Closed thesourcerer8 closed 1 year ago
"CLASS CORE" is a LEF standard---there are only a handful of keywords that can follow "CLASS". What you're referring to is the "SITE" specification, e.g.: SITE GF018hv5v_mcu_sc7
which matches the SITE
definition in the technology LEF file.
What I'm not so sure about is whether "CORE" in "CLASS CORE" is allowed to be lowercase. As far as I know, all LEF keywords are uppercase.
@RTimothyEdwards the case is technically supposed to be uppercase, but OpenROAD accepts lower case.
@thesourcerer8 +1 to @RTimothyEdwards CLASS CORE only signifies the cell type. the SITE
value is what you're after.
Like so many things, I cannot find anywhere in the LEF documentation that says whether LEF/DEF keywords must be uppercase. It is sort of implied by usage, but as far as I can see, not explicitly stated anywhere (although many tidbits of syntax are buried deeply in obscure corners of the document).
@RTimothyEdwards I think it probably makes sense to at some point go back and fix that, but at the end of the day OpenROAD does support the mixed case. Unless there's a tool issue then I think at this point we should fix this a lower priority. Given the other very real issues reported by @thesourcerer8
@QuantamHD : Also, the lowercase core
comes unchanged from the original GF LEF sources, so it presumably works with all the commercial tools.
I think this issue can be closed. Issues #104 and #106 are the important ones.
Agreed
Expected Behavior
I would expect 9t cells to have a different CLASS than 7t cells, since they have a different size and therefore need to be placed in a different grid. E.g. Sky130 is using "unithd" for the high-density cells (similar to 7t) and "unit" for the low-density cells (similar to 9t) I am not sure, which behaviour is the best way to do it.
Actual Behavior
Both 9t cells and 7t cells are using CLASS core, which might cause confusion of cells from both libraries are used in a design.