Hi, hackers!
In the measurements above in the thread, I've been using LIFO wake
queue in a primary lockless patch (and it was attached as the previous
versions of a patch) and an "inverted wake queue" (in faсt FIFO) as
the alternative benchmarking option. I think using the latter is more
fair and natural and provided they show no difference in the speed,
I'd make the main patch using it (attached as v6). No other changes
from v5, though.
Regards,
Pavel.