-
Notifications
You must be signed in to change notification settings - Fork 49
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
tools: sof_perf_analyzer: change module CPC calculation #1136
Conversation
Is this basically hiding very short audio glitches? https://bravenewgeek.com/everything-you-know-about-latency-is-wrong/ |
@marc-hb , I am not able to open this link, could you stick to this PR itself? does above link related with this PR? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
code looks good, let's revise commit message a little bit:
In daily performance test, we have a high and abnormal peak
cycle for some module, recently, we find this abnormal peak
cycle is caused by IRQ.
Previously, module CPC is calculated by multiplying the mean
value of cpu_peak with a margin scalar. With further team
discussion based on the new finding, an agreement is reached
to use the product of module average cycle and a margin as
module CPC.
43bb8b8
to
247247b
Compare
I'm not sure about the copyright sorry. I never had any problem loading that page.
Yes, it explains that averages are generally useless. Even more useless in "real-time" situation like audio . |
247247b
to
22fdfec
Compare
Can we have at lease one team member that agrees with this method review this pull request? |
@lgirdwood has agreed to use average cycles with a fix ratio for CPC to upstream in FW manifest. So that we can use CPC in FW manifest to monitor potential performance regression in daily performance test in the future. So the method in this PR 'CPC = AVG(module) * CPC_MARGIN' with CPC_MARGIN =1.5 looks good to me. @aiChaoSONG @marc-hb |
In daily performance test, we have a high and abnormal peak cycle for some module, recently, we find this abnormal peak cycle is caused by IRQ. Previously, module CPC is calculated by multiplying the mean value of cpu_peak with a margin scalar. With further team discussion based on the new finding, an agreement is reached to use the product of module average cycle and a margin as module CPC. Signed-off-by: Baofeng Tian <[email protected]>
Failure not related |
Previously, module CPC is calculated by multiplying the mean value of cpu_peak with a margin scalar, however, abnormal peak cycles are finally proved that it is caused by IRQ, with further team discussion, an agreement was reached to use module average multiplied with a margin as new way for CPC calculation.