Chandler Carruth
2014-May-22 06:54 UTC
[LLVMdev] Use perf tool for more accurate time measuring on Linux
On Thu, May 22, 2014 at 12:09 AM, Tobias Grosser <tobias at grosser.es> wrote:> On 21/05/2014 22:13, Yi Kong wrote: > >> On 20 May 2014 17:55, Tobias Grosser <tobias at grosser.es> wrote: >> >>> I set it to 2667.000 Mhz on parkas16. You can try if this improves >>> something. >>> >> >> I don't see any sign of improvement. >> >> Since we can't get raw data any better, I think we should get this >> merged and tested on build bots for a while and see how it goes. >> > > My builders are prepared for this (and I can fix upcoming issues in case I > forgot something). Chris, Hal, Renato? What do you think?I should have mentioned this idea previously, but what about making the script gracefully degrade in the absence of these tools? That way builders won't regress but ones with the tools will just get better timing. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140522/3af0c4ab/attachment.html>
Yi Kong
2014-May-22 10:34 UTC
[LLVMdev] Use perf tool for more accurate time measuring on Linux
Updated. On 22 May 2014 07:54, Chandler Carruth <chandlerc at google.com> wrote:> I should have mentioned this idea previously, but what about making the > script gracefully degrade in the absence of these tools? That way builders > won't regress but ones with the tools will just get better timing.-------------- next part -------------- A non-text attachment was scrubbed... Name: 0001-timeit-using-perf.patch Type: text/x-patch Size: 1884 bytes Desc: not available URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140522/78803ebd/attachment.bin>
Yi Kong
2014-May-25 16:07 UTC
[LLVMdev] Use perf tool for more accurate time measuring on Linux
I've been running Hal Finkel's benchmarking mode patch along with perf tool patch, and posted the results on http://146.169.44.158:55080/. For the results I've collected until now, by defaulting to use median instead of minimum, there isn't any same-run comparison false positive result. Cheers, Yi Kong On 22 May 2014 11:34, Yi Kong <kongy.dev at gmail.com> wrote:> Updated. > > On 22 May 2014 07:54, Chandler Carruth <chandlerc at google.com> wrote: >> I should have mentioned this idea previously, but what about making the >> script gracefully degrade in the absence of these tools? That way builders >> won't regress but ones with the tools will just get better timing.
Yi Kong
2014-May-27 12:57 UTC
[LLVMdev] Use perf tool for more accurate time measuring on Linux
As all review comments were taken into account, I believe the patch is ready to be committed. I plan on doing so sometime tomorrow unless any objection is raised by then. On 22 May 2014 11:34, Yi Kong <kongy.dev at gmail.com> wrote:> Updated. > > On 22 May 2014 07:54, Chandler Carruth <chandlerc at google.com> wrote: >> I should have mentioned this idea previously, but what about making the >> script gracefully degrade in the absence of these tools? That way builders >> won't regress but ones with the tools will just get better timing.-------------- next part -------------- A non-text attachment was scrubbed... Name: 0001-Timeit-using-perf-on-Linux.patch Type: text/x-patch Size: 1823 bytes Desc: not available URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140527/e2182d1e/attachment.bin>