OpenROAD-flow-scripts
OpenROAD-flow-scripts copied to clipboard
Evaluate degradation on asap7/riscv32i
Description
Design: asap7/riscv32i
, main metric change finish__timing__setup__ws
+81
-> -5
.
Good:
"run__flow__openroad_version": "v2.0-9791-gea56b56a9",
"run__flow__scripts_commit": "c83d09863e38a4b01c925621f08e472e39e91f6d",
"finish__timing__drv__max_cap_limit": 0.0388632,
"finish__timing__drv__max_slew": 244,
"finish__timing__drv__hold_violation_count": 0,
"finish__timing__drv__setup_violation_count": 0,
"finish__timing__setup__tns": 0,
"finish__timing__setup__ws": 81.6611,
"finish__timing__wns_percent_delay": 5.957017,
Bad:
"run__flow__openroad_version": "v2.0-9822-g4e4293bbe",
"run__flow__scripts_commit": "a059a1ba4711a8e63415a41d9488cbbcd1f6778d",
"finish__timing__drv__max_cap_limit": 0.102151,
"finish__timing__drv__max_slew": 42,
"finish__timing__drv__hold_violation_count": 3,
"finish__timing__drv__setup_violation_count": 1,
"finish__timing__setup__tns": -5.10636,
"finish__timing__setup__ws": -5.10636,
"finish__timing__wns_percent_delay": -0.350328,
Suggested Solution
No response
Additional Context
No response
@openroadie I don't know if this is a rsz issue or something else. Please do an initial triage.
@maliberty @vvbandeira as far as I can tell, this is the PR that degraded performance:
https://github.com/The-OpenROAD-Project/OpenROAD/pull/3834
Aug 8 : 81.6611: e088bf159a892982cce289c74367193da8b602db
Aug 8 : 81.6611: 83770be334d117a755f6e1a3f3a8e4cd332b8191
Aug 9 : 81.6611: 3934c57955502c90cf0147af0f4b238f137e7afb
Aug 9 : 81.6611: a2fc4465a414194db0acc9395620ce3f6274e802
Aug 9 : 81.6611: 4474922bd6d85e9491e7ffebfb00cbab7e9e64c9
Aug 9 : -5.1064: 3e12bbef4d35f3f8f922d33935a03c8f35552640
Aug 9 : -5.1064: 4e4293bbec4c2d2abb40a9cb78c69bbaa2ab58fc
Aug 10 : -5.1064: b31a3b960dfa4f0cf387e53d6218e9e8f89d3aa8
As you can see ... the last SHA before 3e12xx maintained the QOR and then 3e12 has the degraded result.