Closed TheMichaelGionet closed 1 week ago
@gadfort can SiC produce a standalone test case like 'make issue'?
@TheMichaelGionet @maliberty
You should be able to use:
sc-issue -cfg build/blah/route/0/inputs/blah.pkg.json (replace with correct path) -generate
This should generate standalone testcase.
I will wait for you to package one before looking at this.
I apologize for the wonky way this is uploaded, github wanted the file under 25 MB so I split up the tar file, but it also wanted it in a supported format (which includes zip files) so I zipped each of the 3 parts. To reconstruct it just unzip each of these and cat them together (aa first, then ab, then ac).
sc_issue_decode_sc_module_1_job0_route0_20240617-101924.tar.gz_pt_ac.zip sc_issue_decode_sc_module_1_job0_route0_20240617-101924.tar.gz_pt_aa.zip sc_issue_decode_sc_module_1_job0_route0_20240617-101924.tar.gz_pt_ab.zip
I was able to unpack and I'm running it now. Another option is to put the files in gdrive and put a link in GH
@gadfort the issu produces runs the entire OR flow starting from
Reading ODB: inputs/decode_sc_module_1.odb
is it possible to make an issue for a single failing step?
@maliberty what do you mean? this is for the single failing task (it's not starting over from scratch)
Do you consider fill cells + grt + drt a single task?
that is what is run in that task, yes.
On what iteration do you see the failure? This is a large design so it looks like it will take a while to reproduce
With this particular run it was optimization iteration 1 after it hit the 20% mark. The exact iteration and percent has not been consistent across runs, but hitting a signal 11 has been consistent across runs.
My run is already mostly through iteration 2 using OR commit 08c19394f8bc95baa2afdc90de68c7476d60a381
It ran to completion without error for me so I can't reproduce your issue.
I have a partition on my PC for ubuntu, I'll try compiling openroad for that and let you know how that works. If all works on your end then I'm willing to bet it's a support issue with its dependencies (specifically for Linux Mint, since I did a somewhat hacky thing to get it to work).
This particular issue seemed to resolve itself when moving to ubuntu.
Describe the bug
I am using siliconcompiler to do an asic flow, where OpenROAD is used for the floorplanning, placement, cts, and routing steps. During the optimization parts of the routing steps, there might be a crash with a signal 11 attached to it. When I give a design a larger area then a crash is less likely to occur, but if I give it a smaller area then a crash is more likely.
openroad_crash_msg.txt
Expected Behavior
Not crashing
Environment
To Reproduce
pull the branch used for siliconcompiler (https://github.com/siliconcompiler/siliconcompiler.git). I used commit e54dbf73d3627c65e5e609b989ddeb36e5d3dbb9. Install siliconcompiler using the tutorial on their guide https://docs.siliconcompiler.com/en/stable/user_guide/installation.html. Extract the uploaded tar file into some directory and cd into it. Do "python3 run.py". It should then fail when it gets to the optimization step during routing.
siliconcompiler_decoder.tar.gz
Relevant log output
No response
Screenshots
No response
Additional Context
No response