SV-EC Ballot Resolution Committee Meeting Wednesday May 4th 2005, 11:00am - 1:00pm PST [Minutes distributed for approval at next (sv-ec) committee meeting] (0111122220) Day (4125815894) (0000000000) Month (4444444445) (0000000000) Year (5555555555) --------- Attendees ---------- (-AAAAAAAAA) Arturo Salz (AAAAAAAA-A) Brad Pierce (AA--AAAAAA) Cliff Cummings (AAAAAAAAAA) Dave Rich (AAAAAAAAAA) Francoise Martinolle (A-----AAAA) Karen Pieper (AAAAAAAA-A) Mehdi Mohtashemi (AAAAAAAAAA) Neil Korpusik (AAAAAAAAAA) Ray Ryan (AAAAAAA-A-) Steven Sharp (AA-AAA---A) Surrendra Dudani (-AAAAAAAA-) Gordon Vreugdenhil (-----A-A--) Dennis Brophy (-------AAA) Phil Moorby (---------A) Don Mills (------A---) Tom Fitzpatrick ^ ^ ^---- Non-Voting meeting (4/29/2005) ** Minutes taken by Neil Korpusik and Mehdi Mohtashemi Agenda: 1. IEEE patent policy ref: http://standards.ieee.org/board/pat/pat-slideset.ppt Move: Assume it was read - Cliff Second: Brad Abstain: Opposed: 2. Review Meeting minutes, April 25th 2005 and April 28th, 2005, and April 29th, 2005 http://www.eda.org/sv-ec/Minutes.html http://www.eda.org/sv-ec/Minutes/SV-EC_BallotRes_Meeting_April_25_2005_Minutes.txt http://www.eda.org/sv-ec/Minutes/SV-EC_BallotRes_Meeting_April_28_2005_Minutes.txt http://www.eda.org/sv-ec/Minutes/SV-EC_BallotRes_Meeting_April_29_2005_Minutes.txt Move: Brad Second: Dave Abstain: Cliff (because too much) Opposed: approved * P1800 meetings: Karen: deadline, p1800 meeting tomorrow, several to roll out, end of the month. keep up with it as they move over. last minute end of next week. SVDB items for review and closure The mantis items have been checked ACKNOWLEDGED. Karen has requested the following: Please have someone in your committee review each issue and either close the issue or 2) add a bug note indicating the trouble with the editing, reassign the issue to the SV-LRM and send a note to Stu and I indicating that this was done. AI: Mehdi, send out note to all members of sv-ec committee to review each Mantis items which has been approved/closed. (firt pass to those who have proposed the mantis item) 3. Continue on open issues: a) issue 233 (Mantis 317) bit-streams Arturo: needed to put the clause about the cycle, which we discussed, the result will be undefined and an error would be issued. Dave: need to change will to shall, is it not an object an element of array. If dynamic array, create an object to fill that array. Arturo: it does not create the object you have to fill the array with the object. Dave: it will fill up the dynamic array with objects. Arturo: it does not know how to call the constructor. Ray: if dynamic array of ints, elements are objects, Arturo: change it to class object. Dave - suggested the following friendly amendments will-->shall object-->class object Motion: with Dave's friendly amendments, accept 233 (mantis 317) Move: Brad to accept 233 (Mantis 317) with friendly ammendment Second: Arturo abstain: opposed: AI: Arturo to place the modified (with friendly ammendments) on svdb. b) issue 266 (Mantis 695) NOTEs Cliff: Arturo offered some friendly wording amendment, better wording in some places. Issue with 16.13, deleting the note, and place it after the example. Cliff will make the changes that John Havilcek requested (minor change to turn one sentence into two). Move: Cliff, to supercede what was approved in svbc Second: Arturo abstain: opposed: AI: Cliff send the update to the reflector, Brad place it on Mantis Karen: Each committee will approve its part of issue 266, then Karen will bring the whole thing to the P1800. c) issue 7 (Mantis 344) Dave: added the syntax of anonymous prog. into the syntax box. 17.3, anonymous program into multiple program. Everything in that paragraph, it was more usage guidelines. Decided to add new item, -- it should be 17.6. instead 17.3. Calling attention, even though program space is invisible with anonymous program, it shares the same scope the same as the package. Arturo: Normative description that was missing. Dave: not meant to be a scope, it is visible, its existence is there, Arturo: creating a syntactical scope not heirarchical scope. Normal programs a name are real scope. In 17.1, informative section. I would like to keep that as scope. Brad: you can declare a program outside of a module, Arturo: yes Francoise: we should not call it scope, Arturo: what is a module, whatever the module is, is the program, but not the anonymous program. Cliff: sufficient information prodcuing errors for users, Arturo: couple of fridndly ammendment, remove or packages, in before the comma, identifiers declared anonymous program inside a package. $unit disambiguiting element, Francoise: change that you wanted to make, Arturo: that is the note that is related to package. Franciose: last paragraph refers to all. AI: Dave Update the proposal and place in SVDB. put scope back into 17.1 renumber to 17.6 (not 17.3) remove or packages: packages or compilation unit scopes --> compilation units in before the comma: "declared in" --> declared in the Note - change will to shall d) issue 8 (Mantis 370) random stability Ray: main item, threads or objects. all get default seeds, would be desirable to get different seeds. each module/program/package have initiatlization