RE: [sv-cc] DPI unions as parameters

Ralph, thanks for all the spadework on this.  I personally
am happy with the idea of not supporting unpacked unions
-- especially if clients can use context and VPI to get
the information some other way.

I did have a couple of questions.

The Bugnote writeup section "1. Unpacked, Untagged" has
a sentence saying "Thus, it is not possible to pass
packed, tagged DPI unions in a reliable, simulator-independent
way."  Did you really mean "packed, tagged" in this
sentence, or did you mean "unpacked, untagged"?

Sections 3 and 4 state that requiring the user to
do the data marshalling requires "that the C user
knows the simulator's internal (SV) representation
for all relevent members."  Is that strictly true?
After all, aren't we already allowing simulators to
use C-D pairs rather than A-B pairs?  Would it be
better to require a fixed representation of the
equivalent packed array to be passed?

Regards,
Jim V.

--------------------------------------------------------- 
James H. Vellenga                            978-262-6381 
Engineering Director                   (FAX) 978-262-6636 
Cadence Design Systems, Inc.         vellenga@cadence.com 
270 Billerica Rd 
Chelmsford, MA 01824-4179 
"We all work with partial information." 
---------------------------------------------------------- 
  
 

] -----Original Message-----
] From: owner-sv-cc@eda.org [mailto:owner-sv-cc@eda.org] On 
] Behalf Of Duncan, Ralph
] Sent: Tuesday, January 31, 2006 7:59 PM
] To: SV-CC
] Subject: [sv-cc] DPI unions as parameters
] 
] There is a new Mantis item, #1322, which details problems 
] with various species of 
] unions as DPI parameters.
]  
] Most of discussion appears as a Bugnote, where some 
] wrap-around effects make
] the text hard to follow.
]  
] I deliberately haven't prepared a proposal, since 
] . that presupposes the outcome of our discussion,
] . the actual text changes may be less work than deciding 
] what's best to do.
]  
] Ralph Duncan
]