[Bug 270785] Performance and power efficiency regression due to pthread_cond_timedwait() changes

From: <bugzilla-noreply_at_freebsd.org>
Date: Wed, 12 Apr 2023 18:31:56 UTC

--- Comment #1 from Alexander Motin <mav@FreeBSD.org> ---
Depending on what the applicaiton/java actually does, I can easily believe that
behavior could change after the mentioned commit, but it is not necessary an OS
bug, it may be an application bug or a lack of optimization.  Previous code
always waited for at least 2ms in pthread_cond_timedwait(), no matter what
timeout is requested, while after the patch it may wake up much sooner, just in
few microseconds.  It may be that application really requests wake ups that
often, but previously it was not properly honored.  For better understanding it
would be good to collect information about the actual system calls and their
arguments before and after, either with truss or may be dtrace, etc.

You are receiving this mail because:
You are the assignee for the bug.