Date: Thu, 13 Apr 2023 20:15:10 +0000 From: bugzilla-noreply@freebsd.org To: bugs@FreeBSD.org Subject: [Bug 270785] Performance and power efficiency regression due to pthread_cond_timedwait() changes Message-ID: <bug-270785-227-Bn6LeZbspD@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-270785-227@https.bugs.freebsd.org/bugzilla/> References: <bug-270785-227@https.bugs.freebsd.org/bugzilla/>
next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D270785 --- Comment #4 from bllcha013@myuct.ac.za --- (In reply to Mark Johnston from comment #3) I cloned releng/13.2 and reverted commit e21c77f80c3b1cf2bd0a9f874e15e7a8e49f0dba. Reverting this commit fixes the regression. I also did the dtrace output as asked - thank you for providing the dtrace script. releng/13.2 (standard, same as RELEASE): syscall seconds calls errors __sysctlbyname 0.000650415 6 0 getrusage 0.008321625 116 0 sched_yield 1.287887942 29936 0 _umtx_op 41.138906134 30136 30132 ------------- ------- ------- 42.435766116 60194 30132 dtrace: description 'fbt::do_wait:entry ' matched 1 probe CPU ID FUNCTION:NAME 2 18111 do_wait:entry struct _umtx_time { struct timespec _timeout =3D { time_t tv_sec =3D 0xc2 long tv_nsec =3D 0x2ed64521 } __uint32_t _flags =3D 0x1 __uint32_t _clockid =3D 0x4 } 2 18111 do_wait:entry struct _umtx_time { struct timespec _timeout =3D { time_t tv_sec =3D 0xc2 long tv_nsec =3D 0x2ed7d390 } __uint32_t _flags =3D 0x1 __uint32_t _clockid =3D 0x4 } releng/13.2 (with commit e21c77f80c3b1cf2bd0a9f874e15e7a8e49f0dba reverted): syscall seconds calls errors getrusage 0.003569531 58 0 sched_yield 0.168773634 3781 0 _umtx_op 16.840977734 3873 3873 ------------- ------- ------- 17.013320899 7712 3873 dtrace: description 'fbt::do_wait:entry ' matched 1 probe CPU ID FUNCTION:NAME 4 18111 do_wait:entry struct _umtx_time { struct timespec _timeout =3D { time_t tv_sec =3D 0x6f long tv_nsec =3D 0xba6df4a } __uint32_t _flags =3D 0x1 __uint32_t _clockid =3D 0x4 } 3 18111 do_wait:entry struct _umtx_time { struct timespec _timeout =3D { time_t tv_sec =3D 0x6f long tv_nsec =3D 0x8bb061f } __uint32_t _flags =3D 0x1 __uint32_t _clockid =3D 0x4 } So it seems that reverting the commit fixes the issue. It also helps the Un= iFi Controller as well (although that wasn't impacted as drastically as Minecra= ft). This might be a kernel bug as you mentioned? --=20 You are receiving this mail because: You are the assignee for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-270785-227-Bn6LeZbspD>