Subject: Re: 1364.1 pragmas
From: David Bishop (dbishop@server.vhdl.org)
Date: Wed Sep 04 2002 - 18:31:51 PDT
-------- Original Message --------
Date: Wed, 04 Sep 2002 09:29:34 -0700
To: pgraham@cadence.com, cliffc@sunburst-design.com
From: Karen Pieper <Karen.Pieper@synopsys.com>
Subject: Re: 1364.1 pragmas
Cc: vlog-synth@server.eda.org, etf@boyd.com
The long term risk is not with pragmas in use today, but the expanding use
of them which will start.
For example, a number of users are beginning to think about how to annotate
relative placement
info into the RTL source (we can discuss whether or not that is a good idea
in an independent
forum). The issue is that one tool may use lambdas as the measurement,
someone else microns,
and someone else some other unit. The conflicting units going forward is
really the issue.
Additionally, as customers begin to use their own attributes the name space
will quickly explode.
Karen
At 07:33 PM 9/3/02 -0700, Paul Graham wrote:
>Precedence: bulk
>
>Regarding attribute "domains", is it likely that different domains of use
>will require homonymous attributes with different meanings? Is it likely
>that the synthesis "fullcase" attribute will collide with a "fullcase"
>attribute in another domain that means something other than "this case
>statement has all the choices specified"? Is it likely that "one_hot" will
>ever refer to something other than a coding scheme? Will "black_box" mean
>something other than an instantiation of an unknown module?
>
>One good way to answer these rhetorical questions is to ask some experts in
>formal verification and simulation to see if they already use pragmas or
>other directives which conflict with existing synthesis attributes.
>
>The 1364.1 committee looked at existing synthesis pragmas and found only 15
>to convert into pragmas. While there are other pragmas in use which do not
>appear as attributes in the 1361.1 document, the fact is that the synthesis
>pragma name space is extremely sparse. If the pragma name spaces from other
>domains are as sparse, then attribute name collisions among domains will be
>very unlikely.
>
>Paul
This archive was generated by hypermail 2b28 : Wed Sep 04 2002 - 18:41:26 PDT