On Mon, Oct 31, 2005 at 09:26:11AM +0000, Robert Watson wrote: > > On Sun, 30 Oct 2005, David O'Brien wrote: > > >>ACPI-fast requires an ioport read which takes about 1us (according to > >>Google). Do that 1000 times a second and you have each CPU spending 1% > >>of its time doing nothing but reading the clock. Yikes. > > > >But we've lived with using the ACPI timercounter (vs. TSC) for quite a > >while now. Why all of a sudden are the authors of this thread having an > >issue with it now. I know about the recent MySQL thread - but with the > >TSC being untrustable on MP and power managed systems, why is there such > >a desire to use the TSC? > > Because in the past few weeks, several developers have noticed that if you > change timercounters from ACPI-fast to TSC, you get significant > performance boosts on real-world workloads. Setting sysctl kern.timecounter.hardware=tsc isn't a great idea. PHK has mentioned several times how it seems many are wanting to play it fast and loose with the time source to the kernel. Providing some alternate loose time answers to an application such as MySQL that doesn't need accurate timing and presumable can handle variation would be a good thing. But that is different than simply setting kern.timecounter.hardware=tsc -- -- David (obrien_at_FreeBSD.org)Received on Mon Oct 31 2005 - 14:20:46 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:46 UTC