Scan for an idle sibling in a single pass
cover Scan for an idle sibling in a single pass
This series of 5 patches reposts three patches from Peter entitled "select_idle_sibling() wreckage". It only scans the runqueues in a single pass when searching for an idle sibling.
Two patches from Peter were dropped. The first patch altered how scan depth was calculated. Scan depth deletion is a random number generator with two major limitations. The avg_idle time is based on the time between a CPU going idle and being woken up clamped approximately by 2*sysctl_sched_migration_cost. This is difficult to compare in a sensible fashion to avg_scan_cost. The second issue is that only the avg_scan_cost of scan failures is recorded and it does not decay. This requires deeper surgery that would justify a patch on its own although Peter notes that
https://lkml.kernel.org/r/[email protected] is potentially useful for an alternative avg_idle metric.
The second patch dropped converted the idle core scan throttling mechanism to SIS_PROP. While this would unify the throttling of core and CPU scanning, it was not free of regressions and has_idle_cores is a fairly effective throttling mechanism with the caveat that it can have a lot of false positives for workloads like hackbench.
Peter's series tried to solve three problems at once, this subset addresses one problem. As with anything select_idle_sibling, it's a mix of wins and losses but won more than it lost across a range of workloads and machines.
kernel/sched/core.c | 19 +++-- kernel/sched/fair.c | 157 ++++++++++++++++++++-------------------- kernel/sched/features.h | 1 - 3 files changed, 92 insertions(+), 85 deletions(-)
Signature · 1 line
2.26.2
PATCH 2/5 sched/fair: Move avg_scan_cost calculations under SIS_PROP
As noted by Vincent Guittot, avg_scan_costs are calculated for SIS_PROP even if SIS_PROP is disabled. Move the time calculations under a SIS_PROP check and while we are at it, exclude the cost of initialising the CPU mask from the average scan cost.
--- kernel/sched/fair.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-)
kernel/sched/fair.c +8 -6
@@ -6153,6 +6153,8 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tif (!this_sd)return -1 ; +cpumask_and (cpus,sched_domain_span (sd), p->cpus_ptr); +if (sched_feat (SIS_PROP )) {u64 avg_cost, avg_idle, span_avg;
@@ -6168,11 +6170,9 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t nr =div_u64 (span_avg, avg_cost); else nr =4 ; - } - - time =cpu_clock (this); -cpumask_and (cpus,sched_domain_span (sd), p->cpus_ptr); + time =cpu_clock (this); + }for_each_cpu_wrap (cpu, cpus, target) {if (!--nr)
@@ -6181,8 +6181,10 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tbreak ; } - time =cpu_clock (this) - time; -update_avg (&this_sd->avg_scan_cost, time); +if (sched_feat (SIS_PROP )) { + time =cpu_clock (this) - time; +update_avg (&this_sd->avg_scan_cost, time); + }return cpu; }
Signature · 1 line
2.26.2
PATCH 1/5 sched/fair: Remove SIS_AVG_CPU
In reply to Mel GormanSIS_AVG_CPU was introduced as a means of avoiding a search when the average search cost indicated that the search would likely fail. It was a blunt instrument and disabled by commit 4c77b18cf8b7 ("sched/fair: Make select_idle_cpu() more aggressive") and later replaced with a proportional search depth by commit 1ad3aaf3fcd2 ("sched/core: Implement new approach to scale select_idle_cpu()").
While there are corner cases where SIS_AVG_CPU is better, it has now been disabled for almost three years. As the intent of SIS_PROP is to reduce the time complexity of select_idle_cpu(), lets drop SIS_AVG_CPU and focus on SIS_PROP as a throttling mechanism.
--- kernel/sched/fair.c | 20 +++++++++----------- kernel/sched/features.h | 1 - 2 files changed, 9 insertions(+), 12 deletions(-)
kernel/sched/fair.c +9 -11
@@ -6145,7 +6145,6 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t {struct cpumask *cpus =this_cpu_cpumask_var_ptr (select_idle_mask);struct sched_domain *this_sd; -u64 avg_cost, avg_idle;u64 time;int this =smp_processor_id ();int cpu, nr =INT_MAX ;
@@ -6154,18 +6153,17 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tif (!this_sd)return -1 ; -/* -* Due to large variance we need a large fuzz factor; hackbench in -* particularly is sensitive here. -*/ - avg_idle =this_rq ()->avg_idle /512 ; - avg_cost = this_sd->avg_scan_cost +1 ; +if (sched_feat (SIS_PROP )) { +u64 avg_cost, avg_idle, span_avg; -if (sched_feat (SIS_AVG_CPU ) && avg_idle < avg_cost) -return -1 ; +/* +* Due to large variance we need a large fuzz factor; +* hackbench in particularly is sensitive here. +*/ + avg_idle =this_rq ()->avg_idle /512 ; + avg_cost = this_sd->avg_scan_cost +1 ; -if (sched_feat (SIS_PROP )) { -u64 span_avg = sd->span_weight * avg_idle; + span_avg = sd->span_weight * avg_idle;if (span_avg >4 *avg_cost) nr =div_u64 (span_avg, avg_cost);else
kernel/sched/features.h +0 -1
@@ -54,7 +54,6 @@ SCHED_FEAT(TTWU_QUEUE, true)/* * When doing wakeups, attempt to limit superfluous scans of the LLC domain. */ -SCHED_FEAT(SIS_AVG_CPU, false)SCHED_FEAT (SIS_PROP , true)/*
Signature · 1 line
2.26.2
PATCH 3/5 sched/fair: Make select_idle_cpu() proportional to cores
In reply to Mel GormanFrom: Peter Zijlstra (Intel) <[email protected]>
Instead of calculating how many (logical) CPUs to scan, compute how many cores to scan.
This changes behaviour for anything !SMT2.
--- kernel/sched/core.c | 19 ++++++++++++++----- kernel/sched/fair.c | 12 ++++++++++-- 2 files changed, 24 insertions(+), 7 deletions(-)
kernel/sched/core.c +14 -5
@@ -7444,11 +7444,20 @@ int sched_cpu_activate(unsigned int cpu)balance_push_set (cpu, false);#ifdef CONFIG_SCHED_SMT -/* -* When going up, increment the number of cores with SMT present. -*/ -if (cpumask_weight (cpu_smt_mask (cpu)) ==2 ) -static_branch_inc_cpuslocked (&sched_smt_present); +do { +int weight =cpumask_weight (cpu_smt_mask (cpu)); +extern int sched_smt_weight; + +if (weight > sched_smt_weight) + sched_smt_weight = weight; + +/* +* When going up, increment the number of cores with SMT present. +*/ +if (weight ==2 ) +static_branch_inc_cpuslocked (&sched_smt_present); + + }while (0 );#endif set_cpu_active (cpu, true);
kernel/sched/fair.c +10 -2
@@ -6010,6 +6010,8 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *pDEFINE_STATIC_KEY_FALSE (sched_smt_present);EXPORT_SYMBOL_GPL (sched_smt_present); +int sched_smt_weight__read_mostly =1 ; +static inline void set_idle_cores (int cpu,int val) {struct sched_domain_shared *sds;
@@ -6124,6 +6126,8 @@ static int select_idle_smt(struct task_struct *p, struct sched_domain *sd, int t#else /* CONFIG_SCHED_SMT */ +#define sched_smt_weight 1 +static inline int select_idle_core (struct task_struct *p,struct sched_domain *sd,int target) {return -1 ;
@@ -6136,6 +6140,8 @@ static inline int select_idle_smt(struct task_struct *p, struct sched_domain *sd#endif /* CONFIG_SCHED_SMT */ +#define sis_min_cores 2 +/* * Scan the LLC domain for idle CPUs; this is dynamically regulated by * comparing the average scan cost (tracked in sd->avg_scan_cost) against the
@@ -6166,10 +6172,12 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t avg_cost = this_sd->avg_scan_cost +1 ; span_avg = sd->span_weight * avg_idle; -if (span_avg >4 *avg_cost) +if (span_avg > sis_min_cores*avg_cost) nr =div_u64 (span_avg, avg_cost);else - nr =4 ; + nr = sis_min_cores; + + nr *= sched_smt_weight; time =cpu_clock (this); }
Signature · 1 line
2.26.2
PATCH 4/5 sched/fair: Remove select_idle_smt()
In reply to Mel GormanFrom: Peter Zijlstra (Intel) <[email protected]>
In order to make the next patch more readable, and to quantify the actual effectiveness of this pass, start by removing it.
