This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: getting more random
- From: Paolo Carlini <paolo dot carlini at oracle dot com>
- To: Benjamin Kosnik <bkoz at redhat dot com>
- Cc: libstdc++ at gcc dot gnu dot org, Ed Smith-Rowland <3dw4rd at verizon dot net>
- Date: Thu, 02 Apr 2009 10:56:41 +0200
- Subject: Re: getting more random
- References: <20090401112221.75c0141c@milvia.artheist.org>
Hi everyone,
> Paolo:
> - Wants time for algorithm analysis, getting feedback from TR1
> implementation and use, literature review.
>
and sorry for a bit of delay in returning to these issues, but indeed,
now it's stage one, the proper time.
Anyway, give me a few hours to have a closer look to the patch: modulo
the small changes already mentioned by Ed, I think we could certainly
start from that code and then move from it. Personally, my higher
priority would be replacing the algorithm for Gamma variables with "A
Simple Method for Generating Gamma Variables" by Marsaglia and Tang (I
mentioned in that thread in December). But in general, since we are
talking about mainline, we can certainly work incrementally, no hurry to
have something stable from the binary point of view too. One last thing:
I have some decent code implementing actual testing of the
distributions, developed with a friend of mine (he wants to remain
anonymous). I can pass it to Ed for further work (last time we checked
one of the available distributions didn't pass a test and we suspected
the testing infrastructure itself had some remaining glitches)
Paolo.