On 5/28/24 8:34 AM, Jens Axboe wrote: > Obivously some variation in runs in general, but it's most certainly > faster in terms of receiving too. This test case is fixed at doing 100K > messages per second, didn't do any peak testing. But I strongly suspect > you'll see very nice efficiency gains here too, as doing two TWA_SIGNAL > task_work is pretty sucky imho. Example on that same box. Running the same message passer test, sys time used for a 3 second run varies between 300ms and 850ms on the stock kernel (with a lot of variance), after it's around ~150ms and pretty steady. -- Jens Axboe