Re: [Sbcl-devel] Improving RANDOM.

[<lut...@fr...>]

Hi,

thanks to Nikodemus and Christophe for the feedback.
Nikodemus wrote (citing out of order):

> ...but I think only the first item is something that seems even
> vaguely in scope now.

I agree. So, below my comments/questions about your first and second
item:

> - Documentation love. https://bugs.launchpad.net/sbcl/+bug/656839

I plan to do something here, yes; but later.
I hope there is an automatic way to get docstrings into the
documentation? The docstrings of MAKE-RANDOM-STATE and SEED-RANDOM-STATE
would fit nicely, IMO.

> - SSE2ified mersenne twister on x86-64.

Yes, this is supposed to be faster and even "more random" than the
current version, AFAIK. On the other hand, the current code contains
a comment from WHN:

  ;;; Using inline VOP support, only available on the x86 so far.
  ;;;
  ;;; FIXME: It would be nice to have some benchmark numbers on this.
  ;;; My inclination is to get rid of the nonportable implementation
  ;;; unless the performance difference is just enormous.

So his principle seemed to be to prefer portable solutions over machine
specific ones?

Whatever; I currently don't intend to work on this. It can be done later
if desired without causing wasted efforts now.

I will upload a patch to LP #309467 for review and then post some more
specific questions about it.

Greetings,

Lutz