aacenc_is: take absolute coefficient value upon energy calculations

This was an oversight when the IS system was being first implemented.
The ener01 part was largely a result of trial and error and the fact
that the sum of coef0 and coef1 could result in a zero was
overlooked. Once ener01 turns to zero it's used to divide the left
channel energy which doesn't turn out so well as it fills IS[]
with -nan's and inf's which in turn confused the quantize_band_cost.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

1 files changed, 2 insertions, 2 deletions

diff --git a/libavcodec/aacenc_is.c b/libavcodec/aacenc_is.c
index 95ba3f8d50..e983b7548f 100644
--- a/libavcodec/aacenc_is.c
+++ b/libavcodec/aacenc_is.c
@@ -114,8 +114,8 @@ void ff_aac_search_for_is(AACEncContext *s, AVCodecContext *avctx, ChannelElemen
                 }
                 for (w2 = 0; w2 < sce0->ics.group_len[w]; w2++) {
                     for (i = 0; i < sce0->ics.swb_sizes[g]; i++) {
-                        float coef0 = sce0->pcoeffs[start+(w+w2)*128+i];
-                        float coef1 = sce1->pcoeffs[start+(w+w2)*128+i];
+                        float coef0 = fabsf(sce0->pcoeffs[start+(w+w2)*128+i]);
+                        float coef1 = fabsf(sce1->pcoeffs[start+(w+w2)*128+i]);
                         ener0  += coef0*coef0;
                         ener1  += coef1*coef1;
                         ener01 += (coef0 + coef1)*(coef0 + coef1);