Camgal

is a plugin to CAMEL that allows to make the link with the amazing AngPow software (for fast/accurate tomographic computations)

MARDEC installation

The software is already installed under
EXECDIR=/dec/users/plaszczy/cmtlib/camgal/HEAD/amd64_linux26

You probably only want to obtain the Cls from a given cosmology and bins configuration

You have to run

$EXECDIR/writeCl camel-parfile angpow-inifile output.fits

where

  • camel-parfile is a standard camel parameter file specifying the cosmology
  • angpow-inifile is a standard Anpow ini file which specifies the shells (the cosmological part is overwriiten by the camel file)
  • output.fits is a name you supply for your output file

To start with you may copy and play with the following files:

  • CAMEL: /dec/users/plaszczy/cmtlib/camgal/HEAD/par/planck.cml
  • ANGPOW: /dec/users/plaszczy/cmtlib/camgal/HEAD/par/colore.ini

Be carefull that the CAMEL file has an important new parameter as

#Pk smoothing (in Mpc)
R_smooth=8

The CLASS linear spectrum is mulitplied by exp(-k^2 Rsmooth^2) It is always a good idea to use such an explicit smoothing radius.

It is connected to some parameters in the AngPow file:

  • cl_kmax is the kmax used for Cl computation (be carefull it is not the "kmax" value in the camel file which is sometimes usefull for other tasks): there is no sense of putting a too large value if you smoothed. For instance with R_smooth=8 there won't be much power left above kmax=0.4 (which is the value used in colore.ini)
  • Lmax : you don't need to put a too high value since Lmax ~ r(zmax)* cl_kmax, where r(zmax) is the (mean) comoving distance to your higher redshift shell (if you don't know it it will actually be written in the fits file). In the planck.cml cosmology ,

r(0.45)~1800 Mpc, why we used Lmax=750 in colore.ini.

There is nothing fundamental there (you may ue "wrong" values) it is only a way to avoid unnecessary 0 computation and you may try a large Lmax value and then shrink it later to your convenience

The code is multi-threaded. But you won't gain much above 16 threads, then try with

export OMP_NUM_THREADS=16