linux4.14+xenomai3.1 real-time

linux4.14+xenomai3.1 real-time

[夏侯量]

dear Xenomai Team:

When we tested on the Intel(r) Celeron(r) processor 1047ue (third
generation) motherboard (using kernel version: 4.1.18, xenomai version: 2.6.
5, patch: ipipe-core-4.1.18-x86-3. ), The real-time test results meet the
requirements.But the 4.1.18 version of the kernel (xenomai version 2.6.5)
running on the Intel(r) Celeron(r) processor 3955U (sixth generation)
motherboard can not guarantee real-time performance

And on Intel(r) Celeron(r) processor 3955U (sixth generation) motherboard
(kernel version: 4.14.134, xenomai version: 3.1, patch:
ipipe-core-4.14.134-x86-2. Patch ), We have done many configuration
modification tests. However, the maximum jitter is 25u under 1ms, the
real-time test results cannot meet our requirements. Our goal is a maximum
jitter of around 10u. 

We would like to ask what relevant content needs to be modified.

 

Re: linux4.14+xenomai3.1 real-time

[Jan Kiszka]

Hi Xiahou Liang,

On 17.04.20 11:36, 夏侯量 via Xenomai wrote:
> dear Xenomai Team:
> 
> When we tested on the Intel(r) Celeron(r) processor 1047ue (third
> generation) motherboard (using kernel version: 4.1.18, xenomai version: 2.6.
> 5, patch: ipipe-core-4.1.18-x86-3. ), The real-time test results meet the
> requirements.But the 4.1.18 version of the kernel (xenomai version 2.6.5)
> running on the Intel(r) Celeron(r) processor 3955U (sixth generation)
> motherboard can not guarantee real-time performance
> 
> And on Intel(r) Celeron(r) processor 3955U (sixth generation) motherboard
> (kernel version: 4.14.134, xenomai version: 3.1, patch:
> ipipe-core-4.14.134-x86-2. Patch ), We have done many configuration
> modification tests. However, the maximum jitter is 25u under 1ms, the

IOW, the maximum jitter was 975 µs? That would be indeed completely off.

> real-time test results cannot meet our requirements. Our goal is a maximum
> jitter of around 10u.

How did  you measure the latency? Via the Xenomai testsuite (latency) or 
via your own application? If the latter, can you reproduce such 
excessive latencies also with Xenomai tests?

> 
> We would like to ask what relevant content needs to be modified.
> 

First of all, don't use 4.14 anymore when starting a new project. It's 
unmaintained Xenomai-wise. Please try with latest 4.19 patch 
(ipipe-core-4.19.114-cip24-x86-12).

Then we would need your kernel .config for that setup if you can still 
reproduce that high latency.

Jan

-- 
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux

答复: linux4.14+xenomai3.1 real-time

[夏侯量]

ok，thanks.

-----邮件原件-----
发件人: Jan Kiszka [mailto:jan.kiszka@siemens.com] 
发送时间: 2020年4月17日 17:50
收件人: 夏侯量; xenomai@xenomai.org
抄送: huang.c.h@lynuc.cn
主题: Re: linux4.14+xenomai3.1 real-time

Hi Xiahou Liang,

On 17.04.20 11:36, 夏侯量 via Xenomai wrote:
> dear Xenomai Team:
> 
> When we tested on the Intel(r) Celeron(r) processor 1047ue (third
> generation) motherboard (using kernel version: 4.1.18, xenomai version: 2.6.
> 5, patch: ipipe-core-4.1.18-x86-3. ), The real-time test results meet 
> the requirements.But the 4.1.18 version of the kernel (xenomai version 
> 2.6.5) running on the Intel(r) Celeron(r) processor 3955U (sixth 
> generation) motherboard can not guarantee real-time performance
> 
> And on Intel(r) Celeron(r) processor 3955U (sixth generation) 
> motherboard (kernel version: 4.14.134, xenomai version: 3.1, patch:
> ipipe-core-4.14.134-x86-2. Patch ), We have done many configuration 
> modification tests. However, the maximum jitter is 25u under 1ms, the

IOW, the maximum jitter was 975 µs? That would be indeed completely off.

> real-time test results cannot meet our requirements. Our goal is a 
> maximum jitter of around 10u.

How did  you measure the latency? Via the Xenomai testsuite (latency) or via your own application? If the latter, can you reproduce such excessive latencies also with Xenomai tests?

> 
> We would like to ask what relevant content needs to be modified.
> 

First of all, don't use 4.14 anymore when starting a new project. It's unmaintained Xenomai-wise. Please try with latest 4.19 patch (ipipe-core-4.19.114-cip24-x86-12).

Then we would need your kernel .config for that setup if you can still reproduce that high latency.

Jan

--
Siemens AG, Corporate Technology, CT RDA IOT SES-DE Corporate Competence Center Embedded Linux

RE: linux4.14+xenomai3.1 real-time

[Meng, Fino]

>
>dear Xenomai Team:
>
>When we tested on the Intel(r) Celeron(r) processor 1047ue (third
>generation) motherboard (using kernel version: 4.1.18, xenomai version: 2.6.
>5, patch: ipipe-core-4.1.18-x86-3. ), The real-time test results meet the requirements.But the 4.1.18 version of the kernel
>(xenomai version 2.6.5) running on the Intel(r) Celeron(r) processor 3955U (sixth generation) motherboard can not
>guarantee real-time performance
>
>And on Intel(r) Celeron(r) processor 3955U (sixth generation) motherboard (kernel version: 4.14.134, xenomai version: 3.1,
>patch:
>ipipe-core-4.14.134-x86-2. Patch ), We have done many configuration modification tests. However, the maximum jitter is
>25u under 1ms, the real-time test results cannot meet our requirements. Our goal is a maximum jitter of around 10u.
>
>We would like to ask what relevant content needs to be modified.
>
>

Hi houliang,

by our experience, to reach worst case jitter around 10u is hard. 
If your cycle period is 1ms, do u really need 10u in your application?

BR / Fino (孟祥夫)
Intel – IOTG Developer Enabling