Parco 2019 debug

Related !919 (merged).

That needs working on. At the moment it includes much too much debugging code and will break other things.

changed the description

Added a link to !919 (merged), which is an MR for the synchronous redistribute code.

Yeah I know, I just created this to see the differences with master. I'll remove the [WIP] when its ready to be reviewed and merged.

Note: @jwillis I also need to merge in some changes from the parco_2019 branch also.

!919 (merged) (the synchronous redist) has now been merged.

added 460 commits

• 6254c920...93464a24 - 458 commits from branch master
• 77dc92d3 - Merge remote-tracking branch 'origin/redist-sync' into parco_2019_debug
• 63daa00d - Merge remote-tracking branch 'origin/master' into parco_2019_debug

Compare with previous version

added 29 commits

• 63daa00d...3068e6ad - 28 commits from branch master
• 2996b977 - Merge branch 'master' into parco_2019_debug

Compare with previous version

added 1 commit

• 78b91162 - Removed debug statements.

Compare with previous version

added 1 commit

• 22dbd425 - Removed profiling messages by default. Added missing qsort back in.

Compare with previous version

added 5 commits

• 846c490a - Remove duplicate setting of default group ID for all particles.
• 5f957a1f - Fixes a bunch of warnings in some calls to sprintf() taht trigger…
• e7780133 - Applied code formatting tool.
• 02ccc0e2 - Merge branch 'parco_2019' of gitlab.cosma.dur.ac.uk:swift/swiftsim into parco_2019
• db2de652 - Merge branch 'parco_2019' into parco_2019_debug

Compare with previous version

So I've tidied this MR up now. I've removed all of the debug statements and merged in the changes from the parco_2019 branch.

I've tested it with EAGLE-6 on 6 MPI ranks as a standalone FOF. Do you have another test that runs the FOF with SWIFT?

Also, I could move the parallel_replicate_ICs.py to the tools directory if you'd like?

added 1 commit

• dec4b91a - Moved and documented parallel_replicatye_ICs.py script.

Compare with previous version

added 1 commit

• 03964eb8 - Moved and documented parallel_replicatye_ICs.py script.

Compare with previous version

unmarked as a Work In Progress

changed title from [WIP]: Parco 2019 debug to Parco 2019 debug

I've tested it with EAGLE-6 on 6 MPI ranks as a standalone FOF. Do you have another test that runs the FOF with SWIFT?

Pretty much any of the EAGLE examples will run fof on-the-fly to produce black holes.

I guess you noticed that a lot of functions have changed file such that the diffs are tricky. Hope that means nothing got lost in the MR.

Any reason to not always define that function?

It's not called when WITH_FOF_TIMING is defined so throws a compiler error if the macro is not there.

changed this line in version 10 of the diff

Also, thinking more of it, I am not a big fan of the def WITH_FOF_TIMING since it does more than activating timing. It actually deactivates quite a bit of code. Maybe some name that reflects better what it does would be good. WITHOUT_GROUP_PROPERTIES_CALCULATION ? that's a bit long...

What are the proper config options and cmd line arguments to run SWIFT with black holes? And what should the frequency of FOF timesteps should I use?

I knew that the functions had changed files but I didn't see that when I merged with master strangely. I think the only major changes that we made were the mapper functions for setting initial values and at which point they are called.

I agree with changing the name of WITH_FOF_TIMING. I'll have a think about what to change it too.

Config: --with-subgrid=EAGLE

Run: --with-black-holes --fof

Set a frequency such that it runs often if it is for testing. Otherwise, any of the runs in the EAGLE_ICs directory have a physically motivated frequency set but they start in a regime where no BHs form for a long time.

How about WITHOUT_GROUP_PROPS instead of WITH_FOF_TIMING?

This should be ready to merge after that change. I have tested EAGLE-50 with FOF enabled and it ran for 24 hours no problem. Although I did have to disable black hole physics.

added 1 commit

• 09de4943 - Renamed WITH_FOF_TIMING macro to WITHOUT_GROUP_PROPS.

Compare with previous version

Sounds good to me. How many FOF calls did you get?

6 calls to FOF.

Great thanks.

merged

mentioned in commit fec911c1