Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Sloooow Graphics


From   David Kantor <dkantor@jhu.edu>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Sloooow Graphics
Date   Tue, 29 Jun 2004 17:16:23 -0400

At 10:18 AM 6/29/2004 -0400, Jean-Philippe C. Stijns wrote:

Hi All, Does anyone know why Stata is so slow to produce graphs. I have
already set my mem to 200m. Is there any other potential issue? Thank you.
First, Graphics were completely rewritten as of Stata 8, and they are slower than before. But you get much more capability and customizability.

For my own use, I do not use graphics often, and I have not learned much about the new graphics. So, for the rare times I need graphics, I will often use the old version -graph7- or -gr7-. While this is retro technology, and you could criticize me for not keeping up-to-date, it has two advantages: I already know how to use it (as far as I need it), and it is faster. Often, the speed advantage alone is compelling enough to make me prefer it.

See -help graph7- or -help gr7- for more details, including links to other version 7 graphing programs.

The other matter is that you set mem to 200m. You should be aware that at a certain point, setting mem to a high value will cause major slowdown -- a quantum leap downward in speed. This is when you run out of actual memory and you start to use virtual memory. When this happens depends on several factors, some of which are a bit unpredictable. These are: amount of physical memory on the machine, how much memory is allocated to other processes, and what operating system you use. (My experience with Windows is that, with no other user processes running, I can take about 85% of the machine's physical memory before virtual memory kicks in. This is my experience, whereas some system experts have told me to expect this to be more like 50%. It may be that the portion needed by the system is relatively constant -- rather than proportional.) It's actually more complicated, as what you really need is contiguous memory, if I understand it correctly. There may be some FAQs or archived Statalist messages on this matter.

And I should add that, as I understand it, prior to the point of invoking virtual memory, taking more memory will not speed up your processing; but it will make the difference in whether your processing can run at all.

I hope this helps.
-- David

David Kantor
Institute for Policy Studies
Johns Hopkins University
dkantor@jhu.edu
410-516-5404

*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* 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   |   What's new   |   Site index