Steven, That's good data. I feel the 15% number you've been using may overrepresent the Verilog community as a whole, but on the other hand it's not all that close to zero either. We've been shipping software with these keywords in place for a year and a half and have certainly had many reports of keyword conflict, but neither those reports nor our inventory of customer designs would indicate 15% (this is just feel, but probably more like half of that). So let's say your number is high and mine is low....10% is still enough to worry about so I haven't worried overly much about the 15% number you've been using. What I wasn't sure of, though, was whether 'cell' or 'config' was the main culprit. We've seen both but I hadn't had a reason to track which was the greater issue.I accept your view that 'config' is around 2/3 of the issue, so I concur with your assessment that the lexical hack is too messy to be justified if it only solves a third of the problem. That at least simplifies the decision to all or nothing. The middle ground doesn't seem worth doing. -randy. Steven Sharp wrote: >Randy, > >I will get back to the rest of your email later, but I wanted to >respond to one item now. > > > >>In the spirit of compromise, I'd be willing to >>consider separating the two lists if we could retain the keyword, >>config, in the Verilog source keywords. This would at least allow us to >>implement a messy lexical analyzer mode change based on this keyword. >> >> > >I appreciate the spirit. If this would have resolved the problem, I >might have suggested something similar myself (even though I strongly >dislike context sensitive lexing). Unfortunately, the keyword "config" >is the main cause of keyword conflicts. > >In our suite of 67 real customer designs that we use for testing, "config" >is used as an identifier in 7 designs (or 10.5%). An additional 3 designs >(4.5%) used "cell" as an identifier. None of the other config keywords >appeared . So eliminating all of the config keywords except "config" >wouldn't help much. > >Changing to less-used keywords to replace "config" and "cell" would solve >the main problem, but would not be backward compatible. > >Steven Sharp >sharp@cadence.com > > >Received on Thu Apr 21 18:07:19 2005
This archive was generated by hypermail 2.1.8 : Thu Apr 21 2005 - 18:09:02 PDT