fmtconvert: Add a new method, int32_to_float_fmul_array8

This is similar to int32_to_float_fmul_scalar, but
loads a new scalar multiplier every 8 input samples.
This enables the use of much larger input arrays, which
is important for pipelining on some CPUs (such as
ARMv6).

Signed-off-by: Martin Storsjö <martin@martin.st>

2 files changed, 26 insertions, 0 deletions

diff --git a/libavcodec/fmtconvert.c b/libavcodec/fmtconvert.c
index 54f70305df..63f62c396a 100644
--- a/libavcodec/fmtconvert.c
+++ b/libavcodec/fmtconvert.c
@@ -32,6 +32,15 @@ static void int32_to_float_fmul_scalar_c(float *dst, const int32_t *src,
         dst[i] = src[i] * mul;
 }
 
+static void int32_to_float_fmul_array8_c(FmtConvertContext *c, float *dst,
+                                         const int32_t *src, const float *mul,
+                                         int len)
+{
+    int i;
+    for (i = 0; i < len; i += 8)
+        c->int32_to_float_fmul_scalar(&dst[i], &src[i], *mul++, 8);
+}
+
 static av_always_inline int float_to_int16_one(const float *src){
     return av_clip_int16(lrintf(*src));
 }
@@ -81,6 +90,7 @@ void ff_float_interleave_c(float *dst, const float **src, unsigned int len,
 av_cold void ff_fmt_convert_init(FmtConvertContext *c, AVCodecContext *avctx)
 {
     c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_c;
+    c->int32_to_float_fmul_array8 = int32_to_float_fmul_array8_c;
     c->float_to_int16             = float_to_int16_c;
     c->float_to_int16_interleave  = float_to_int16_interleave_c;
     c->float_interleave           = ff_float_interleave_c;
diff --git a/libavcodec/fmtconvert.h b/libavcodec/fmtconvert.h
index a28af36c21..bd833efec6 100644
--- a/libavcodec/fmtconvert.h
+++ b/libavcodec/fmtconvert.h
@@ -39,6 +39,22 @@ typedef struct FmtConvertContext {
                                        float mul, int len);
 
     /**
+     * Convert an array of int32_t to float and multiply by a float value from another array,
+     * stepping along the float array once for each 8 integers.
+     * @param c   pointer to FmtConvertContext.
+     * @param dst destination array of float.
+     *            constraints: 16-byte aligned
+     * @param src source array of int32_t.
+     *            constraints: 16-byte aligned
+     * @param mul source array of float multipliers.
+     * @param len number of elements to convert.
+     *            constraints: multiple of 8
+     */
+    void (*int32_to_float_fmul_array8)(struct FmtConvertContext *c,
+                                       float *dst, const int32_t *src,
+                                       const float *mul, int len);
+
+    /**
      * Convert an array of float to an array of int16_t.
      *
      * Convert floats from in the range [-32768.0,32767.0] to ints