@@ -154,8 +154,6 @@ 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,

                                 int len);

-void ff_vector_fmac_scalar_neon(float *dst, const float *src, float mul,

-                                int len);

 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,

@@ -329,7 +327,6 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)

     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;

     c->butterflies_float          = ff_butterflies_float_neon;

     c->scalarproduct_float        = ff_scalarproduct_float_neon;

     c->vector_fmul_reverse        = ff_vector_fmul_reverse_neon;

@@ -682,54 +682,6 @@ NOVFP   vdup.32         q8,  r2

         .unreq          len

 endfunc

-function ff_vector_fmac_scalar_neon, export=1

-VFP     len .req r2

-VFP     acc .req r3

-NOVFP   len .req r3

-NOVFP   acc .req r2

-VFP     vdup.32         q15, d0[0]

-NOVFP   vdup.32         q15, r2

-        bics            r12, len, #15

-        mov             acc, r0

-        beq             3f

-        vld1.32         {q0},     [r1,:128]!

-        vld1.32         {q8},     [acc,:128]!

-        vld1.32         {q1},     [r1,:128]!

-        vld1.32         {q9},     [acc,:128]!

-1:      vmla.f32        q8,  q0,  q15

-        vld1.32         {q2},     [r1,:128]!

-        vld1.32         {q10},    [acc,:128]!

-        vmla.f32        q9,  q1,  q15

-        vld1.32         {q3},     [r1,:128]!

-        vld1.32         {q11},    [acc,:128]!

-        vmla.f32        q10, q2,  q15

-        vst1.32         {q8},     [r0,:128]!

-        vmla.f32        q11, q3,  q15

-        vst1.32         {q9},     [r0,:128]!

-        subs            r12, r12, #16

-        beq             2f

-        vld1.32         {q0},     [r1,:128]!

-        vld1.32         {q8},     [acc,:128]!

-        vst1.32         {q10},    [r0,:128]!

-        vld1.32         {q1},     [r1,:128]!

-        vld1.32         {q9},     [acc,:128]!

-        vst1.32         {q11},    [r0,:128]!

-        b               1b

-2:      vst1.32         {q10},    [r0,:128]!

-        vst1.32         {q11},    [r0,:128]!

-        ands            len, len, #15

-        it              eq

-        bxeq            lr

-3:      vld1.32         {q0},     [r1,:128]!

-        vld1.32         {q8},     [acc,:128]!

-        vmla.f32        q8,  q0,  q15

-        vst1.32         {q8},     [r0,:128]!

-        subs            len, len, #4

-        bgt             3b

-        bx              lr

-        .unreq          len

-endfunc

-

 function ff_butterflies_float_neon, export=1

 1:      vld1.32         {q0},[r0,:128]

         vld1.32         {q1},[r1,:128]

@@ -27,6 +27,7 @@

 #include <stdio.h>

 #include "libavutil/common.h"

+#include "libavutil/float_dsp.h"

 #include "libavutil/intmath.h"

 #include "libavutil/intreadwrite.h"

 #include "libavutil/mathematics.h"

@@ -383,7 +384,7 @@ typedef struct {

     int profile;

     int debug_flag;             ///< used for suppressing repeated error messages output

-    DSPContext dsp;

+    AVFloatDSPContext fdsp;

     FFTContext imdct;

     SynthFilterContext synth;

     DCADSPContext dcadsp;

@@ -1865,8 +1866,8 @@ static int dca_decode_frame(AVCodecContext *avctx, void *data,

             float *back_chan = s->samples + s->channel_order_tab[s->xch_base_channel]     * 256;

             float *lt_chan   = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;

             float *rt_chan   = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;

-            s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);

-            s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);

+            s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);

+            s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);

         }

         if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {

@@ -1908,7 +1909,7 @@ static av_cold int dca_decode_init(AVCodecContext *avctx)

     s->avctx = avctx;

     dca_init_vlcs();

-    ff_dsputil_init(&s->dsp, avctx);

+    avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);

     ff_mdct_init(&s->imdct, 6, 1, 1.0);

     ff_synth_filter_init(&s->synth);

     ff_dcadsp_init(&s->dcadsp);

@@ -2401,14 +2401,6 @@ static void vector_fmul_scalar_c(float *dst, const float *src, float mul,

         dst[i] = src[i] * mul;

 }

-static void vector_fmac_scalar_c(float *dst, const float *src, float mul,

-                                 int len)

-{

-    int i;

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

-        dst[i] += src[i] * mul;

-}

-

 static void butterflies_float_c(float *restrict v1, float *restrict v2,

                                 int len)

 {

@@ -2904,7 +2896,6 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)

     c->butterflies_float = butterflies_float_c;

     c->butterflies_float_interleave = butterflies_float_interleave_c;

     c->vector_fmul_scalar = vector_fmul_scalar_c;

-    c->vector_fmac_scalar = vector_fmac_scalar_c;

     c->shrink[0]= av_image_copy_plane;

     c->shrink[1]= ff_shrink22;

@@ -417,17 +417,6 @@ typedef struct DSPContext {

     void (*vector_fmul_scalar)(float *dst, const float *src, float mul,

                                int len);

     /**

-     * Multiply a vector of floats by a scalar float and add to

-     * destination vector.  Source and destination vectors must

-     * overlap exactly or not at all.

