Skip to content

Adding dtype option to reduce storage

Emmanuel Branlard requested to merge wtlib/pyconturb:f/dtype into master

Adding option to store main matrices in np.float32, to reduce storage. The computation of the coherence and sigma is slightly faster, but not the Cholesky decomposition. If !45 (closed) is used, then the computational time is about twice as fast when using float32 instead of float64.

Edited by Jenni Rinker

Merge request reports