The nonpolarized SHT performance tests for Healpy, fastSHT-CPU, and fastSHT-GPU with N _side = 128, ﹩{{\ell }}_{\max }=383﹩ and N _maps that varies from 1000 to 8000. The upper/lower panels show the case without/with iterations. The colored bars (corresponding to the left y-axis) give the time costs per map, and the stars/dots (corresponding to the right y-axis) show the acceleration factors relative to Healpy. The CPU computations are done with 24 cores.