On 04/25/2011 12:21 PM, Aℓex Converse wrote: > On Sat, Apr 23, 2011 at 1:35 PM, Ronald S. Bultje <[email protected]> wrote: >> Hi, >> >> On Sat, Apr 23, 2011 at 4:00 PM, Justin Ruggles >> <[email protected]> wrote: >>> On 04/23/2011 03:55 PM, Justin Ruggles wrote: >>> >>>> >>>> Based on patches by clsid2 in ffdshow-tryout. >>>> --- >>>> configure | 2 ++ >>>> libavcodec/aac.h | 2 +- >>>> libavcodec/aacdec.c | 37 ++++++++++++++++++++++++++++++------- >>>> libavcodec/aacsbr.c | 7 +++++-- >>>> libavcodec/ac3dec.c | 33 ++++++++++++++++++++++++++++++++- >>>> libavcodec/dca.c | 28 +++++++++++++++++++++++++--- >>>> libavcodec/vorbis_dec.c | 23 ++++++++++++++++++++++- >>>> 7 files changed, 117 insertions(+), 15 deletions(-) >>> >>> >>> This updated patch fixes decoding AAC w/ SBR. But one of the fate aac >>> samples doesn't pass due to being slightly off. I don't know the best >>> way to fix that... >>> >>> TEST aac-am00_88 >>> stddev: 1.91 PSNR: 90.69 MAXDIFF: 64 bytes: 11563008/ 11563008 >> >> For float, just increase the amount by which it is allowed to be off >> in the test specs. >> > > 64 seems fairly large. Are you sure scale isn't off somewhere in the > main predictor?
Yeah, I thought 64 was kind of large too. I'll do some more analysis on that sample. -Justin _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
