It was stated earlier in the thread that tip was using xorshift32+multiplication. I assumed that was what you were saying that PCG was not faster than. I have not tested myself but I have no reason to disbelieve the results on the PCG site. It's a red herring to compare against xorshift128+ or xoroshiro128+, as that is not what's (said to be) used at tip currently. I mistakenly left out the "32" when referring to it. Admittedly the specific kind of PCG was not stated either, which would be important for any comparison.
This is not an area that I claim to have much knowledge of; it just appeared to me that you were dismissing PCG without citing evidence when my cursory review suggested that it did have advantages. On Fri, Dec 29, 2017 at 10:00 PM Юрий Соколов <funny.fal...@gmail.com> wrote: > I've said nothing about "multiplication". Where you take "multiplication" > from? Xorshift+ is already good enough for most usages. > > PCG performance page didn't include results neither for xorshift128+, nor > for xoroshiro128. > > But xoroshiro page has comparisons: http://xoroshiro.di.unimi.it/#speed > > And, did you measure by yourself? > > 30 дек. 2017 г. 6:44 AM пользователь "Matt Harden" <matt.har...@gmail.com> > написал: > > PCG isn't faster than xorshift+multiplication? Do you have a rebuttal to > their website that indicates it is? (called xorshift* on that page) > http://www.pcg-random.org/rng-performance.html > > Or are you just saying it's not meaningfully faster for the purpose of > channels because the other channel overheads are high enough to make the > difference irrelevant? Or maybe that it isn't faster with current Go > compliers? > > They also claim it's better statistically, but perhaps if we only care > about fairness it's essentially the same. > > On Mon, Oct 30, 2017 at 12:34 AM Sokolov Yura <funny.fal...@gmail.com> > wrote: > >> > it would be great to move to pcg. >> >> Why? PCG isn't faster, isn't simpler and isn't better (though not worse >> either). >> >> -- >> You received this message because you are subscribed to the Google Groups >> "golang-nuts" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to golang-nuts+unsubscr...@googlegroups.com. >> >> For more options, visit https://groups.google.com/d/optout. >> > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
