Looks good to me. You should consider proposing it to
Code Review if you are interested in actual comments. Just two quick ones though: 1/
omp master
doesn't imply any barrier at the end so be careful that other threads might get started on what follows before the master parts is done; and 2/ which is strongly connect with my 1st point, be careful that whenever a non-blocking communication has been initiated, you
must not use the communication buffers until the communication is completed. I didn't check if it is the case here, but just saying. —
Gilles 19 secs ago