[PATCH] D32881: Read public symbol names for LazyObjectFiles in parallel.

Rui Ueyama via llvm-commits llvm-commits at lists.llvm.org
Thu May 25 19:47:10 PDT 2017


Here is the table of your benchmark results. It seems they aren't different
to me.

--threads

Index          w/o Index
3.128589132    3.128069636    1.000
3.302902865    3.294943688    1.002
4.842033728    4.852418117    0.997
5.031806070    5.007721261    1.004
5.061450822    5.067567389    0.998
5.223625773    5.213661213    1.001
2.777111220    2.770632998    1.002
2.870435296    2.876185518    0.998
2.041417779    2.043296975    0.999
2.120939637    2.121668866    0.999
1.543028971    1.551262595    0.994
1.556195605    1.559701231    0.997

--no-threads

Index          w/o Index
4.648823604    4.634666491    1.003
4.656633102    4.656090134    1.000
7.298663474    7.288234015    1.001
7.310040628    7.303389055    1.000
7.508681533    7.496336533    1.001
7.517153977    7.512991823    1.000
4.089264337    4.097596474    0.997
4.082382992    4.089621870    0.998
2.202735522    2.210372640    0.996
2.205530082    2.216226573    0.995
3.156375368    3.165926995    0.996
3.158119913    3.166431410    0.997

On Mon, May 22, 2017 at 9:36 AM, Rafael EspĂ­ndola <
rafael.espindola at gmail.com> wrote:

> I finally got the time to setup a local machine. It is an mac pro
> running linux. I used the tests in
> https://drive.google.com/open?id=0B7iRtublysV6RjFBcVJhSjg1eHM, but
> excluded the ones that use thin archives.
>
> The results are attached.  The results are as stable as it gets and
> they still show a regression.
>
> Cheers,
> Rafael
>
> On 5 May 2017 at 13:06, Rui Ueyama <ruiu at google.com> wrote:
> > VPS isn't a good environment to collect benchmark numbers as it is noisy
> due
> > to other virtual machine sharing the same machine.
> >
> >
> > On Fri, May 5, 2017 at 12:34 PM, Rafael Avila de Espindola
> > <rafael.espindola at gmail.com> wrote:
> >>
> >> Rui Ueyama via Phabricator <reviews at reviews.llvm.org> writes:
> >>
> >> > ruiu added a comment.
> >> >
> >> > I think the testcase is the same. Probably that's due to difference of
> >> > machines. My workstation is beefy -- 2-socket Xeon with 10 physical
> cores
> >> > each. What's your machine and your numbers?
> >>
> >> I tested on a c4.8 large on ec2, which looks to have 18 cores.
> >>
> >> Cheers,
> >> Rafael
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170525/6dd563c1/attachment.html>


More information about the llvm-commits mailing list