[MITgcm-support] pickup after changing number of processors
Martin.Losch at awi.de
Thu Jan 7 08:41:33 EST 2010
I think this only works if you have "globalFiles" or
"useSingleCPUio=.true.", so that you get only pickup file (instead of
one per processor). If you have multiple pickup files, you probably
have to combine them into one (simple with matlab, but there's are
also a perl script somewhere in utils/scripts).
On Dec 30, 2009, at 3:31 AM, Yuan Lian wrote:
> Hi David,
> You don't need to use pickupStrictlyMatch=.FALSE. to increase number
> of processors.
> When you change number of processors, make sure the total number of
> in longitude and latitude unchanged. For instance, Nx=sNx*nSx*nPx,
> if nPx
> is doubled, sNx will be reduced to half of the original value
> (assuming you use
> same number of tiles).
> After modifying SIZE.h, you will need to recompile the code. Make
> sure the number of
> physical processors you are using matches nPx*nPy. For example, if
> nPx=8 and nPy=2,
> you need 16 processors. Hope this helps.
> David Hebert wrote:
>> I have more processors available during the middle of a run. So I
>> changed SIZE.h accordingly and resubmitted. I get the error message
>> in STDERR.0000, but the simulation seems to be running...
>> PID.TID 0000.0001) *** ERROR *** S/R EEBOOT_MINIMAL: No. of
>> processes not equal to nPx*nPy 256 128
>> Since simulation seems to be running, is this something I need to
>> be concerned with? Do I need to specify pickupStrictlyMatch=.FALSE.
>> in data file in order to use the increase in processors?
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
More information about the MITgcm-support