Add hard decoder gates (nand, inv, pnan)
Add conditions for routing using LI layer in s8.
Generalize bus layers for decoders.
Move custom cells to own directory.
Fixed via directions, etc.
Add 4x16 hierarchical decoder and test case
Changes to allow decoder height to be a 2x multiple of bitcell height.
Split of control logic tests.
Fixed track spacing in SRAM and channel router
PEP8 cleanup.
Convert to use li layer in pgates.
Fix multifinger devices with li layers.
Simplify wordline driver input routing.
Fix power pin direction option update.
PEP8 cleanup
Changes to simplify metal preferred directions and pitches.
Split of control logic tests.
this was the remainder of applying a diff using "patch". To avoid this
mistake, add the filetypes created by "patch" to the .gitignore.
Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
we don't want to propagate the sense amp's bl/br names out of the
sense_amp_array. Thus the sense_amp_array gets them named as
"bl"/"br" again.
Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
all pin names should be wrapped into a function/property. This ensures
that there is exactly one place to change the name.
Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
we don't want to propagate the write driver bl/br names out of the
write_driver_array. Thus the write_driver_array gets them named as
"bl"/"br" again.
Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
all pin names should be wrapped into a function/property. This ensures
that there is exactly one place to change the name.
Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
both functions share a lot of code and are passing around a lot of data
under similar names (inst1, inst1_start_bit, inst1_bl_name, ...). Thus
we group all these elements in a named tuple to ease passing around
these elements.
All callers of channel_route/connect_bitlines() either pass in the bl/br
names or rely on "br_{}"/"bl_{}" as defaults. These hard coded values
should be determined by the instances. Thus we get the bitline names
based on the instances passed in. The callers only provide a template
string, to take care of the case that bitlines are called "bl_out_{}".
Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
if we rely on the names of the submodules (sense_amp_array,
write_driver_array, etc.) for port_data's pins, we get into trouble on
multiport SRAMs. To avoid this we use explicit names for br/bl depending
on the port number in port_data. Now each submodule does no longer need to
figure out the right name depending on the port number.
Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
before the names of bl/br from the bitcell were assumed. If we want to
allow renaming of bl/br from bitcells, we have to seperate the other
modules from that. Note, that we don't touch every occurence of bl/br,
but only the once necessary that pin renaming of the bitcell works.
Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
each module should be able to state how their bl/br lines are named. Here we
always connect port_data with the bitcell_array, so port_data needs function
that return the names of bl/br.
Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>