[Xenomai-help] Xenomai latency measurements

[Xenomai-help] Xenomai latency measurements

[roderik.wildenburg]

[-- Attachment #1: Type: text/plain, Size: 682 bytes --]

I performed some measurements with the Xenomai latency test.
In case somebody is interested I attached the results for our PPC-System
(5200B 400MHZ).
In this document I also compared latencies of kernel 2.4 with kernel
2.6.
Any comments are welcome, especially whether the values I measured are
reasonable for a 400MHZ PPC.

Roderik

--------------------------------------------------------
manroland AG
Vorsitzender des Aufsichtsrates: Hanno C. Fiedler
Vorstand: Gerd Finkbeiner (Vorsitzender), Dr. Ingo Koch, Dr. Markus Rall, Paul Steidle   
Sitz der Gesellschaft: Offenbach am Main, Registergericht: Amtsgericht Offenbach HRB-Nr. 42592
USt-Ident-Nr. DE 250200933

[-- Attachment #2: XenomaiLatencyEng.pdf --]
[-- Type: application/octet-stream, Size: 529375 bytes --]

Re: [Xenomai-help] Xenomai latency measurements

[Philippe Gerum]

On Wed, 2011-06-01 at 13:14 +0200, roderik.wildenburg@domain.hid
wrote:
> I performed some measurements with the Xenomai latency test.
> In case somebody is interested I attached the results for our PPC-System
> (5200B 400MHZ).
> In this document I also compared latencies of kernel 2.4 with kernel
> 2.6.
> Any comments are welcome, especially whether the values I measured are
> reasonable for a 400MHZ PPC.
> 

Yes, they look like ok but the CPU frequency is not what makes the
difference, the performance of the memory subsystem does - particularly
since ppc MMU management is software-assisted.

We would have more significant information by comparing Xenomai 2.5.6
over Linux 2.4.25.

> Roderik
> 
> --------------------------------------------------------
> manroland AG
> Vorsitzender des Aufsichtsrates: Hanno C. Fiedler
> Vorstand: Gerd Finkbeiner (Vorsitzender), Dr. Ingo Koch, Dr. Markus Rall, Paul Steidle   
> Sitz der Gesellschaft: Offenbach am Main, Registergericht: Amtsgericht Offenbach HRB-Nr. 42592
> USt-Ident-Nr. DE 250200933
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help

-- 
Philippe.

Re: [Xenomai-help] Xenomai latency measurements

[roderik.wildenburg]

Just one more question about the latency test:
Is the "Timer Handler" test an ISR so that the latency of this test tells us the interrupt latency or do I have to run an other test to determine the interrupt latency?.
If Timer Handler is an ISR could you tell me why "ordinary" Linux load has influence on this latency?

I tried to run the latency test also on a kernel 2.4.25 with Xenomai 2.5.6 but I get the error message " Xenomai: mmap local sem heap: No such device or address" (see below). Do you have an idea what is wrong? Google says something about a missing /dev/rtheap, but this device is available on my system.

Many thanks in advance
Roderik

---------------------------------------------------------------------------------------
mrconfig:/fat/sbin # latency -t0
Xenomai: mmap local sem heap: No such device or address


mrconfig:/fat/sbin # strace latency -t0
execve("/fat/sbin/latency", ["latency", "-t0"], [/* 37 vars */]) = 0
:
:
access("/dev/rtheap", F_OK)             = 0
rt_sigaction(SIGILL, {0xff23cb8, [ILL], SA_RESTART}, {SIG_DFL}, 8) = 0
rt_sigaction(SIGILL, {SIG_DFL}, {0xff23cb8, [ILL], SA_RESTART}, 8) = 0
open("/dev/rtheap", O_RDWR)             = 3
ioctl(3, 0, 0xcfbf4210)                 = 0
mmap(NULL, 12288, PROT_READ|PROT_WRITE, MAP_SHARED, 3, 0) = -1 ENXIO (No such device or address)
close(3)                                = 0
dup(2)                                  = 3
fcntl64(3, F_GETFL)                     = 0x2 (flags O_RDWR)
fstat64(3, {st_mode=S_IFCHR|0600, st_rdev=makedev(4, 64), ...}) = 0
ioctl(3, TCGETS, {B38400 opost isig icanon echo ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x30016000
_llseek(3, 0, 0x7fffeac8, SEEK_CUR)     = -1 ESPIPE (Illegal seek)
write(3, "Xenomai: mmap local sem heap: No"..., 56Xenomai: mmap local sem heap: No such device or address
) = 56
close(3)                                = 0
munmap(0x30016000, 4096)                = 0
exit(1)                                 = ?

