Re: [PATCH v4] devfreq: qcom: bwmon: Reduce zone1 threshold count to 3

From: Krzysztof Kozlowski

Date: Fri Nov 20 2026 - 06:51:18 EST


On 20/11/2026 12:01, Pushpendra Singh wrote:
>
>
> On 11/19/2026 5:37 PM, Krzysztof Kozlowski wrote:
>> On 19/11/2026 12:44, Pushpendra Singh wrote:
>>> Reduce zone1_thres_count from 16 to 3 so the driver can lower the bus
>>> vote after 3 sample windows instead of waiting for 16. The previous
>>> 16‑window delay (~64 ms) is too long at higher FPS workloads,
>>> causing delayed decision making and measurable power regression.
>>>
>>> Power and performance measurements were collected for
>>> zone1_thres_count = 3 and 16, with the following results:
>>>
>>> Usecase zone1_cnt=16 zone1_cnt=3 Delta
>>> YouTube streaming (power) 4126.49 mW 3991.30 mW -3.28%
>>> Google Meet audio (power) 3058.79 mW 2929.73 mW -4.22%
>>> Speedometer 3.1(perf) 25.8 26.5 +2.7%
>>> Geekbench 6.5 ST(perf) 3907 3905 -0.1%
>>> Geekbench 6.5 MT(perf) 18636 18680 +0.2%
>>>
>>> The reduced threshold therefore allows BWMON to react sooner to falling
>>> bandwidth demand, providing measurable power savings in the tested
>>> workloads without a meaningful performance regression.
>>
>> I do not see how you resolved my comment, so same comment.
>
> Thanks for the feedback. I had addressed the previous comment by rerunning the power and performance measurements on the latest kernel (v7.3.rc5) and shared the updated results in my previous reply[1]. I have also included these results in the commit message in this respin.
>
> Could you please clarify if there is any additional concern that I may have missed? I’ll be happy to address it.

OK, I missed then the part of commit msg explaining which hardware.

Best regards,
Krzysztof