[Ffmpeg-devel] Using Intel's quantizer

[g.]

I'm looking into how to integrate Intel's IPP Quantizer to see if it is faster 
than what we have in ffmpeg.

Looking at ffmpeg's dct_quantize_ routine it has the variables "bias" 
and "qmat". Would these be equivalent to Intel's "Quantizer" and "Pointer to 
the matrix of inverted quantization coefficients"?

The full function description is:

IppStatus ippiQuant_MPEG2_16s_C1I(Ipp16s *pSrcDst, Ipp32s QP, const Ipp32f
*pQPMatrix, Ipp32s *pCount);

Parameters
pSrcDst: Pointer to the block of DCT coefficients.
QP: Quantizer.
pQPMatrix: Pointer to the matrix of inverted quantization coefficients.
pCount: Number of the non-zero coefficients after quantization.

Also, in the ffmpeg code there is

    if(s->dct_error_sum)
        s->denoise_dct(s, block);

What is that about?

g.