[FFmpeg-devel] [PATCH] avcodec/aacenc_is: replace pow(x, 0.75) by x/sqrtf(sqrtf(x))

[Ganesh Ajjanagadde]

This is quite an accurate approximation; testing shows ~ 2ulp error in
the floating point result. Tested with FATE.

Alternatively, if one wants "full accuracy", one can use powf, or sqrt
instead of sqrtf. With powf, one gets 1 ulp error (theoretically should be 0, as
0.75 is exactly representable) on GNU libm, with sqrt, 0 ulp error.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde at gmail.com>
---
 libavcodec/aacenc_is.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libavcodec/aacenc_is.c b/libavcodec/aacenc_is.c
index 4f7ec19..846b8ec 100644
--- a/libavcodec/aacenc_is.c
+++ b/libavcodec/aacenc_is.c
@@ -54,7 +54,7 @@ struct AACISError ff_aac_is_encoding_err(AACEncContext *s, ChannelElement *cpe,
         FFPsyBand *band0 = &s->psy.ch[s->cur_channel+0].psy_bands[(w+w2)*16+g];
         FFPsyBand *band1 = &s->psy.ch[s->cur_channel+1].psy_bands[(w+w2)*16+g];
         int is_band_type, is_sf_idx = FFMAX(1, sce0->sf_idx[(w+w2)*16+g]-4);
-        float e01_34 = phase*pow(ener1/ener0, 3.0/4.0);
+        float e01_34 = phase*(ener1/ener0)/sqrtf(sqrtf(ener1/ener0));
         float maxval, dist_spec_err = 0.0f;
         float minthr = FFMIN(band0->threshold, band1->threshold);
         for (i = 0; i < sce0->ics.swb_sizes[g]; i++)
-- 
2.7.0