Subject: Re: [boost] [random] new threefry random engine
From: Thijs (M.A.) van den Berg (thijs_at_[hidden])
Date: 2014-04-20 04:07:41
On Apr 20, 2014, at 5:23 AM, Nat Goodspeed <nat_at_[hidden]> wrote:
> On Apr 19, 2014, at 6:35 PM, Thijs van den Berg <thijs_at_[hidden]> wrote:
>> I have addressed most other points youâve mentioned, but the performance issue of a generic rounds version has failed me.
> Have you considered compile-time recursion to emit an unrolled loop with some max rounds?
Thanks, I was indeed thinking about that.
I've tried too little yet, I was too puzzled by the factor 2.5 an I thought I made a mistake somewhere in my code.
So far I've used the performance tool in random with default settings, so I'll first do my own benchmark and try various compiler options, perhaps look at assembler, and various compilers.
I think recursion is a very good option. It will prevent run time loop counters and the computations that depend on it and make the code less predictable.
I'll try that first!
> Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk