Undefined symbol error with ACT DR4 likelihood

[margaret-ikape]

Hi @joezuntz ,

I have just installed the new cosmosis version and included the ACT likelihood (which runs fine), but when I try to run the test sampler on the ACT likelihood, I get

File "/home/margaret/env/lib/python3.9/site-packages/cosmosis/runtime/module.py", line 103, in __init__
    self.library, language = self.load_library(filename)
  File "/home/margaret/env/lib/python3.9/site-packages/cosmosis/runtime/module.py", line 291, in load_library
    raise SetupError("You specified a path %s for a module. "
cosmosis.runtime.module.SetupError: You specified a path /home/margaret/cosmosis-standard-library/likelihood/actpolfull_dr4.01/actpol.so for a module. File exists, but could not be opened. Error was /home/margaret/cosmosis-standard-library/likelihood/actpolfull_dr4.01/actpol.so: undefined symbol: __cosmosis_modules_MOD_datablock_put_double

Could you please tell me what I'm doing wrong?

Thank you!

[joezuntz]

Hi @margaret-ikape ,

ACTPol isn't included with cosmosis, so I think this is a module you've installed separately?

I can see the actpolfull_dr4.01 file is publicly available, but I'm not sure what interface file you're using (i.e. the file with the setup and execute functions), or the Makefile. If you're able to share those I can have a look.

In general, though, the error you're seeing happens when flags looking something like this are missing from a Makefile:

USER_FFLAGS+=-Wl,-rpath,${PWD}
USER_LDFLAGS+=-Wl,-rpath,${PWD} -lcosmosis

best,
Joe

[margaret-ikape]

Thanks, @joezuntz. I added the two lines above and it works fine now.

Thanks again,

[joezuntz]

Great, glad this worked for you! If you're able to share the likelihood code interface to cosmosis so I could package it with cosmosis then that would be really great (I have a vague memory of discussing this with Renee Hlozek at one point). No worries if not, or not yet.

I'll close this issue now.
Cheers,
Joe

[margaret-ikape]

Absolutely. I’m going to confirm the likelihood is installed properly by reproducing the results from the ACT paper, and then I can share it with so it can be added. Thanks again, Margaret

…

On Sep 29, 2022, at 1:05 PM, joezuntz ***@***.***> wrote: Great, glad this worked for you! If you're able to share the likelihood code interface to cosmosis so I could package it with cosmosis then that would be really great (I have a vague memory of discussing this with Renee Hlozek at one point). No worries if not, or not yet. I'll close this issue now. Cheers, Joe — Reply to this email directly, view it on GitHub <#47 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFQNFJ3JAMPRZT2EIL34IRLWAXD5JANCNFSM6AAAAAAQY3MLLA>. You are receiving this because you were mentioned.

[margaret-ikape]

Hi Joe, I hope this message meets you well. I’ve recently come back to the ACT likelihood and I’m running some checks to reproduce the results from https://arxiv.org/pdf/2007.07289.pdf <https://arxiv.org/pdf/2007.07289.pdf>. I’m finding that when I set NonLinear in the cosmosis ini file to ‘both’ or ‘lens’, I get the same result. Same goes for NonLinear = ’none’ and ‘pk’. I tried to look through the CAMB code in cosmosis and found the following code. In env/lib/python3.9/site-packages/camb/model.py In def set_non_linear_lensing, if nonlinear: if self.NonLinear in [NonLinear_pk, NonLinear_both]: self.NonLinear = NonLinear_both else: self.NonLinear = NonLinear_lens else: if self.NonLinear in [NonLinear_pk, NonLinear_both]: self.NonLinear = NonLinear_pk else: self.NonLinear = NonLinear_none I guess this means that if nonlinear is True, that is NonLinear != none, and if the user chooses either pk or both, then code defaults to both, otherwise, code defaults to lens (This probably explains why NonLinear = both and lens give same result (?) ) otherwsie, if nonlinear is not true, then code defaults to either pk or none (which again explains why NonLinear = none and pk give same reult) Is this correct? am I reading the code correctly? Thanks, Margaret

[joezuntz]

Hi @margaret-ikape - that's interesting, I hadn't seen this. It's more a camb question - I'm not sure if it's intended or not.