ALF meeting, March 29, 1999 at Fort Collins Attendees: Mike Andrews Mentor Graphics Tim Ehrler VLSI Technology Ted Elkind Cadence Tim Jennings Motorola Darshan Rauniyar Mentor Graphics Wolfgang Roethig NEC Andres Teene LSI Logic Jacob Xiao LSI Logic Jay Abraham SI2 (by phone) Jose DeCastro LSI Logic (by phone) Sergei Sokolov Sente (by phone) Next meeting: May 4 (ALF), May 5, 6 (OLA) in Austin (SI2) Physical library update ----------------------- Representatives in different groups: PAL - Tim Jennings PDEF - Steve Mecham and others A.I. Give an educational presentation about PDEF -- Steve Mecham Status: Steve Mecham presented PDEF in the OLA meeting on March 30. LEF-API meeting: Tue April 27, Wed 28 at Cadence, San Jose Representatives from different work groups (ALF, OLA, OPEF, PDEF) will be invited. EDA vendor update ----------------- Sergei: ALF will be native library format for Sente WW, especially for COT customers. Plan for OLA support in future release. Doc. about WW ALF support available under NDA. Mike: ALF library generation tool for COT planned by Mentor Graphics. Ted: Cadence studies use of ALF for simulation model support through OLA. Roadmap is published, no product plans yet. Jay: ASIC vendor has redistribution rights for OLA compiler from SI2, reduced cost license also available. Development of OLA compilers by independent 3rd party would be a good thing for the market, like multiple Verilog simulators etc. Some startup companies are looking at this. Mike: Issue with platform for OLA compiler, especially for support of early Solaris versions. Solaris 2.5 and upwards should be supported. ASIC vendor update ------------------ Wolfgang: NEC has new version of ALF library with power data available under NDA. Tim Ehrler: VLSI library and data generation flow for all downstream libraries from ALF source are under development. Tim Jennings: No update from Motorola due to reorg. Jose DeCastro: LSI Logic has ALF library available. It contains timing, function, power. VSIA demo --------- Andres: main purpose of VSIA demo is to demonstrate the flow, to integrate IP models from different sources, to identify issues. Mike: OLAworx may not be ready to generate function graph by DAC demo. We may need to generate some function graphs by hand and use RESTRICT_CLASS for tools which need the function graph. Wolfgang: If it is easy to write the function graph by hand, why is it difficult for OLAworx? Jay: OLAworx phase 2 spec will be ready in 2-3 weeks. The answer may be there. Wolfgang: Goal should be at least a prototype function graph generation capability by DAC timeframe. A.I. Test OLAworx for function parsing -- Mike Mike: still need downloadable wireload models A.I. recycle the wireload models from last year's demo -- Jay Abraham Mike: need single port synchronous memory with separate data I/O 256x8 and a ROM 4Kx8 A.I. find some memories -- Jacob for LSI Logic, Wolfgang for NEC OVI approval status ------------------- Wolfgang: ALF 1.0.11 is to be approved as ALF 1.1 at OVI board meeting on Thursday April 8. OVI board will also discuss when and how ALF should be transfered to IEEE. Tim requests the following last-minute corrections: - Some examples in chapter 4 contain "SEL" instead of "S" for ALF_MUX intatiations. - BNF must allow arithmetic_model_container to contain all_purpose_item | arithmetic_model A.I. send those items to Yatin for correction -- Tim Ehrler misc. discussion ---------------- Support for function graph. Issue, need to push OLAWorx to implement this item as early as possible. OLA support for multiple compiled libraries - yes, but issue with name collisions possible. A.I. ask John Beatty in OLA meeting Physical library and process parameter discussion ------------------------------------------------- Do we want physical library standardization? - yes Does standardization of the source matter? - yes for shortterm Jose: coherent semantics are most important. This can be achieved either by single source language (candidate ALF) or by semantic mapping between different languages, i.e. mapping of data models. This does not mean translation between different sources. ALF group is very open to collaboration. Does standardization of the API matter? - yes for longterm Coherent data representation is the main issue, the issue of API versus file is secondary. In principle, API has two benefits over ASCII file: - API reduces storage requirement for large file. This applies to design files (DEF) rather than to library file (LEF). - API hides proprietary data. This does not apply for physical data, as opposed to OLA. The API for physical data will give the same data as the source. There is no proprietary modeling in-between. SIPPS discussion ---------------- Do we want process parameter (SIPP) standardization? - yes Extraction tools need a small number of process parameters in order to build a large extraction rule file. The contents of the rule file can be considered part of the extraction tool itself. Jacob: There are different fundamental approaches for extraction rules, therefore standardization of the rule file or of an API replacing the rule file is very difficult if not impossible. On the other hand, standardization of the set of process parameters is relatively easy. Simulation discussion --------------------- Wolfgang explained chapter 3.12 and Verilog example Semantics for timing vectors are not explained here, but in chapter 4.15, and they are implemented in OLAWorx. The implementation is somehow ahead of the specification, since ALF adopted well-known existing semantics for timing analysis without explicitely restating them. A.I. Draft a document for timing semantics in next future ALF version for formal reasons -- Ted Elkind Parasitics discussion --------------------- Wolfgang presented a proposal for parasitic description in ALF. Feedback: Proposed features seem too advanced for short-term considerations. Short-term requirements can be fulfilled with SPEF and/or SPICE. Recommendation: look again at the proposal in 6 months. Timing for cores ---------------- Wolfgang revisited the proposal from last meeting and illustrated the description of cycle-dependent timing arcs in already existing ALF language on whiteboard. Recommendation: Gather a complete set of modeling requirements before considering extensions to ALF. A.I. Update the core timing modeling proposal -- Wolfgang