-     * @param dst result vector, 16-byte aligned

-     * @param src input vector, 16-byte aligned

-     * @param mul scalar value

-     * @param len length of vector, multiple of 4

-     */

-    void (*vector_fmac_scalar)(float *dst, const float *src, float mul,

-                               int len);

-    /**

      * Calculate the scalar product of two vectors of floats.

      * @param v1  first vector, 16-byte aligned

      * @param v2  second vector, 16-byte aligned

@@ -26,7 +26,11 @@

 void ff_vector_fmul_neon(float *dst, const float *src0, const float *src1, int len);

+void ff_vector_fmac_scalar_neon(float *dst, const float *src, float mul,

+                                int len);

+

 void ff_float_dsp_init_neon(AVFloatDSPContext *fdsp)

 {

     fdsp->vector_fmul = ff_vector_fmul_neon;

+    fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_neon;

 }

@@ -62,3 +62,51 @@ function ff_vector_fmul_neon, export=1

 3:      vst1.32         {d16-d19},[r0,:128]!

         bx              lr

 endfunc

+

+function ff_vector_fmac_scalar_neon, export=1

+VFP     len .req r2

+VFP     acc .req r3

+NOVFP   len .req r3

+NOVFP   acc .req r2

+VFP     vdup.32         q15, d0[0]

+NOVFP   vdup.32         q15, r2

+        bics            r12, len, #15

+        mov             acc, r0

+        beq             3f

+        vld1.32         {q0},     [r1,:128]!

+        vld1.32         {q8},     [acc,:128]!

+        vld1.32         {q1},     [r1,:128]!

+        vld1.32         {q9},     [acc,:128]!

+1:      vmla.f32        q8,  q0,  q15

+        vld1.32         {q2},     [r1,:128]!

+        vld1.32         {q10},    [acc,:128]!

+        vmla.f32        q9,  q1,  q15

+        vld1.32         {q3},     [r1,:128]!

+        vld1.32         {q11},    [acc,:128]!

+        vmla.f32        q10, q2,  q15

+        vst1.32         {q8},     [r0,:128]!

+        vmla.f32        q11, q3,  q15

+        vst1.32         {q9},     [r0,:128]!

+        subs            r12, r12, #16

+        beq             2f

+        vld1.32         {q0},     [r1,:128]!

+        vld1.32         {q8},     [acc,:128]!

+        vst1.32         {q10},    [r0,:128]!

+        vld1.32         {q1},     [r1,:128]!

+        vld1.32         {q9},     [acc,:128]!

+        vst1.32         {q11},    [r0,:128]!

+        b               1b

+2:      vst1.32         {q10},    [r0,:128]!

+        vst1.32         {q11},    [r0,:128]!

+        ands            len, len, #15

+        it              eq

+        bxeq            lr

+3:      vld1.32         {q0},     [r1,:128]!

+        vld1.32         {q8},     [acc,:128]!

+        vmla.f32        q8,  q0,  q15

+        vst1.32         {q8},     [r0,:128]!

+        subs            len, len, #4

+        bgt             3b

+        bx              lr

+        .unreq          len

+endfunc

@@ -28,9 +28,18 @@ static void vector_fmul_c(float *dst, const float *src0, const float *src1,

         dst[i] = src0[i] * src1[i];

 }

+static void vector_fmac_scalar_c(float *dst, const float *src, float mul,

+                                 int len)

+{

+    int i;

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

+        dst[i] += src[i] * mul;

+}

+

 void avpriv_float_dsp_init(AVFloatDSPContext *fdsp, int bit_exact)

 {

     fdsp->vector_fmul = vector_fmul_c;

+    fdsp->vector_fmac_scalar = vector_fmac_scalar_c;

 #if ARCH_ARM

     ff_float_dsp_init_arm(fdsp);

@@ -35,6 +35,22 @@ typedef struct AVFloatDSPContext {

      */

     void (*vector_fmul)(float *dst, const float *src0, const float *src1,

                         int len);

+

+    /**

+     * Multiply a vector of floats by a scalar float and add to

+     * destination vector.  Source and destination vectors must

+     * overlap exactly or not at all.

+     *

+     * @param dst result vector

+     *            constraints: 16-byte aligned

+     * @param src input vector

+     *            constraints: 16-byte aligned

+     * @param mul scalar value

+     * @param len length of vector

+     *            constraints: multiple of 4

+     */

+    void (*vector_fmac_scalar)(float *dst, const float *src, float mul,

+                               int len);

 } AVFloatDSPContext;

 /**