gen_fabric_macro quietly fails at magic-write-lef on WSL, but not Alma 8

[mtetrault]

I've run the gen_fabric_macro command using the demo project without any modifications (including pdk). The flow runs to completion on an Alma 8 environment, but it fails on windows WSL (debian).

On WSL, looking closer in step 46 (magic-writelef), it seems to complete and begins step 47, however in fact the lef file is empty. The log file's last entry is "Diagnostic: Writing LEF output for cell eFPGA".

On Alma 8, the log file has 2 extra lines:
"Diagnostic: Scale value is 0.005000"
"[INFO] LEF Write Complete."

I'm wondering if this error in WSL might be caused by insufficient RAM in my WSL setup. Reported ressources for step 46 for WSL are as follows:
"peak_resources": {
"cpu_percent": 109.6,
"memory_rss": "7GiB",
"memory_vms": "9GiB",
"threads": 1
},
"avg_resources": {
"cpu_percent": 97.52441314553985,
"memory_rss": "3GiB",
"memory_vms": "3GiB",
"threads": 0.9990610328638497
}

and for the Alma 8 successful run:
"peak_resources": {
"cpu_percent": 118.2,
"memory_rss": "12GiB",
"memory_vms": "12GiB",
"threads": 1
},
"avg_resources": {
"cpu_percent": 98.38653683319278,
"memory_rss": "6GiB",
"memory_vms": "6GiB",
"threads": 0.9991532599491956
}

Could this be the source of the error?

[KelvinChung2000]

Insufficient RAM is possible. Are you doing the write under the Linux file system or windows mounted drive? As this could make a difference as well.

[mtetrault]

I'm running under the linux file system, a subdirectory of $HOME. I'll see with colleagues if anyone has a 32Gb windows system, but most systems have 16Gb.

[KelvinChung2000]

It feels like a memory problem. One thing to check is the amount of memory given to the WSL VM.

[mtetrault]

I reduced the size of the macro to 4 lines of tiles, and the lefout was successful. So memory is very likely. I got feedback on an available 32 Gb system. I will request access or ask someone to run the flow there.