http://defect.opensolaris.org/bz/show_bug.cgi?id=4616
Bill Holler <bill.holler at sun.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ACCEPTED
--- Comment #2 from Bill Holler <bill.holler at sun.com> 2008-11-25 20:34:37
---
The fix introduces a platform function pointer which is set by cpudrv
to a cstate aware idle cpu selection function.
A test system with this change appears to save a few watt at idle.
It is hard to tell because the power reading bounces around a lot.
Unfortunately a fork stress test regresses 25%. Powertop shows the system
spends more time in C3 and less in C2 during the stress test with this
fix. The system also used about 5 to 7 less watts during the stress test.
It is possible the performance regression is due to a slowdown in
cmt_balance(), or the slowdown is due to CPUs entering deeper C-states
due to being left in C3 longer (because C2 idle CPUs are chosen before the C3
idle CPUs). If the latter is the case, a new bug we are about to file may
fix the performance regression.
Experiments will be run to see if this change slowed down cmt_balance().
--
Configure bugmail: http://defect.opensolaris.org/bz/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
