Sdk11.1 cyclictest #352

bryanbrattlof · 2025-07-09T14:03:15Z

No description provided.

praneethbajjuri · 2025-07-09T14:50:53Z

Since there are multiple platform variants for AM62x.
Please document clearly on which board the performance numbers are captured on.
Under
"The latencies observed with this SDK are summarized below:"
may be update as
"The latencies observed with this default yocto SDK image on https://www.ti.com/tool/SK-AM62B-P1 and are summarized below:"

cshilwant · 2025-07-09T16:12:01Z

source/devices/AM62AX/linux/RT_Linux_Performance_Guide.rst

-   "Average (us)","5","6","6","5"
-   "Maximum (us)","68","31","34","28"
+   "Average (us)","6","6","6","6"
+   "Maximum (us)","45","30","39","64"


@praneethbajjuri AM62A coreSDK RC for 11.1 would still be +2 weeks from now.
Most likely nothing in vendor kernel should impact the RT Latency numbers but any chances of
stable update from branch linux-6.12.y of git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux into ti-linux-6.12.y-cicd ?

@cshilwant this numbers are ok for now. Will re-do AM62A and AM62L RT test closer to final week of release again and update the numbers if there is a delta.

Okay, ACK from my end once #352 (comment) is fixed

@bryanbrattlof @praneethbajjuri our target for 62A is to keep below 50us. 64us seems a regression ?

r-vignesh · 2025-07-10T04:28:27Z

source/devices/AM62AX/linux/RT_Linux_Performance_Guide.rst

-   "Average (us)","5","6","6","5"
-   "Maximum (us)","68","31","34","28"
+   "Average (us)","6","6","6","6"
+   "Maximum (us)","45","30","39","64"


@bryanbrattlof @praneethbajjuri our target for 62A is to keep below 50us. 64us seems a regression ?

r-vignesh · 2025-07-10T04:29:16Z

source/devices/AM62PX/linux/RT_Linux_Performance_Guide.rst

-   "Average (usec)","6","6","5","6"
-   "Maximum (usec)","30","33","31","35"
+   "Average (usec)","6","6","6","6"
+   "Maximum (usec)","34","74","35","39"


74us needs to be investiagated. This is more than x2 the original number

yeah these numbers are all over the place for all platforms. I've opened a ticket internally to follow up on this

bryanbrattlof · 2025-07-10T14:27:10Z

Yeah I agree @praneethbajjuri . I've updated that sentence to be a little more specific.

StaticRocket · 2025-07-10T21:20:09Z

Could you address the vale comments?

source/devices/AM62AX/linux/RT_Linux_Performance_Guide.rst

praneethbajjuri · 2025-07-10T21:43:23Z

source/devices/AM62PX/linux/RT_Linux_Performance_Guide.rst

-   "Average (usec)","6","6","5","6"
-   "Maximum (usec)","30","33","31","35"
+   "Average (usec)","6","6","6","6"
+   "Maximum (usec)","34","74","35","39"


this usec vale warning can be ignored.

praneethbajjuri

lets hold off on integration of this . Till the investigation completes on why latency numbers changed.
Might need to expand this PR to include the rootcause and workaround details to get better perf numbers.

cshilwant · 2025-07-11T16:20:28Z

moved it to draft based on #352 (review)

cshilwant · 2025-08-01T08:16:13Z

source/devices/AM62AX/linux/RT_Linux_Performance_Guide.rst

At line no. 2,

RT-linux 11.01 Performance Guide

Update the real time Linux latency numbers for SDK 11.1 Signed-off-by: Bryan Brattlof <bb@ti.com>

Update the real-time Linux latency numbers for SDK 11.1 Signed-off-by: Bryan Brattlof <bb@ti.com>

github-actions · 2025-10-02T13:35:54Z

New warnings found with rstcheck:

source/devices/AM64X/linux/RT_Linux_Performance_Guide.rst:584: (SEVERE/4) Title level inconsistent:
source/devices/AM64X/linux/RT_Linux_Performance_Guide.rst:595: (SEVERE/4) Title level inconsistent:
source/devices/AM64X/linux/RT_Linux_Performance_Guide.rst:607: (SEVERE/4) Title level inconsistent:
source/devices/AM64X/linux/RT_Linux_Performance_Guide.rst:619: (SEVERE/4) Title level inconsistent:
source/devices/AM64X/linux/RT_Linux_Performance_Guide.rst:652: (SEVERE/4) Title level inconsistent:
source/devices/AM64X/linux/RT_Linux_Performance_Guide.rst:665: (SEVERE/4) Title level inconsistent:
source/devices/AM64X/linux/RT_Linux_Performance_Guide.rst:680: (SEVERE/4) Title level inconsistent:
source/devices/AM64X/linux/RT_Linux_Performance_Guide.rst:694: (SEVERE/4) Title level inconsistent:

praneethbajjuri · 2025-10-02T18:31:19Z

@dao-qiu please review

praneethbajjuri · 2025-10-02T19:11:11Z

@bryanbrattlof can you update in commitmsg and histogram as well on what is the exact tag used for a given device to get this numbers.
Example:
AM62A 11.1 is RC7
AM62P/62/64/ 11.1 is RC5
AM62L 11.1 and AM62P 11.1.1 is RC16 ( Wait till this PR is merged #459 )

