make -C $GYRE_DIR test on m1 macs

Bug/problem reports for any of the GYRE executables (gyre_ad, gyre_nad, etc)
Post Reply
fxt
Posts: 2
Joined: Thu Jul 14, 2022 2:08 pm

make -C $GYRE_DIR test on m1 macs

Post by fxt » Sun Jul 17, 2022 12:26 pm

hi everyone,

i have gyre-6.0.1 installed on a variety of intel and m1 macs.
i'm using mesasdk-x86_64-macos-22.6.1.pkg on the intel machines
and mesasdk-aarch64-macos-22.6.1.pkg on the m1 machines.

% make -C $GYRE_DIR test

on the intel machines passes all tests.
on the m1 machines, all tests involving hd5 fail, for example,

TEST numerics (DZIEM variables)...
dataset: </E_norm> and </E_norm>
28 differences found
...failed: summary.h5 and ref/summary.h5 differ

i'm curious if other have seen the same thing on their m1 macs,
or if its my-local-problem, or something else.

thanks!

fxt

User avatar
rhtownsend
Site Admin
Posts: 397
Joined: Sun Mar 31, 2013 4:22 pm

Re: make -C $GYRE_DIR test on m1 macs

Post by rhtownsend » Thu Aug 18, 2022 7:45 am

(realize I never answered this).

This, ultimately, is due to the new gcc 12 compiler that ships with the SDK for MacOS on ARM (M1) processors. The compiler breaks bit-for-bit compatibility with previous compilers, hence the errors.

Importantly, however, they are reproducible and nothing to worry about. So, carry on GYRE-ing on the M1!

fxt
Posts: 2
Joined: Thu Jul 14, 2022 2:08 pm

Re: make -C $GYRE_DIR test on m1 macs

Post by fxt » Sun Aug 21, 2022 12:38 pm

hi rich,

thanks for this info.

So, carry on GYRE-ing on the M1!

will do!

https://ui.adsabs.harvard.edu/abs/2022A ... C/abstract

fxt

Post Reply