Pavel Begunkov

One-off hack would be terrible, so I'd think of a way to submit requests bypassing userspace sync, even if it's relatively slow. So, there are two parts to it: 1....

Hey, we've slightly changed poll updates, now there are IORING_OP_POLL_REMOVE command. Makes much more sense, doesn't require file/fd and will even have less overhead. This won't work with latest 5.13...

> Ok, so let me know when I can test them ("This won't work with latest 5.13 branch", so what should I use?) :-) Apologies, I wasn't clear enough. The...

I can fix them up myself if you want, just wanted to make it explicit about this ABI change.

> Ok, I started having a look at this - updated to kernel > b2f284ce46c86e9fa309cd3565df8d8599ce5b6b > and validated it was running: > Linux ip-172-31-25-161 5.12.0-rc7-5.13-uring-210416 #1 SMP Fri Apr 16...

@mohsenomidi, I always had problems with that test. Apart from not testing anything useful, the numbers are least strange... Just run it on my laptop, 5.14. The number of iteration...

p.s. the second result (42 vs 2) looks weird, may be the test being buggy.

> > I have tried using register buffers and register files, but they don't seem to help much with performance. As well as enabling the polling mode, but that only...

> I have tried using register buffers and register files, but they don't seem to help much with performance. As well as enabling the polling mode, but that only seems...

@MarcoF1, it needs to be investigated, but my guess is that you don't do enough of batching. I.e. submitting several requests at once, and if I'm reading your code correctly...