On Sep 12, 2011, at 8:51 AM, Jeff Squyres wrote:
> *** Remember that although the opal_hwloc_topology global variable will
> always be available, ##IT MAY BE NULL## on platforms where hwloc was compiled
> out / not supported. Therefore, you MUST protect access to hwloc API calls
> with #if OPAL_HAVE_HWLOC! See the original RFC text below.
Oops! Ralph just reminded me that this was slightly inaccurate.
If hwloc is not present, then the global variable opal_hwloc_topology won't be
present at all (because its corresponding hwloc type won't be available).
Hence, the example in the original RFC isn't quite right:
>> if (NULL != opal_hwloc_topology) {
>> #if OPAL_HAVE_HWLOC
>> // ...use hwloc API, etc.
>> #endif
>> }
This really should be:
#if OPAL_HAVE_HWLOC
if (NULL != opal_hwloc_topology) {
// ...use hwloc API, etc.
}
#endif
Sorry for the confusion!
--
Jeff Squyres
[email protected]
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
