Summary of discussion of harvester/random locking and performance optimization

Andrea Campi andrea+freebsd_current at
Sat Aug 21 11:05:20 PDT 2004

On Sun, Aug 15, 2004 at 05:46:16PM +1000, Peter Jeremy wrote:
> >systems may not have a TSC, and insead read the system clock (ouch!).  We
> >may want to investigate what approaches we can use to mitigate this,
> >especially if systems like soekris boxes don't have TSC.
> in efficiently supporting the i486.  If the various low-end iA32 clones
> do support the TSC, there seems little point in changing the status quo.

Since nobody chimed in yet, my Soekris box has:

kern.timecounter.stepwarnings: 0
kern.timecounter.nbinuptime: 76822456
kern.timecounter.nnanouptime: 2
kern.timecounter.nmicrouptime: 2728
kern.timecounter.nbintime: 2659850
kern.timecounter.nnanotime: 596
kern.timecounter.nmicrotime: 2659254
kern.timecounter.ngetbinuptime: 15352306
kern.timecounter.ngetnanouptime: 3901
kern.timecounter.ngetmicrouptime: 278552
kern.timecounter.ngetbintime: 0
kern.timecounter.ngetnanotime: 0
kern.timecounter.ngetmicrotime: 15
kern.timecounter.nsetclock: 5
kern.timecounter.hardware: i8254
kern.timecounter.choice: ELAN(-2000) i8254(0) dummy(-1000000)
kern.timecounter.tick: 1

This is the ELAN-based model, which is a 486 class CPU. More powerful
models might or might not have the TSC. I haven't investigated the
cost/benefit of using ELAN timecounter instead of i8254.


               I believe the technical term is "Oops!"

More information about the freebsd-current mailing list