[Ffmpeg-devel] int vs. float profiler, take 2

Michael Niedermayer michaelni
Sat May 21 02:44:45 CEST 2005


Hi

On Saturday 21 May 2005 02:23, Mike Melanson wrote:
> Michael Niedermayer wrote:
> > Hi
> >
> > On Saturday 21 May 2005 01:36, Mike Melanson wrote:
> >>Falk Hueffner wrote:
> >>>It is completely futile to try to describe the performance of an
> >>>operation by a single number. You need at least 3:
> >>
> >>	All benchmarks are flawed.
> >
> > no, thats not true, a benchmark tells you how good some specific code in
> > a specific enviroment in relation to some scoring metric is
> > its flawed only if applied to a different situation without considering
> > the difference
>
> 	Therefore, my benchmarks are perfectly valid, assuming a certain
> environment and set of constraints.

yes absolutely, just dont forget that its not too representative for a general 
int vs float comparission
a comparission of a (i)mdct or similar optimized for float (by you maybe) and 
int (by rich maybe) would be more representative

-- 
Michael





More information about the ffmpeg-devel mailing list