Re: JAVA LINPAC

[johnhazel@xxxxxxxxxxxxx]

On Sat, 16 June 2001, bob@acm.org wrote:

> >All the Linux flavors I have experimented with were noticably slower than
> >an equivalent task run on Windows. This is particularily true of screen/window
> >redraws.
> Interesting, I would want to see this, and the overall problem.

Do you have a dual boot Lin<>Win machine to play with?
It's quite apparent in redraws of windows after they have been moved around the desktop? 
There is also evidence that the order of magnitude performance gap on JAVA LINPAC that I experienced is not unique.
Go to:
http://www.netlib.org/benchmark/linpackjava/timings_list.html
Notice that on comparable machines the results are very much in favor of the Win machines. (save me tux! save me!)

Re slow redraw: 
> >Could the problem be there in the screen redraw?
> Perhaps. It may also be misconfig of the Xserver/card combo.

This has been the case on 486 33, 66, 100Mhz with many different video cards to and with PCI and AGP machines 350-800MHz using Storm, Caldera, Redhat, Mandrake, Turbo, and Suse vs Win 95, 98, 200ME all on default configurations. It seems unlikely that all these combinations would always misconfig the Xserver/card combo. I hate to blurt this out in the open but I'm not one to run from a problem... 

> >(BTW I'm looking for a reasonable Excel substitution, all candidates so far have been terribly slow and lacking features.StarOffice, Gnumeric etc. That led me to look at JAVA)
> Excel, etc are applications, JAJVA is a programming language; you can't really 
> draw equivalences here. A spreadsheet written in JAVA may be great, or it 
> may be a poor performer based on the skills (or lack) and needs of the spread-
> shhet implementor.

Yes what I meant was that no turn-key solutions (spreadsheets) would perform well enough so I was contemplating writing my own "Liftroll" wing analysis program to get the flexibility-performance that I need. So I looked at JAVA not to write a SS but to write a wing analysis program.


> You need to find a JRE/JVM guru, I'm none such, but they exist.. maybe even 
> here.

I'm going to post to some other places. This huge performance problem ought to get any of the commercial  Linux providers attention.
A run at the JAVA LINPAC authors also might help. The Xfree group should also panic over it.

BTW thanks for you kind responses.

John Hazel

Liftroll is at: http://www.geocities.com/jebbushell/COOKBOOK.htm 
Find the best deals on the web at AltaVista Shopping!
http://www.shopping.altavista.com