Subject: Re: Checking for dimensional consistency in VHDL-AMS models
From: Alain Vachoux (alain.vachoux@xemics.ch)
Date: Tue Dec 21 1999 - 00:44:51 PST
Dear working group member,
You'll find below a reply from Alex Zamfirescu that bounced. Alex asked me
to repost his mail so do I.
Best regards, Alain Vachoux
---------------------------------------------------------------------
Dear Experts:
It is fine to think about solving the dimension
problem in a general way.
However, what I see and understood that was proposed is
a system where only real subtypes will be used. In that case
we DO NOT NEED TO SOLVE THE POLYMORPHISM
PROBLEM YET, since operator resolution is based on BASE
type. The proposed solution exploits this "feature" that might
look like "polymorphism." The attributes attached to the
subtypes could be simply checked against by calls through
PLI (when that will be standard) at the beginning of simulation
(after elaboration).
There is a set of choices for the mechanisms to trigger the check.
With this VHDL could have a system that checks the units
that is complementary to what will come out of the Rosetta
effort. Note that we do not use phisical types (which are
discrete types) here, and I hope that Rosetta will address this
problem too, not only the discrete types.
My conclusion is that a package of subtypes and a
STANDARD mechanism for checking will be adequate for
some time, and it is better than what we have without
any check.
Dave, I allways admired your wide view, and am sure that
Rosetta will solve the problem (right this time), but may be
smaller (local) solutions have room too. So, my question
to you Dave is:
1. What is wrong with a system that checks (in a standard way)
that subtypes match for dimension?
The other questions are:
2. Is such system feasible with what
VHDL provides?
3. If not, what are the minimal changes or required additions?
In a paper presented this Fall at VIUF we proposed adding
a validation phase after elaboration and before simulation.
We used that phase for BDD based verification that the
logic of the gates compute the right boolean functions. We used
PLI calls. Same technique in the same phase could be used
for static dimension checking. In fact for the dimension problem
there is no need for waiting until elaboration, but waiting pays off
by giving the ability to use PLI calls that will soon be standard.
I wish you all a "multi-dimensional" pleasure in enjoying
the holidays, and the change of y into k :-)
Regards,
Alex Z
---------------------------------------
Alex N. D. Zamfirescu
VP Engineering
Alternative System Concepts, Inc
Telephone: 650 473 1067
Confidential Fax: 650 323 4643
E-mail: mno1@ricochet.net
---------------------------------------
This archive was generated by hypermail 2b28 : Wed Jan 26 2000 - 15:58:31 PST