[FFmpeg-devel] [PATCH 1/2] avutil: add float_dsp.vector_dmul

[James Almer]

On 9/12/2018 7:42 AM, Paul B Mahol wrote:
> Signed-off-by: Paul B Mahol <onemda at gmail.com>
> ---
>  libavutil/float_dsp.c |  9 +++++++++
>  libavutil/float_dsp.h | 16 ++++++++++++++++
>  2 files changed, 25 insertions(+)

Ah, i had a patch like this lying around but never got to send it.
I still have the x86 simd implementation, though, so I'll send that later.

> 
> diff --git a/libavutil/float_dsp.c b/libavutil/float_dsp.c
> index 1d4911d815..6e28d71b57 100644
> --- a/libavutil/float_dsp.c
> +++ b/libavutil/float_dsp.c
> @@ -32,6 +32,14 @@ static void vector_fmul_c(float *dst, const float *src0, const float *src1,
>          dst[i] = src0[i] * src1[i];
>  }
>  
> +static void vector_dmul_c(double *dst, const double *src0, const double *src1,
> +                          int len)
> +{
> +    int i;
> +    for (i = 0; i < len; i++)
> +        dst[i] = src0[i] * src1[i];
> +}
> +
>  static void vector_fmac_scalar_c(float *dst, const float *src, float mul,
>                                   int len)
>  {
> @@ -131,6 +139,7 @@ av_cold AVFloatDSPContext *avpriv_float_dsp_alloc(int bit_exact)
>          return NULL;
>  
>      fdsp->vector_fmul = vector_fmul_c;
> +    fdsp->vector_dmul = vector_dmul_c;
>      fdsp->vector_fmac_scalar = vector_fmac_scalar_c;
>      fdsp->vector_fmul_scalar = vector_fmul_scalar_c;
>      fdsp->vector_dmac_scalar = vector_dmac_scalar_c;
> diff --git a/libavutil/float_dsp.h b/libavutil/float_dsp.h
> index 2c24d93471..9c664592bd 100644
> --- a/libavutil/float_dsp.h
> +++ b/libavutil/float_dsp.h
> @@ -173,6 +173,22 @@ typedef struct AVFloatDSPContext {
>       * @return sum of elementwise products
>       */
>      float (*scalarproduct_float)(const float *v1, const float *v2, int len);
> +
> +    /**
> +     * Calculate the entry wise product of two vectors of doubles and store the result in
> +     * a vector of doubles.
> +     *
> +     * @param dst  output vector
> +     *             constraints: 32-byte aligned
> +     * @param src0 first input vector
> +     *             constraints: 32-byte aligned
> +     * @param src1 second input vector
> +     *             constraints: 32-byte aligned
> +     * @param len  number of elements in the input
> +     *             constraints: multiple of 16

Why not 8?

> +     */
> +    void (*vector_dmul)(double *dst, const double *src0, const double *src1,
> +                        int len);
>  } AVFloatDSPContext;
>  
>  /**
> 

LGTM.