13.11.2014 01:16, Andrey Semashev пиÑ?еÑ?: > On Wednesday 12 November 2014 20:03:48 Alexander E. Patrakov wrote: >> 12.11.2014 14:26, Alexander E. Patrakov пиÑ?еÑ?: >>> I will recheck the quality separately later today, in order to verify >>> that it is still as good as in the previous tests. Please don't merge >>> the patches until this is done. >> Done. The -mq, -hq and -vhq variants of the resampler never produce >> audible distortions. The -lq variant sometimes does, by means of >> suppressing very high frequencies, but this is relevant to artificial >> tests only, and only if the listener knows that these frequencies are >> supposed to be there. Thus, quality is on par with speex-float-5, the >> CPU consumption is even better than with speex-float-1. Conclusion: >> >> *** the patches are generally acceptable *** > Great! And thanks a lot for the quality data and information. I will send v2 > patches in a day or two. > >> However, because the low-quality and high-quality versions eat very >> similar amount of CPU time, I'd just expose a single (high or very high) >> quality setting. > Given that -lq is actually slower than -mq in some cases and has worse > quality, I agree there is no point in keeping it. > > However, the other three presets do have different performance and quality. In > my test results [1] -mq is about 2 times faster than -vhq, and -hq is > somewhere in between. Performance wise, there should be no problem with -vhq > on modern CPUs, but maybe the little extra would be desired in embedded domain > to conserve battery. Do you think we could keep the three presets: -mq, -hq > and -vhq? > > [1]: http://lastique.github.io/src_test/ OK, let's just drop -lq. -- Alexander E. Patrakov
