Friday, August 26, 2011

two dudes spotted inside a house fighting with the lights on: random!

Date: August 26, 2011 8:29:50 AM EDT
Subject: Beowulf Digest, Vol 90, Issue 18

The "beoboys" have been chatting about random numbers for a few weeks now. Turns out that not only has the beowulf list fallen into an utter shambles of testosterone fueled rants with hardly anything to do with HPC, more recently it has actually become really, really pretty funny ;-)

Check out this one:
> If we assume that reality of life represents randomness, which is another
> rather good question in how far that theory is plausible, then using that
> assumption i'm very sure that the RNG's i investigated so far have a
> distribution which is too perfect, more perfect than i have seen in any
> reality.

That's because you live in a different reality than everybody else, Vincent.

> In fact most RNG's fill all slots faster than O ( n log n ), yet it's O ( n
> log n ) that they follow.

In fact, they don't.

> This is RNG's that have come through all tests as being a good and
> very acceptabe RNG to be used.

No, it's not.

> Realize i'm no RNG expert, so all the names of all those tests.
> For me it's just push button technology. I just designed a test
> and found it very odd that all RNG's have such perfect distributions
> that they don't even miss a single slot.

It's odd because your test is broken.

It gets way, way better later on, in addition to some pretty awesome references to "space alien" type posts and some other pretty scorching, hard-code vitriolic stuff this utter gem pops up at the end:
Seriously, dude. Ask yourself "Am I really smarter and better informed
than Pascal, Fermat, Laplace, Bayes, not to mention all of those
contemporary humans who have been devoting entire well-educated
careers to random numbers as if all of modern e-commerce depended on
them (it does) or is it just barely possible that I've made a mistake?"

Come on, you can do it. I know it is difficult for you, but try..

*coffee goes everywhere!*

The full thread is here:

In summary beowulf list =~ dodo =~ slashdot

c'est fini!

Thursday, August 18, 2011

O for the memories

My first memory of real scientific computing was working at Manchester University under Prof. John Helliwell with Alwyn Jones's O package.

So just had to go old school on our iMac (now running Linux due to the heat) and see if I could remember those glory days... Yes I know - a long way from my old SGI tucked away in a dark corner of our crystallography lab in Manchester, but it was worth a shot, just to see if this "manager" can still hack it...

All credit Alwyn:

T.A. Jones, J-Y. Zou, S.W. Cowan & M. Kjeldgaard (1991) Improved methods for the building of protein models in electron density maps and the location of errors in these models. Acta Cryst. A47 110-119.

[any opinions here are all mine, and have absolutely nothing to do with my employer]
(c) 2011 James Cuff