On Jul 5, 2010, at 10:33 PM, <[email protected]> wrote:
> But here's a patch:
>
> --- opal/include/opal/sys/cache.h.orig 2010-07-06 14:29:44.000000000 +1200
> +++ opal/include/opal/sys/cache.h 2010-07-06 14:32:34.000000000 +1200
> @@ -30,7 +30,9 @@
> *
> * For now hardwire this to a reasonable value, and automate later - RLG
> */
> +#ifndef CACHE_LINE_SIZE
> #define CACHE_LINE_SIZE 128
> +#endif
I did a little better than that. Given that we'll soon be moving to using
hwloc information (does hwloc work on NetBSD and FreeBSD, btw?), I converted
OMPI's use of CACHE_LINE_SIZE to opal_cache_line_size -- an int, rather than a
#define -- and removed this opal/include/opal/sys/cache.h file altogether.
(a header file for a single #define? What were we thinking?)
See the trunk commit here:
https://svn.open-mpi.org/trac/ompi/changeset/23349
If that works for you, I'll move it over to v1.5.
Do you need a corresponding change in v1.4?
--
Jeff Squyres
[email protected]
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
