RE: [PATCH v4 1/6] x86,sched: Add support for frequency invariance

[Doug Smythies]

On 2019.11.25 08:06 Giovanni Gherdovich wrote:
> On Sat, 2019-11-23 at 23:49 -0800, Doug Smythies wrote:
>> ...
>> Kernel 5.4-rc8 + revert, intel_cpufreq/schedutil: 3899 seconds
>> Kernel 5.4-rc8 + gg 6 + revert, intel_cpufreq/schedutil: 2740.7 seconds
>> Ratio: 0.70 (as expected)
>> Kernel 5.4-rc8, intel_cpufreq/schedutil: 2334.7 seconds (faster than expected)
>> Kernel 5.4-rc8 + gg 6 patch set, intel_cpufreq/schedutil: 2275.0 seconds (faster than expected)
>> Ratio: 0.97 (not as expected)
>> Kernel 5.4-rc8, intel_cpufreq/performance: 2215.3 seconds
>> Kernel 5.4-rc8, intel_cpufreq/ondemand: 3286.3 seconds
>> Re-stated from previous e-mail:
>> Kernel 5.3-rc8, intel_cpufreq/schedutil: ratio: 0.69 (I don't have the original times)
>
> Hello Doug,
>
> schedutil in 5.4 going a lot faster than in 5.3 would be a surprise. I'm
> running that same test too to check if I can see it as well.

Great, thanks. But see below.

>
> Besides, as it's already been said this patchset adds frequency
> scale-invariance to scheduler metrics such as load and utilization and that's
> useful also in areas other than frequency scaling (most notably the scheduler
> load balancer).

The issue with the schedutil governor not working properly in the 5.4 RC series
appears to be hardware dependant.

My test computer is Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz., Sandy Bridge.
On a temporary basis, I acquired a computer with an
Intel(R) Core(TM) i5-4460 CPU @ 3.20GHz, Haswell,
and schedutil governor behaviour with the exact same kernels is fine:

That "gitsource" test, "make test" 6 times, first run thrown out:

Kernel 5.4 intel_cpufreq/schedutil: 3411.8 seconds
Kernel 5.4 + gg 6 intel_cpufreq/schedutil: 1696.7 seconds
Ratio: 0.49
Recall you got a ratio of 0.49 with 5th generation, Broadwell.

... Doug