Linux 6.12 To Drop Old Code That Slows Down CPU Frequency Polling

[phoronix]

Phoronix: Linux 6.12 To Drop Old Code That Slows Down CPU Frequency Polling

The Linux 6.12 kernel cycle later this year has a change coming that will impact users of the "Schedutil" CPU frequency scaling governor. This change is dropping the "LATENCY_MULTIPLIER" that has been within the kernel code the past two decades to slowdown how frequent the CPU frequency evaluation occurs. In turn the revised logic can allow for that CPUFreq frequency re-evaluation to occur more often...

https://www.phoronix.com/news/Linux-Dropping-Latency-Multi

[rene]

but, did anyone benchmark that?

[emansom]

but, did anyone benchmark that?

They knew beforehand you would.

[rene]

They knew beforehand you would.

did not really made a difference on my Ryzen 7950x re-compiling the linux kernel - If I had better numbers I would have said so. Anyone else? Maybe at least for interactive loads? ;-) In the good old days people includes benchmark results in their patch proposals ;-)

[NotMine999]

did not really made a difference on my Ryzen 7950x re-compiling the linux kernel - If I had better numbers I would have said so. Anyone else? Maybe at least for interactive loads? ;-) In the good old days people includes benchmark results in their patch proposals ;-)

You forget that we now live in the world of DEVOPS and Meta - "Move fast & break stuff"

[MorrisS.]

Optimization is welcome always. I don't know if it would have been possible to get it earlier.

[WileEPyote]

Awesome. Off to recompile mainline.

[MastaG]

6.12 will be a very nice kernel with all of its performance improvements across the stack.

The only remaining things on my wishlist would be:
- RT support
- Winesync/ntsync (forgot the name) fully merged

[mrg666]

I backported the patch to 6.10.3 and benchmarked kernel build with an i7-13700, 64 GB RAM. It seems there is an insignificant improvement with the patch applied.

no patch 6.10.3
real    6m45.433s
user    87m3.404s
sys     12m30.118s

with patch 6.10.3
real    6m45.898s
user    87m2.458s
sys     12m23.791s
​

[PluMGMK]

I'm really confused about this… Does this patch increase the polling frequency? How does that improve performance?
Why did the 1000× factor make sense in the past? Will this kill performance on older CPUs?