minor bug when asking for nonlinear spectra

[emiliobellini]

Hi all, in the latest version of Class (after 2.9) there is a minor bug with the value of "k_max_for_pk" when you ask for nonlinear spectra.

Basically, whenever the user specify P_k_max_1/Mpc or P_k_max_h/Mpc, this value overwrites the "nonlinear_min_k_max" value.

This is beacuse k_max is first inferred in the input_read_parameters_nonlinear function, and only after that the parameters P_k_max_1/Mpc and P_k_max_h/Mpc are read in input_read_parameters_spectra.

A clean and quick workaround is to replace if (flag1 == TRUE){ ppt->k_max_for_pk=param1*pba->h; } if (flag2 == TRUE){ ppt->k_max_for_pk=param2; }

with if (flag1 == TRUE){ ppt->k_max_for_pk=MAX(ppt->k_max_for_pk,param1*pba->h); } if (flag2 == TRUE){ ppt->k_max_for_pk=MAX(ppt->k_max_for_pk,param2); }

Unless you want to read P_k_max_h/Mpc in input_read_parameters_nonlinear or even in input_read_parameters_general (it could make sense to have there the initial scale factor, together with k_min and k_max, but this depends on you)

Cheers, Emilio

[lesgourg]

Indeed, this was annoying! We also saw it recently, and this will be fixed in the minor release 3.1.2 due this afternoon. Thanks for noticing! Now, the values of ppt->k_max_for_pk and ppr->nonlinear_min_k_max will be kept separate in the code. This allows another change: the function .get_pk_and_k_and_z() will return the P(k) only up to ppt->k_max_for_pk, and not up to the MAX(,) of the previous two numbers. It makes more sense this way. Thanks for noticing!

[emiliobellini]

Fantastic!