You'll undoubtedly get a more detailed response from an expert - but
basically on SPARC, it is able to determine the number of lgroups from
the platform code, so it doesn't need to do the probing tricks that it
has to do on the Opteron.
cheers,
steve
Mike Marty wrote:
Hello,
I've noticed that the OpenSolaris seems to support NUMA memory allocation
(memory placement optimization) for Opteron machines. Specifically the lgroups
are created in this platform-specific file:
/usr/src/uts/i86pc/os/lgrpplat.c
However the corresponding Sparc platform file appears *not* to probe memory
latencies and assumes UMA memory allocation:
/usr/src/uts/sun4/os/lgrpplat.c
In the above file, the lgrp_plat_probe() function is stubbed out whereas the i86pc file appears to have full functionality.
I know that the Sun Enterprise 25000 machine is NUMA.
I would like NUMA in-order to simulate future NUMA chip-multiprocessors using Virtutech Simics and the Wisconsin GEMS toolkit.
Why does OpenSolaris lack the Lgroup memory-probing functionality for the Sparc
platform?
Thanks,
Mike Marty
This message posted from opensolaris.org
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
--
stephen lau // [EMAIL PROTECTED] | 650.786.0845 | http://whacked.net
opensolaris // solaris kernel development
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
