Closed stefanottili closed 2 months ago
The input data still has the macros marked as RING. Could you provide the test case with the exact configuration that lead you to the crash?
gunzip -c RISC2/lef/risc2.lef.gz | grep -in BLOCK
15531: CLASS BLOCK ;
19167: CLASS BLOCK ;
21165: CLASS BLOCK ;
I also commented out the VIARULES for TURN via that only have one layer of metal.
387 #VIARULE TURNmetal1 GENERATE
388 # LAYER metal1 ;
389 # DIRECTION vertical ;
390 #
391 # LAYER metal1 ;
392 # DIRECTION horizontal ;
393 #END TURNmetal1
...
432 # DIRECTION horizontal ;
433 #END TURNM6
'make' will coredump immediately.
run:
openroad -gui -threads max -log log risc2.or
Describe the bug
rtl_macro_placer coredumps after only reading lef/def.
see https://github.com/The-OpenROAD-Project/OpenROAD/issues/5284 for the input data.
I did replace CLASS RING with CLASS BLOCK to get macro_placement to work, I'm assuming rtl_macro_placer will need this too.
Expected Behavior
no coredump
Environment
To Reproduce
openroad -gui -threads max risc2.or
Relevant log output
Screenshots
No response
Additional Context
No response