[Users] Meeting minutes

[Frank Loeffler]

On Tue, Apr 12, 2011 at 09:30:59AM -0400, Roland Haas wrote:
> * RotatingSymmety90/180: Intel compilers seem to create faulty code when
> using OpenMP for the count and icount reduction operations, the counts
> end up being number_of_threads*total_number_of_points

I tracked this down to the compiler options "-ip". After removing this
option, the testsuites pass. Apparently, this option has been the cause
of trouble several times in the past already (reading from several
option list comments now), so be very careful when using it.

I am not sure if I should trust the icc help:

-[no-]ip  enable(DEFAULT)/disable single-file IP optimization
          within files

Because it states that -ip would be on by default, but I do see, using
the numrel-intel.cfg from simfactory, that the results change when I
don't use the option.

Frank

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
Url : http://lists.einsteintoolkit.org/pipermail/users/attachments/20110412/35ef1c8d/attachment.bin