Re: Connect-ResolveTo stmts


Subject: Re: Connect-ResolveTo stmts
From: Kevin Cameron (Kevin.Cameron@nsc.com)
Date: Fri Mar 08 2002 - 21:34:54 PST


[ I don't think I'll be attending the meeting on Monday (if there is one) since
I'll be attending HDL-Con - I presume others will be going too.]

Srikanth Chandrasekaran wrote:

> Hi all,
>
> Problem:
> With the current version of the LRM there is lot of ambiguity with regards to
> which of the rules apply when trying to do discipline resolution
>
> connect x y b resolveTo b
> connect x y resolveTo x
> connect x y a resolveTo a
>
> In the above set of rules it is not clear, when two nets having disciplines x
> & y will be resolved to b, z, or a.
>
> Also it is not clear in the LRM whether the resolved discipline should be part
> of the list specified.

Shouldn't be a requirement.

> Solution: Best Match Fits
> When there is an exact match for the set of disciplines specified, the
> resolved discipline would be as per the rule specified in the exact match. In
> the above case x,y would resolve to discipline x. It shall be an error to
> specify more than one rule with the exact fit. The resolved discipline need
> not be one of the disciplines specified in the discipline list.
>
> When there is no exact fit, then the resolved discipline would be based on the
> subset of the rules specified. If there is more than one subset matching a set
> of disciplines, the simulator shall give an error message on ambigous connect-resolveto
> specification saying that there is more than one rule that matches for resolving the
> disciplines.
>
> Example 1:
> ----------
> connect x y a resolveTo a
> connect x y resolveTo x
>
> x,y would resolve to x.
> x,y,a would resolve to a
> y,a would resolve to a
>
> Example 2:
> ----------
> connect x y a resolveto a
> connect x y resolveTo x
> connect x y b resolveto z
>
> x,y would resolve to x
> x,y,b would resolve to z
> y,a would resolve to a
>
> Example 3:
> ----------
> connect x y a resolveto a
> connect x y b resolveto b
>
> x,y would resolve to "ERROR: Ambigous"
> y,a would resolve to a
> y,b would resolve to b
>
> Example 4:
> ----------
> connect x y resolve to x
> connect x y resolve to y // ERROR: Conflict
>
> cheers,
> Sri

I'd agree that missing rules should be an error, but over-specifying the rules
should (in my opinion) just generate a warning. It could be difficult to fix the
rules in practice when re-using IP. For now I would just take the "first seen"
rule or leave it as vendor specific.

There is no implication that either rule is "wrong" when two apply.

Kev.



This archive was generated by hypermail 2b28 : Fri Mar 08 2002 - 17:37:37 PST