Luca Bertagna
Luca Bertagna
This was completed in #2750
It is possible that this is NOT the bad point. I looked at the code, and it seems clean (there's a separate issue, as I pointed out on slack, but...
Actually, the SPA pr that just got merged changes the init sequence of SPA. We no longer call the `get_global_min_dof_gid` in SPA, so we should re-run with current master, to...
TBC, I don't expect the SPA pr to fix any bug that was in the code. At least not very likely. But we may as well continue the debug effort...
I quickly went through the code (before the SPA pr merge), and I _think_ the last place where MPI happens before the SPA::set_grid call is when the AD builds the...
Since we can't prove that the pb is gone by simply not seeing it, we need to use some statistical approach. @ndkeen how often did you notice the hang before...
``` - sfc_alb_dir_vis - sfc_flux_lw_dn - sfc_flux_sw_net ``` are in our default output, and our nightly tests pass on pm-gpu/crusher. The `monolithic_vs_restarted` test passes on weaver as well. Is this...
@AaronDonahue can we close this?
Since we can't query whether a var attribute exists or not, this issue entails adding such capability to our IO interfaces.
Interesting scenario. I'll think about a solution that is robust and self-explanatory. Side topic: the fact that some process may be subcycled but "indirectly" (that is, it is not subcycled,...