Performance of 4.x vs 5.x (Re: Lifetime of FreeBSD branches)

[Krzysztof Kowalik]

Kris Kennaway [kris at obsecurity.org] wrote:
> [...]  One obvious guess is that it's due to VFS being under Giant,
> which causes lots of contention with other subsystems that also
> require Giant, and therefore introduces latency.  If so, you'd see a
> substantial performance improvement on 6.0 with debug.mpsafevfs=1.

I didn't see any visible difference, in the given scenario of
uncompressing firefox's sources, when tried mpsafevfs's patches when
they got announced on current at .

The funny thing is, that I saw it on my Athlon XP box *very* visibly,
while I it's quite ok on my current workstation, which is Celeron based
(and yes, it's much slower, CPU wise, than the Athlon machine).

Perhaps some funny chipset issue?

Old box:
FreeBSD 5.3-RELEASE #2: Wed Nov 17 00:19:56 CET 2004
[...]
ACPI APIC Table: <ASUS   A7V333  >
Timecounter "i8254" frequency 1193182 Hz quality 0
CPU: AMD Athlon(TM) XP 2000+ (1668.71-MHz 686-class CPU)
  Origin = "AuthenticAMD"  Id = 0x662  Stepping = 2
  Features=0x383fbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,MMX,FXSR,SSE>
  AMD Features=0xc0400000<AMIE,DSP,3DNow!>
real memory  = 1073725440 (1023 MB)
avail memory = 1041154048 (992 MB)
[...]
emu10kx0: <Creative SBLive! 5.1 [SB0100]> port 0x9400-0x941f irq 19 at
device 12.0 on pci0
pcm0: <EMU10Kx PCM Interface> on emu10kx0
pcm0: <eMicro EM28028 AC97 Codec>
[...]
atapci0: <Promise PDC20276 UDMA133 controller> port
0xa400-0xa40f,0xa800-0xa803,0xb000-0xb007,0xb400-0xb403,0xb800-0xb807
mem 0xe6800000-0xe6803fff irq 19 at device 6.0 on pci0
ata2: channel #0 on atapci0
ata3: channel #1 on atapci0
atapci1: <VIA 8233A UDMA133 controller> port
0x8400-0x840f,0x376,0x170-0x177,0x3f6,0x1f0-0x1f7 at device 17.1 on pci0
ata0: channel #0 on atapci1
ata1: channel #1 on atapci1
ad0: 152627MB <SAMSUNG SP1614N/TM100-24> [310101/16/63] at ata0-master UDMA100
ad1: 190782MB <WDC WD2000JB-00FUA0/15.05R15> [387621/16/63] at
ata0-slave UDMA100
ar0: 57220MB <ATA RAID1 array> [7294/255/63] status: READY subdisks:
 disk0 READY on ad4 at ata2-master
 disk1 READY on ad5 at ata2-slave
[...]

New box:
FreeBSD 5.4-RELEASE-p1 #0: Sat May 14 18:43:25 CEST 2005
[...]
CPU: Intel(R) Celeron(TM) CPU                1200MHz (1202.73-MHz
686-class CPU)
  Origin = "GenuineIntel"  Id = 0x6b1  Stepping = 1
  Features=0x383f9ff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,MMX,FXSR,SSE>
real memory  = 536805376 (511 MB)
avail memory = 515629056 (491 MB)
[...]
atapci0: <Intel ICH2 UDMA100 controller> port
0xf000-0xf00f,0x376,0x170-0x177,0x3f6,0x1f0-0x1f7 at device 31.1 on pci0
ata0: channel #0 on atapci0
ata1: channel #1 on atapci0
[...]
pcm0: <Intel ICH2 (82801BA)> port 0xe000-0xe03f,0xdc00-0xdcff irq 7 at
device 31.5 on pci0
pcm0: <SigmaTel STAC9700/83/84 AC97 Codec>
[...]
ad0: 76319MB <ST380021A/3.19> [155061/16/63] at ata0-master UDMA100
ad1: 152627MB <SAMSUNG SP1614N/TM100-24> [310101/16/63] at ata0-slave
UDMA100
[...]

So they are, unfortunately, a little bit different machines. And no, I had
no chance to try 5.4-RELEASE on the amd one.

In general, I find 5.4-RELEASE performing better, if I can say that
without doing any real benchmarking, than any previous 5.x.
-- 
Krzysztof Kowalik               |  () ASCII Ribbon Campaign
Computer Center, AGH UST        |  /\ Support plain text e-mail