This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Test 3 (Was: Lock elision test results)
- From: Dominik Vogt <vogt at linux dot vnet dot ibm dot com>
- To: libc-alpha at sourceware dot org
- Date: Wed, 3 Jul 2013 11:13:23 +0200
- Subject: Re: Test 3 (Was: Lock elision test results)
- References: <20130614102653 dot GA21917 at linux dot vnet dot ibm dot com> <1372767484 dot 22198 dot 4505 dot camel at triegel dot csb>
- Reply-to: Libc-Alpha at sourceware dot org
On Tue, Jul 02, 2013 at 02:18:04PM +0200, Torvald Riegel wrote:
> On Fri, 2013-06-14 at 12:26 +0200, Dominik Vogt wrote:
> > Test 3 (cacheline pingpong)
> > ======
> > Test execution
> > --------------
> >
> > Identical to test 1.
> >
> > Result
> > ------
> >
> > (1) unpatched : 100.00%
> > (2) old glibc : 103.94%
> > (3) elision off: 76.25%
> > (4) elision on : 373.38%
>
> So in this case we scale almost linearly, nice. Do you know why it is
> different in test 1? Just the abort ratio as cause of this would be
> kind of surprising, given that abort rates differ not that much
> (or aborts would have to be really costly).
Aborts _are_ expensive.
> > The abort ratio in (4) in all threads is < 0.01%.
>
> In general, I think it would be good if you could port these tests to
> the performance benchmark framework that Siddhesh has been working on,
> and contribute them to glibc. This way, we can also work on avoiding
> performance regressions as the one you reported for the 2.15 vs. current
> case without the elision patches.
I can look into this when I have the clearancy to post source
code.
Ciao
Dominik ^_^ ^_^
--
Dominik Vogt
IBM Germany