[AMD Official Use Only - General]
Hi Rene:
> -----Original Message-----
> From: Rene Rebe <[email protected]>
> Sent: Tuesday, August 29, 2023 5:35 PM
> To: Meng, Li (Jassmine) <[email protected]>
> Cc: Huang, Ray <[email protected]>; [email protected]; linux-
> [email protected]; [email protected]; [email protected]; Shuah
> Khan <[email protected]>; [email protected];
> Fontenot, Nathan <[email protected]>; Sharma, Deepak
> <[email protected]>; Deucher, Alexander
> <[email protected]>; Limonciello, Mario
> <[email protected]>; Huang, Shimmer
> <[email protected]>; Yuan, Perry <[email protected]>; Du,
> Xiaojian <[email protected]>; Viresh Kumar <[email protected]>;
> Borislav Petkov <[email protected]>; Meng, Li (Jassmine) <[email protected]>;
> Karny, Wyes <[email protected]>
> Subject: Re: [PATCH V4 3/7] cpufreq: amd-pstate: Enable AMD Pstate
> Preferred Core Supporting.
>
> Caution: This message originated from an External Source. Use proper
> caution when opening attachments, clicking links, or responding.
>
>
> Dear Meng Li and team,
>
> thank you so much for working on finally bringing AMD preferred core
> scheduling to mainline Linux!
>
> > The initial core rankings are set up by AMD Pstate when the system
> > boots.
>
> I tested this patch on our Ryzen 7950x and 5950x systems and could
> unfortunatlely not find any performance differences. I therefore took a
> closer look and as far as I can tell the conditional for the initial preferred
> performance priorities appears to be reversed. I marked them down below. I
> also attached a patch for the fix. With that fixed I can measure a 0.7%
> improvement compiling Firefox on 7950x. I wonder slightly how this ever past
> testing before, ...
>
> I think it would be a good idea to always expose the hw perf values in sysfs to
> help users debugging hardware issues or BIOS settings even with percore not
> enabled and therefore not using the unused 166 or 255 values anyway.
>
[Meng, Li (Jassmine)]
I will add a new sysfs attribute for the highest performance of preferred core.
> With that fixed, however, Linux is still not always scheduling to preferred
> cores, but that appears to be an independant limitation of the current linux
> scheduler not strictly using the priority for scheduling, yet. With manual
> taskset guidance I could further improve the Firefox build time by some
> more seconds to over 1% overall performance improvement, if the linux
> scheudler would more reliably schedule minute long running rust lto link
> tasks to the preferred cores and not some mediocre ones.
>
>
> > - highest_perf = amd_get_highest_perf();
> > - if (highest_perf > AMD_CPPC_HIGHEST_PERF(cap1))
> > - highest_perf = AMD_CPPC_HIGHEST_PERF(cap1);
> > -
> > - WRITE_ONCE(cpudata->highest_perf, highest_perf);
> > + if (prefcore)
> > + WRITE_ONCE(cpudata->highest_perf,
> AMD_PSTATE_PREFCORE_THRESHOLD);
> > + else
> > + WRITE_ONCE(cpudata->highest_perf,
> > + AMD_CPPC_HIGHEST_PERF(cap1));
>
> Conditional reversed, assigns THRESHOLD if enabled!
[Meng, Li (Jassmine)]
The highest_perf is used to calculate the frequency value.
The values between 166 and 255 represent priority, not the highest perf.
>
> > WRITE_ONCE(cpudata->nominal_perf,
> AMD_CPPC_NOMINAL_PERF(cap1));
> > WRITE_ONCE(cpudata->lowest_nonlinear_perf,
> > AMD_CPPC_LOWNONLIN_PERF(cap1)); @@ -318,17 +322,15 @@ static int
> > pstate_init_perf(struct amd_cpudata *cpudata) static int
> > cppc_init_perf(struct amd_cpudata *cpudata) {
> > struct cppc_perf_caps cppc_perf;
> > - u32 highest_perf;
> >
> > int ret = cppc_get_perf_caps(cpudata->cpu, &cppc_perf);
> > if (ret)
> > return ret;
> >
> > - highest_perf = amd_get_highest_perf();
> > - if (highest_perf > cppc_perf.highest_perf)
> > - highest_perf = cppc_perf.highest_perf;
> > -
> > - WRITE_ONCE(cpudata->highest_perf, highest_perf);
> > + if (prefcore)
> > + WRITE_ONCE(cpudata->highest_perf,
> AMD_PSTATE_PREFCORE_THRESHOLD);
> > + else
> > + WRITE_ONCE(cpudata->highest_perf,
> > + cppc_perf.highest_perf);
>
> Same here. Not using highest_perf if enabled, ...
>
> Signed-off-by: Ren? Rebe <[email protected]>
>
> --- linux-6.4/drivers/cpufreq/amd-pstate.c.vanilla 2023-08-25
> 22:34:25.254995690 +0200
> +++ linux-6.4/drivers/cpufreq/amd-pstate.c 2023-08-25
> 22:35:49.194991446 +0200
> @@ -282,9 +282,9 @@
> * the default max perf.
> */
> if (prefcore)
> - WRITE_ONCE(cpudata->highest_perf,
> AMD_PSTATE_PREFCORE_THRESHOLD);
> - else
> WRITE_ONCE(cpudata->highest_perf,
> AMD_CPPC_HIGHEST_PERF(cap1));
> + else
> + WRITE_ONCE(cpudata->highest_perf,
> + AMD_PSTATE_PREFCORE_THRESHOLD);
>
> WRITE_ONCE(cpudata->nominal_perf,
> AMD_CPPC_NOMINAL_PERF(cap1));
> WRITE_ONCE(cpudata->lowest_nonlinear_perf,
> AMD_CPPC_LOWNONLIN_PERF(cap1)); @@ -303,9 +303,9 @@
> return ret;
>
> if (prefcore)
> - WRITE_ONCE(cpudata->highest_perf,
> AMD_PSTATE_PREFCORE_THRESHOLD);
> - else
> WRITE_ONCE(cpudata->highest_perf, cppc_perf.highest_perf);
> + else
> + WRITE_ONCE(cpudata->highest_perf,
> + AMD_PSTATE_PREFCORE_THRESHOLD);
>
> WRITE_ONCE(cpudata->nominal_perf, cppc_perf.nominal_perf);
> WRITE_ONCE(cpudata->lowest_nonlinear_perf,
>
>
> --
> Ren? Rebe, ExactCODE GmbH, Lietzenburger Str. 42, DE-10789 Berlin
> https://exactcode.com | https://t2sde.org | https://rene.rebe.de