Flow fails at PDN during macro hardening

[KHermanUBB]

Description

The flow fails at PDN during macro hardening. Design revised by @msaligane and suggested to be reported here as a bug.

Environment

• Operating System Information: Linux CalcMachine 4.15.0-159-generic #167-Ubuntu SMP `
• Docker Version: 20.10.7, build 20.10.7-0ubuntu1~18.04.2
• OpenLane Tag in use: 2021.09.23_03.17.13

Reproduction Material

config.tcl config.zip

Expected behavior

A clear and successful flow

Logs

[WARNING PSM-0039] Unconnected instance _772_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _773_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _774_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _775_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _776_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _777_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _778_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _779_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _780_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _781_ at location (5.400um, 13.600um) layer: 2.
[WARNING PSM-0039] Unconnected instance _782_ at location (5.400um, 13.600um) layer: 2.
[ERROR]: PDN generation failed.
[ERROR]: You may need to adjust your macro placements or PDN  offsets/pitches to power all standard cell rails (or other PDN stripes)  in your design.

[Colin-Holehouse]

The PSM messages are PDNSim detailing the instances that it sees as not being connected to power and/or ground.

There are several possible root causes:

1. The floorplan is too small to accomodate power/ground straps
2. The floorplan contains narrow channels that do not have a connection to power/ground
3. There are no global power/ground connections to the reported cells
4. It's somewhat suspicious that all these instances are at the same location, is there a placement problem, are the stdcell rows missing?

From the dimensions of DIE_AREA in the config.tcl it seems like we should be able to rule out the first of these, but its not clear if one of thes others is the root cause in this case, or if it's due to something else.

To identify the root cause in this case it will be necessary to examine the floorplan after pdngen and/or placement and check the connection to these cells to determine the best way to resolve the issue.

If it's possible to share the full log files as well, then this could help with a diagnosis of the issue.

[vijayank88]

@KHermanUBB Is the issue still valid? If yes please attach a test case generated by flow.