Minutes of the OOVHDL meeting at the Spring VIUF, San Jose, CA, on March 19, 1997

Minutes taken by David Barton, Chair.

Discussion on connection between other efforts and OOVHDL

Dave reported first on the DASC Steering committee charge.

Kamal stated that he doesn't want to connect up with the SID group; there is a distinction between systems design and how it is expressed (OO). John agreed; a system can even be in assembly language.

Greg disagreed. OO structures seem useless at gate or RTL level; the only place that it seems useful is in the systems area. It may be useful to keep things separate, but eventually come together.

Wolfgang said as well that OOVHDL structures are useful at the software / hardware codesign, and at the interface between hardware and the system. Thus, there needs to be a close relationship between OOVHDL and SID. We are a mechanism for the interface definition. There may be other mechanisms.

Kamal chimed in that, even though Greg is correct that OOVHDL is more useful at a high level, that OO is not restricted to a specific level. Object orientation makes the basic infrastructure of the language better (higher), and this can effect everything. John then said that he agreed that OO is not visible at the structural or RTL level (by definition). This is not an argument against object orientation. Just because something supports an interface process that we need to be absorbed in the systems group.

John's point was that to combine the OOVHDL and systems level group is to combine two separate goals: systems design with a kind of language structure. This is a kind error that we need to avoid.

Greg took the approach of looking at specific efforts. SLDL is separate, at a much higher level. VHDL 2000 Systems is also going to be at a higher level. On the other hand, Greg is in favor of the present DASC charge to pin down what the difference between the two groups is. Greg says that the SID stuff is concentrating on interfaces, and that adding OO features to this might be very worthwhile in spite of the fact that they might want to be separate later. Greg thinks the SID group is more a recommended practice than a set of language practices.

Peter said that they are taking a dual approach to OOVHDL: data modeling and architecture modeling. SID seems to do both, where Suave is more data oriented. Peter suggests that the groups might be able to separate along this axis, while talking very closely. It seems that there is an intersection in the architectural area in particular. Kamal agreed with this basic separation.

John asked if any had a problem with a VHDL that supports all known language constructs? Laughter --- all these constructs may not be necessary for any one project. Perhaps (John says) VHDL should be the union of all constructs, with recommendations for various applications.

Peter said that we need to avoid multiple solutions for specific design situations. This complicates training, tools, and design portability. John agreed with this. John noted that we have separate standards now: .1, .2, and whatever. These may not be consistent. Designing the language so that *some* compiler can do the whole thing might be a good idea. Greg said that solving this is a combination of two thing: the VASG / DASC needs to maintain control, and making sure the different efforts talk.

Kamal said that most languages are used in subsets; this is certainly true for VHDL. But Kamal does not want this to get too far out of hand. At least if people use the same language, they should be able to recognize it.

John asked if we need new language constructs to support OO? General yes.

Wolfgang then stated that the discussion seems to be iterating over a specific round of discussions. The group has made little overall progress; however, there is a lot of specific technical progress. Wolfgang suggested that we now enter a new kind of discussion by entering into a new round of reviewing specific examples written in the differing OOVHDL candidates. Dave asked if this is oriented to a broad design community or this study group, and Wolfgang said the former. This may not be practical, but Wolfgang think it might be doable. Siemens is interested in doing a comparison, for example.

John agrees with sampling the user community; however, he wants to sample based on existing constructs rather than on whole languages. Examples of source code seem important to John here. Kamal responded that this is the way that SID is working now, and it seems fairly successful. Wolfgang said that this is the process that we had, in fact, been doing and that we have moved past this. John said that we did not do this with clear goals. Wolfgang said that this may be true for the study group, but is not true for the two groups with proposals.

Wolfram said that it is difficult to ask designers for their conceptual wishes. Designers tend to prefer the concrete. If a modeler is not familiar with object orientation, then we cannot expect him to give us good requests. John took this further, asking if we know better than the designer what they need. Peter said this may be true, trying to avoid arrogance.

Dave warned against acrimony in discussions about competing proposals, and everyone in general agreed.

John asked about a sanitized version of the results of the designer survey, and Wolfgang took this as an action item. Wolfram says that the version of the document on the reflector is the latest version, and Wolfgang said that he was referring to the more detailed version. To be determined later.

Dave tried to give a summary after a change of rooms. In the summary, Dave mentioned the two proposals, and Greg mentioned that the University of Evry have a specific proposal for object orientation as well. Peter and Phil asked if they have a concrete proposal, and Greg was not sure.

John suggested that we go for a PAR, and then go for a conduit. Kamal suggested that we can't really decide yet whether to go for a PAR or as a conduit. Dave asked what our goal is for the next several months. Dave suggested that our goal is therefore a set of Language Change specifications that can either be collected as a PAR or fed into the appropriate organization.

Greg suggested that we revisit the Design Objectives as soon as possible. Dave then delegated Greg to chair the subcommittee to review the Design Objectives and report at the next meeting.

Dave also asked Wolfgang and Peter to co-chair a subcommittee to prepare a set of LCS suggestions. They will at least attempt to scope the effort at the next meeting.

Greg suggested that some sort of implementation impact should be included in the LCS. John disagreed; he believes that we should decide on the language feature before we decide how expensive it is to implement. Greg said that a cost benefit analysis is necessary, and that therefore some indication of the cost should be in the LCS. John said we can estimate the cost on a construct, not on a suggested implementation. Greg agreed with that.

John suggested we do "function point count". Dave asked what this is, and John responded that this is a construct to determine what the effort is to implement something by estimating the lines of code from the requirements. You look at the functionality in terms of inputs, outputs, data stores, and others to see if it works. Peter said he is not aware of anyone using this technique in estimating compilers or engineering systems. Peter suggested that the person to do this is someone who can do it for compilers. If we can find someone who knows how to do it, we will do it.

Dave has an action item to distribute the notes.

The preprocessor versus new compiler question was brought up again. Greg said that we don't want to constrain ourselves to it, but it might be one possibility.

Next meetings: there will be a meeting at the DAC. Someone asked about meetings in conjunction with FDL or with VIUF. FDL seems to be very crowded; however, we might be able to work something. Our next meeting will be determined at the DAC meeting.

Greg asked for members of the subcommittee. They are: Phil Wilsey, Wolfram, John, and Kamal.

Greg asked about patents or licensing issues. Wolfgang said he knew of none associated with the language (there may be some associated with the tools). Greg suggested that Wolfgang or Peter contact Ron or someone at the University of Evry.