|
|
Magic Mailing List |
|
From: Jeff Sondeen (sondeen AT ISI DOT EDU) Date: Mon Mar 24 2003 - 00:13:35 EST
Hi, Charles, yes a 1/2 lambda grid would solve the snapping problem for the Tanner library, but then you're on a 1/2 lambda grid which is a bit more tedious. but this won't address the diagonal spacing drc rule problem, eg. if contact spacing is 3 lambda, an array of contacts spaced diagonally apart by (2,3) lambda will meet it (since sqrt(9+4) > 3) but magic will require (3,3) spacing. /jeff cfk writes: > Well, I'm a relative newcomer to this scene, but it seems to me that all of > this is accomplishable in the technology file. I could understand a > different technology file for Magic, Tanner, Mentor or Cadence tools but the > GDS or CIF file itself should not be peculiar to any particular tool if the > technogy file itself is complete. I believe the 1/2 lambda spacing is done > by having the lambda grid be itself one half of the design rules so that the > technology files multiplies all of the DRC checks by a constant (like two). > So, with that said, I toss it out to the key magic hackers themselves to > work out as I believe that GDS or CIF files, like Gerber files for printed > circuit boards should be usuable by any software package that follows the > rules for the GDS/CIF or Gerber file format with the appropriate rosetta > stone. Maybe someone more knowledgeable about technology files them myself > can consider putting one together for the Mosis library. > > Charles > > > ----- Original Message ----- > From: "Jeff Sondeen" <sondeen AT ISI DOT EDU> > To: "cfk" <cfk AT pacbell DOT net> > Cc: "Jeff Sondeen" <sondeen@ISI.EDU>; <magic-dev AT csl DOT cornell.edu> > Sent: Sunday, March 23, 2003 6:53 PM > Subject: Re: mosis cell library > > > > > > if i recall correctly, Tanner laid out many cells taking advantage of > > diagonal spacings (that is, drc was just met considering true > > non-orthogonal distance), but magic can only maintain orthogonal drc > > spacings. in addition, Tanner used 1/2 lambda spacings whenever > > allowed by mosis (magic must round up to nearest lambda), and also > > often used a 1/2 lambda grid, so many vertices as well as cell > > instance locations have to be snapped by magic to the lambda grid. > > > > also, since every contact and via was separate cell instance, my (old) > > magic was consuming tons of memory and taking ages to handle all the > > cell instances. > > > > in general, lots of edits will be required and you probably cannot > > maintain the pitch that Tanner achieved using non-orthogonal drcs as > > well as 1/2 lambda rules. > > > > /jeff > > > > > > cfk writes: > > > A little more progress after I read Tutorial #9 for the fourth time is > an > > > innocent sentence on page 2 that says "Stream files do not identify a > > > top-level cell, so you wont see anything on the screen after youve used > the > > > calma read command. So, I saved all the cells with writeall and now I > have a > > > myriad of .mag files in this directory like Inv, Inv2, Nand2, Nor2 and > the > > > like. When I startup magic again on the invertor (magic -w -T > > > SCN5M_DEEP.12.TSMC) with say Inv.mag and run "drc find", I get about 34 > drc > > > errors with what looks like some but not all features. So, I can > actually > > > see a portion of the invertor from mosis, just not completely. > > > > > > So, gds reads with some problems, and cif segfaults as of this morning. > > > > > > > Let me try again with a little more detail. I can read the pads under > this > > > > directory with no problem. I saved the technology file under this > > > directory: > > > > SCN5M_DEEP.12.TSMC.tech27 and I can startup magic with "magic -w -T > > > > SCN5M_DEEP.12.TSMC". From magic's prompt, I can go ;cif istyle and > see the > > > > style is "lambda=0.12(p)", so far so good. The problem occurs when I > do a > > > > "cif read mTSMd025DL.cif" in that magic segfaults. I did a little > research > > > > yesterday and I can see it appears unhappy before the segfault with > the > > > > statements "DS 1 6 1000" indicating a scale of 6/1000 (centimicrons, > I > > > > think, but perhaps millimicrons and here might be the rub). After > reading > > > > and printing out errors from each and every "DS x 6 1000" statment, > magic > > > > segfaults. > > > > > > > > When reading the similar gds file with "calma read , magic doesnt > > > segfault, > > > > but does display a myriad of errors of the form "# cols doesnt divide > > > > displacement ref pt". Quitting magic ":quit" indicates there are 58 > cells > > > > which have been modified, but there are no visible cells on the > screen > > > > (perhaps they are 1000 times smaller then I am setup to display, or > > > > something like that). > > > > > > > > It seems to me that I am missing an invocation or a setup under > magic. It > > > > should be possible to read a cif or gds file created by the Tanner > L-Edit > > > > program as both tanner and magic should use similar rules for reading > and > > > > writing cif and gds files. > > > > > > > > And just to make it incredibly frustrating, electric reads and > displays > > > both > > > > the cif and gds libraries just fine as downloaded from the mosis web > site. > > > > > > > > > > > > > Hi, why don't you start with the README's under > > > > > > > > > > ftp://ftp.isi.edu/pub/sondeen/magic/tanner_pads_ported_to_magic/ > > > > > > > > > > altho they may be out-of-date. > > > > > > > > > > /jeff > > > > > > > > > > cfk writes: > > > > > > Gentlemen: > > > > > > I've spent a while studying the mosis cell libraries today > at > > > > > > > > > > > http://www.mosis.org/Technical/Designsupport/std-cell-library-scmos.html > > > and > > > > > > downloaded both the TSMC 0.25 micron and Agilent 0.50 micron > > > libraries. > > > > In > > > > > > attempting to load the gds files with "calma read > mTSMd025DL.gds" or > > > > "calma > > > > > > read mhps05dl.gds" from either with default technology or > scmos-sub, > > > I > > > > get a > > > > > > "#cols doesnt divide displacement ref pt" error on many cells > and my > > > > screen > > > > > > is blank. If I try to read the CIF versions, I get a seg fault > and > > > > Magic > > > > > > exits. > > > > > > > > > > > > I am hoping that there is some combination of technology > > > > machinations > > > > > > and clever button pushing that would actually let me read and > > > > appreciate > > > > > > these cell libraries from MOSIS. Any suggestions would be > greatly > > > > > > appreciated. > > > > > > > > > > > > Charles
|
|
|
|