Recent Linux Kernel Patch

[sepp256]

I read this highlight on slashdot today about how a recent kernel patch will have a huge difference in the scheduler performance, and I was wondering if this would make a big difference in performance for games or emulators running on the Pandora. Surely it will at least make the UI more responsive when doing simple things like browsing the web.

http://linux.slashdot.org/story/10/11/16/1330233/The-200-Line-Linux-Kernel-Patch-That-Does-Wonders

http://www.phoronix.com/scan.php?page=article&item=linux_2637_video&num=1

Any thoughts?

 

[sebt3]

I read this highlight on slashdot today about how a recent kernel patch will have a huge difference in the scheduler performance, and I was wondering if this would make a big difference in performance for games or emulators running on the Pandora.

Disclaimer : I'm no kernel developper nor a scheduler specialist,
but my understanding is that this patch improve the responsivness of the whole by grouping some process together, this group is now scheduled as a whole thus limiting the CPU available to each of these process and then indirectly giving this CPU pack to the rest of the system.

So psx4pandora will have more CPU power if you play on your pandora while compiling the kernel with -j6 than without (hint in this scenario, psx4pandora will still be too slow to be playable)
In the end of the day, you still want your pandora away of "server load" while playing.
In the best case scenario, this patch will have no impact on your game performance. In the worst, your game get grouped with an other process, and then it finally get less time on CPU, but XFCE get smoother (hided by the fact your app is fullscreen anyway)

Surely it will at least make the UI more responsive when doing simple things like browsing the web.

Err, Chromium get more cpu cycle by using a multi-process design. It would loose this advantage if you group back these process in the scheduler imho

 

[Prometheus]

Just a quick note for reference's sake: There's already a thread on this in the Developers' Corner.

 

[sepp256]

but my understanding is that this patch improve the responsivness of the whole by grouping some process together, this group is now scheduled as a whole thus limiting the CPU available to each of these process and then indirectly giving this CPU pack to the rest of the system.

Well, I guess I am just speculating, but I recall that there was a 'background process killer' app that was recommended to be run prior to running emulators in order to kill unnecessary things like wifi/etc so that it wouldn't cause hiccups during emulation. Would this patch eliminate the need for such a background app killer if it is reducing the maximum latency for task switching by an order of magnitude (as stated in the article)? Would the UI also be more responsive when you are multitasking?

 

[sepp256]

Just a quick note for reference's sake: There's already a thread on this in the Developers' Corner.

Sorry about that. I only looked through the General talk forum before posting...

 

[Prometheus]

^ That's fine, it's not a problem. I just thought it would be worth providing the link so that the other postings about it aren't overlooked.