On 10.06.14 22:03, Eric J Korpela wrote:
The only thought I've had recently is to have a "calibration" plan class that has a non-SIMD non-threaded unoptimized CPU-only app_version that gets sent out once out of every N (~100,000) results. This (as the least efficient app_version) could set the pfc_scale. Again, it would require project participation, so it wouldn't work.
That reminds me of the idea of "pilot" applications that would test a particular host for the feasibility of a particular app (version), e.g. hardware properties (motion sensors?) or installed software (versions) (like libc version on Linux) that the Client doesn't detect and report. Such applications also could run project application specific benchmarks, that are (presumably) much closer to the actual performance of project applications on a given host.
However setting up these applications would require rather more work from the projects, than less (which was the original intention). Best, Bernd _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
