Re: [cairo] Re: Better coverage from cairo performance suite (and some results)



On Fri, 06 Oct 2006 15:42:57 -0700, Carl Worth wrote:
> I just ran it and it took over 34 minutes.

Which is way too long for my impatience, (not to mention it was
several times slower on the 770). So I've just implemented an adaptive
mechanism for executing fewer iterations per test when possible.

The approach I took was to set a desired threshold value for the
standard deviation (I chose 3%) and to stop iterating as soon as that
threshold was reached by a consecutive number of iterations (I chose
5).

So on my x86 laptop with the XAA server that results in an average of
about 9.5 iterations per test (rather than the previous 100 iterations
per test). And the complete suite completes in about 3 minutes
compared to 34 minutes.

So I'm pretty happy with that. But if anybody has any suggested
refinements for determining when we can declare that we are getting
stable results, I would be glad to hear them.

-Carl

Attachment: pgpuVqBCrxG2p.pgp
Description: PGP signature



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]