Re: [Gretl-users] Loop recursion: Speed comparison with Matlab

Am 25.10.2017 um 14:17 schrieb Allin Cottrell: > On Wed, 25 Oct 2017, Sven Schreiber wrote: > >> Am 25.10.2017 um 08:15 schrieb Riccardo (Jack) Lucchetti: >> >>> I took a different approach, which avoids loops completely. >> ... >>> This appears to be quite fast. >> >> This is of course fantastic for the SB package, but still leaves loops >> slower than with Matlab. I'm getting more and more convinced that they are >> doing just-in-time compilation there, FWIW. > > Yes, that's surely the case. > > Allin I ran my MATLAB example code for the following 4 cases (T=10000, nboot=1000) <RESULTS> Elapsed time is 0.315147 seconds. <-- f. accel 'on', f. JIT 'on' Elapsed time is 19.185277 seconds. <-- f. accel 'off', f. JIT 'on' Elapsed time is 0.658673 seconds. <-- f. accel 'on', f. JIT 'off' Elapsed time is 19.029025 seconds. <-- f. accel 'off', f. JIT 'off' </RESULTS>