RNG, seeded with default seed, implementation dependenet. thread stability, static thread get the next. Arturo: it defines parent thread would be when there is none. Ray: it defines parent random number generators when there is none defined, all default seed. Francoise: what about dynamic thread. Ray: it is seeded with the next number from the parent thread. Move: Ray Second: Arturo abstain: opposed: approved e) issue 188 (Mantis 595) Brad: legal syntax for those expression, should be like port connect, a signal can be both input and output Arturo: the bnf is ok, this covers it well Francoise: can not have clocking declaration be assigned another clocking declration. Ray: Is it legal in port connection. Arturo: it should be legal, it may not be useful thing, Francoise: do we want to allow it. does expression covers it. Arturo: it should cover it, we should not disallow it. Move: Brad Second: Arturo Abstain: opposed: approved. f) issue 199, Mantis 607 Accessing values of clocking variables Arturo: reading an output only signal is not allowed, assigning to input. Charles said, it was implied but not explicity said. Doug: I did find that in 16.3, inputs with explicit, clocking block outputs, CHANGE THE FOLLOWING: next to last paragraph in 16.3 in paragraph, 16.3, are sampled --> shall be sampled, in later sentence: likewise, are driven --> shall be driven add a proposal to end of 16.2. Move: Arturo to accept these addtions/modifications Second: Brad Abstain: opposed: AI: Doug - add a bug note for his friendly amendment. AI: Brad - update mantis with a formal proposal. g) issue 30/31/32 scheudling semantics (Mantis 549, 551, 553) and the newly added Mantis 707/708/709/710/711/712/713 Francoise: We are in disagreement with the changes that have been proposed had good discussions with experts on program block, they think that we are making a mistake by adding time queues. they think we should change the prgoram block, it is risky to add such big change now, we would like to defer the discussion of the issues to the next version of the standards, it is not ready at this time, i would like to have experts such as Jay to be part of discussion. the way the algorithm is defined is wrong. we should not have an inactive region to handle #delay. Arturo: adding regions was not prohibited, such as vhdl, you need to put it in somewhere. The LRM does not forbid any time queue by any mandate. When we discuss with Jay, tacit approval to add time-queue, Francoise: we should not be adding more regions. we do not have our experts, at this time, can we withdraw the proposal. we can not come back to the solution that come back to it. the change that you are proposing too much, we would like to have it fixed, we want to have more time than two or three days. We are not in agreements, there is a proposal that we have been reviewing is not ready. Mehdi: This Was brought up by balloting issues, Do not want to defer to next version of LRM since this would leave users in limbo. Phil: if we do not do this now, the implementations will go their own way, the rules will be set so different, we may not be able to get out of it. Doug: it is difficult if not done anything now. Dave: to be fair, this was done as well with 240 Neil: every testbench has multithreaded, it uses pli with the design, you get a natural barrier. we do not have that clean separation, yet, constantly going back and forth, we want to see atomic program update block. we have a different model, than Vera model, users have many threads in the testbench not just one thread, if not resolved it limits. Francoise: if they do not change Karen: IEEE will allow changes from LRM to another version, we minimize it Cliff: anything that we do here could possibly result in a negative vote. Arturo: Jay was in agreement on vendors adding extra regions from the beginning. without an extra region, #0 will behave differently Francoise: #0 inactive region was a mistake. Francoise: withdrawing issues 30,31, 32 in from the spreadsheet. 32 553 - sharp - program block semantics 31 551 - sharp - program block interaction with time queues 30 549 - sharp - Section 15 "Scheduling Semantics" *** Francoise request was assumed to have taken place. *** Based on the feedback from Karen we decided to move ahead with the mantis items that had proposals even though the original Positive/High ballot issues (30/31/32) were requested to be withdrawn. Mantis 707: Arturo: pre-active region, never updated reference, outside of iterative regions, the pre-active region is now part of reference algorithm. Move: Arturo Second: Cliff abstain: opposed: Mantis 708: feedback arrows on pre-NBA, post-NBA, post-Observed Arturo: we ommitted, the arrows from each of PLI regions, Ray: the algorithm always MOve: Arturo Second: Cliff abstain: opposed; AI: Brad create the actual .html/pdf proposal out of the bugnotes Mantis 709: preponed region - no writes allowed Arturo: specifically for pli callbacks, changed to prepone region provides it is illegal to add variables in preponed region Move: Arturo Second: cliff abstain: opposed Mantis 710: no pli callbacks in observed region Arturo: pli callbacks are not allowed in observed Move: ARturo Second: cliff abstain: opposed Mantis 711: added