Paul R. C. Kent
Paul R. C. Kent
Ugh. I think there are two issues here: (1) the DMC is unstable for whatever reason (2) a failure in one twist brings down the ensemble. This latter one should...
Following discussions with Jaron, I attempted to reproduce this over the weekend, using a CPU non-mpi build. Each individual twist in this run could be run on a single node....
Update: I don't think I have been able to reproduce this problem, including above. Occasional fluctuations down to ~1800 walkers (10% difference from target) but have not seen a divergence....
@aannabe do you want to comment on the strategy here? All the other files created around the same time have the "expected" cutoff of O(1).
Check with Mark?
Thanks for the update. Another reason to have the e4s build run with offload is that it will highlight the current cuda
Before doing any work on this I would like to understand how significant an issue it is for actual science runs. Then we can discuss and assign a priority, decide...
Ideal for v4.0, but I don't think it is a blocker since there is at least a working route in the current code. v4.1 might be a better target.
Yes, if this warning is hit repeatedly the run will likely be biased. An occasional warning is often OK if you have enough additional statistical samples to dilute a small...
Thanks for putting this up Kayahan. Note that -- if you can -- it is better to break these up into many smaller PRs. It will be easier to review...