LSSTDESC / SprintWeek2020

Meeting repository for the LSST DESC 2020 Sprint Week

Apache License 2.0

2 stars 0 forks source link

[SPRINT] cluster 2pt using cosmoDC2/DC2 #60

Open humnaawan opened 3 years ago

humnaawan commented 3 years ago

Measuring the Cluster 2pt using cosmoDC2/DC2

Contacts: @humnaawan @cavestruz Day/Time: TBD Main communication channel: #desc-cl GitHub repo: None (yet)

Goals and deliverable

The goal here is to measure the Cluster 2pt function using cosmoDC2/DC2 with the motivation to 1) figure out the various working parts of the analysis pipeline, and 2) validate DC2 sims.

The reach goal would be to

Measure the Cluster 2pt function
Figure out further investigations using the cluster 2pt, both for cluster joint analysis and joint probes analyses
Draft a DESC Note

The stretch goal would be to

Have a clear sense of the next project(s)
Have a finalized DESC Note

Resources and skills needed

Experience with DC2, interest in measuring cluster 2pt function.

Detailed description

Please see the project confluence page for more details.

humnaawan commented 3 years ago

thanks @katrinheitmann @mishakb for a fruitful conversation earlier. just to recap some of the discussion points, there are a few outstanding questions before we can really measure cluster 2pt with DC2:

what would it take for the cluster finder to be run on DC2 data? @erykoff can you please comment on this?
@erykoff can you also say how long the cluster finding takes and whether we can expect redmapper to be runnable by anyone (e.g. aside from you)? i ask this since its likely that we will be statistics-limited with DC2 and will need SkySim5000 for our purposes.
does CCL already have the functionality to do cluster 2pt predictions? @marina-ricci can you please comment on this (or tag someone who would know)?

@katrinheitmann please let me know if i'm missing something.

mishakb commented 3 years ago

@humnaawan, sure. I beleive that @pennalima may also have some comments on the CCL side.

erykoff commented 3 years ago

1) So I set redMaPPer running on DC2 DR6 before I took last week off, and there was some sort of problem that I am going to track down this afternoon (hopefully). 2) The calibration part was taking a few days on my wimpy old machine, it could have been faster at nersc potentially but I didn't feel like dealing with nersc (and there's also something strange that it sometimes hiccups and needs to be restarted from where it left off, and that has been more of a headache). This calibration part only needs to be done once on a subset. And then the time it takes scales with how many cores you throw at it. With 12 cores for DC2 DR6 it should take about a day (but I need to fix this problem before I can confirm.) 3) It should be runnable by other people, and I know it has been run by folks in the wild (who submit feature requests and bug reports).

humnaawan commented 3 years ago

thanks @erykoff! it'll be great to get progress updated on DC2 DR6 once its available.

also, i assume that the code is not set up on NERSC (please correct me if im wrong) - do you think this is something that can be achieved as a sprint during this week?

erykoff commented 3 years ago

It does support nersc via docker/shifter, but now I can't find my documentation on that...which would be something useful to do (properly document running at nersc!) this week after I fix the DC2 DR6 catalog generation.

humnaawan commented 3 years ago

@erykoff hope you're well. im just following up regarding the status of the redMaPPer output for DC2. can you please let me know if/where i can find it?