Recommend charm version 8 #6759
Conversation
| Parallel::number_of_procs<size_t>(cache) > 1) { | ||
| ERROR_NO_TRACE( | ||
| "Dynamically creating elements in parallel executables is broken " | ||
| "until charm 8."); |
There was a problem hiding this comment.
It's not quite this bad, is it? I've been using AMR with h-refinement in a separate phase in elliptic executables for a while now with no issues.
There was a problem hiding this comment.
@kidder Do you know what operations can trigger the hang?
There was a problem hiding this comment.
I believe it occurs when one of the initial elements is destroyed and then recreated in a later iteration on a PE other than PE 0. If the current elliptic code usage only refines, and never coarsens, this condition won't occur.
See https://github.com/kidder/charm_spectre_tests/tree/main/dynamic_array_hang
There was a problem hiding this comment.
The elliptic code only refines, yes.
There was a problem hiding this comment.
okay, then probably only erroring in CreateParent would be adequate
There was a problem hiding this comment.
I've dropped the changes to CreateChild.
nilsdeppe
left a comment
There was a problem hiding this comment.
Larry and Nils, any followup on the discussion so we can merge this?
This does not update containers, CI, or clusters, just the docs and cmake files.
Proposed changes
Upgrade instructions
Charm version 8 is now recommended over version 7, but 7 is still supported for applications not using dynamic h-refinement or similar post-startup element creation.
Code review checklist
make docto generate the documentation locally intoBUILD_DIR/docs/html.Then open
index.html.code review guide.
bugfixornew featureif appropriate.Further comments