[Carpet] flesh timers and Carpet

[Erik Schnetter]

On Apr 15, 2008, at 12:36:26, Thomas Radke wrote:
> Hi Erik,
>
> thorn AlphaThorns/MPIClock makes the assumption that the number of  
> flesh
> timers created is the same on all processors. Is this assumption true
> for the timers created in Carpet ?
>
> The context for this: Bruno has a run which crashes in an
> MPI_Allreduce() call inside MPIClock because processor 0 wants to  
> reduce
> the values of 366 timers whereas all the other processors only know
> about 365 timers.


That is interesting.  It could be that some ASCII I/O thorn creates a  
timer only on the root processor.

I suggest to switch off MPIClock; you can instead use thorn  
TimerReport to see timing results from all processors.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from www.keyserver.net.



-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 194 bytes
Desc: This is a digitally signed message part
Url : /archives/developers/attachments/20080415/08ff05e2/attachment.pgp