Hi, On Mon, Jul 16, 2007 at 11:04:11PM -0700, hirakendu das wrote: > Hmm, seems that I can reproduce it frequently (in fact everytime :-). > The lspci outputs are attached. Although there is no difference (diff > -ruN). Yes, there is no clue which program etc is causing it or whats > happening. Don't mind, but I even suspect some powertop flaw, because > doing a cat /proc/acpi/processor/CPU0/power shows only C2 or C3 as > active (starred). And frankly, there is no difference in the way CPU > fan behaves. (Would we not expect the fans to whine if it were indeed > in C0 mostly ?).
Umm, excuuuse me, but didn't you just say that the system felt more responsive in bad state? ;) In that case it definitely wouldn't be a powertop flaw. Maybe watching /proc/interrupts provides some clue? And booting with idle=poll provides higher performance about equal to the one the "bad state" provides, right? (due to "bad state" keeping CPU in non-Cx states almost as much as idle=poll does. Just to further nail down whether it's a "powertop flaw" or whether it's a real issue... Maybe running oprofile on an idle system in good and bad state shows some characteristic differences which could nail it down? (but probably most extra "bad state" activity is interrupt-based, thus it could be difficult to trace via oprofile) Andreas Mohr _______________________________________________ Power mailing list [email protected] http://www.bughost.org/mailman/listinfo/power
