Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update case members in case_setup #1414

Merged
merged 5 commits into from
Apr 25, 2017
Merged

Update case members in case_setup #1414

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
de5b988
PET must halve TASKS when doubling THRDS
jgfouca Apr 24, 2017
0ed54b2
Double nodes instead of halving tasks
jgfouca Apr 25, 2017
975958d
Re-initialize key case values upon case.setup
jgfouca Apr 25, 2017
4cb3075
Order of operations was not correct
jgfouca Apr 25, 2017
56aba43
PIO settings need to happen before init_derived_attributes
jgfouca Apr 25, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions scripts/lib/CIME/SystemTests/pet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,10 @@ def _case_one_setup(self):
if self._case.get_value("NTHRDS_%s"%comp) <= 1:
self._case.set_value("NTHRDS_%s"%comp, 2)

# Subtle: machines like titan will not let you overload the number of threads
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't understand why there was any overloading going on - the PET uses full nodes on the first test and undersubscribes on the second one. Isn't this just a matter of updating the aprun command correctly for each of (between) the two runs?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@jedwards4b if you double the threads without reducing the tasks, you'd expect to have double the total threads. On a machine that allows you to put 2 threads on a single core, this isn't a problem, but apparently aprun/titan doesn't let you do that. Keep in mind that the num_nodes has already been computed at this point in the case object, so that fact the more threads requires more nodes on titan is lost unless we update by hand like we do here.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In pet.py _case_one_setup you set NTHRDS = 2 for each component then you run case_setup to recreate the case.run and case.test scripts for the correct node and task setup
In _case_two_setup you set NTHRDS = 1 and MAX_TASKS_PER_NODE to half it's original value
and run case.setup again to recreate the case.run and case.test scripts for the same number of nodes but undersubscribed. I think that the issue that you are having is that the aprun command line is not getting updated to be consistent with these changes.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@jedwards4b , I've been thinking and you're right. The real issue is that critical values that are set up in case._initialize_derived_attributes are not updated when case_setup is called, even though the values that went in to compute those derived attributes may have changed. I'm going to test a different fix.

# on a node.
self._case.num_nodes *= 2

# Need to redo case_setup because we may have changed the number of threads
case_setup(self._case, reset=True)

Expand Down
1 change: 1 addition & 0 deletions scripts/lib/update_acme_tests.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,7 @@
"ERP.f45_g37_rx1.A",
"SMS_D_Ln9.f19_g16_rx1.A",
"DAE.f19_f19.A",
"PET.f19_f19.A",
"SMS.T42_T42.S",
"PRE.f45_g37_rx1.ADESP")
),
Expand Down