[Fwd: Re: random variate from power exponential distribution: continue]

[Giulio Bottazzi]

Hi all,
I'm sorry to interfere with this enjoyable and, at least for me,
extremely instructive discussion about architectures comparison but I
would like to resubmit to your attention the problem of the generation
of power exponential random variates in gsl.

Using the modified program by Olaf (thanks a lot for your suggestions
!)I tested again on two machines (Mobile Pentium 4 and Opteron) adding
my new findings to Olafs' original result.txt (attached).

If I can dare a tentative summary, it seems Intel, Alpha and Power4+
share more or less the same sequence of "best" algorithm
when b is increased, i.e. GS-ED-EN-GS (analogously to the original paper
by Tadikamalla) while Sun is better with a GS-ED-GS sequence and on AMD
GS is almost always preferable.

Do you think that, in a hypothetical rewriting of the actual
implementation of the power exponential random variable generator in GSL
(randist/exppow.c), one should take in consideration the different
architectures, maybe via architecture-specific pieces of code toggled by
preprocessor directives? Or it would be better to try to cook up a
"reasonable" routine valid on "the average machine" (whatever that
means... probably nothing)

I would be happy if someone can provide me with some opinion/suggestion
on this point.

Best,
	Giulio.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: results.txt
URL: <http://sourceware.org/pipermail/gsl-discuss/attachments/20041020/37c19043/attachment.txt>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/gsl-discuss/attachments/20041020/37c19043/attachment.sig>