Segmentation fault in openroad

tgingold commented 4 years ago

I have just updated to rc3, but I cannot build any design as I get a segmentation fault in openroad (for example on spm).

Maybe I am alone with this issue, but rc2 was working correctly.

Do you have any tips to investigate such as crash ? In particular, any easy way to get all the environment variables so that a binary can be executed under a debugger ?

agorararmard commented 4 years ago

@tgingold Can you share any of the errors (segfault and what came before it)? At what part of openroad does it break? Did you rebuild the pdk to use rc3?

ax3ghazy commented 4 years ago

@tgingold: The relevant variables should be found under config.tcl under the run directory if you'd like to source them and run the openroad binary under a debugger.

tgingold commented 4 years ago

That's during cts (for spm):

[INFO]: Running TritonCTS...
[ERROR]: during executing: "openroad -exit /openLANE_flow/scripts/openroad/or_cts.tcl"
[ERROR]: Last 10 lines:
Warning: Liberty cell sky130_fd_sc_hd__xor3_2 pin VPB missing from LEF macro
Warning: Liberty cell sky130_fd_sc_hd__xor3_4 pin VNB missing from LEF macro
Warning: Liberty cell sky130_fd_sc_hd__xor3_4 pin VPB missing from LEF macro
Notice 0: 
Reading DEF file: /openLANE_flow/designs/spm/runs/29-09_18-06/results/placement/spm.placement.def
Notice 0: Design: spm
Notice 0:     Created 36 pins.
Notice 0:     Created 636 components and 3304 component-terminals.
Notice 0:     Created 405 nets and 1120 connections.
Notice 0: Finished DEF file: /openLANE_flow/designs/spm/runs/29-09_18-06/results/placement/spm.placement.def

[ERROR]: Please check openroad  log file
[ERROR]: Dumping to /openLANE_flow/designs/spm/runs/29-09_18-06/error.log

    while executing
"try_catch openroad -exit $::env(SCRIPTS_DIR)/openroad/or_cts.tcl"
    (procedure "run_cts" line 13)
    invoked from within
"run_cts"
    (procedure "run_non_interactive_mode" line 14)
    invoked from within
"run_non_interactive_mode {*}$argv"
    invoked from within
"if { [info exists flags_map(-interactive)] || [info exists flags_map(-it)] } {
    puts_info "Running interactively"
    if { [info exists arg_values(..."
    (file "./flow.tcl" line 160)

And yes, I have updated the pdk too.

agorararmard commented 4 years ago

Could you share the full run directory?

tgingold commented 4 years ago

Here is the zip.

spm-bug.zip

agorararmard commented 4 years ago

I can see nwell and pwell layers in the LEF file, which indicates that something went wrong with the pdk generation. So, can you check the commit hashes for open_pdks and skywater-pdk as well as the magic version you have installed locally?

tgingold commented 4 years ago

$ magic --version
8.3.61

[skywater-pdk]$ git describe
v0.0.0-187-g5cd70ed

[open_pdks]$ git describe --long
1.0.40-0-g48db3e1

nwell and pwell are not present in libraries/sky130_fd_sc_hd/latest/tech/sky130_fd_sc_hd.tlef. Is it expected ?

[latest]$ git describe
v0.0.2-32-g300eb6f

tgingold commented 4 years ago

I will retry to generate the pdk.

tgingold commented 4 years ago

No success.

Can you briefly explain how the nwell and pwell are expected to be handled by open_pdk ? Thanks!

agorararmard commented 4 years ago

@tgingold, I'm not sure if this is caused by open_pdks or the fact that the commit of the skywater-pdk that you have doesn't add up with the version of the sky130_fd_sc_hd/latest. if you have v0.0.0-187-g5cd70ed, then the commit hash of the sky130_fd_sc_hd should be 74c2ab

Did you try re-cloning the skywater-pdk from scratch?

Also, to answer your question, they are in the cells lef, not the tech lef (they shouldn't exist in both).

tgingold commented 4 years ago

Ah thanks. I wrongly checkout the master of sky130_fd_sc_hd. Now it works.

Thank you again.

The-OpenROAD-Project / OpenLane

Segmentation fault in openroad #45