Re: [ng-spice-devel] Some suggestions for ng-spice


To ng-spice-devel@ieee.ing.uniroma1.it
From Erik de Castro Lopo <erikd@zip.com.au>
Date Wed, 24 May 2000 08:47:24 +0000
Delivered-To mailing list ng-spice-devel@ieee.ing.uniroma1.it
Mailing-List contact ng-spice-devel-help@ieee.ing.uniroma1.it; run by ezmlm
Organization Erik Conspiracy Secret Labs
References <Pine.LNX.3.96.1000524081329.20739A-100000@ieee.ing.uniroma1.it > <392B81AF.515069EB@lpmo.univ-fcomte.fr >
Reply-To ng-spice-devel@ieee.ing.uniroma1.it
Sender erikd@zip.com.au

Nicolas Ratier wrote:
> 
> Dear ng-spice group
> 
> I feel very concern with your project :
> 
> But I don't understand why you really need to used the spice3f4
> code. 

Actually ng-spice uses the spice3f5.

> From a software point of view it is often easy to rebuild
> a project from scratch than to modified (bad written) existant one.

Ng-spice works now. I have written a filter program which re-writes
spice2/xspice models as spice3 code (I'll be releasing it after I
do some more testing). How long is it going take to write a new 
simulator from scratch? When can I start using it?


> To my mind "bad written" mean than it was write with a 30 years
> technologie code. But be sure I admire the scientific work made
> by the Berekley team.
> 
> That says, may I suggest some things :
> 
> 1) The scanner and parser should be written with modern tool
> like flex, bison and gentle.

The Spice "language" is not nice and regular and hence cannot easily
be parsed by machine generated parsers.

> 2) The integration and iteration methods should be completly
> isolated from the physical models.
> 
> 3) The "pure mathematics" part should used powerfool tool
> already build for that : blas, lapack, ... etc. don't reinvent
> the weel.
> 
> 4) Used Object-Oriented Scientific Computing. Now, when
> using C++ and OOP, is is possible to make code as fast
> as f77 code. Some code C++ library have been developp
> is that goal. I think to Blitz++ code (GPL !!!!) :
> http://oonumerics.org/blitz

You should be aware that the current stated goals of the ng-spice team
is to work in C rather than C++.

Personally I think C++ is an abomination but I don't want to
start a language war here. 

> 5) The documentation shoult be very expensive and pedagogical.
> I known that is a boring part of the job, but it is
> this part that users will most used ...
> What does a new developper must understand to participate ?
> What is the stucture of the program.
> What are exactly the model used ?
> (I fell very frustrated with the Berekley spice documentation)
> 
> That all !!! I hope very much you take all this, not for
> critisism, but for positive suggestion. May be, it will be possible
> to meet together during a few days to speak of all of that in
> an informal way. Most people taking part in the ng-devel seems
> to coming from Europe, so it should not be too expensive.

I'm all in favour of a new simulator. I for one however will continue
to maintain and USE the current simulator for CURRENT work. When and if 
a new simulator comes along that works as well as ng-spice I will be
happy to switch to it as long as its open source and hackable.

Just my $0.02,
Erik
-- 
+-------------------------------------------------+
     Erik de Castro Lopo     erikd@zip.com.au
+-------------------------------------------------+
Q. What is the difference between Jurassic Park and Microsoft?
A. One is an over-rated high tech theme park based on prehistoric
   information and populated mostly by dinosaurs, the other is a 
   Steven Spielberg movie.

Partial thread listing: