Speaker
Description
In the context of soon to be released data from large-scale structure (LSS) Stage IV surveys that offer extremely precise and broad measurements, I will present $\texttt{SwiftC}_\ell$: a fast, accurate and differentiable Python pipeline for the beyond Limber computation of the angular power spectrum. $\texttt{SwiftC}_\ell$ includes all relevant contributions to wide-angle surveys that are computed exactly up to arbitrary precision. We compare our pipeline to the N5K challenge, a challenge aimed at finding a suitable pipeline for the data analysis of the Rubin Observatory Legacy Survey of Space and Time (LSST). In this frame of reference, $\texttt{SwiftC}_\ell$ performs the computation of the 120 different angular power spectra over 103 angular separation bins in 2 ms on one GPU core. About 100x faster than the previous best code, $\texttt{SwiftC}_\ell$ delivers the accuracy and speed required in near-future data analyses. Furthermore, all $\texttt{SwiftC}_\ell$'s outputs are auto-differentiable, facilitating gradient-based sampling that will be crucial in the near future.