Subject: RE: [sv-cc] Meeting minutes, July 16, 2003
From: Warmke, Doug (doug_warmke@mentorg.com)
Date: Thu Jul 17 2003 - 12:56:18 PDT
Hi Andrzej and others,
Thanks for taking the time to communicate more details on
this issue over the reflector. I'll be able to contribute
more effectively if I can see this kind of detail. Sorry
for my earlier misunderstanding. From the meeting minutes,
it appeared that the detailed analysis and suggestions I had
posted were not considered. Thanks for doing so.
What appears to have happened is that there is a philosophical
question about maximizing re-use in the grammar between DPI
functions and native functions. Actually, I would prefer this
approach myself, since it is much more elegant and it is more
in the spirit of keeping DPI import functions as close as
possible to native functions.
However, when I tried to make that work out, it was not neat,
nor elegant. Thus I decided on the "bifurcated grammar"
approach I proposed in my earlier mail.
We discussed this internally, and after weighing pros and cons,
we prefer the bifurcated grammar approach.
The main reason is that the re-use-with-semantic-restrictions
approach involves providing incomplete and misleading information
at the one place where formal means of expression should maximize
accuracy. The LRM reader is required to perform an ad hoc search
throughout the text sections, hunting for caveats that contradict
the formal grammar. While this isn't totally unreasonable, it is
more obtuse and hard-to-use.
Another reason to bifurcate the grammar is that it gives us
(and the other SV committees) a more stable base from which to
make changes, without perpetrating surprise side-effects on
other parties.
The cost of this bifurcation is technically minor. Just a half
dozen or so extra (and simple) nonterminals. The true cost seems
to be our concept of the "elegant" parallels between import
functions and native functions. I still think this concept can
apply at a high level. But I believe we should take advantage
of BNF's expressiveness in order to reflect the true syntax
of the import function construct.
Anyways, that is our vote. Either way can work, mainly I will
be happy to see the currently inadequate grammar put to rest.
Regards,
Doug
-----Original Message-----
From: Andrzej Litwiniuk
To: Warmke, Doug
Cc: sv-cc@eda.org
Sent: 7/16/2003 1:48 PM
Subject: Re: [sv-cc] Meeting minutes, July 16, 2003
> I'm curious why no mention was made of my email
> on the BNF issue, nor my suggested changes in
> the grammar. I don't think the changes were quite
> ridiculous enough to deserve no discussion at all.
Doug,
Although your email on the BNF issue is not explicitly mentioned
in the minutes, it was your email because of which the issue was
brought up today.
Joao and I had a short discussion before todays meeting
and in our opinion the way to go is to have a single common set
of productions both for native function headers and for import DPI
headers
with semantical restrictions.
We believe that it would be easier and more convenient to add semantical
checks rather than to duplicate virtually a lot of productions.
Actually, this is a typical approach in language definitions,
for example, a syntax for expressions typically allows much more than
is actually permitted in a language.
I don't think any binding decisions have been made today, nevertheless
my impression is that no one objected the approach "common productions
+ semantical restrictions"
What do you think on that approach?
Thanks,
Andrzej
> Even if they were, I'd like to hear the reasons -
> then I would learn something important.
>
> Thanks and regards,
> Doug
>
>
> -----Original Message-----
> From: Francoise Martinolle
> To: sv-cc@eda.org
> Sent: 7/16/2003 9:32 AM
> Subject: [sv-cc] Meeting minutes, July 16, 2003
>
> Approval of minutes: Andrzej
> Michael seconded
>
> Attendees: Swapnajit, Michael, Joao Ghassan, Francoise
>
> Swapnajit:
> role of the champion:
> Our champion (Joao) would be involved for any changes of the sv 3.1
> LRM.
> He would take take part in any discussion affecting the 3.1 LRM. If
> anything needs to be changed, the champions would be looking at other
> committees sections and resolve the issue.
> The chair has to approve the change.
>
> How erratas will be done? same mode of working as before
>
> sv list of issues compilled
>
> garbage for chandles: was editorial mistake
> Joao to submit the editorial change.
>
> Who is in charge to integrate erratas in the LRM? Probably Stuart.
>
> Dave Smith is collecting erratas.
>
> Question on the BNF:
> Andrzej: two specialized productions
> or uniform set of production with semantical restrictions : on
> the
> type of formal arguments and passing argument mode not permitted for
> DPI.
> Andrzej and Joao by Friday will send to the cc committee a proposed
set
> of
> changes to correct this problem. By next meeting have a proposed
errata.
> 1
> week review period set by Accellera shall be obeyed.
>
> Joao prefers the uniform production.
> Swapnajit: if any changes in the set of rules inthe BNF, does it have
> any
> impact on the other parts of the bnf.
> Changes to the basic production defining function will have an impact.
> single change in a single place.
>
> Joao stated that this change needs to be agreed by other committees.
>
> deadline for next meeting to review list of issues which was sent
out
>
> Meeting ajourned.
>
This archive was generated by hypermail 2b28 : Thu Jul 17 2003 - 12:57:54 PDT