You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
PZ(1:KLON, 1) = 0.
DO JLON = 1, KLON
DO JLEV = 1, KLEV
....
ENDDO
ENDDO
Running the pool_allocator transformation will give you something like
DO JLON = 1, KLON
PZ(1:KLON, 1) = 0.
DO JLEV = 1, KLEV
....
ENDDO
ENDDO
I assume this has nothing to do with the pool_allocator and is due to the transformation that tries to merge jlon loops failing to replace the range index with a jlon?
Small example can be found in ac_cloud_model2.F90 here: https://github.com/rolfhm/minimal_test/tree/pool_infinite_recursion
The text was updated successfully, but these errors were encountered:
Hi Rolf,
thanks for this, I share your assessment. I think what you need here is to apply the resolve_vector_notation utility as part of the housekeeping in the beginning:
That should replace the vector notation by an explicit horizontal loop, and, as a consequence, then be handled correctly in the de- and re-vectorizing during SCC.
If you have something like
Running the pool_allocator transformation will give you something like
I assume this has nothing to do with the pool_allocator and is due to the transformation that tries to merge jlon loops failing to replace the range index with a jlon?
Small example can be found in ac_cloud_model2.F90 here:
https://github.com/rolfhm/minimal_test/tree/pool_infinite_recursion
The text was updated successfully, but these errors were encountered: