sched/fair: Clear SMT siblings after determining the core is not idle
authorMel Gorman <mgorman@techsingularity.net>
Mon, 30 Nov 2020 14:40:20 +0000 (14:40 +0000)
committerIngo Molnar <mingo@kernel.org>
Fri, 11 Dec 2020 09:30:38 +0000 (10:30 +0100)
The clearing of SMT siblings from the SIS mask before checking for an idle
core is a small but unnecessary cost. Defer the clearing of the siblings
until the scan moves to the next potential target. The cost of this was
not measured as it is borderline noise but it should be self-evident.

Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20201130144020.GS3371@techsingularity.net
kernel/sched/fair.c

index f5dcedacc104342b10761dd0c2049ff2969a08bd..efac224c703fd143009cf0e69eebab70f7a7bf4a 100644 (file)
@@ -6086,10 +6086,11 @@ static int select_idle_core(struct task_struct *p, struct sched_domain *sd, int
                                break;
                        }
                }
-               cpumask_andnot(cpus, cpus, cpu_smt_mask(core));
 
                if (idle)
                        return core;
+
+               cpumask_andnot(cpus, cpus, cpu_smt_mask(core));
        }
 
        /*