| Message ID | 20240215180708.891086-1-philip.cox@canonical.com |
|---|---|
| Headers | show |
| Series | performance: Scheduler: Improve cache bouncing on tg->load_avg | expand |
On 15.02.24 19:07, Philip Cox wrote: > > BugLink: BugLink: https://bugs.launchpad.net/bugs/2053251 > > SRU Justification: > > [Impact] > On high core count systems, the updating of the load average can happen so > often that it impacts performance. > > [Fix] > Reduce the updating of the load average to at most once per millisecond. > > > [Test Plan] > I have tested this patch, as has Intel. > > [Where problems could occur] > The risk for regression is low as it is a fairly targetted change, and is just reducing the number > of times that the load average is updated. > > -- > > > Aaron Lu (1): > sched/fair: Ratelimit update to tg->load_avg > > kernel/sched/fair.c | 13 ++++++++++++- > kernel/sched/sched.h | 1 + > 2 files changed, 13 insertions(+), 1 deletion(-) > Again the question: what about Mantic? The fix is from v6.7-rc1. Acked-by: Stefan Bader <stefan.bader@canonical.com>
On Thu, Feb 15, 2024 at 01:07:07PM -0500, Philip Cox wrote: > > BugLink: BugLink: https://bugs.launchpad.net/bugs/2053251 > > SRU Justification: > > [Impact] > On high core count systems, the updating of the load average can happen so > often that it impacts performance. > > [Fix] > Reduce the updating of the load average to at most once per millisecond. > > > [Test Plan] > I have tested this patch, as has Intel. > > [Where problems could occur] > The risk for regression is low as it is a fairly targetted change, and is just reducing the number > of times that the load average is updated. > > -- > > > Aaron Lu (1): > sched/fair: Ratelimit update to tg->load_avg > > kernel/sched/fair.c | 13 ++++++++++++- > kernel/sched/sched.h | 1 + > 2 files changed, 13 insertions(+), 1 deletion(-) > > -- > 2.34.1 > > > -- > kernel-team mailing list > kernel-team@lists.ubuntu.com > https://lists.ubuntu.com/mailman/listinfo/kernel-team In addition to Stefan's comment, please include the SRU Justificaiton (Impact, Fix, Test Plan etc.) also in the public Launchpad bug. Acked-by: Manuel Diewald <manuel.diewald@canonical.com>
On 15/02/2024 19:07, Philip Cox wrote: > BugLink: BugLink: https://bugs.launchpad.net/bugs/2053251 > > SRU Justification: > > [Impact] > On high core count systems, the updating of the load average can happen so > often that it impacts performance. > > [Fix] > Reduce the updating of the load average to at most once per millisecond. > > > [Test Plan] > I have tested this patch, as has Intel. > > [Where problems could occur] > The risk for regression is low as it is a fairly targetted change, and is just reducing the number > of times that the load average is updated. > > -- > > > Aaron Lu (1): > sched/fair: Ratelimit update to tg->load_avg > > kernel/sched/fair.c | 13 ++++++++++++- > kernel/sched/sched.h | 1 + > 2 files changed, 13 insertions(+), 1 deletion(-) > Applied to jammy master-next branch. Thanks!
On Thu, 2024-02-15 at 13:07 -0500, Philip Cox wrote: > > BugLink: BugLink: https://bugs.launchpad.net/bugs/2053251 > > SRU Justification: > > [Impact] > On high core count systems, the updating of the load average can > happen so > often that it impacts performance. > > [Fix] > Reduce the updating of the load average to at most once per > millisecond. > > > [Test Plan] > I have tested this patch, as has Intel. > > [Where problems could occur] > The risk for regression is low as it is a fairly targetted change, > and is just reducing the number > of times that the load average is updated. > > -- > > > Aaron Lu (1): > sched/fair: Ratelimit update to tg->load_avg > > kernel/sched/fair.c | 13 ++++++++++++- > kernel/sched/sched.h | 1 + > 2 files changed, 13 insertions(+), 1 deletion(-) >