[Bug nptl/33704] RFE: Relax failing pthread_mutex_trylock to load with relaxed MO.

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=33704

--- Comment #10 from Florian Weimer <fweimer at redhat dot com> ---
Eh, bug summary is not correct? There is no difference between relaxed and
acquire MO on x86-64.

The question really is whether to do a relaxed MO load first, which is totally
redundant if the trylock is going to succeed. I think it will increase latency
for the success case by quite a bit because the cache line is first shared, and
then goes to exclusive for the CAS. With the current implementation, the CAS
goes to exclusive immediately on success.

-- 
You are receiving this mail because:
You are on the CC list for the bug.