[GiNaC-list] Variations of CPU time for the *same* calculation

Alexei Sheplyakov varg at metalica.kh.ua
Mon Aug 10 21:29:07 CEST 2009

Hi Luigi,

On Sun, Aug 09, 2009 at 05:19:52PM +0200, Luigi Capozza wrote:
> >> With more complicated expressions also the run time can vary from
> >> execution to execution up to a factor of 2 or more.
> > 
> > What exactly is a `run time'? CPU time? Wall clock time? Something else?
> > 
> I mean CPU time, sorry.

I haven't experienced such a weird variations, neither with `real-life' code
nor with benchmarks (which try to make term ordering even more random *on
purpose*). So you definitely found something abnormal unless some of
the following conditions are met:
- you have other CPU-bound task(s) running
- you have real-time processes running (although these guys typically don't
  consume much CPU time they can preempt other processes very frequently)
- you use power saving utility (either user-space or in-kernel).

Best regards,

More information about the GiNaC-list mailing list