Re: [sv-ec] package classes and anonymous programs

From: Steven Sharp <sharp_at_.....>
Date: Tue Nov 21 2006 - 15:26:46 PST
>From: Gordon Vreugdenhil <gordonv@model.com>

>Perhaps.  We already have exceptions for systf calls in a constant
>function context as one example.

Some comments on that:

1. I opposed allowing systf calls in constant functions and ignoring
them.  That doesn't change the LRM, but does show that my viewpoint
is self-consistent.

2. Since those contexts evaluate the function during elaboration rather
than at run-time, that weakens the expectation that they should behave
the same way.  We don't even evaluate them in remotely the same way:
we use an interpreter during elaboration, and compiled code at run-time.

3. Overriding of built-in systfs with user-defined ones can change the
behavior of systf calls from run to run.  We allow dynamically loading
the user PLI routines.  You could argue that ignoring the systf calls
at elaboration time is equivalent to dynamically linking them to no-ops,
instead of to the built-in definitions which could be meaningless at
elaboration time.  Since you may not be running in the same executable
at elaboration and run-time, there is no reason to expect the binding of
the systfs to be the same.

If this is the only example, I think it is too weak of an analogy to
provide much precedence.

Steven Sharp
sharp@cadence.com
Received on Tue Nov 21 15:26:57 2006

This archive was generated by hypermail 2.1.8 : Tue Nov 21 2006 - 15:27:06 PST