Introduces three choices for handling large arrays in atomic transfer.
So far, it applies only on continuous opacity.
limit_mem = 0: continuous opacities are computed on the non-LTE wavelength grid and stored on the ram (large ram consumption, faster approach)
limit_mem = 1 (Actual implementation): continuous opacities are computed on a small continuum wavelength grid and stored on the ram (moderate ram consumption, good compromise between speed and memory usage)
limit_mem = 2: continuous opacities are computed on a small continuum wavelength grid on-the-fly (slowest, cheap in ram)
in case 1 and 2, the continuum opacity is linearly interpolated locally on the non-LTE wavelength grid
cpinte
commented
11 months ago
Introduces three choices for handling large arrays in atomic transfer. So far, it applies only on continuous opacity. limit_mem = 0: continuous opacities are computed on the non-LTE wavelength grid and stored on the ram (large ram consumption, faster approach) limit_mem = 1 (Actual implementation): continuous opacities are computed on a small continuum wavelength grid and stored on the ram (moderate ram consumption, good compromise between speed and memory usage) limit_mem = 2: continuous opacities are computed on a small continuum wavelength grid on-the-fly (slowest, cheap in ram)
in case 1 and 2, the continuum opacity is linearly interpolated locally on the non-LTE wavelength grid