On Mon, 29 Sept 2025, 16:04 Lynne via ffmpeg-devel, <[email protected]> wrote:
> PR #20627 opened by Lynne > URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20627 > Patch URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20627.patch > > Nothing major. > > > >From 485939f5ba2051d3091bb130f0879d3b1ed83e6a Mon Sep 17 00:00:00 2001 > From: Lynne <[email protected]> > Date: Tue, 30 Sep 2025 00:02:19 +0900 > Subject: [PATCH 1/2] aacdec_usac: memset uncoded coeffs instead of > iterating > > Microoptimization. > --- > libavcodec/aac/aacdec_usac.c | 6 ++---- > 1 file changed, 2 insertions(+), 4 deletions(-) > > diff --git a/libavcodec/aac/aacdec_usac.c b/libavcodec/aac/aacdec_usac.c > index c4b821bbba..38d5328029 100644 > --- a/libavcodec/aac/aacdec_usac.c > +++ b/libavcodec/aac/aacdec_usac.c > @@ -670,10 +670,8 @@ static int decode_spectrum_ac(AACDecContext *s, float > coef[1024], > > ff_aac_ac_finish(state, i, N); > > - for (; i < N/2; i++) { > - coef[2*i + 0] = 0; > - coef[2*i + 1] = 0; > - } > + /* Zero out uncoded coeffs */ > + memset(coef + 2*i, 0, N*sizeof(*coef))); > This looks wrong. It is only correct if i is guaranteed to be zero before the for loop. The number of elements to zero is not N*sizeof(...) otherwise. Kieran > _______________________________________________ ffmpeg-devel mailing list -- [email protected] To unsubscribe send an email to [email protected]
