[lld] r316731 - De-template EhFrameSection. NFC.
Galina Kistanova via llvm-commits
llvm-commits at lists.llvm.org
Fri Jan 5 12:44:11 PST 2018
The bot is running here against a local LNT server yet.
Could you take a look to see if the measurements consistent with what you
see locally, and if we would need to add or change something there, please?
On Fri, Dec 1, 2017 at 10:24 AM, Galina Kistanova <gkistanova at gmail.com>
> > ideally with the last released version of clang to make the results
> easier to reproduce.
> I was thinking along the same lines.
> And it might also make sense to have the clang version in the data schema,
> since it will be changing every now and then.
> On Thu, Nov 30, 2017 at 9:11 AM, Rafael Avila de Espindola <
> rafael.espindola at gmail.com> wrote:
>> Galina Kistanova <gkistanova at gmail.com> writes:
>> > Hi Rafael,
>> > The machine is ready.
>> > Did you decide what do you want to run there? I guess this one -
>> > https://reviews.llvm.org/D39506?
>> Correct. That script has to send results to a lnt instance, so we might
>> need to set one up with the link.yaml schema.
>> An example run looks like:
>> $ benchmark.py --wraper sudo,cset,shield,-e,-- --machine <name>
>> --revision <rev> <directory_with_lld_tests>
>> I have attached the script I use in my machines to setup the benchmark
>> environment. It basically implements the suggestions on
>> https://www.llvm.org/docs/Benchmarking.html: disable turbo, smt and
>> address space randomization. It then uses cset to remove a core from the
>> scheduler regular list so that we can run the benchmark on it.
>> The ld.lld being benchmarked should be build with -DLLVM_BUILD_STATIC=ON
>> -DLLVM_ENABLE_PIC=OFF and ideally with the last released version of
>> clang to make the results easier to reproduce.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the llvm-commits