On Tue, Mar 29, 2016 at 10:51 PM, Michael Niedermayer
<mich...@niedermayer.cc> wrote:
> This is a hotfix and not a real fix of the underlaying bug
> The underlaying bug is ATM not fully understood
>
> iam not sure if we should apply this or not
>
> Signed-off-by: Michael Niedermayer <mich...@niedermayer.cc>
> ---
> libavcodec/aacenc_quantization.h | 13 +++++++++++--
> 1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/libavcodec/aacenc_quantization.h
> b/libavcodec/aacenc_quantization.h
> index 4250407..d367be0 100644
> --- a/libavcodec/aacenc_quantization.h
> +++ b/libavcodec/aacenc_quantization.h
> @@ -141,8 +141,17 @@ static av_always_inline float
> quantize_and_encode_band_cost_template(
> if (BT_ESC) {
> for (j = 0; j < 2; j++) {
> if (ff_aac_codebook_vectors[cb-1][curidx*2+j] == 64.0f) {
> - int coef = av_clip_uintp2(quant(fabsf(in[i+j]), Q,
> ROUNDING), 13);
> - int len = av_log2(coef);
> + float a = fabsf(in[i+j]) * Q;
> + double f = sqrtf(a * sqrtf(a)) + ROUNDING;
> + int coef, len;
> +
> + if (f > INT_MAX || f < 16) {
> + av_log(NULL, AV_LOG_ERROR, "f %f is out of range
> this is a internal error\n", f);
> + f = INT_MAX;
> + }
> +
> + coef = av_clip_uintp2(f, 13);
> + len = av_log2(coef);
>
> put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2);
> put_sbits(pb, len, coef);
Actually I just understood the underlying bug and am testing a fix.
Basically, scalefactors need to be bound by (roughly)
coef2minsf(maxval), which isn't being done atm, and some signals
prompt the encoder to pick lower and lower scalefactors trying to
consume unspent bits that cannot really be consumed.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
