {\
     int     i, j;\
 \
diff --git a/libavcodec/vorbisdec.c b/libavcodec/vorbisdec.c
index 6785105460..65bfb218ea 100644
--- a/libavcodec/vorbisdec.c
+++ b/libavcodec/vorbisdec.c
@@ -30,6 +30,7 @@
 #include <math.h>
 
 #define BITSTREAM_READER_LE
+#include "libavutil/float_dsp.h"
 #include "avcodec.h"
 #include "get_bits.h"
 #include "dsputil.h"
@@ -128,6 +129,7 @@ typedef struct vorbis_context_s {
     AVFrame frame;
     GetBitContext gb;
     DSPContext dsp;
+    AVFloatDSPContext fdsp;
     FmtConvertContext fmt_conv;
 
     FFTContext mdct[2];
@@ -987,6 +989,7 @@ static av_cold int vorbis_decode_init(AVCodecContext *avccontext)
 
     vc->avccontext = avccontext;
     ff_dsputil_init(&vc->dsp, avccontext);
+    avpriv_float_dsp_init(&vc->fdsp, avccontext->flags & CODEC_FLAG_BITEXACT);
     ff_fmt_convert_init(&vc->fmt_conv, avccontext);
 
     if (avccontext->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
@@ -1609,7 +1612,7 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
     for (j = vc->audio_channels-1;j >= 0; j--) {
         ch_floor_ptr = vc->channel_floors   + j           * blocksize / 2;
         ch_res_ptr   = vc->channel_residues + res_chan[j] * blocksize / 2;
-        vc->dsp.vector_fmul(ch_floor_ptr, ch_floor_ptr, ch_res_ptr, blocksize / 2);
+        vc->fdsp.vector_fmul(ch_floor_ptr, ch_floor_ptr, ch_res_ptr, blocksize / 2);
         mdct->imdct_half(mdct, ch_res_ptr, ch_floor_ptr);
     }
 
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index b7188d1fdc..a4e3239fb8 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -2581,11 +2581,6 @@ int  ff_add_hfyu_left_prediction_sse4(uint8_t *dst, const uint8_t *src,
 
 float ff_scalarproduct_float_sse(const float *v1, const float *v2, int order);
 
-void ff_vector_fmul_sse(float *dst, const float *src0, const float *src1,
-                        int len);
-void ff_vector_fmul_avx(float *dst, const float *src0, const float *src1,
-                        int len);
-
 void ff_vector_fmul_reverse_sse(float *dst, const float *src0,
                                 const float *src1, int len);
 void ff_vector_fmul_reverse_avx(float *dst, const float *src0,
@@ -2915,7 +2910,6 @@ static void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx, int mm_flags)
     c->vorbis_inverse_coupling = vorbis_inverse_coupling_sse;
     c->ac3_downmix             = ac3_downmix_sse;
 #if HAVE_YASM
-    c->vector_fmul         = ff_vector_fmul_sse;
     c->vector_fmul_reverse = ff_vector_fmul_reverse_sse;
     c->vector_fmul_add     = ff_vector_fmul_add_sse;
 #endif
@@ -3077,7 +3071,6 @@ static void dsputil_init_avx(DSPContext *c, AVCodecContext *avctx, int mm_flags)
         }
     }
     c->butterflies_float_interleave = ff_butterflies_float_interleave_avx;
-    c->vector_fmul = ff_vector_fmul_avx;
     c->vector_fmul_reverse = ff_vector_fmul_reverse_avx;
     c->vector_fmul_add = ff_vector_fmul_add_avx;
 #endif
diff --git a/libavcodec/x86/dsputil_yasm.asm b/libavcodec/x86/dsputil_yasm.asm
index 9e598b2a9c..2c3b6cc2bb 100644
--- a/libavcodec/x86/dsputil_yasm.asm
+++ b/libavcodec/x86/dsputil_yasm.asm
@@ -1130,38 +1130,6 @@ VECTOR_CLIP_INT32 6, 1, 0, 0
 %endif
 
 ;-----------------------------------------------------------------------------
-; void vector_fmul(float *dst, const float *src0, const float *src1, int len)
-;-----------------------------------------------------------------------------
-%macro VECTOR_FMUL 0
-cglobal vector_fmul, 4,4,2, dst, src0, src1, len
-    lea       lenq, [lend*4 - 2*mmsize]
-ALIGN 16
-.loop
-    mova      m0,   [src0q + lenq]
-    mova      m1,   [src0q + lenq + mmsize]
-    mulps     m0, m0, [src1q + lenq]
-    mulps     m1, m1, [src1q + lenq + mmsize]
-    mova      [dstq + lenq], m0
-    mova      [dstq + lenq + mmsize], m1
-
-    sub       lenq, 2*mmsize
-    jge       .loop
-%if mmsize == 32
-    vzeroupper
-    RET
-%else
-    REP_RET
-%endif
-%endmacro
-
-INIT_XMM sse
-VECTOR_FMUL
-%if HAVE_AVX
-INIT_YMM avx
-VECTOR_FMUL
-%endif
-
-;-----------------------------------------------------------------------------
 ; void vector_fmul_reverse(float *dst, const float *src0, const float *src1,
 ;                          int len)
 ;-----------------------------------------------------------------------------