Subject: FW: BOUNCE vhdl-200x-tbv@eda.org: Non-member submission from [Kevin Cameron
-----Original Message-----
>From bhasker Fri Oct 10 13:11:57 2003
On Fri, 10 Oct 2003, Tim Schneider wrote:
> ** At 10:18 AM 10/10/2003, Jayaram Bhasker wrote: **
The only dangerous thing about pointers is dereferencing them
Personnally I think VHDL (and Verilog) should have standard calling
Kev.
>
This archive was generated by hypermail 2b28
: Fri Oct 10 2003 - 13:23:35 PDT
From: Jayaram Bhasker (JBhasker@esilicon.com)
Date: Fri Oct 10 2003 - 13:18:33 PDT
From: owner-vhdl-200x-tbv@eda.org [mailto:owner-vhdl-200x-tbv@eda.org]
Sent: Friday, October 10, 2003 4:12 PM
To: owner-vhdl-200x-tbv@eda.org
Subject: BOUNCE vhdl-200x-tbv@eda.org: Non-member submission from [Kevin
Cameron <grfx@grfx.com>]
Received: from typhoon.he.net (typhoon.he.net [64.62.229.2])
by server.eda.org (8.12.0.Beta7/8.12.0.Beta7) with SMTP id h9AKBt6O028393
for <vhdl-200x-tbv@eda.org>; Fri, 10 Oct 2003 13:11:57 -0700 (PDT)
Received: from localhost ([127.0.0.1]) by typhoon.he.net for <Tim.Schneider@synopsys.com>; Fri, 10 Oct 2003 13:11:43 -0700
Date: Fri, 10 Oct 2003 13:11:43 -0700 (PDT)
From: Kevin Cameron <grfx@grfx.com>
X-X-Sender: grfx@typhoon.he.net
To: Tim Schneider <Tim.Schneider@synopsys.com>
cc: vhdl-200x-tbv@server.eda.org, "" <vhdl-200x@server.eda.org>
Subject: RE: [vhdl-200x] Re: VHDL 200x request
In-Reply-To: <6.0.0.14.0.20031010104125.02799c38@pophost.synopsys.com>
Message-ID: <Pine.LNX.4.50.0310101253070.28324-100000@typhoon.he.net>
References: <6.0.0.14.0.20031010104125.02799c38@pophost.synopsys.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
>
> >Using VHPI is a non-trivial task for a typical designer.
>
> Agreed...
>
> A direct "C" mechanism definitely has usefulness. I've seen folks on the Verilog
> side (designers) gravitate to this much more easily than writing such in
> PLI/VPI For a "quick" model its really useful, access would be either
> port or process/task based.
>
> Keeping it modular and not allow passing of anything that is not allowed in
> standard VHDL (i.e. no pointers and such) helps keep it somewhat 'safe'.
> and would avoid language bloat.
inappropriately, since VHDL (and Verilog) doesn't have to allow you to
dereference them there isn't much harm in allowing them to be passed from
one piece of foreign code to another. SystemVerilog is a bit stupid about
this since although it has sufficiently C-like syntax to allow foreign
(C) code prototypes it doesn't allow anything other than a void (handle)
type for pointers, so type checking is limited.
interfaces that match C/C++ so that cross calling is simpler than the
current schemes - it all ends up going through the same assemblers
and linkers (and sometimes C/C++ compilers) anyway. Things like PLI
only exist because the original implementation of Verilog was an
interpreter rather than compiled code, similarly VPI would be better
implemented directly in C++.
> -tim
>
>
>