-
Notifications
You must be signed in to change notification settings - Fork 378
Description
Subject
[Stage]: Other. Please describe below.
Describe the bug
Found during an asap7 cva6 autotuner run. In three of the fifty trials, ORFS fails in the 6_report step with VSS connectivity check failures. The issue might be a PDN generation issue, but it's detected in the 6_report step.
The log snippet is below, but here's what I think I see after looking at the final ODB for one of the instances that it complains about:
- The warning messages reference unconnected nodes and instances
- The tap cell's VSS pin looks connected at the M1/M2 level since there are M2 and M1 straps over the top of the instance's VSS pin plus the V1 via. The location referenced in the log message is the center of one of the V1 vias
- But, it looks like the M2 strap might not be connected to the power grid on the upper levels. When I trace back to the left side of the die, I see M5 straps that are connected to M6, but nothing that connects M5 down to M2. So, that might be the root of the problem
Expected Behavior
Correctly connected power grid or error out earlier.
Environment
804e222b1d9123817209d1f2a39dc9723ac1c76b - git commit tag
To Reproduce
make issue tarball: https://drive.google.com/file/d/1jJFth_bWmw0YpVgpEmhNG4MZSifi41M0/view?usp=drive_link
or update asap7 cva6:
- config.mk
- utilization: 73
- core margin: 2.1564801575253996
- place density: 0.7644920567209246
- cts cluster diameter: 21
- cts cluster size: 51
- clock period: 1131.1165760413621
Relevant log output
Log file snippet:
Percentage drop : 18.29 %
######################################
[WARNING PSM-0038] Unconnected node on net VSS at location (115.902um, 104.760um), layer: M1.
[WARNING PSM-0038] Unconnected node on net VSS at location (115.902um, 105.300um), layer: M1.
<lots of M1 warning messages>
[WARNING PSM-0038] Unconnected node on net VSS at location (115.902um, 104.760um), layer: M2.
[WARNING PSM-0038] Unconnected node on net VSS at location (115.902um, 105.300um), layer: M2.
<lots of M2 warning messages>
[WARNING PSM-0039] Unconnected instance PHY_EDGE_ROW_379_4_Left_1829/VSS at location (115.938um, 104.760um).
[WARNING PSM-0039] Unconnected instance PHY_EDGE_ROW_380_4_Left_1830/VSS at location (115.938um, 104.760um).
<lots of unconnected instance warning messages for endcaps>
[WARNING PSM-0039] Unconnected instance FILLER_460_2108/VSS at location (116.154um, 126.360um).
[WARNING PSM-0039] Unconnected instance FILLER_466_2108/VSS at location (116.154um, 127.980um).
<lots of unconnected instance warning messages for fill cells>
[WARNING PSM-0039] Unconnected instance gen_cache_hpd.i_cache_subsystem.i_cva6_icache.gen_sram\[1\].data_sram.gen_techno_cut.genblk1.i_tc_sram_wrapper_3552/VSS at location (116.505um, 129.060um).
[WARNING PSM-0039] Unconnected instance gen_cache_hpd.i_cache_subsystem.i_cva6_icache.gen_sram\[1\].data_sram.gen_techno_cut.genblk1.i_tc_sram_wrapper_3550/VSS at location (116.505um, 129.600um).
<more unconnected instance warning messages for the SRAMs, fillers, etc>
[ERROR PSM-0069] Check connectivity failed on VSS.
Error: final_report.tcl, 62 PSM-0069
Command exited with non-zero status 1
Elapsed time: 2:38.38[h:]min:sec. CPU time: user 176.49 sys 52.03 (144%). Peak memory: 2675580KB.
Screenshots
Location referenced in the log file. The tap cell has a M1 pin for VSS and we place M1 and M2 straps plus three V1 vias on top of the tap cell M1 pin
Left side of the core area. M6 -> M5 connection exists (teal and blue areas at the bottom of the pic). But, there aren't any vias between M5 and M2 (magenta to red nearer the top of the pic)
Additional Context
No response