Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Overriding a loop if 0 observations using tabstat


From   Robert Picard <picard@netbox.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Overriding a loop if 0 observations using tabstat
Date   Wed, 28 Apr 2010 22:18:04 -0400

Vince,

Thanks for the cogent explanation. I really appreciate all the juicy
details. Everything makes sense when the extra free space is taken
into consideration. It's great to know that we can induce an extra
speed bump by the judicial use of a small memory allocation for those
special cases that combine cache size data and long run times. I'm
relieved that for everything else, the size of the memory allocation
does not matter unless you venture into VM issues.

Thanks for taking the time to enlighten us,

Robert

On Wed, Apr 28, 2010 at 5:27 PM, Vince Wiggins, StataCorp
<vwiggins@stata.com> wrote:
> Short question
> --------------
>
> In a thread that started out about looping then became about relative speed
> with differing amounts of memory, Richard Goldstein <richgold@ix.netcom.com>,
> Robert Picard <picard@netbox.com>, and Kit Baum <baum@bc.edu> have written
> about substantial differences in runtimes when performing many iterations of
> some very simple computations on a very small dataset.  They attribute the
> difference in runtimes to differing amounts of memory allocated to Stata.
>
>
> Short answer
> ------------
>
> First, I believe their timings.  Second, I am not concerned.
>
> There is nothing to worry about here.  These timings are an artifact of a
> toy-size example combined with the way modern computers handle cache memory.
> You will see timing differences of this sort only on small datasets and when
> when doing easy computations that are performed repeatedly -- Robert, Rich,
> and Kit (RRK) are computing means and counts over a 100,000 observation
> dataset comprised of a single variable.  The only reason they observe much
> timing passing at all is that they are iterating the task 10,000 times.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index