You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm seeing this warning a lot for the TDB branch of S3D when running with onepool:
[0 - 7fefa8bfd780] 26.369478 {4}{runtime}: LEGION WARNING: WARNING! Leaf task compute_adjacent (UID 192176) attempted to allocate a DeferredBuffer/Value/Reduction of 1103 bytes in Z_COPY_MEM memory but no space was reserved for dynamic allocations during the lifetime of this task. Legion has managed to procure for you an allocation this time but there is no guarantee that you will be so lucky the next time. We strongly encourage all users to place tight upper bounds on the required memory for all leaf tasks either statically at the point of task variant registration or dynamically at the point that the task is mapped. (from file /lustre/orion/cmb103/scratch/seshuy/legion_s3d_tdb/legion/runtime/legion/legion_context.cc:25512)
The text was updated successfully, but these errors were encountered:
Yes, you'll get these messages until Regent is updated to tell Legion what allocations it needs. Which unfortunately I'm not sure when I'll have time to do at this point.
With sufficient memory it should run anyway, though it's always possible that Legion will generate too much parallelism, causing things to overflow.
Right, these are the warnings you'll get from now on using the "legacy" path for deferred buffers. If it works you'll get the warning, but if it doesn't work it will give an error to avoid hanging on a cyclic Realm deferred allocation dependence. At the moment there's not a good way to suppress the warnings.
I'm seeing this warning a lot for the TDB branch of S3D when running with onepool:
The text was updated successfully, but these errors were encountered: