cvs commit: src/sys/i386/i386 apic_vector.s src/sys/i386/isa atpic_vector.s

[Andy Farkas]

Bruce Evans wrote:

> Bah, the benchmark that gave these results was very buggy.  Here is a
> better version.  Results first (for a random pattern to simulate 75% of
> interrupts coming drom userland):

Thank you Bruce for a most excellent explanation. You're indepth knowledge
is invaluable!

It seems that the best "way" depends on your cpu.

I ran your benchmark on several boxes I have here. Here are the results in
case you're interested:

box 1 - dual Pentium 133MHz (old -current, SCHED_ULE, WITNESS, INVARIANTS)

FreeBSD 5.2-CURRENT #1: Wed Jan 14 01:01:48 EST 2004
Timecounter "i8254" frequency 1193182 Hz quality 0
CPU: Pentium/P54C (132.00-MHz 586-class CPU)
  Origin = "GenuineIntel"  Id = 0x52c  Stepping = 12
  Features=0x3bf<FPU,VME,DE,PSE,TSC,MSR,MCE,CX8,APIC>
FreeBSD/SMP: Multiprocessor System Detected: 2 CPUs
Intel Pentium detected, installing workaround for F00F bug

	newway		oldway		newwayfix1	newwayfix2

run 1:	81632862	76581818	80754520	83012440
run 2:	81615364	76597496	80690176	83021998
run 3:	81602834	76905848	80702088	83024774
state average 0.750


box 2 - quad Pentium Pro 200MHz (-current, SCHED_ULE, no WITNESS/INVARIANTS)

FreeBSD 5.2-CURRENT #0: Tue Feb  3 07:20:56 EST 2004
Timecounter "i8254" frequency 1193182 Hz quality 0
CPU: Pentium Pro (198.95-MHz 686-class CPU)
  Origin = "GenuineIntel"  Id = 0x619  Stepping = 9
Features=0xfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV>
FreeBSD/SMP: Multiprocessor System Detected: 4 CPUs

run 1:	195867711	196066980	196889051	196618417
run 2:	195794028	196157055	196451690	196151350
run 3:	195784692	195967221	195984615	196410787
state average 0.750


box 3 - single Pentium Pro 200MHz (4.9-release)

FreeBSD 4.9-RELEASE-p1 #1: Fri Nov 28 22:05:48 EST 2003
Timecounter "i8254"  frequency 1193182 Hz
Timecounter "TSC"  frequency 199740088 Hz
CPU: Pentium Pro (199.74-MHz 686-class CPU)
  Origin = "GenuineIntel"  Id = 0x619  Stepping = 9
Features=0xf9ff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,SEP,MTRR,PGE,MCA,CMOV>

run 1:	185314554	186588693	177427347	177493554
run 2:	185846418	185793670	178317400	176717967
run 3:	186362963	185661694	178136376	176700259
state average 0.750

%%%

Comparing box 2 and 3 (both PPro-200) must show the difference between 4.9
and 5.2, but why does newwayfix[12] have an effect only on 4.9?

--

 :{ andyf at speednet.com.au

        Andy Farkas
    System Administrator
   Speednet Communications
 http://www.speednet.com.au/