CCL_ERROR_SPLINE when using CCL with CAMB in Firecrown

[bclevine]

I was recently running a chain using Firecrown when I encountered the following error: pyccl.errors.CCLError: Error CCL_ERROR_SPLINE: ccl_correlation.c: ccl_tracer_corr_fftlog(): failed to create spline. Upon request (@tilmantroester), I am posting this issue here!

The full report in the context of Firecrown is described in detail here: LSSTDESC/firecrown#567.

As far as I have been able to determine, the issue only arises very rarely, with certain parameter combinations. In the case I encountered, P(k) is given to CCL by CAMB (with Firecrown/Cosmosis passing the P(k) to CCL), and the error is caused by the GGL bin combination Lens 4 x Source 2 (using 0-indexing). I'm using the DES Y3 redmagic data here.

The error occurred with the following parameters:

Omega_m = 2.96911371e-01
Omega_b = 4.05897605e-02
n_s = 8.93974434e-01
A_s = 7.20033942e-10
h = 8.41633373e-01
Omega_nuh2 = 2.35114043e-03
ia_bias = 7.44524660e-01
alphaz = 1.23387072
z_piv = 0.62
lens4_bias = 2.44415947
src2_delta_z = 1.09596651e-03
lens4_delta_z = -8.39691869e-03
lens4_sigma_z = 1.15323136

When I try to recreate this in pure CCL, using those parameters, I don't get any failures — so I know part of the problem is the P(k) that's passed from CAMB. To recreate this problem in Firecrown, you can use the instructions I described here: LSSTDESC/firecrown#567, but modify factory.py to only use the relevant bin combination and then replace debug_list.ini with the following:

#cosmological_parameters--omega_m	cosmological_parameters--omega_b	cosmological_parameters--n_s	cosmological_parameters--a_s	cosmological_parameters--h0	cosmological_parameters--omnuh2	firecrown_two_point--ia_bias	firecrown_two_point--alphaz	firecrown_two_point--lens4_bias	firecrown_two_point--src2_delta_z	firecrown_two_point--lens4_delta_z	firecrown_two_point--lens4_sigma_z

2.96911371e-01	4.05897605e-02	8.93974434e-01	7.20033942e-10	8.41633373e-01	2.35114043e-03	7.44524660e-01	1.23387072e+00	2.44415947e+00 1.09596651e-03	-8.39691869e-03	1.15323136e+00

I hope this is enough relevant information, but please let me know if there's anything else I can provide for this report!