On Tue, Dec 26, 2006 at 06:40:59PM +0100, Toon Moene wrote:
A run of HIRLAM for which the grid was carefully tuned (half a year ago)
to the RAM size of my laptop suddenly started to swap two weeks ago when
using gfortran 4.3 as the compiler.
I'm not aware of anything that would do this. In fact,
Roger Sayle committed a few changes that may reduce
memory usage. Richi did some work on the memory allocation,
but I doubt that that would effect HIRLAM. There's the
commit by Richi and Diego on the 2006-12-14. It was my
impression that a smaller value for MAX_ALIASED_VOPS causes
a compile time issue not runtime.
On the library side, the two suspects are Thomas Koenig's
patches on 2006-12-01 and 2006-12-06. I would guess that
the first patch could have increased the memory usage.