Problems with cosmomc as a generic sampler

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Diego Herranz
Posts: 4
Joined: April 20 2015
Affiliation: Instituto de Fisica de Cantabria (CSIC-UC)

Problems with cosmomc as a generic sampler

Post by Diego Herranz » July 08 2015

Dear all,

I'm trying to use COSMOMC (January 2012 version) to run a generic likelihood with 5 parameters. I face the following problem: if I set generic_mcmc = .TRUE., num_hard = 5 and num_initpower = 0, num_norm = 0, the code won't compile, returning array size errors in places such as:

>> P%InitPower%rat(in) = CMB%norm(norm_amp_ratio)

Even when I disable array size check on compilation, mcmc runs sending messages such as:

>> Computing tensors: F
>> Doing CMB lensing: F
>> lmax = 2100
>> Number of C_ls = 4

which suggest me that cosmomc is trying to use CMB likelihood even if I have set generic_mcmc = .TRUE.

It's obvious I'm missing something.

I've tried also to leave the original values of num_norm and num_initpower unchanged and to use a params.ini file with aditional parameters set to zero:

param6 = 0 0 0 0 0

In this case, the code compiles and runs, but it seems to be trying to do something "CMB-ish" and the results don't make sense.

Please, does anybody have any clue?

Thank you,

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Problems with cosmomc as a generic sampler

Post by Antony Lewis » July 08 2015

Try a more recent version.

Post Reply