Minutes of P1364-2005 teleconference 10-May-2005 Attendees: Tom Fitzpatrick Neil Korpusik Cliff Cummings Stu Sutherland Steve Sharp Brad Pierce Mark Hartoog Alec Stanculescu 1. 680 - config Steve - keyword is main concern Cliff - suggesting that configs are "worthless" as they are today. Mark - configs don't work well with -v and -y and then replacing uselibs with configs Cliff - didn't standardize uselibs since configs are there now. Steve - not a one to one mapping. Mark - separate compile model - don't need libmap file for this case. Cliff - would like to replace -v and -y Mark - could add more to libmap file for this - there are a lot of gory details on how -v and -y work. There probably are no two exact implementations. Cliff - current use model - don't need -v -y, libmap file specifies where the files are, config - specifies top level design and libraries. Use model not useful since we still need to compile all of the files separately using command line options. If call a module more than once the configs can be useful. Why build these things if can already use -v and -y. Can do most with -y and -v. Rules are that complex for most users. Mark - subtleties when modules in libraries instantiate other modules. There are some rules about not yet seeing a module name yet if not found in same file - or something like that. Cliff - librescan can be used. Tom - what is the big problem - using libmaps and configs give an unambiguous mapping from modules to source files. - configs exist today. Lets make them consistent. Mark - configs are used today. Tom - would like to focus on making configs unambiguous. Steve - major changes are possible except that we need to contact existing customers. - keywords - config keywords exist in legacy verilog code. People don't like having to specify directives. 1995 code may not compile properly with these new keyword definitions. - allow configs in verilog source files. keywords - have a separate set of dialect keywords - adding new keywords was rejected as too much of a change. Mark - we should make the smallest change possible. The new dialect specification would be the smallest change. Prefers not to change the keywords. Steve - not adding any new keywords, just turning off the existing config keywords. instance design config -- 15% of legacy code use this cell -- 15% of legacy code use this 1364-2001-noconfigs // people prefer this to +configs option 1364-2001 // all v2k keywords All the keywords used by libmap files and configs. Stu - likes this option but wants to be clear about which set of keywords we are talking about. Alec - why need to turn these off? steve - legacy code uses some of these keywords. Stu - more consistent to list the full list for each option. - prefers to have a text description that shows which are removed. Steve - list of keywords cell config endconfig design incdir include instance liblist library use Brad - prefers to not change the ordering of the BNF sections. Syntax box updates. Stu - contents of syntax boxes are done by hand but xrefs are automatic. Cliff - Adam wanted to require the semicolons. AI/Cliff - update proposal #4 to proposal #5. AI/Brad - add it to mantis AI/All - review update #5 Motion: add 1364-2001-noconfigs Keep config_declaration in A.1.1 (from Cliff's proposal #4) library_descriptions --> library_description Note to Stu - search for and make all others of this text. semicolons - added to config_rule_statement clauses. Steve Second: Cliff Abstain: Alec - no voting rights - good solution (turning off keywords) - allowing in source files - rushed - prefers to deal with all of the issues at once. Oppose: Passed Thanks for your efforts, cooperation and congeniality -Tom & Neil