Subject: RE: [sv-cc] Updated version of SV VPI extensions (dated Jan 12)
From: Joao Geada (Joao.Geada@synopsys.com)
Date: Wed Jan 14 2004 - 08:55:36 PST
Doug:
thanks for numbering the questions. Makes answering them easier.
1- note that there is also a class of VPI objects to describe "scopes"
(see page 15, section 26.6.3)
Objects have both a scope and a containing instance. In many cases the scope
and the instance are the same, but for objects in named blocks etc this
may not be the case.
Note that this distinction was already there in 1364-2001.
2- compilation units are treated exactly as packages, but with system-dependent
names.
We need to add a note to the diagram on page 1 to make this explicit.
3- right
4- yes, the diagram on page 11 needs some further cleanup (the arrows from struct,
union, bit and logic to range are not drawn quite right). Note that this relation
describes the packed dimensions (or packed array-like access for packed structs/unions)
5- should not be there (also refer to answer 4 above)
6- unnamed scopes come about because SV now permits unnamed blocks to contain variable
declarations, eg
always @(foo) begin
int i;
for (i = 0; i < 10; i++) something();
end
7- hm, needs further cleanup. And we need both a DPIImport and a DPIExport clause,
as the VPI needs to be able to identify both the functions/tasks coming in from
the outside world as well as the functions/tasks made visible to the outside world.
8- right
9- I think that relation went out drinking. It will be sober next time ;-)
10- good point. No, we have not incorporated tagged unions.
Joao
==============================================================================
Joao Geada, PhD Principal Engineer Verif Tech Group
Synopsys, Inc TEL: (508) 263-8083
377 Simarano Drive, Suite 300, FAX: (508) 263-8069
Marlboro, MA 01752, USA
==============================================================================
> -----Original Message-----
> From: owner-sv-cc@eda.org [mailto:owner-sv-cc@eda.org]On Behalf Of
> Warmke, Doug
> Sent: Wednesday, January 14, 2004 11:32 AM
> To: 'Joao.Geada@synopsys.com'; Sv-Cc
> Cc: Avinash G Mani; Charles Dawson; Francoise Martinolle; Duncan, Ralph
> Subject: RE: [sv-cc] Updated version of SV VPI extensions (dated Jan 12)
>
>
> Joao, Team,
>
> Here are some questions/comments on the latest version of the VPI spec.
> Sorry we didn't have time to do a more thorough job on this.
>
> 1. vpiInstance is described on pages 1 and 4 as
> (package, module, interface, program).
> On page 4, there is reference to an "immediately enclosing scope".
> What about named generate scopes? It seems to me that since those
> are valid scopes in the simulation image, a query to vpiInstance
> should return a handle to that scope, rather than its surrounding
> (package, module, interface, program).
>
> 2. What about Compilation unit scope? ($unit)
> I don't see any reference to that.
> $unit is *almost* exactly like a package.
> Some differences are:
> - The scope automatically runs out at the end of a translation unit
> (normally a single file, but can be the set of files presented
> to the tool at once as per a user-selectable option)
> - You can declare modules and interfaces in $unit scope but
> not in a package
>
> 3. Typo on page 3, "prgrams" should be "programs"
>
> 4. On the bottom of page 11 there is a red "Needs Note" that should
> be fixed up. There is some half-tone overlaid text to the right
> of that which looks unusual and should be cleaned.
>
> 5. To the right of bit Tspec is a dangling line segment.
>
> 6. In the Scope (26.6.3) section, how does an "unnamed scope"
> come about? I think we should be more explicit in this note.
> The only example I can think of is an unnamed generate block.
> However, the IEEE is leaning towards considering such scopes
> as fully collapsed into their next-outer enclosing scope
> (post-elab view).
>
> 7. In Task, Function Declaration (26.6.18), Note 2 refers to
> vpiInterfaceTask/vpiInterfaceFunction. However, those are not
> defined anywhere in the diagram above. Also, what is "extern"
> and "DPiExtern"? Shouldn't the latter be "DpiImport"?
>
> 8. Starting on page 27, the font describing the various "str:", "int:"
> etc. changes from the font normally used for that job. Should be
> cleaned up to be consistent.
>
> 9. On page 29, the "multiclock sequence expr" balloon has weird looking
> astigmatic font problems(!)
>
> 10. Did you have time to look over the tagged unions feature that was
> approved by SV-BC last month?
>
> Thanks for all the good work.
> It's really nice to have VPI up near to the feature set of the
> primary language, rather than lagging far behind.
>
> Regards,
> Doug
>
>
>
> > -----Original Message-----
> > From: owner-sv-cc@server.eda.org
> > [mailto:owner-sv-cc@server.eda.org] On Behalf Of Joao Geada
> > Sent: Tuesday, January 13, 2004 2:16 PM
> > To: Sv-Cc
> > Cc: Avinash G Mani; Charles Dawson; Francoise Martinolle;
> > Duncan, Ralph
> > Subject: [sv-cc] Updated version of SV VPI extensions (dated Jan 12)
> >
> >
> >
> >
> > Updated SV VPI extension diagrams
> >
> > Joao
> >
>
This archive was generated by hypermail 2b28 : Wed Jan 14 2004 - 08:57:10 PST