❯ uname -a
Linux cachyos-x8664 6.11.4-rt7-1-cachyos-gcc-lazy #1 SMP PREEMPT_DYNAMIC Mon, 21 Oct 2024 20:47:32 +0000 x86_64 GNU/Linux
Check it out at CachyMod.
Some insights into PREEMPT_LAZY at The long road to lazy preemption
❯ uname -a
Linux cachyos-x8664 6.11.4-rt7-1-cachyos-gcc-lazy #1 SMP PREEMPT_DYNAMIC Mon, 21 Oct 2024 20:47:32 +0000 x86_64 GNU/Linux
Check it out at CachyMod.
Some insights into PREEMPT_LAZY at The long road to lazy preemption
I’ve tried out PREEMPT_LAZY and generally it sucks at responsiveness under load. I may revisit it in 6.12 but if the results are the same as 6.11, we won’t consider it.
I’ve tried it out again in 6.12. It performs well and may even be a contender for the default option in our default kernel. I had a bad experience with it in 6.11 due to me not cherrypicking all the patches necessary, i.e. I’ve only grabbed patches from [PATCH 0/7] Lazy preemption bits and not clearmod/patches/sched_eevdf_allow_shorter_slices_to_wakeup-preempt.patch at main · marioroy/clearmod · GitHub which is needed for EEVDF to actually preempt tasks.
In case there are misunderstandings, the CachyMod project is great at enabling users to use PREEMPT_LAZY on 6.11 kernel. The 2nd patch I mentioned above is already upstreamed in 6.12 and I had a great experience with lazy here, which was not the case when I tried out PREEMPT_LAZY in 6.11 for myself.