• Previous message: John Griessen: "installing magic-7.2.30"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Dear Charles,

   I found the files you were referring to; I just had to find the root
cell of your email, which due to the email problems I'm having, all came
to me in reverse order.  I've downloaded the CIF and GDS files, and
will get back to you on the errors.

   I'm repeating the previous email again, because last time I meant
to CC to magic-dev, and forgot to.
						---Tim

---------------------------------------------------------------------
Dear Charles,

   Sorry I couldn't respond earlier than today;  APL, maybe for war/
security reasons, seems to be holding all my email and forwarding to me
in batches.  Or maybe its just a network setup glitch.  Who knows?
Anyway, I'm concerned especially about the segmentation fault on the
CIF read.  Magic-7.2 is *supposed* to read CIF and GDS files independently
of any scalefactors present in the technology file.  The purpose of this
is to ensure that (except for certain problems related to boolean functions,
shrinks, and grows in the cifinput section of the tech file) CIF and GDS
reads produce exactly what was in the file, without screwing up the
original layout by forcing it to fit to a coarser lambda grid.  Magic
will subdivide its internal grid, switching from its usual 1 lambda = 1
internal unit setup to a 1 lambda = n internal units.
   I recall one problem that was mentioned to me earlier which I have
not yet had time to look into, and it may be relevant here.  The problem
is that for very, very fine scales in combination with large designs,
the positions overflow the 32-bit integer boundary for certain scale        
manipulations in the CIF-to-magic plane conversion.  Havoc naturally
ensues.  I can confirm this pretty quickly, and maybe even send you a
fix sometime today.

   As for not seeing the top-level cell in the GDS read:  Magic version
7.2 compiled under Tcl, and started with the wrapper (-w) option, and
with the Tcl package "BLT" installed on the system (whew, that's a lot
of prerequisites. . .) has the "Cell hierarchy view" window which is
similar to L-Edit and provides a complete list of cells in a tree
structure which lets you find the root cell of the GDS file very quickly.
Equivalently, there is the command "cellname top" which replaces the
command "listtop" that Philippe wrote which I incorporated into version
7.1.  The commands "cellname" and "instance" provide a convenient way to
search the hierarchical structure of a layout.
---------------------------------------------------------------------

• Next message: cfk: "magic-7.2.31"
• Previous message: John Griessen: "installing magic-7.2.30"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]