Add new memory runLog info to memoryProfiler.py #1970
Conversation
| @@ -101,11 +129,30 @@ def interactEveryNode(self, cycle, node): | |||
| mpiAction.broadcast().invoke(self.o, self.r, self.cs) | |||
|
|
|||
| def interactEOL(self): | |||
| r"""End of life hook. Good place to wrap up or print out summary outputs.""" | |||
| """End of life hook. Good place to wrap up or print out summary outputs.""" | |||
| if self.cs["debugMem"]: | |||
There was a problem hiding this comment.
I'm just wondering... should this same settings check be done at interactBOL and interactEveryNode?
I could be totally off-base here, I just can't tell why this is only done at the one time node.
There was a problem hiding this comment.
I'm not following. Can you elaborate?
There was a problem hiding this comment.
Well, this memory profiler does nothing in interactEOL unless the setting debugMem is set to True.
So.... I would EXPECT that setting would control all the time points, not just the one.
If that's not the case cool, I'm just wondering what the difference is.
There was a problem hiding this comment.
That setting does get used in interactBOL and interactEveryNode. These two just also do other stuff too.
Granted I don't know the original intent. The git blame for the vast majority of this module go well beyond my time with TerraPower.
Co-authored-by: John Stilley <[email protected]>
There was a problem hiding this comment.
I'm a fan!
Note that we realize this code probably shouldn't be in ARMI but to push towards a deadline we have a separate cleanup ticket to bring parts of the memory profiler into internal tooling, where it belongs.
* main: Fixing HexBlock docstrings (#1981) Removing a duplicate hex rotation impl tag (#1979) Avoiding closing plots that are meant to be interactive (#1978) Ensuring HexBlock.rotate updates child spatial locators (#1943) Add new memory runLog info to memoryProfiler.py (#1970) Fixing various doc build issues (#1974) Hiding sphinx-needs warnings during doc build (#1973) Fixing warnings in CLI startup (#1972)
What is the change?
Adds some potentially useful
runLogstatements to the memory profiler interface on how much memory is is use , how much is available, and the total available for a simulation. The first two are dynamic (i.e., change with (cycle, node)) while the third is static.Why is the change being made?
Some memory intensive downstream analysis found this useful, so we thought we'd push it up to ARMI.
Checklist
docfolder.pyproject.toml.