Hi,
On 2022-08-27 18:02:40 -0700, Andres Freund wrote:
> FWIW, I did notice that netbsd does have working unnamed semaphores. I don't
> know how long ago they were added, but they apparently didn't work quite right
> in 2018 [1]. No meaningful performance chance in the main regression tests,
> I'll run a concurrent check world comparison in the background...
Unnamed ones are substantially worse unfortunately. On an 8 core netbsd 9.3
VM:
sysv:
real 4m39.777s
user 7m35.534s
sys 7m33.831s
unnamed posix
real 5m44.035s
user 7m23.326s
sys 11m58.946s
The difference in system time is even more substantial than the wall clock
time. And repeated runs were even worse.
I also had the ecpg tests hang in one run with unnamed posix semas, until I
killed 'alloc'. Didn't reproduce since though.
So clearly we shouldn't go and start auto-detecting unnamed posix sema
support.
Greetings,
Andres Freund