boot-time slowdown for measure_migration_cost
In-Reply-To: <[email protected]>
Quoted 7 lines · Tony Luck
Might it be wise to see whether the 2% variation that I saw can be repeated on some other architecture? Bjorn's initial post was just questioning whether we need to spend this much time during boot to acquire this data. Now we have *one* data point that on an ia64 with four cpus with 9MB cache in a single domain that we can speed the calculation by a factor of three with only a 2% loss of accuracy. Can someone else try this patch and post the before/after values for
migration_costfrom dmesg?
Before:
messages.1:Jan 24 01:19:45 d2 kernel: [ 6.377117] migration_cost“52 messages.1:Jan 27 21:07:55 d2 kernel: [ 6.384871] migration_cost“29 messages.1:Jan 28 11:00:32 d2 kernel: [ 6.384215] migration_cost“38 messages.1:Jan 28 12:55:03 d2 kernel: [ 6.389189] migration_cost“64
After:
messages:Jan 31 07:55:07 d2 kernel: [ 1.859359] migration_cost’74
This was on a dual PII Xeon with 2MB L2 cache. About 3.5x as fast and only 1% change.
Maybe the default could be to run the quick test with an option to run the more-accurate one?
Signature · 1 line
Chuck