mrconfig:/fat/sbin # ll /dev/rtheap
crw-rw-rw-    1 root     root      10, 254 May 19 14:16 /dev/rtheap
mrconfig:/fat/sbin #
mrconfig:/fat/sbin # uname -r
2.4.25
mrconfig:/fat/sbin # cat /proc/xenomai/version
2.5.6

---------------------------------------------------------------------------------------

> -----Ursprüngliche Nachricht-----
> Von: Philippe Gerum [mailto:rpm@xenomai.org]
> Gesendet: Mittwoch, 1. Juni 2011 14:09
> An: Wildenburg, Roderik RAEK1 MRA
> Cc: xenomai@xenomai.org
> Betreff: Re: [Xenomai-help] Xenomai latency measurements
> 
> On Wed, 2011-06-01 at 13:14 +0200, roderik.wildenburg@domain.hid
> wrote:
> > I performed some measurements with the Xenomai latency test.
> > In case somebody is interested I attached the results for our PPC-System
> > (5200B 400MHZ).
> > In this document I also compared latencies of kernel 2.4 with kernel
> > 2.6.
> > Any comments are welcome, especially whether the values I measured are
> > reasonable for a 400MHZ PPC.
> >
> 
> Yes, they look like ok but the CPU frequency is not what makes the
> difference, the performance of the memory subsystem does - particularly
> since ppc MMU management is software-assisted.
> 
> We would have more significant information by comparing Xenomai 2.5.6
> over Linux 2.4.25.
> 
> > Roderik
> >
> > --------------------------------------------------------
> > manroland AG
> > Vorsitzender des Aufsichtsrates: Hanno C. Fiedler
> > Vorstand: Gerd Finkbeiner (Vorsitzender), Dr. Ingo Koch, Dr. Markus Rall, Paul
> Steidle
> > Sitz der Gesellschaft: Offenbach am Main, Registergericht: Amtsgericht
> Offenbach HRB-Nr. 42592
> > USt-Ident-Nr. DE 250200933
> > _______________________________________________
> > Xenomai-help mailing list
> > Xenomai-help@domain.hid
> > https://mail.gna.org/listinfo/xenomai-help
> 
> --
> Philippe.
> 
>

--------------------------------------------------------
manroland AG
Vorsitzender des Aufsichtsrates: Hanno C. Fiedler
Vorstand: Gerd Finkbeiner (Vorsitzender), Dr. Ingo Koch, Dr. Markus Rall, Paul Steidle   
Sitz der Gesellschaft: Offenbach am Main, Registergericht: Amtsgericht Offenbach HRB-Nr. 42592
USt-Ident-Nr. DE 250200933

Re: [Xenomai-help] Xenomai latency measurements; supplementation

[roderik.wildenburg]

[-- Attachment #1: Type: text/plain, Size: 924 bytes --]

> We would have more significant information by comparing Xenomai 2.5.6
> over Linux 2.4.25.

I added the measurements for the combination Xenomai 2.5.6 on Linux 2.4.25.

As I would prefer to skip the interrupt latency measurements I am reasoning whether the kernel timer handler latencies are in any way comparable with interrupt latencies (e.g. "interrupt latencies are 5 times smaller than kernel timer latencies", or "interrupt latencies are always (much) smaller than kernel timer latencies" ). Does anybody have experience or knowledge about this issue (Jan?)?

Roderik

--------------------------------------------------------
manroland AG
Vorsitzender des Aufsichtsrates: Hanno C. Fiedler
Vorstand: Gerd Finkbeiner (Vorsitzender), Dr. Ingo Koch, Dr. Markus Rall, Paul Steidle   
Sitz der Gesellschaft: Offenbach am Main, Registergericht: Amtsgericht Offenbach HRB-Nr. 42592
USt-Ident-Nr. DE 250200933

[-- Attachment #2: XenomaiLatencyEng.pdf --]
[-- Type: application/octet-stream, Size: 594570 bytes --]