Intel P-State Driver Ready To Take On Alder Lake Hybrid Processors

[phoronix]

Phoronix: Intel P-State Driver Ready To Take On Alder Lake Hybrid Processors

Headlining the power management updates for the Linux 5.14 merge window is the Intel P-State CPU frequency scaling driver now being adapted to handle hybrid processors...

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.14-Power-Management

[Mitch]

Hybrid x86 processors have me so intrigued and excited. Especially when you throw Hyperthreading and Turbo Boosting into the mix. A scheduler has so much to take into effect. I'd love to see a visual of what the scheduler is doing against various workloads with all the above technologies combined.

At the same time, I really want to try an Alter Lake laptop and see how much one could stretch the battery. Think about all the little bits running all the time on your machine and how much better your battery life could be with all those background tasks. For me, I have a lot of sync software plus little graphs on my Latte Dock. I'm also a tab monster when I browse so PowerTop often gives me a hard time about that too.

Once hybrid processors get to x86, fun exercises might be:
1: Test performance, cooling, and battery with all little cores disabled.
2: Test performance, cooling, and battery with all big cores disabled.

Atom processors have always had incredible power efficiency from what I remember. I had a second-gen Atom netbook that could game over road trip, browse the web a lot, then game all the way back home on a single charge. This was Sims 2, which maxed the system out but was playable.

[skeevy420]

Hybrid x86 processors have me so intrigued and excited. Especially when you throw Hyperthreading and Turbo Boosting into the mix. A scheduler has so much to take into effect. I'd love to see a visual of what the scheduler is doing against various workloads with all the above technologies combined.

Don't know about the scheduler itself, but this is a visualization of how the programmer making the scheduler feels.

[Mitch]

Don't know about the scheduler itself, but this is a visualization of how the programmer making the scheduler feels.

Oh gosh yes! What I'd give just to be a fly on the wall in a scheduling discussion / review. I do some odd programming now and then. There are far more skilled programmers out there than me. I'd like to think a good approach would be to expose tunable parameters with good defaults. That way, it's easier to test and apply various optimizations without having to change code.

Thinking they could have migration thresholds (e.g. if little core at 80% util, promote task to a vacant bigger core). Different parameters may trade performance, power-efficiency, and thermals of various various workloads depending on their settings. Hopefully, they document this so that a layperson like me can understand. I haven't been this interested in new hardware in so long.

*Sidenote: skeevy, I just noticed you and I made our accounts the same month and year.

[skeevy420]

Oh gosh yes! What I'd give just to be a fly on the wall in a scheduling discussion / review. I do some odd programming now and then. There are far more skilled programmers out there than me. I'd like to think a good approach would be to expose tunable parameters with good defaults. That way, it's easier to test and apply various optimizations without having to change code.

Thinking they could have migration thresholds (e.g. if little core at 80% util, promote task to a vacant bigger core). Different parameters may trade performance, power-efficiency, and thermals of various various workloads depending on their settings. Hopefully, they document this so that a layperson like me can understand. I haven't been this interested in new hardware in so long.

*Sidenote: skeevy, I just noticed you and I made our accounts the same month and year.

29 Posts vs 4432 Posts.

I need a hobby.

[intelfx]

29 Posts vs 4432 Posts.

I need a hobby.

I guess you're looking at it.