Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!decwrl!hplabs!hpfcdc!stroyan
From: stroyan@hpfcdc.HP.COM (Mike Stroyan)
Newsgroups: comp.windows.x
Subject: Re: X benchmarking
Message-ID: <6910008@hpfcdc.HP.COM>
Date: 5 May 89 22:31:10 GMT
References: <8905050502.AA03307@xuucp.ch.apollo.com>
Organization: HP Ft. Collins, Co.
Lines: 18

> Right? If so, how can a benchmark client tell when the
> graphics requests are really finished?

> Joe Bowbeer

Firing from the hip, it seems that XGetImage of a pixel would have to
wait for rendering to complete.  The time required for the XGetImage
might distort measurements a bit.  A benchmark could compare a set of
operations followed by XGetImage with lone XGetImage.  The difference
should be close to the true time for all but the most bizarre server
implementations.

The time based on XQueryPointer should be of interest as well.  If a
pipelined graphics processor can perform rendering in parallel, then the
shorter time required to put data in the pipe will reflect the actual
performance of some applications.

Mike Stroyan, stroyan@hpfcla.hp.com