Re: [ng-spice] spice3f5 benchmark


To ng-spice@ieee.ing.uniroma1.it
From "GLAO S. Dezai" <dezai@hotbot.com>
Date Wed, 22 Mar 2000 04:49:38 -0800
Delivered-To mailing list ng-spice@ieee.ing.uniroma1.it
Mailing-List contact ng-spice-help@ieee.ing.uniroma1.it; run by ezmlm
Organization HotBot Mail (http://mail.hotbot.com:80)
Reply-To ng-spice@ieee.ing.uniroma1.it

 
--

> -1- you mention that the spice3f5 code is quite bad. Can you quantify this?
>     What about the kernel code: stamp generation, sparse matrix code, 
>analysis code?
>     Is this easy to structure and optimize these specific parts? 


My opinion is that the worst part of the spice3 code is the front-end. The 
code written by Quarles (ckt, ni,inp,dev)and Kundert(sparse) is not so bad on 
the point of view of numerical  algorithms and software design. The code is 
easier to maintain. An homogenous coding style (Quarles) has been used(names, 
pointers,...). A small number (?!? I think) of memory leaks should be found 
in these modules. However Some effort have to be made on the analysis parts 
to integrate modern analysis algorithms to enhance spice3's possibilities.
The front-end has been coded with a more free style. Pointer are allocated 
and not freed. Strings are copied every where etc...
I am not an expert of the software policies questions(GPL LGPL,...) but a 
good question is:
Should we spend our time to track and fix the enormous amount of memory leaks 
in fte, cp and misc and other ugly things or completly rewrite these modules 
? I think that is the way followed by the new simulator project. I think that 
professional designers like Reid should help to build or complete the new 
simulator design chart. All the specifications needed for a good and modern 
simulator.
Dezai 




HotBot - Search smarter.
http://www.hotbot.com

Partial thread listing: