AMD Zen1/Zen2/Zen3 PowerCap RAPL Support Queued For Linux 5.11

[phoronix]

Phoronix: AMD Zen1/Zen2/Zen3 PowerCap RAPL Support Queued For Linux 5.11

The work reported on back in October for RAPL PowerCap patches for AMD Zen CPUs from Zen 1 through Zen 3 are set to arrive with Linux 5.11 in early 2021...

http://www.phoronix.com/scan.php?pag...rCap-RAPL-5.11

[birdie]

Along related lines and for those not aware, Linux 5.8 added the "AMD_Energy" driver for exposing CPU energy information to user-space via the hardware monitoring (HWMON) interfaces. Though at the moment that driver does not yet support Zen 3 processors.

Only supports Epyc CPUs.

This work led by Google engineers allow the AMD Family 17h/19h RAPL (Runtime Average Power Limiting) support within the Linux kernel's power-capping framework in the PowerCap driver.

Meanwhile Linux still doesn't properly support advanced power features of Zen 2 in any shape or form: https://patchwork.kernel.org/[email protected]/

So much for "excellent" open source support for its products. And we already have Zen 3 CPUs.

[CochainComplex]

Only supports Epyc CPUs.



Meanwhile Linux still doesn't properly support advanced power features of Zen 2 in any shape or form: https://patchwork.kernel.org/[email protected]/

So much for "excellent" open source support for its products. And we already have Zen 3 CPUs.

could you elaborate your statement a bit? I'm not much into advanced power features so I don't know which ones are missing.

[Volta]

Only supports Epyc CPUs.



Meanwhile Linux still doesn't properly support advanced power features of Zen 2 in any shape or form: https://patchwork.kernel.org/[email protected]/

So much for "excellent" open source support for its products. And we already have Zen 3 CPUs.

Mental illness. That's it. It's rather clear from the link an idiot provided it's completely something else than the idiot tries to insinuate:

So in general I think it is a huge mistake to expose all that to userspace. Before you know it, there's tools that actually rely on it, and then inhibit the kernel from doing anything sane with it.

> Profiles will be documented in the performance/optimization guides. I don't think userspace can really do anything sane with this; it lacks much if not all useful information.
> Note:
> * AMD systems will not have a policy applied in the kernel at this time.

And why the heck not? We're trying to move all cpufreq into the scheduler and have only a single governor, namely schedutil -- yes, we're still stuck with legacy, and we're still working on performance parity in some cases, but I really hope to get rid of all other cpufreq governors eventually. And if you look at schedutil (schedutil_cpu_util in specific) then you'll see it is already prepared for CPPC and currently only held back by the generic cpufreq interface. It currently only sets desired freq, it has information for min/guaranteed, and once we get thermal intergrated we might have sensible data for max freq too.

> TODO: > * Create a linux userspace tool that will help users generate a CPPC profile
> for their target workload.

Basically a big fat NAK for this approach to cpufreq.

> * Create a general CPPC policy in the kernel.

We already have that, sorta.

[pal666]

Meanwhile Linux still doesn't properly support advanced power features of Zen 2 in any shape or form: https://patchwork.kernel.org/[email protected]/

So much for "excellent" open source support for its products. And we already have Zen 3 CPUs.

moron, could you link to windows source adding support for so called "advanced power features of zen 2"?

[hotaru]

PowerCap RAPL

isn't that the latest Intel security vulnerability? why would anyone want that on AMD?

[andre30correia]

they wil miss the next lts kernel? Amd need to start to hired more people for their driver support

[birdie]

I'm so glad I don't see messages from Volta. Never seen such an aggressively illiterate person boasting about his lack of intellect, indesire to learn anything and lacking basic respect towards others.

1. The driver Michael referrers to only supports AMD Epyc CPUs.
2. The driver which I mentioned allows for instance to set any frequencies in the supported range: from 2200Mhz to the boost frequency, while the standard ACPI driver for AMD CPUs only allows to set very few frequencies, e.g. for my Ryzen 7 3700X CPU these are 3600, 2800 and 2200MHz, see: https://bugzilla.kernel.org/show_bug.cgi?id=206307

Unlike belligerent absolute pieces of shit on Phoronix who want to believe they are "open source fans" I normally know what I'm talking about. In real life however they are too often just fanatics who make Linux look like an elitist OS for the chosen.

[Volta]

I'm so glad I don't see messages from Volta. Never seen such an aggressively illiterate person boasting about his lack of intellect, indesire to learn anything and lacking basic respect towards others.

I'm glad I'm seeing messages from birdie. I've never seen such sick shit who compromises himself whenever he talks. The above is great example of this. The clueless moron.

[polarathene]

Yo, can you lot tone it down? Nobody cares about your opinions of other users here. Just talk about the actual article and leave the resentment to private messages if you must.

I'm so glad I don't see messages from Volta.

What context in this thread gave you any information that Volta had posted a comment then?

isn't that the latest Intel security vulnerability? why would anyone want that on AMD?

That issue was due to permissions to RAPL being accessible by any user to get access to metrics. Like you would for monitoring other parts of your system, someone just found a way to leverage that for an exploit that for some attack I recall required a few hundred hours to pull off.

One attack was rather specific that was performed on hardware that wouldn't usually be paired with the software that was exploited, and which has since been patched since Feb this year. You probably shouldn't use a web browser in that case, since JS could be running in the background to exploit some vulnerability not yet known and compromise your system when the stars are perfectly aligned. Human's often have an unpatched vulnerability too, a $5 wrench.