linux-kernel - [QUESTION] The delay between sampling for CPPC

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <0453c753-681e-4d42-65e5-a5967697f6ca@os.amperecomputing.com>
Date:   Mon, 20 Mar 2023 10:21:51 -0700
From:   Yang Shi <yang@...amperecomputing.com>
To:     rafael@...nel.org, viresh.kumar@...aro.org,
        george.cherian@...ium.com, pprakash@...eaurora.org
Cc:     scott@...amperecomputing.com, linux-pm@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: [QUESTION] The delay between sampling for CPPC

Hi folks,

When testing CPPC cpufreq on our platform, we noticed the error may be 
quite high and it (high error) may happen quite often. For example, our 
platform max freq is 2.8GHz, when the CPUs were fully loaded (100% 
load), we saw cpuinfo_cur_freq may show 4GHz, it means > 40% error. And 
the high error (> 1%) happened 256 times out of 2127 samples (sampled 
every 3 seconds) in 2hrs test.

We tried to change the delay to 100us, things got improved although we 
still saw high error, but it happened much less often (22 times out of 
3623 samples).

So why was 2us used as the delay? Does it make sense to use a larger 
delay, for example, 100us used by our test, or even larger, for example, 
ms granularity?

Thanks,
Yang