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
For HPX and HPX-5 implementations, the profile_listener object should replicate the current local state from each locality into the HPX AGAS, and provide mechanisms for the policies to extract remote locality APEX state from other localities. This would provide built-in globally-aware support.
The initial thought is that when the profiler listener finishes processing the current profiler queue (or after N profiler objects are processed, whichever comes first) the current profile state would be replicated in the HPX AGAS. Could be replicated periodically also/instead.
I'm not sure how this would be supported for other runtimes (i.e. MPI+X).
For OCR, this would be similar to HPX.
This code should live in the respective runtime glue-code - in HPX3, that's easy because we can implement it in the XPRESS-APEX code that will be submoduled into the HPX project. HPX-5 already has some code for performing something similar, we should modify that to implement this for HPX-5.
The text was updated successfully, but these errors were encountered:
For HPX and HPX-5 implementations, the profile_listener object should replicate the current local state from each locality into the HPX AGAS, and provide mechanisms for the policies to extract remote locality APEX state from other localities. This would provide built-in globally-aware support.
The initial thought is that when the profiler listener finishes processing the current profiler queue (or after N profiler objects are processed, whichever comes first) the current profile state would be replicated in the HPX AGAS. Could be replicated periodically also/instead.
I'm not sure how this would be supported for other runtimes (i.e. MPI+X).
For OCR, this would be similar to HPX.
This code should live in the respective runtime glue-code - in HPX3, that's easy because we can implement it in the XPRESS-APEX code that will be submoduled into the HPX project. HPX-5 already has some code for performing something similar, we should modify that to implement this for HPX-5.
The text was updated successfully, but these errors were encountered: