On Tue, Feb 08, 2005 at 12:31:13PM -0600, Paul Leppert wrote: > were dropped (if any)? > - What's a good way to determine cpu "idleness" during the test?
I have been considering that mythcommflag might be a good benchmark to test systems. We would put out a couple of standard videos, a few minutes long, in 1080i and 720p that take lots of CPU. Start them playing and run mythcommflag on another standard video, and it prints out the frames/second it is getting in the spare CPU. That is a good measure of the spare CPU because it does it trying a task we actually want to use the spare CPU for. One could also try transcoding. Mythcommflag currently does a cumulative fps, it could be modified for an fps averaged over the last 5 seconds or something. There are two kinds of benchmarks. Synthetics are easy to do and run but may not tell you enough about real world results. Real benchmarks on code you want to run tell you a lot more. Unfortunately the machine idle time on hyperthreaded Pentiums is somewhat synthetic. The CPU says it has 60% idle CPU available but it could be lying.
_______________________________________________ mythtv-users mailing list mythtv-users@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users