Liberty/ALF harmonization meeting, Oct. 20, 2003 at Synopsys, Mountain View Participants ------------ Greg Hackney, Mentor John Williams, Markanix wolfgang Roethig, NEC Rahul Deokar, Cadence Rajesh Kumar, Synopsys Rohit Sharma, Circuit Semantics Alex Zamfirescu, ASC Andrea Cosmin, Artisan Kevin Kranen, Synopsys Subramanian Venkateswaran, Sun Robert Jones, Magma Maddhu Karunaratne, V-cube Martin Folten, HP (phone) V. Chandramouli, HP (phone) Sam Kwan, Synopsys (phone) Explanation of purpose, overview of work done so far ---------------------------------------------------- Presented by Wolfgang using the refered document lib2alf.2003_10_16.pdf Expectations from participants ------------------------------ The following opinions/expectations were expressed: - Create ontology (dictionary), identify trivial and non-trivial constructs for translation (Alex) - More user acceptance of a standard library, less fragmentation (Rahul, Robert) - Better characterization efficiency, less redundancy (Andrea) - Consistent interpretation of data by tools (Rohit) - Make liberty as stable as ALF, less frequent version changes (Maddu) - Define semantics of new modeling items not yet covered by liberty (Martin) - Focus on library contents rather than format (Kevin) Overview of work to do ---------------------- The group agreed that a deliverable should be a requirement specification for library contents and characterization. Each item should be explained. Both liberty and ALF descriptions should be provided for each item. Attention needs to be paid to finding the right correspondence, which is not always evident. For example "area" in liberty corresponds to "size" in ALF, not to "area" in ALF. The explanation should highlight semantic differences, wherever they appear. For example, "noise" is normalized in ALF, but de-normalized in liberty. For "energy", capacitive charge-discharge power is included in ALF, but excluded, i.e., separately accounted for, in liberty. PR plan ------- Defered to next meeting Technical discussion -------------------- Wolfgang explained the refered document lib2alf.workdoc.2003_10_20 which basically gives a list of work items and a subjective ranking. The purpose of the ranking is mainly to facilitate a schedule wherein the most important items are done first. The items related to the topic "timing" were reviewed in more detail, and the workdoc was updated as follows: * Delay, slewrate, setup, hold Delay, slewrate, setup and hold are the basic timing characterization data. For setup and hold there exist methods of independent as well as interdependent characterization (not supported in liberty). Definition of thresholds for timing measurement should be represented and used consistently in the library. Should include all SDF 3.0 measurements: recovery, removal, pulsewidth, skew, period, retain. Conditional timing arcs. * Data = f(voltage, temperature) Timing data (and to a slightly lesser extend) power, signal integrity and reliability data should be characterized as a function of supply voltage(s) and/or environmental temperature. This leads typically to 4 or more characterization dimensions (input slew and output load cap being the two basic characterization dimensions). Another "characterization dimension", albeit without interpolation or curve fitting, is the process corner. Selection of significant characterization points and appropriate fitting into polynomials or accurate interpolation are important aspects in this context. Other parameters: related to vt, back bias ... On-chip process variation parameters: geometry ... Next steps ---------- The group agreed to work initially together before breaking into subgroups for different modeling topics. Therefore, we will work on the topics sequentially, starting with timing. The next meeeting is scheduled for Wed. Nov 19, 10AM to 12 noon. Action items ------------ Write a draft for a DoD (Design Objective Document) by next meeeting Owner: Wolfgang Roethig Expected completion: by next meeting Write a proposal for ontology, suitable as a chapter in DoD Owner: Alex Zamfirescu Expected completion: TBD by Alex Give a pointer to the relevant liberty information for this project Owner: Kevin Kranen Expected completion: per email, earliest convenience Give pointer to ISO xpress tutorial on ontology/dictionary Owner: Alex Zamfirescu Expected completion: per email, earliest convenience Review the workdoc and give feedback, especially on ranking. Point out missing or misrepresented topics, if any Owner: all Expected completion time: per email before next meeting Find out if Magma can host next meeting, if yes, prepare logistics Owner: Robert Jones Expected completion time: 2 weeks before next meeting