Isca
Isca copied to clipboard
[WIP] working version on gfortran11 on macOS v11.2
I am opening the PR as a way to share and discuss but shouldnt be merged in its current form
This follows up on https://github.com/ExeClim/Isca/issues/210 and are modifications which eventually allowed me to compile and run held_suarez_test_case.py
on a Mac (running OS 11.2).
There were several types of issues as you'll see, some of which I uncovered after upgrading to the latest gfortran compiler (v11)
- linux system calls which error out on Mac; some I just commented out for now; the affinity part is switched off via a
MACOS
cpp flag; more or less as done by @jamesp in 2018 (https://github.com/jamesp/Isca/tree/mac) - the
-fallow-argument-mismatch
and-fallow-invalid-boz
part is related to warnings that have recently been turned into errors (the options revert to warnings). Would need a switch like in https://github.com/MITgcm/MITgcm/pull/480 probably for gfortran < v10 - in
create_xgrid.c
I needed to add a definition ofisHeadNode_b
(same as inmosaic_util.c
but for the name) - added
gfortran_mac
option file
Maybe this could get reviewed and merged after some revision. Please chime in
@gaelforget Quick question: do you have an intuition for whether or not the additional changes you've made (i.e., beyond those in Jamie's branch) would affect compilation on Mac setups for which Isca currently does compile (i.e., what Jamie had working)?
@gaelforget Quick question: do you have an intuition for whether or not the additional changes you've made (i.e., beyond those in Jamie's branch) would affect compilation on Mac setups for which Isca currently does compile (i.e., what Jamie had working)?
Except for the -fallow-argument-mismatch
and -fallow-invalid-boz
part that need to be skipped with gfortran < 10 (see https://github.com/MITgcm/MITgcm/pull/480) I would think so. Would be awesome if you could double check that
Great to see progress being made on this @gaelforget, @ntlewis!
@gaelforget how is the performance when you run on mac hardware?