On Monday 07 November 2005 08:38 pm, Vaibhave Agarwal wrote: > On Mon, 7 Nov 2005, John Baldwin wrote: > > And even then it can't be used for any device interrupts since there > > aren't any I/O APICs. On a UP machine without I/O APICs, it's actually > > probably more optimal to just use irq0 and irq8 for clocks rather than > > the lapic timer anyway. The only real possible gain is the ability to > > use the profiling interrupt from the local APIC. > > I got access to the BIOS of the Pentium 3 machine I am using, but it has > no option to enable/disable the local APIC. Yes, I've not seen any BIOSen that do. > Joseph Koshy is right, Linux enables the local APIC timer while booting > up. I got the following in the bootup log of Linux 2.4 kernel on the same > machine. > > ------------------------- > Local APIC disabled by BIOS -- reenabling. > Found and enabled local APIC! > > Using local APIC timer interrupts. > calibrating APIC timer ... > ------------------------- Just because Linux does for UP doesn't mean it is more optimal for FreeBSD. :) On FreeBSD with the lapic timer you have 2 * hz interrupts per second. With the irq0/irq8 combo you have hz + stathz interrupts per second. The difference is 2000 vs 1128. Granted, the lapic timer interrupt handler doesn't have to talk to hardware out on the LPC bus.. > Though there is no I/O apic in the UP machines, but I only wanted to use > local APIC timer in the lapic_timer_oneshot() mode to schedule few timers > accurately. You can increase the rate of the rtc timer. We run it at profhz (1024) when profiling is enabled for example. -- John Baldwin <jhb_at_FreeBSD.org> <>< http://www.FreeBSD.org/~jhb/ "Power Users Use the Power to Serve" = http://www.FreeBSD.orgReceived on Tue Nov 08 2005 - 15:11:09 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:47 UTC