StaticRocket · 2025-10-03T21:21:55Z

https://github.com/TexasInstruments/processor-sdk-doc/pull/352/files/b72f3fb58404e5e18a43712b21d10b6f30d5a17d#r2247290922

dao-qiu · 2025-10-03T22:03:59Z

test comment

dao-qiu · 2025-07-14T15:30:57Z

source/devices/AM64X/linux/RT_Linux_Performance_Guide.rst

-   "Average (usec)","8","8"
-   "Maximum (usec)","77","105"
+   "Average (usec)","10","8"
+   "Maximum (usec)","85","64"


What is the targeted max threshold for AM64x that we are choosing? I'm hearing no more than AM62x < 75us and AM62a/am62p <50us, what about for AM64x and AM62Lx?

dao-qiu · 2025-10-03T21:52:00Z

source/devices/AM62AX/linux/RT_Linux_Performance_Guide.rst

+.. note::
+
+   A known issue in this SDK release is affecting this benchmark.
+   Applying `this patch`_ on Yocto's meta-ti layer will restore the


Not all customers are likely going to have ability to rebuild Yocto to restore the switch latencies. Could there be instructions added that could be an alternative to having to rebuild Yocto for those who don't have the hardware/ability to do so?

bryanbrattlof requested review from cshilwant, StaticRocket, praneethbajjuri, uditkumarti, aniket-l and VeeruPrudhvi as code owners July 9, 2025 14:03

github-actions bot assigned bryanbrattlof, DhruvaG2000 and r-vignesh Jul 9, 2025

cshilwant reviewed Jul 9, 2025

View reviewed changes

r-vignesh reviewed Jul 10, 2025

View reviewed changes

bryanbrattlof force-pushed the sdk11.1-cyclictest branch from 0c611af to f8839d9 Compare July 10, 2025 14:26

praneethbajjuri approved these changes Jul 10, 2025

View reviewed changes

praneethbajjuri reviewed Jul 10, 2025

View reviewed changes

praneethbajjuri requested changes Jul 11, 2025

View reviewed changes

cshilwant marked this pull request as draft July 11, 2025 16:19

cshilwant requested changes Aug 1, 2025

View reviewed changes

source/devices/AM62AX/linux/RT_Linux_Performance_Guide.rst

Copy link

Member

cshilwant Aug 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

At line no. 2,

RT-linux 11.01 Performance Guide

bryanbrattlof added 5 commits October 2, 2025 08:34

perf(linux): AM62AX: Update PREEMPT_RT perf numbers for 11.1

67ce8a0

Update the real time Linux latency numbers for SDK 11.1 Signed-off-by: Bryan Brattlof <bb@ti.com>

perf(linux): AM62LX: Update PREEMPT_RT perf numbers for 11.1

5e83920

Update the real-time Linux latency numbers for SDK 11.1 Signed-off-by: Bryan Brattlof <bb@ti.com>

perf(linux): AM62PX: Update PREEMPT_RT perf numbers for 11.1

ce3c183

Update the real-time Linux latency numbers for SDK 11.1 Signed-off-by: Bryan Brattlof <bb@ti.com>

perf(linux): AM62X: Update PREEMPT_RT perf numbers for 11.1

0229a79

Update the real-time Linux latency numbers for SDK 11.1 Signed-off-by: Bryan Brattlof <bb@ti.com>

perf(linux): AM64X: Update PREEMPT_RT perf numbers for 11.1

b72f3fb

Update the real-time Linux latency numbers for SDK 11.1 Signed-off-by: Bryan Brattlof <bb@ti.com>

bryanbrattlof force-pushed the sdk11.1-cyclictest branch from f8839d9 to b72f3fb Compare October 2, 2025 13:34

bryanbrattlof marked this pull request as ready for review October 2, 2025 13:35

praneethbajjuri self-requested a review October 2, 2025 18:29

praneethbajjuri requested a review from cshilwant October 2, 2025 18:29

praneethbajjuri approved these changes Oct 2, 2025

View reviewed changes

dao-qiu suggested changes Oct 3, 2025

View reviewed changes

Sdk11.1 cyclictest #352

Are you sure you want to change the base?

Sdk11.1 cyclictest #352

Uh oh!

Conversation

bryanbrattlof commented Jul 9, 2025

Uh oh!

praneethbajjuri commented Jul 9, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bryanbrattlof commented Jul 10, 2025

Uh oh!

StaticRocket commented Jul 10, 2025

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

praneethbajjuri left a comment

Choose a reason for hiding this comment

Uh oh!

cshilwant commented Jul 11, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Oct 2, 2025

Uh oh!

praneethbajjuri commented Oct 2, 2025

Uh oh!

praneethbajjuri commented Oct 2, 2025

Uh oh!

StaticRocket commented Oct 3, 2025

Uh oh!

dao-qiu commented Oct 3, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!