Subject: Re: [sv-cc] Minutes for the SV-CC Conference Call @ 22-Apr-2003
From: Kevin Cameron x3251 (Kevin.Cameron@nsc.com)
Date: Wed Apr 23 2003 - 11:18:27 PDT
> From Andrzej.Litwiniuk@synopsys.com Wed Apr 23 10:21:37 2003
>
> > As National Semiconductor's representive I'd like to record a "NO" vote for the CC committee's
> > portion of the LRM - National's needs were not addressed by the committee.
> > [...]
> >
> > Regards,
> > Kev.
>
> Dear Kevin,
>
> I do respect your opinion and your right to judge whether National's needs have been addressed.
> I also understand that you would much favour a more object-oriented C++ style solution.
>
> Do you think it may be conceivable to define C++ layer of DPI,
> while sticking to the same SV layer?
>
> I believe that SV side of DPI is very SystemVerilog-ish since it virtually
> introduces no new concepts and keeps external function calls undistinguishable
> from native calls. On the other hand the actual implementation, linkage
> and function call and argument passing protocols are transparent to SV.
> So it seems that an alternative, more C++ oriented implementation of the foreign
> side of DPI is not unfeasible.
> Surely not in 3.1 timeframe, but in the future? Why not?
> It might be welcome by SystemC fans ... What do you think?
>
> Regards,
> Andrzej
It's certainly possible to do a C++ "DPI" - a more direct implementation of VPI would
probably do.
I currently have a renewed enthusiasm for SystemC. I think it was a big mistake not
to go with C++ syntax for the class mechanisms of SV. Just building the bridges from
SV 3.0 to C++ (SystemC) would probably have been less effort and more useful than
a lot of what has been added for 3.1.
Regards,
Kev.
This archive was generated by hypermail 2b28 : Wed Apr 23 2003 - 11:20:28 PDT