On Thu, Jun 11, 2009 at 11:40 AM, Bill Hart<[email protected]> wrote:
> These timings look suspiciously close, but I don't think they should > be. Perhaps we are wrong and the time complexity of this benchmark > doesn't depend much on MPIR at all, but other things? Or is it > possible we've picked up a system wide GMP by accident when linking in > this benchmark? > > Even untuned the FFT should be significantly faster in MPIR 1.2, so I > doubt tuning will be the issue. How repeatable are these timings? And > can you check with ldd which library it is actually linking against, > just to make sure. I'm linking the GMP and MPIR libraries statically so there isn't any accidental system-wide by accident. I just did a make tune (without the secondary long FFT tuning, just the one built into the tuneup app) and it didn't make a difference, the speed is the same. On the core2 side, the timings are the same between 1.1 and 1.2. This is just for the BPSW code not the other parts of the benchmark Brian is working on. The only thing that is different between the tests is the GMP/MPIR libraries, everything else is the same. The benchmarks are completely repeatable, you see the same time difference +/- 0.5 seconds, MPIR 1.1.1 is always faster than 1.2 by about 6%. I will try to use Brian's benchmark code to see if it highlights any differences. Jeff. --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en -~----------~----~----~----~------~----~------~--~---
