This is the mail archive of the
mailing list for the libc-ports project.
Re: [patch, mips] Improved memset for MIPS
- From: "Carlos O'Donell" <carlos at redhat dot com>
- To: Steve Ellcey <sellcey at mips dot com>
- Cc: libc-ports at sourceware dot org
- Date: Fri, 06 Sep 2013 13:12:24 -0400
- Subject: Re: [patch, mips] Improved memset for MIPS
- Authentication-results: sourceware.org; auth=none
- References: <93a232b5-9d0b-4a27-bbb5-16e3ae7c4b89 at BAMAIL02 dot ba dot imgtec dot org> <522957A4 dot 2030400 at redhat dot com> <1378483403 dot 5770 dot 307 dot camel at ubuntu-sellcey>
On 09/06/2013 12:03 PM, Steve Ellcey wrote:
> On Fri, 2013-09-06 at 00:18 -0400, Carlos O'Donell wrote:
>> Two things really:
>> (a) Testing details?
>> Could you please elaborate more on "some standalone performance
>> What specific benchmarks did you run?
> Basically, I just wrote and used a test program that does a bunch of
> memset's. Nothing fancy or very intricate.
Are you able to post this test program for posterity along with
>> What does the glibc microbenchmark show about your changes? Do they
>> show a benefit?
> I didn't try this, but I can. Is there anything on the glibc web page
> about how to run this benchmark? Does it happen as part of the standard
> 'make check'?
Just run `make bench', wait a while, and compare results before and after.
Look at bench/README for more details.
>> Steve, I trust your experience with MIPS, but I'd like to see all
>> of us drive a little more detail into these performance related
>> patches. I'm also curious if the microbenchmark shows a performance
>> progression. The glibc community is trying hard to add some objectivity
>> to our performance measurements, prevent performance regressions, and
>> use the tests to experiment with new implementations.
> That sounds reasonable. I just need a bit of help on where this is and
> how to run it.
My pleasure. Ask if you get stuck.
>> (b) the code formatting isn't in line with the project requirements.
> I'll fix these up and resubmit when I have the changes (and some more
> performance data).
Thanks for being accommodating.