jemalloc 3 performance vs. mozjemalloc

Bradley C. Kuszmaul bradley at mit.edu
Mon Feb 9 22:12:58 PST 2015


On Tue, Feb 10, 2015 at 1:00 AM, Mike Hommey <mh at glandium.org> wrote:
>
> ... except I'm talking about arm and arm has very different performance
> properties.
>
>  I'd be interested in hearing what happens on ARM (I was responding to
Don's statement that x86 lock prefixes are very expensive.  I wouldn't
characterize them that way.).  Attached is a program I wrote that measures
it.  It may require some changes to run on ARM (for example, the
__mm_pause() primitive seems to be intel-specific rather than gcc-specific)

-Bradley
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://jemalloc.net/mailman/jemalloc-discuss/attachments/20150210/71368b9d/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: lock-overhead.cc
Type: text/x-c++src
Size: 3747 bytes
Desc: not available
URL: <http://jemalloc.net/mailman/jemalloc-discuss/attachments/20150210/71368b9d/attachment.cc>


More information about the jemalloc-discuss mailing list