Jeff:

I ran a post-processor yesterday on shavano (interactive)
that iterated 135 time steps and ended up using over 14Mw
of memory -- way more than it should have.

I use fortran pointers (hpalloc, etc) extensively, so I
looked over my code to be sure I was deallocating appropriately.
I found no problems, so I ran 3 shorter jobs on shavano this
morning, iterating 5, 10, and 15 times. The jobs used HWM
2.232, 2.795, and 2.959 Mw respectively. This confirmed that
memory use was increasing with number of iterations, altho 
apparently not linearly.

Next I ran the same 3 cases on paiute, and memory use there
held steady at 2.425 Mw! It used exactly the same memory for
5, 10, and 15 steps.

Why would the memory increase on shavano but remain steady
on paiute??  I know they are running unicos 8 and 9 respectively.
I am using PrgEnv on both machines. 

--Ben