This is the mail archive of the mailing list for the libc-ports 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, mips] Improved memset for MIPS

On Thu, Dec 12, 2013 at 5:19 PM, Andrew Pinski <> wrote:
> On Fri, Sep 20, 2013 at 10:32 AM, Steve Ellcey <> wrote:
>> On Fri, 2013-09-20 at 16:43 +0000, Joseph S. Myers wrote:
>>> On Wed, 18 Sep 2013, Steve Ellcey wrote:
>>> > 2013-09-18  Steve Ellcey  <>
>>> >
>>> >     * sysdeps/mips/memset.S: Change prefetching and add loop unrolling.
>>> >     * sysdeps/mips/mips64/memset.S: Remove.
>>> OK, though I believe "proAptive" in a comment should be "proAptiv".
>> You are right.  I fixed that and checked in the patch.
> I noticed this patch causes some performance regressions on Octeon due
> to having 128 byte cache lines.
> Changing PREFETCH_CHUNK/PREFETCH_FOR_STORE to assume 128 byte cache
> line gives us the performance back and improves over the original code
> at least 15%.
> That is:
> #  define PREFETCH_CHUNK 128
> #  define PREFETCH_FOR_STORE(chunk, reg) \
>     pref PREFETCH_STORE_HINT, (chunk)*128(reg);

Submit a patch for that?

We have microbenchmarks now, but the next hardest
part is going to be archiving data by device so that
the community can help track performance and point
out regressions like this.


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