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 1/2] Fix strrchr regression.

On Mon, Aug 05, 2013 at 07:54:14PM +0200, Liubov Dmitrieva wrote:
>    I see you didn't check for Haswell. Can you please contribute your
>    profiler to glibc to let everyone reproduce measurements of your good
>    benchmarks. I want to be able to get data for Haswell, Atom and Silvermont
>    for your benchmark to be sure that we didn't miss something. There was the
>    discussion already that it is important to improve glibc benchmarks.

Certainly I want to add it, just did not have time to make code of it pretty. 

For now I put  profiler that I used to do tests here: 

One thing you could help is that I place results to my page for now because I
did not have time to find/setup suitable hosting and it would be good
to have one.

I modified a script that I use for automating connecting to various hosts in my
department for wider usage.

It will now automatically send me results to my email. A better way with hosting
would be for example to have scponly/ssh accounts for commiters so pages can be
updated directly (easily done by changing mutt command to scp one).

To use it you need to copy profiler to directory ~/benchmark_string/strrchr_profile

Then you need to fill file bench_list with machines to test, my file is

bench i7_nehalem
bench i7_sandy_bridge
bench i7_ivy_bridge
After that run attached benchmarks_strrchr file. 


Attachment: benchmarks_strrchr
Description: Text document

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