<@U016HU5HK8V> do you mean where cells are prefixe...
# openlane-2
a
@tnt do you mean where cells are prefixed with a random
XX_
two-character sequence? @Mitch Bailey might be able to answer that if @donn isn't otherwise available. Or do you mean something else Sylvain?
Actually I meant @David Lindley (or @David Lindley ) instead of @Mitch Bailey, but either might know :)
k
If this is about the XX_, try setting MAGIC_MACRO_STD_CELL_SOURCE to PDK
t
Yes, I meant where they end up with the prefix.
@Kareem Farid But I want to do it selectively only for some cells.
k
Not possible at the moment, through the flow. I am wondering what is the idea behind selecting some cells
d
@tnt We do this in the flow to prevent name collision when we combine the user project wrapper into the harness.
t
@David Lindley I know. And I want that most of the time.
But if I want to whitelist SRAM cells, I needs them to keep their names ...
d
The renaming happens in
read_macro_gds_blackbox
- sadly it is one of those things magic does automagically and is opaque about it.
k
If you change the section above to ready the gds file instead of setting the property, based on the macro name, it should accomplish what you want
t
TBH I still don't see exactly where / what triggers the rename in the above.
Is that the
LEFview
property ?
m
@tnt
TBH I still don’t see exactly where / what triggers the rename in the above.
Me either. The
LEFview
property combined with the
GDS_*
properties, tell magic to copy the gds data exactly to the output instead of converting from the mag data.
Looking at the
tapeout/logs/ship_truck.log
, the prefixed cells are not listed. Diving (digging?) into the code, I think the prefix is added in the
calmaDumpStructure
routine of
magic/calma/CalmaWrite.c
which adds a hash based on the gds library if the library is different than the original library for a cell. However, AFAICT, the modified cell names do not appear in any of the log files.