extra regions in text for last para 15.3 Arturo: pospone region provides pli callback points, nba observed reactive Move: Arturo Second: cliff abstain: opposed Mantis 712: add a pre-postponed pli region Arturo: cbatendtime placed, agreed to new pli region ,pre-postponed region the last writable regions Move: Arturo Second: cliff abstain: opposed Mantis 713: One word change in last para of page 188 - #0 delay requires --> #0 delay control Arturo: explicit #0, delay control instead control, this was proposed by Steve sharp Move: Arturo Second: cliff abstain: Francoise (not the simulation expert here) opposed Mantis 722 - section 15 update - The latest mail from Arturo, May 4th 2005 to the reflectro Arturo: very much close to what was Gordon proposal 2., It also provided for separation between program and design, what Neil is asking for. two changes, addition of re-enactive region, and looping around. - the proposal provides for separation and atomic updates - 1. re-inactive region 2. looping around reactive, re-inactive Dave: re-inactive would seem strange. pending re-active, inactive to pending active Cliff: it would be ok Neil: I would rather stick with the names Arturo proposed Francoise: why do we need a re-inactive region? Arturo: if you do not have re-enactive region, in program #0 block it would go into the existing inactive region. This would cause an inter-mingling of the design updates and the testbench updates. With the proposal we can move some testbench code from the design to the program block it will still work as expected. Mehdi: this is a crucial element that we would like to send to p1800, and champions to review. Francoise: if we are executing in re-active regions, all previous regions have been exhausted, Arturo: not true, you could be in NBA, region, you could have #0 in design domain, idea for re-inactive is to have #0 in program, similar to as #0 in design. Adding another region is how you suspend to the next delta-cycle Francoise: there is already ordering, Arturo: regions have oreders, but the queues within a region do not have any ordering, you can not talk about putting it at the end of queue or beginning Francoise: why not move it to the next cycle. why do we have to treat it differently with #0 not the same as #delay. Arturo: the regions what gives you ordering, it is not strictly needed if you want #0 in program would behave in a different way. Francoise: I do believe it should. I am opposing to adding another region. Arturo: if we want to go about anmolious behaviour for #0 we do not really need to add it. Franocise: objects to adding the re-inactive region. I am not sure this is the right way to fix this, also not sure that we need the extra loop. Ray: there are two changes, the second is the loop, Francoise: I do not think we need that loop, Arturo: the proposal is formalizing with what you just said, The proposal was written to achieve consensus with minimal changes. Francoise: i think we need to have something like currently executing region Arturo: this would have been a better way to do it from the beginning, it would have broken the backward compatiblity. Phil: the implementation for verilog-xl was different than the standard, we need to get the various mechanisms right with minimal change, AI: Brad create the mantis 722 Mantis for this and place the proposal in. Move: Arturo to accept propoasl for mantis 722 Second: Neil Abstain: Cliff 1. my gut feeling is that the proposal is good 2. a semi-complex late-stage change 3. may diminish consensus (e.g. Cadence may vote no) Opposed: Dave - not enough time to review internally at Mentor Francoise - against re-inactive region, and also wants to get rid of the inactive region as well. Passed with 2 no and one abstain Vote summary: ------------ 6 YES votes Neil, Phil, Arturo, Karen, Brad, Ray 1 abstain Cliff 2 NO Dave, Francoise Mastis 714 - proposal wasn't in the proper form Mantis 715 - proposal wasn't in the proper form 4) Next meeting Next Meeting Leave Monday's normal time slot open. Monday May 9th, 2005. 11:00- 1:00 pm ============================================================================== Action items: (5/4/2005): 1) AI(issue 233-Mantis 317) Arturo to place the modified (with friendly ammendments) on svdb. 2) AI(issue 266, Mantis 695) Cliff send the update to the reflector, Brad place it on Mantis 3) AI(issue 7, Mantis 344) Dave Update the proposal and place in SVDB. put scope back into 17.1 renumber to 17.6 (not 17.3) remove or packages: packages or compilation unit scopes --> compilation units in before the comma: "declared in" --> declared in the Note - change will to shall 4) AI(issue 199- Mantis 607) Doug - add a bug note for his friendly amendment. Brad - update mantis with a formal proposal. 5) AI(Mantis 708) Brad create the actual .html/pdf proposal out of the bugnotes 6) AI(Mantis 722) Brad create the mantis 722 Mantis for this and place the proposal in. 7) AI: REVIEW ALL APPROVED/CLOSED MANTIS items for editorial correction Mehdi, send out note to all members of sv-ec committee to review each Mantis items which has been approved/closed. (firt pass to those who have proposed the mantis item) ==============================================================================