[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Run time for GLLAMM multilevel modeling
That's true, the multiprocessor feature thus far is relatively useless
for Stata, although I heard that Stata Corp. has been looking into
multithreading some of their code. A UNIX cluster however can provide
the reliability and uptime measured in months rather than in hours for
a Windows machine. Besides, if one is estimating five or ten models in
-gllamm-, that can easily be put into separate do-files thus using
multiple processors more efficiently.
I also know that our university fastest cluster is about three times
faster than my laptop, which is a considerable factor here. But all
that needs exploring for one's particular computational opportunities,
On Tue, 23 Nov 2004 10:24:49 +0100, Michael Ingre
> On 2004-11-22, at 22.12, Stas Kolenikov wrote:
> > Figure out if you can run your problem on your university computational
> > cluster or something.
> I'm not so sure that a cluster will make -gllamm- or even Stata faster.
> A few months ago there was a discussion about performance on this list
> and Alan Riley replied "Apple's dual G5, for an application like Stata
> which only makes use of a single processor at a time, will not run
> gllamm (or any other Stata code) significantly faster than the same
> computer with only a single processor would"
> (http://www.stata.com/statalist/archive/2004-04/msg00471.html). The
> fastest computer at that time to run Stata code was the Opteron running
> 64-bit Linux and 64-bit Stata.
* For searches and help try: