[Xenomai-help] interrupt latencies (irqloop/irqbench)

[Xenomai-help] interrupt latencies (irqloop/irqbench)

[Markus Franke]

[-- Attachment #1: Type: text/plain, Size: 571 bytes --]

Dear all,

I am doing some latency benchmarks with irqbench/irqloop.
Everything works fine with irqloop in testmode (1=user, 2=kernel and
3=irq handler) But when trying to start with testmode 4=hard IRQ
handler, I get the following message:

---snip---
irqloop: error starting test: Resource temporarily unavailable
---snap---

This occurs somewhere when calling:

	ioctl(benchdev, RTTST_RTIOC_IRQBENCH_START, &config)

Any quick suggestions how to solve this problem?
By the way: I read already the provided document in "txt" folder.


Thanks in advance,

Markus Franke

[-- Attachment #2: Markus.Franke.vcf --]
[-- Type: text/x-vcard, Size: 245 bytes --]

begin:vcard
fn:Markus Franke
n:Franke;Markus
adr;quoted-printable:;;Vettersstra=C3=9Fe 64/722;Chemnitz;Saxony;09126;Germany
email;internet:Markus.Franke@domain.hid
x-mozilla-html:FALSE
url:http://www.tu-chemnitz.de/~franm
version:2.1
end:vcard

Re: [Xenomai-help] interrupt latencies (irqloop/irqbench)

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1359 bytes --]

Markus Franke wrote:
> Dear all,
> 
> I am doing some latency benchmarks with irqbench/irqloop.
> Everything works fine with irqloop in testmode (1=user, 2=kernel and
> 3=irq handler) But when trying to start with testmode 4=hard IRQ
> handler, I get the following message:
> 
> ---snip---
> irqloop: error starting test: Resource temporarily unavailable
> ---snap---
> 
> This occurs somewhere when calling:
> 
> 	ioctl(benchdev, RTTST_RTIOC_IRQBENCH_START, &config)
> 
> Any quick suggestions how to solve this problem?

This test mode exploits the I-pipe support for multiple domains
intensively. By default, Xenomai is configured to be the topmost domain
and is optimised for this use case. But it can also be "second-best" and
let someone else be first in the interrupt pipeline - e.g. xeno_irqloop.
To achieve this, switch off CONFIG_XENO_OPT_PIPELINE_HEAD (nucleus
option) and retry the test.

But while watching the (probably) impressive numbers, keep the
restricted usage scenario of such stacked domains in mind. This kind of
super domains lacks any RTOS environment that Xenomai provides. It's
only useful for simple jobs that can be done within an IRQ handler and
do not need an RT-scheduler.

> By the way: I read already the provided document in "txt" folder.
> 

Then there is a bug - in the doc. :)

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 249 bytes --]

Re: [Xenomai-help] interrupt latencies (irqloop/irqbench)

[Markus Franke]

[-- Attachment #1: Type: text/plain, Size: 640 bytes --]

Jan Kiszka wrote:
> This test mode exploits the I-pipe support for multiple domains
> intensively. 

I know....I read the nice txt-document :-)

> By default, Xenomai is configured to be the topmost domain
> and is optimised for this use case. But it can also be "second-best" and
> let someone else be first in the interrupt pipeline - e.g. xeno_irqloop.
> To achieve this, switch off CONFIG_XENO_OPT_PIPELINE_HEAD (nucleus
> option) and retry the test.

Ok I understand. It would be good if such a hint would be in this txt-file.

> Then there is a bug - in the doc. :)

Why? The above information is simply missing. :-)

Regards,
Markus

[-- Attachment #2: Markus.Franke.vcf --]
[-- Type: text/x-vcard, Size: 245 bytes --]

begin:vcard
fn:Markus Franke
n:Franke;Markus
adr;quoted-printable:;;Vettersstra=C3=9Fe 64/722;Chemnitz;Saxony;09126;Germany
email;internet:Markus.Franke@domain.hid
x-mozilla-html:FALSE
url:http://www.tu-chemnitz.de/~franm
version:2.1
end:vcard

Re: [Xenomai-help] interrupt latencies (irqloop/irqbench)

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 765 bytes --]

Markus Franke wrote:
> Jan Kiszka wrote:
>> This test mode exploits the I-pipe support for multiple domains
>> intensively. 
> 
> I know....I read the nice txt-document :-)
> 
>> By default, Xenomai is configured to be the topmost domain
>> and is optimised for this use case. But it can also be "second-best" and
>> let someone else be first in the interrupt pipeline - e.g. xeno_irqloop.
>> To achieve this, switch off CONFIG_XENO_OPT_PIPELINE_HEAD (nucleus
>> option) and retry the test.
> 
> Ok I understand. It would be good if such a hint would be in this txt-file.
> 
>> Then there is a bug - in the doc. :)
> 
> Why? The above information is simply missing. :-)

I consider missing information as a bug too. However, fixed now.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 249 bytes --]