Hmm, I use just a super-naive LCG
unsigned long hi, lo;
lo = 16807 * (pmseed & 0xffff);
hi = 16807 * (pmseed >> 16);
lo += (hi & 0x7fff) << 16;
lo += hi >> 15;
pmseed = (lo & 0x7fffffff) + (lo >> 31);
return ((pmseed & 0xffff) * max) >> 16;
in Pachi. Frankly, I think I could get away with just
pmseed = pmseed * 16807 + 12345;
return pmseed % max;
Do you really need high quality RNG in your program? I'd be interested
if someone using a sophisticated RNG would compare it with using just
the above wrt. Elo strength and performance.
(BTW, for floats, there's this neat trick to take advantage of the
internal representation
union { unsigned long ul; floating_t f; } p;
p.ul = (((pmseed *= 16807) & 0x007fffff) - 1) | 0x3f800000;
return p.f - 1.0f;
see also http://rgba.org/articles/sfrand/sfrand.htm)
On Sun, Mar 29, 2015 at 05:58:56PM +0800, remco.bloe...@singularityu.org wrote:
> I switched to SFMT [0]. But that was some years ago, there might be faster
> options now.
>
> I also generated it in megabyte batches and consume it from there, generating
> a new megabyte as needed.
>
> Lastly, I had some code to make sure I did not consume more bits of entropy
> than required. Two uniform choices, one bit. Three choices: fractional
> bits.
>
> [0]
>
> http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/SFMT/
>
> — Remco
>
> -----Original Message-----
> From: folkert <folk...@vanheusden.com>
> To: computer-go@computer-go.org
> Sent: Sun, 29 Mar 2015 17:50
> Subject: [Computer-go] fast + good RNG
>
> Hi,
>
> I measured that std::mt19937_64 (the mersenne twister from the standard
> c++ libraries) uses about 40% cpu time during playouts.
>
> So I wonder: is there a faster prng while still generating good enough
> random?
>
>
> Folkert van Heusden
>
> --
> Nagios user? Check out CoffeeSaint - the versatile Nagios status
> viewer! http://www.vanheusden.com/java/CoffeeSaint/
> ----------------------------------------------------------------------
> Phone: +31-6-41278122, PGP-key: 1F28D8AE, www.vanheusden.com
> _______________________________________________
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
> _______________________________________________
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
--
Petr Baudis
If you do not work on an important problem, it's unlikely
you'll do important work. -- R. Hamming
http://www.cs.virginia.edu/~robins/YouAndYourResearch.html
_______________________________________________
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
