preempt_rt on BeagleV-Fire

preempt_rt on BeagleV-Fire

[Tobias Schaffner]

Hi all,

has anyone already tested preempt_rt on the BeagleV-Fire?

OSADL has evaluated preempt_rt on both the PolarFire Icicle Kit
(MPFS250T) and the PolarFire Discovery Kit (MPFS095T) and has reported
rather problematic latency results so far[1].

Since the BeagleV-Fire is also based on a PolarFire SoC (MPFS025T), I
suspect we might run into similar issues. Any experiences, measurements,
or insights would be greatly appreciated.

Best,
Tobias

[1] https://www.osadl.org/Latency-plot-of-system-in-rack-7-slot.qa-latencyplot-r7s2.0.html?shadow=1

Re: [cip-dev] preempt_rt on BeagleV-Fire

[Pavel Machek]

[-- Attachment #1: Type: text/plain, Size: 1105 bytes --]

Hi!

> has anyone already tested preempt_rt on the BeagleV-Fire?
> 
> OSADL has evaluated preempt_rt on both the PolarFire Icicle Kit
> (MPFS250T) and the PolarFire Discovery Kit (MPFS095T) and has reported
> rather problematic latency results so far[1].
> 
> Since the BeagleV-Fire is also based on a PolarFire SoC (MPFS025T), I
> suspect we might run into similar issues. Any experiences, measurements,
> or insights would be greatly appreciated.

Those OSADL results actually look good to me :-).

Ok, so let me explain. BeagleV-Fire seems to have slow CPUs. Like
"Nokia N900" slow. That's TI OMAP 3430 SoC, 600 MHz ARM Cortex-A8 CPU
from 2010 or so.

It is certainly less than 10% of speed of Core i9-9900K, and that one
reaches 100usec...

https://www.osadl.org/Latency-plot-of-system-in-rack-0-slot.qa-latencyplot-r0s5.0.html?latencies=&showno=&shadow=1&slider=360

...so 350usec is actually not that bad, AFAICT.

Not sure what our requirements are, sometimes people put the
time-critical stuff into FPGA.

Best regards,
								Pavel
-- 
In cooperation with Nabla.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

Re: [cip-dev] preempt_rt on BeagleV-Fire

[Tobias Schaffner]

Hi Pavel,

On 1/9/26 10:58, Pavel Machek via lists.cip-project.org wrote:
> Hi!
> 
>> has anyone already tested preempt_rt on the BeagleV-Fire?
>>
>> OSADL has evaluated preempt_rt on both the PolarFire Icicle Kit
>> (MPFS250T) and the PolarFire Discovery Kit (MPFS095T) and has reported
>> rather problematic latency results so far[1].
>>
>> Since the BeagleV-Fire is also based on a PolarFire SoC (MPFS025T), I
>> suspect we might run into similar issues. Any experiences, measurements,
>> or insights would be greatly appreciated.
> 
> Those OSADL results actually look good to me :-).
> 
> Ok, so let me explain. BeagleV-Fire seems to have slow CPUs. Like
> "Nokia N900" slow. That's TI OMAP 3430 SoC, 600 MHz ARM Cortex-A8 CPU
> from 2010 or so.
> 
> It is certainly less than 10% of speed of Core i9-9900K, and that one
> reaches 100usec...
> 
> https://www.osadl.org/Latency-plot-of-system-in-rack-0-slot.qa-latencyplot-r0s5.0.html?latencies=&showno=&shadow=1&slider=360
> 
> ...so 350usec is actually not that bad, AFAICT.

I cannot see any CPU isolation in the profile data for the Core 
i9-9900K. Maybe they are using cgroups here, and this is not visible.

Comparing the isolated cores of the VisionFive2 @ 1500 MHz [1] to the 
MPFS250T @ 600 MHz [2], it is more like 45 µs to 250 µs.

Your point is still valid. The gap may not be as large as it looks at 
first glance, but I don’t think the difference in frequency explains it 
completely.

Best,
Tobias

[1] 
https://www.osadl.org/Latency-plot-of-system-in-rack-3-slot.qa-latencyplot-r3s2.0.html?shadow=1
[2] 
https://www.osadl.org/Latency-plot-of-system-in-rack-7-slot.qa-latencyplot-r7s2.0.html?shadow=1

> Not sure what our requirements are, sometimes people put the
> time-critical stuff into FPGA.
> 
> Best regards,
> 								Pavel
> 
> 
> 
> -=-=-=-=-=-=-=-=-=-=-=-
> Links: You receive all messages sent to this group.
> View/Reply Online (#21493): https://lists.cip-project.org/g/cip-dev/message/21493
> Mute This Topic: https://lists.cip-project.org/mt/117149606/7247900
> Group Owner: cip-dev+owner@lists.cip-project.org
> Unsubscribe: https://lists.cip-project.org/g/cip-dev/unsub [tobias.schaffner@siemens.com]
> -=-=-=-=-=-=-=-=-=-=-=-
>