Re: [Gretl-devel] [Gretl-users] When Julia is faster than C and Gretl

Am 18.02.2018 um 10:37 schrieb Sven Schreiber: So Allin had the great idea to enable binary-mode transfer between Python and Gretl, and I tried it. (Attention, the binary transfer won't work on Windows yet even with the latest snapshot due to the issue mentioned in the other thread. You would have to edit the gretl_io.py file on your machine first.) And here are new results. (Don't compare the absolute speeds with my previous post, they are from different machines.) SB 1.3, 2017-11-14 (F. Di Iorio, S. Fachin, A. Tarassow, Riccardo "Jack" Lucchetti) Gretl native (SB package) This took = 7.587454 sec. Python / Numba pure calculation This took = 1.887286 sec. Python / Numba with text transfer This took = 6.591593 sec. Python / Numba with gzipped text transfer This took = 26.816650 sec. Python / Numba with binary transfer This took = 2.087111 sec. Now we're talking! However, it just occurs to me that "Gretl native" is misleading, because the SB package is obviously hansl-native, but it's not a native libgretl C function. Anyway. thanks, sven