Avi Kivity

@xemul please review

MADV_POPULATE_WRITE is probably better than MADV_COLLAPSE. We could just issue both. 1. If MADV_POPULATE_WRITE is available, issue it, followed by MADV_COLLAPSE which is likely a no-op 2. Otherwise do the...

We don't have a fragmented unordered map (don't think we ever saw a need - to exceed 128k one would need 16k buckets, no? In general I don't want Seastar...

Thanks, I'll update my patch. But unfortunately I still see a 10% slowdown, I don't think my changes did anything.

v2: applies patch from @tchaikov to bump the uring version dependency

Client: ``` ab -k -c 100 -n 1000000 http://localhost:10000/ ``` Server: ``` ./build/release/apps/httpd/httpd --smp 1 --cpuset 0 -m 1G --reactor-backend linux-aio ``` Requests per second: 110300.38 [#/sec] (mean) Server: ```...

The most I was able to see is a large difference in IPC, but I have no idea how that happens.

I did find a difference: ```console [avi@avi seastar (uring-poll-first)]$ sudo perf stat -a -e irq_vectors:reschedule_entry ./build/release/apps/httpd/httpd --smp 1 --cpuset 0 -m 1G --reactor-backend io_uring INFO 2024-02-11 18:18:18,608 seastar - Reactor...

I thought the patch addresses those self-interrupts, but maybe not)

Ah that run was without the patch.