Chris, Daniel, it looks like you both volunteered in this thread to be the
"random" maintainer. Whoever it ends up being -- if you would like any
backup I'd be happy to be listed as a co-maintainer. I have some interest
in the topic because I was doing a bit of work recently on splittable RNG.
By the way, what is the feeling about using the FFI and foreign code in core
libraries' implementations (aside from unix)? The reason I ask is that it
would be nice to fix the statistical weakness in System.Random.split and I
think the best way to do it is:
- (1) use a cryptographic RNG technique
- (2) use native code and possibly HW acceleration to make up for the
performance hit in (1)
I did a prototype of this in http://hackage.haskell.org/package/intel-aes,
and it seemed to have the potential to make a decent stdGen. First, this is
because on the hardware I tested there was no performance regression
relative to System.Random -- the portable, C, AES-based implementation
performs a little better than the pure Haskell System.Random. Second, newer
Intel and AMD machines have AESNI, which at least doubles the performance
over the portable software version.
So it's a win for both performance and correctness; BUT there's more work to
ensure portability across hardware and across Haskell implementations -- so
does that tradeoff have any chance for a Core library?
-Ryan
On Wed, Jun 8, 2011 at 4:14 AM, Chris Dornan
Congrats on the award by the way – you have been winning my own superhero award but it’s cool to see some objective recognition.
I had volunteered for random (and anything else you needed) – I thought you were being diplomatic (and maybe you were).
Just made the major delivery on my contract so definitely feeling like some good works are in order…
Cheers,
Chris
*From:* libraries-bounces@haskell.org [mailto: libraries-bounces@haskell.org] *On Behalf Of *Simon Peyton-Jones *Sent:* 07 June 2011 10:43 PM *To:* Haskell Libraries *Cc:* Simon Peyton-Jones *Subject:* New libraries process
Friends
Last month Simon M and I put forward a proposal to revise the process for developing the Core Haskell libraries: http://www.haskell.org/haskellwiki/Library_submissions/NewDraft
There was some discussion, leading to improvements now incorporated in the draft, but I believe that there was general support; indeed no one opposed the change.
Several weeks have gone by, so I suggest that we adopt the new process forthwith. Unless I hear otherwise I'll replace http://www.haskell.org/haskellwiki/Library_submissions with the new process in a day or two.
We still need volunteers to become the maintainer of mtl random [though I think someone maybe did volunteer, Daniel perhaps?] Win32 Please!
best wishes
Simon
_______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries ------------------------------
No virus found in this message. Checked by AVG - www.avg.com Version: 10.0.1375 / Virus Database: 1511/3686 - Release Date: 06/07/11
_______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries