ISAC Minutes 27 July 1995 Analogy Chair: Clive Charlwood (crc@synopsys.com) Minutes: Clive Charlwood (crc@synopsys.com) Attendees: Clive Charlwood Daniel Barclay (2nd Day only) Jacques Rouilliard Bill Paulsen Alex Zamfirescu Chuck Swart We started by thanking Chuck for 1) Migrating the old IRs and 2) Hosting the meeting. His (part time) participation is a great help to the team. Work Items: 1115: Bill will write up the analysis and present at the next VAS (AI) The LRM already states that this is illegal (Page 56, line 196). 1010: (?) Jacques will write up the analysis for consideration at next ISA (AI) ------------------------------- See page 106, line 530 There are 4 options here: 1) Move towards the "Old Ada" definition. IE There is no double meaning for the div operators. 2) You count the conversion and the take the smallest count. If two counts are equal then you error out. 3) Follow the proposed resolution 4) Rework the type system (a la ADA95). To use the concept of "First Subtype". There is also an issue with "for I in -1 to 10" because "-1" is not a literal Chuck states that an IR should/may exist. Anyway, he will create one. newIR: Chuck will create a new IR covering the ""for I in -1 to 10" issue. (AI) ------------------------------- 1081: Bill will update the IR to say attr specs follow 4.3.1.(AI) Section 4.3.1 says that signal s1, s2... is equivalent to signal s1...; signal s2 ... If you want to call the function once you can always assign the result for a constant and then associate the constant value with the object. Therefore we decided to follow 4.3.1. i.e. the function is called once for each attributed object. This is done in the order that the objects are listed. IN the case of all/others the attributes are applied in "textual order". Finally, we noticed that pieces of IR40 did not make it into the LRM. Once example it that attr specs of all/others do not apply to interface objects. No such restriction appears in the LRM. newIR: Bill review IR40, pieces not in vhdl93 will be placed to a new IR (AI) ------------------------------- We then spent the afternoon dealing with analog time. Chuck kept track of this discussion. In discussing this we managed to convince ourselves that postponed processes were not correctly defined in vhdl93. Daniel arrived the next day and set us straight. ------------------------------- 1086: Jacques will insert the info for Serge's paper (AI) After a short discussion on the general anxiety with generalized aliases we decided to punt until we have a better list of the issues. =============================== Day Two: - We started by deciding when we will have the next meeting. [Since then I have been thinking. I don't this we should have it with a conference. It is much more difficult to set up. Instead I would like to have it in October at Viewlogic in Fremont. What do people think? - Clive ------------------------------- 1069: Clive will update the IR to agree with the proposal (AI) After a long (too long) discussion on whether zero can/does have a sign. We decided we didn't care :) - the section number if now 7.2.6 ------------------------------- 1088: Jacques will add update IR for the VASG As stated previously, several of the issues in this IR should be moved to an LCS dealing with design libraries and the visibility outside of DUs. dealing with each issue separatly: 1) This is legal and should continue to be legal. - A of E is (technically) illegal. because there is no use clause to make E visible. This should be cleaned up in the LCS 2&3) These are real problems and form the bulk of the LCS. However, the IR should warn that this behavior is non-portable and should be avoided. 4) This is legal and should remain so. newLCS: Jacques will migrate remaining issues in to a new LCS. (AI) ------------------------------- 1047: Clive will update and then present at the VASG meeting (AI) We agree with the IR. We feel that the "future" section should include discussion on 1) Making the entity class optional for non all/other cases. or 2) Making both type and subtype synonyms for ALL named (sub)types. ------------------------------- 1072: Clive will update and then present at the VASG meeting (AI) We agree with the proposed resolution. A process does indeed get "created" :) ------------------------------- 1074: Daniel will review the IR and LCS and bring it back to ISAC (AI) LCS32 deals with this issue. However, we are not sure it was correctly implemented in 93. The pieces that are missing should be added to this IR. Also, a note should be added that the proposed "fix" limits some recursion. ------------------------------- 1075: Alex update the IR and present it at VASG (AI) After much debate we decided to follow the EXACT wording from ADA. [My notes didn't say who owned this one. So I gave it to Alex because he didn't have any action items yet :)]