WIP: Filler Script

[dnltz]

Work in progress!

This script should automatically generate the correct amount of filler cells into a design.

Please ignore all .lym files as they're get converted into the script later.

[dnltz]

@klayoutmatthias can you please help me with this klayout extension to add metal fill? I'm trying to convert .lym files into native Python since it's in my impression easier to check the density after metal fill and re-fill under some conditions.

My problem is klayout just hangs forever with the TilingProcessor. See the output from this script. Looks like it hangs inside self.top_cell.fill_region(...). Any idea how to fix that?

klayout -n sg13g2 -zz -r /home/daniel/work/aesc/i2c-gpio-expander/pdks/IHP-Open-PDK/ihp-sg13g2/libs.tech/klayout//tech/scripts/filler.py \
	-rd output_file=/home/daniel/work/aesc/i2c-gpio-expander/tools/OpenROAD-flow-scripts/flow/results/ihp-sg13g2/SG13G2Top/base/6_final.gds \
	/home/daniel/work/aesc/i2c-gpio-expander/tools/OpenROAD-flow-scripts/flow/results/ihp-sg13g2/SG13G2Top/base/6_final.gds
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
Warning: Box.destroy: already destroyed!
hi
ho
(-275000,-275000;525000,525000)
(56203,39403;56203,43603;52003,43603;52003,47803;47803,47803;47803,49000;525000,49000;525000,39403);(43603,52003;43603,56203;39403,56203;39403,62000;45000,62000;45000,52003);(320000,65000;320000,143000;192000,143000;192000,156000;176000,156000;176000,169000;160000,169000;160000,182000;152000,182000;152000,195000;144000,195000;144000,204254;204257,143997;322997,143997;322997,66997;399003,66997;399003,143997;404997,143997;404997,66997;481003,66997;481003,143997;486997,143997;486997,66997;525000,66997;525000,65000);(216943,175503;175503,216943;175503,219696;219696,175503);(236388,203003;203003,236388;203003,245971;245971,203003);(258663,234503;234503,258663;234503,261416;261416,234503);(288718,277003;277003,288718;277003,295996;295996,277003);(305688,301003;301003,305688;301003,306136;306136,301003);(423693,448623;423693,452000;525000,452000;525000,448623);(64000,312000;64000,525000;66997,525000;66997,486997;143997,486997;143997,481003;66997,481003;66997,404997;143997,404997;143997,399003;66997,399003;66997,322997;143997,322997;143997,312000)...
134 (0,0;5000,10000) None
^Cmake: *** [Makefile:69: sg13g2-filler] Interrupt

[klayoutmatthias]

Hi @dnltz,

I don't see some obvious issue in the code, but I have not used this combination myself yet.
The tiling processor basically can run on multiple threads, but Python is single-threaded. This involves a mutex, so maybe we have a deadlock problem here. Is there is difference if you use a single CPU?

Essentially, you should be able to put the filler code inside the expressions (using variables for the arguments). This would allow multi-threaded execution of the filler step.

Can you point me to some example file I can use to reproduce the issue?

Thanks,

Matthias

[sergeiandreyev]

@dnltz, should we close this PR? as the data is already pushed to the repo

[dnltz]

@sergeiandreyev please keep it open. The idea is to rework the filler scripts and move away from .lym files.