This is the mail archive of the mailing list for the glibc project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Optimize strstr, strcasestr and memmem

On Mon, May 21, 2012 at 2:23 PM, OndÅej BÃlka <> wrote:
> I expanded benchmark to cover more cases. I added graph creation. For
> now I host it on my page
> It additionaly tests a planted strings -it gets in file distribution of
> prefixes so with distribution
> 50 0
> 40 1
> 10 8
> will character be with probabilities 50% not in needle, 40% first in
> needle, 10% member of sequence consisting of first 8 characters of
> needle.
> This distribution is one of extremes for my alg;orithm as 16-byte is
> matched with probability 40% due of last byte.

This looks really nice, thank you for working on this!

Could you please add titles and labels to the X and Y axis of each
graph (including units)?

> On Sun, May 20, 2012 at 09:57:13AM -0400, Carlos O'Donell wrote:
>> Would you be willing to work with the community to integrate such a
>> test into glibc?
>> Would you be able to help maintain the micro-benchmarks?
> Yes, I am willing.


Could you write up a page on the glibc wiki describing how to run the
benchmarks and interpret the results?

Does that sound like a good first step?

The step after that is going to be to ask the machine maintainers to
look at the benchmark for their machine.

Is the benchmark code portable? I noticed the use of rdtsc, can you
adjust to code to use something else as a fallback for HP timing if
the code is not run on x86?

Could you also looking into formatting the code following the GNU
Coding Standard such that inclusion into any GNU package could be made

Thank you again for helping out here!


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]