arm: Add VFP-accelerated version of int32_to_float_fmul_scalar

               Before           After
               Mean    StdDev   Mean    StdDev  Change
This function   1175.0   4.4      366.2  18.3   +220.8%
Overall        19285.5 292.0    18420.5 489.1     +4.7%

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

1 files changed, 10 insertions, 0 deletions

diff --git a/libavcodec/arm/fmtconvert_init_arm.c b/libavcodec/arm/fmtconvert_init_arm.c
index 4f36729b00..9aab7bbac7 100644
--- a/libavcodec/arm/fmtconvert_init_arm.c
+++ b/libavcodec/arm/fmtconvert_init_arm.c
@@ -28,6 +28,9 @@
 void ff_int32_to_float_fmul_scalar_neon(float *dst, const int32_t *src,
                                         float mul, int len);
 
+void ff_int32_to_float_fmul_scalar_vfp(float *dst, const int32_t *src,
+                                       float mul, int len);
+
 void ff_float_to_int16_neon(int16_t *dst, const float *src, long len);
 void ff_float_to_int16_interleave_neon(int16_t *, const float **, long, int);
 
@@ -38,6 +41,13 @@ av_cold void ff_fmt_convert_init_arm(FmtConvertContext *c, AVCodecContext *avctx
     int cpu_flags = av_get_cpu_flags();
 
     if (have_vfp(cpu_flags) && have_armv6(cpu_flags)) {
+        if (!have_vfpv3(cpu_flags)) {
+            // This function doesn't use anything armv6 specific in itself,
+            // but ff_float_to_int16_vfp which is in the same assembly source
+            // file does, thus the whole file requires armv6 to be built.
+            c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_vfp;
+        }
+
         c->float_to_int16 = ff_float_to_int16_vfp;
     }