Hi Gianluca, On Wed, Aug 30, 2000 at 10:27:29AM +0200, Gianluca Cornetta wrote: > Hi, > > I had a look to the sxlib and I noticed that the cell pitch changes, however > this is only a minor problem. Yes the pitch is 5 lambdas for sxlib and 6 for sclib. > > My design is full-custom and does not use neither sclib or sxlib cells. > I told you that I implemented a set of CPL differential gates with > precharge/evaluate signal. > > Now, changing the abutement box size and the pitches of the cells from 6n to > 5n is not a major problem; the problem is now GLOP. > > Yesterday I tried to run GLOP on a netlist of mine and I realized that the > program cannot perform any buffer insertion if the cell has more than one > output as in my case. In addition some signals such as the precharge/evaluate > (that acts like a clock) require large driving capabilities but it is very > hard to do these optimizations by hand. Do you know how can I tackle this > problem? Glop doesn't perform good optimizations ... You'd better use LooN (asim.lip6.fr/~ludo/alliance/download/loon/) Glop is not supported any more, and Loon will replace it in the new Alliance release. > > Just some more question about sxlib. I have noticed that a generic signal S > has multiple connectors named S_xx where xx is the pitch from the south side > of the abutment > box; however in the VBE description of the circuit appears only a connector > named S that does not exist in the cell layout. I suppose that this shoul be > the format that I have to use for my own cells as well. Am I right? Yes, all connectors are modelized with special objects called "ref_con" and as you said, they are named S_xx. > > Is there any other constraints for the cell? for example minimum number of > connectors per signal. > > I have noticed that the internal connections of a cell are only in metal-1, > should I avoid metal-2 for internal routing or I am allowed to do that. > As far as i know there is no other constraints for a cell. You can ask Franck Wasjburt (Franck.Wajsburt@lip6.fr) for details. What i know is that in the new Alliance release, ref_con objects won't exist anymore. There will be only vertical metal segments with a particular layer. For example a connector of ALU1 will be modelized with a vertical segment of CALU1 (C is for Connector ;) ) SCR as well should disappeared, but there is no other standard cell router for the moment. Anyway i think the best for you is to keep and follow the sclib topology, because sxlib constraints are not really fixed and it can change at any time. > > Best regards, > > --Gianluca Regards, Ludo. (_) ___ Ludovic JACOMME _ _ ( ) ( ) ( ) ( 6 ) Laboratoire LIP6, Equipe ASIM ( ) (_) ( _ ) Couloir 55-65, 2eme etage, ( )___ ( ) Universite P. et M. Curie (P6) (_____) (_) 4 place Jussieu, 75252 Paris Cedex 05 Tel: (33) 01.44.27.54.15 Fax: (33) 01.44.27.72.80 ICQ: 62526530 mailto: Ludovic.Jacomme@asim.lip6.fr http: //asim.lip6.fr/~ludo