[CELT-dev] Optimisation Help

Gregory Maxwell gmaxwell at gmail.com
Fri Feb 12 04:43:00 PST 2010

On Fri, Feb 12, 2010 at 2:03 AM, Chen-Po Sun <chenpo at fmod.org> wrote:
> Hi,
> I have been looking into optimising the CELT decoder for speed to make it
> acceptable for use in games, we would need it to be at least twice as fast
> as it currently is for this. I was hoping to be able to crunch some things
> down with some SIMD but there doesn't seem to be any good candidates for
> that.

Profiling 0.7.1's decoder here I see the IMDCT (kiss_fft.c and mdct.c)
at the top of the profile with 23.93% of the estimated cycle count. It
can be simd-ized, especially for the short block case. Because of the
reduced overlap window in CELT some operations could probably also be

Doing that wouldn't get you to your 2x alone, but if you are using
only a single frame size it's probably the easiest target for
improvement especially as optimized FFT like algorithms are a fairly
widely studied subject.

More information about the celt-dev mailing list