Many AMD CPU Feature Additions Land In Linux 6.13

[phoronix]

Phoronix: Many AMD CPU Feature Additions Land In Linux 6.13

The in-development Linux 6.13 kernel is bringing a lot of exciting improvements for AMD Linux customers...

https://www.phoronix.com/news/Linux-6.13-AMD-CPU

[wangi]

Would there be any benefit in turning on AMD P-State for the third gen (Milan) EPYC CPUs?

[David Huang]

It's really frustrating to see AMD kept shipping code that's not gonna actually work at all.

The 3D V-cache optimizer driver requires x86_sched_itmt_flags() to be returned in x86_die_flags() to function (i.e. let the kernel prioritize threads to 3D V-cache die), otherwise the core prioritizing scheduling would only gonna work within a single CCD rather than selecting the preferred CCD first, then selecting the highest performance cores. 7950X3D definitely needs that, it's also the whole point of this driver. But that flag is only enabled for heterogeneous configurations like Strix Point because AMD added a feature check for X86_FEATURE_AMD_HETEROGENEOUS_CORES there before returning the flag in x86_die_flags().

The whole preferred core scheduling thing on Linux is just a few AMD kernel engineers following Intel's implementation without a clue how they worked, and how to test them to ensure they work as expected.

[pWe00Iri3e7Z9lHOX2Qx]

It's really frustrating to see AMD kept shipping code that's not gonna actually work at all.

The 3D V-cache optimizer driver requires x86_sched_itmt_flags() to be returned in x86_die_flags() to function (i.e. let the kernel prioritize threads to 3D V-cache die), otherwise the core prioritizing scheduling would only gonna work within a single CCD rather than selecting the preferred CCD first, then selecting the highest performance cores. 7950X3D definitely needs that, it's also the whole point of this driver. But that flag is only enabled for heterogeneous configurations like Strix Point because AMD added a feature check for X86_FEATURE_AMD_HETEROGENEOUS_CORES there before returning the flag in x86_die_flags().

The whole preferred core scheduling thing on Linux is just a few AMD kernel engineers following Intel's implementation without a clue how they worked, and how to test them to ensure they work as expected.

Oof. Thanks for the heads up. I need a sad panda emoji for my 7900X3D.

[David Huang]

Oof. Thanks for the heads up. I need a sad panda emoji for my 7900X3D.

It's a relatively easy 1-line patch for users that are willing to build their own kernel though. Besides, setting WINE_CPU_TOPOLOGY in Proton can do a better job than letting the kernel decide where to schedule the threads, so it's not really a huge problem. It's just disappointing to see AMD missing the last steps to get something working out of the box.

[halo5]

It's really frustrating to see AMD kept shipping code that's not gonna actually work at all.

I just created a simple bash script for running stuff on the X3D cores (mostly CFD simulations):

x3d:
#!/bin/bash
taskset -c 0-7,16-23 $1

Works in a pinch...