Review phiprof calls and add the MPI tag to relevant places
Opened this issue · 1 comments
When going through some things with Jonas I noticed that some phiprof groups doing MPI things are not tagged as such in our phiprof calls. At least, it looked like it in the field solver. I guess we should add the MPI tag to the timers for ghost cell updates at least.
Opinions?
Not a bad idea as such. Not sure how much we've really spent on looking at timer groups - it might be nice to add them also to e.g. boundaries (with the new ionosphere incoming) etc.
Also, at that point, we might want to add barriers inside 3D translation to correctly get the wait amounts after imbalanced trans, because currently it's a bit of a confusing sherlock task to figure out efficiency there. But, it's all just logging, so I think low on the priority list.