Date: Wed, 10 Oct 2018 16:22:46 +0000 From: bugzilla-noreply@freebsd.org To: bugs@FreeBSD.org Subject: [Bug 232153] Processes without explicit cpuset only use half the SMP threads (on a Ryzen) Message-ID: <bug-232153-227@https.bugs.freebsd.org/bugzilla/>
next in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D232153 Bug ID: 232153 Summary: Processes without explicit cpuset only use half the SMP threads (on a Ryzen) Product: Base System Version: CURRENT Hardware: amd64 OS: Any Status: New Severity: Affects Only Me Priority: --- Component: kern Assignee: bugs@FreeBSD.org Reporter: greg@unrelenting.technology (AMD Ryzen 7 1700: 1 package(s) x 2 cache groups x 4 core(s) x 2 hardware threads) Somewhere in the 12-ALPHA cycle, the following started happening: All processes are only being scheduled on half the hardware threads =E2=80= =94 either CPUs 0-7 or 8-15 (rebooting can change which half it is). Launching a benchmark or build process with 16 threads only uses 8 CPUs; it= 's clearly visible in sysutils/htop that half the CPUs always sit at 0.0% with occasional 0.5%-1% tasks while the other half is constantly getting at least 1-3% utilization for the random background activity on the system, etc. etc. But launching a process explicitly with 'cpuset -l0-15 some_command' DOES u= se all 16 CPUs just fine!! It's just the default cpuset that's avoiding half of the CPU. But cpuset(1) does not report anything unusual, the default cpusets include all 16 CPUs: ~ % cpuset -s 0 -g=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20 =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20 cpuset 0 mask: 0= , 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 cpuset 0 domain policy: first-touch mask: 0 ~ % cpuset -s 1 -g=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20 =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20 cpuset 1 mask: 0= , 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 cpuset 1 domain policy: first-touch mask: 0 ~ % cpuset -s 2 -g=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20 cpuset 2 mask: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 cpuset 2 domain policy: interleave mask: 0 --=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-232153-227>