Subject: Minutes for the meeting on 9/24 Date: Tue, 24 Sep 2002 21:42:58 -0700 From: Tarak Parikh To: Yatin Trivedi , SVCC , tarak@athdl.com Here are the minutes of the SV-CC meeting for 9/24/02 Date : 9/24/02 Time: 9:00 am (PST) Attendance Yatin Trivedi Tarak Parikh John Stickly Swapanjit Mitra Mike Rohleder Stuart Swan Alain Reynaud Kevin Cameron Joao Geada Ghassan Khoory Francoise Martinole Start of the meeting, Yatin introduced 3 new participants to the committee (John, Swapanjit and Tarak). Minutes of the last meeting were approved. One question that came up was answered. Currently the assertions API was focussed to work with simulators. Then Yatin, presented the details of the meeting that was held on Sept. 17 for the full SV committee. More than 30 people attended and a schedule was worked out to generate the LRM, which is as follows Committee Release Draft Final Release Nov/Dec 2002 Mar '03 May/Dac '03 The next full committee meeting will be sometimes in November '02 A final note on the full SV meeting. Since no proposal was received by the committee by the Sept. 15th deadline, we will not be working on the VPI enhancements. This was followed by a presentation by Joao Geada of Synopsys on their donations to the committee. Synopsys has donated various API's to the committee for forming the basis of the C-API's for the System Verilog 3.1. The presentation was supposed to cover the following API Direct-C API Assertion API Coverage API Before the presentation, there was some discussion on the question of what constitues a donation and what can one do with a donation. The conclusion was " A donation acts as a starting point for any future work. It can be modified, rejected, bits can be extracted and used, modified etc." Following questions arose during Joao's presentation - How does this account for C-blend (I am not sure of this term) that was the API available in co-design. Since there are new features in the language compared to what direct-C is presenting, would that not be a better starting point - Again, what does a donation imply - Why not use VPI. Joao's answer was that API's like VPI imply that a lot more data has to be available to the user. This implies a significantly higher burden on the tool and will result in sub-optimum performance. A direct API is directly mapped onto the tools's data structure. - Why not use a class library? - There was significant confusion and discussion about the 2 modes of operation within the direct-C Api. The direct-C api has an abstract mode which has more checks and protection built in (argument checking and such) and a direct mode which can cause corruption of the actual simulator database. Synopsys recommends a usage model where the user try out the abstract mode and then once the code is stable, switch to a direct mode. The confusion involved the mode of switching. Would it require re-write of certain pieces of code, re-linking with different libraries, or setting some mode flag in the header files. Certainly, re-writing pieces of the actual C code would defeat some of the purpose. - Is there a plan to have verilog calls available from C code After, this phase of the presentation, we ran out of time for the other sections and so decided to adjourn for the day. Actions: 1. Ghassan to send out the proposal by email for voting 2. Voting to be accomplished by email, to be finished by Sunday 3. Yatin to send out the guidelines for the voting after gathering such information from Vassilios. Currently, it stands at one vote per person and simple majority required. The vote is for the full donation. 4. Yatin to send presentation at the SV meeting to SV web site. Next meeting : Next Tuesday (10/1/02) Tarak ---------------------------- Tarak Parikh @HDL, Inc tarak@athdl.com (408) 441-1317 x 111 (408) 219-4658