--- kernel/sched/fair.c | 30 ------------------------------ 1 file changed, 30 deletions(-)
kernel/sched/fair.c +0 -30
@@ -6103,27 +6103,6 @@ static int select_idle_core(struct task_struct *p, struct sched_domain *sd, int return -1; } -/* -* Scan the local SMT mask for idle CPUs. -*/ -static int select_idle_smt (struct task_struct *p,struct sched_domain *sd,int target) -{ -int cpu; - -if (!static_branch_likely (&sched_smt_present)) -return -1 ; - -for_each_cpu (cpu,cpu_smt_mask (target)) { -if (!cpumask_test_cpu (cpu, p->cpus_ptr) || - !cpumask_test_cpu (cpu,sched_domain_span (sd))) -continue ; -if (available_idle_cpu (cpu) ||sched_idle_cpu (cpu)) -return cpu; - } - -return -1 ; -} -#else /* CONFIG_SCHED_SMT */ #define sched_smt_weight 1
@@ -6133,11 +6112,6 @@ static inline int select_idle_core(struct task_struct *p, struct sched_domain *s return -1; } -static inline int select_idle_smt (struct task_struct *p,struct sched_domain *sd,int target) -{ -return -1 ; -} -#endif /* CONFIG_SCHED_SMT */ #define sis_min_cores 2
@@ -6331,10 +6305,6 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target)if ((unsigned )i < nr_cpumask_bits)return i; - i =select_idle_smt (p, sd, target); -if ((unsigned )i < nr_cpumask_bits) -return i; -return target; }
Signature · 1 line
2.26.2
PATCH 5/5 sched/fair: Merge select_idle_core/cpu()
In reply to Mel GormanBoth select_idle_core() and select_idle_cpu() do a loop over the same cpumask. Observe that by clearing the already visited CPUs, we can fold the iteration and iterate a core at a time.
All we need to do is remember any non-idle CPU we encountered while scanning for an idle core. This way we'll only iterate every CPU once.
--- kernel/sched/fair.c | 97 +++++++++++++++++++++++++++------------------ 1 file changed, 59 insertions(+), 38 deletions(-)
kernel/sched/fair.c +59 -38
@@ -6006,6 +6006,14 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *preturn new_cpu; } +static inline int __select_idle_cpu (struct task_struct *p,int core,struct cpumask *cpus,int *idle_cpu) +{ +if (available_idle_cpu (core) ||sched_idle_cpu (core)) +return core; + +return -1 ; +} +#ifdef CONFIG_SCHED_SMT DEFINE_STATIC_KEY_FALSE (sched_smt_present);EXPORT_SYMBOL_GPL (sched_smt_present);
@@ -6066,40 +6074,34 @@ void __update_idle_core(struct rq *rq)* there are no idle cores left in the system; tracked through * sd_llc->shared->has_idle_cores and enabled through update_idle_core() above. */ -static int select_idle_core (struct task_struct *p,struct sched_domain *sd,int target) +static int select_idle_core (struct task_struct *p,int core,struct cpumask *cpus,int *idle_cpu) { -struct cpumask *cpus =this_cpu_cpumask_var_ptr (select_idle_mask); -int core, cpu; +bool idle = true; +int cpu;if (!static_branch_likely (&sched_smt_present)) -return -1 ; - -if (!test_idle_cores (target, false)) -return -1 ; - -cpumask_and (cpus,sched_domain_span (sd), p->cpus_ptr); +return __select_idle_cpu (p, core, cpus, idle_cpu); -for_each_cpu_wrap (core, cpus, target) { -bool idle = true; - -for_each_cpu (cpu,cpu_smt_mask (core)) { -if (!available_idle_cpu (cpu)) { - idle = false; -break ; +for_each_cpu (cpu,cpu_smt_mask (core)) { +if (!available_idle_cpu (cpu)) { + idle = false; +if (*idle_cpu ==-1 ) { +if (sched_idle_cpu (cpu) &&cpumask_test_cpu (cpu, p->cpus_ptr)) { + *idle_cpu = cpu; +break ; + } +continue ; } +break ; } - -if (idle) -return core; - -cpumask_andnot (cpus, cpus,cpu_smt_mask (core)); +if (*idle_cpu ==-1 &&cpumask_test_cpu (cpu, p->cpus_ptr)) + *idle_cpu = cpu; } -/* -* Failed to find an idle core; stop looking for one. -*/ -set_idle_cores (target,0 ); +if (idle) +return core; +cpumask_andnot (cpus, cpus,cpu_smt_mask (core));return -1 ; }
@@ -6107,9 +6109,18 @@ static int select_idle_core(struct task_struct *p, struct sched_domain *sd, int#define sched_smt_weight 1 -static inline int select_idle_core (struct task_struct *p,struct sched_domain *sd,int target) +static inline void set_idle_cores (int cpu,int val) { -return -1 ; +} + +static inline bool test_idle_cores (int cpu,bool def) +{ +return def; +} + +static inline int select_idle_core (struct task_struct *p,int core,struct cpumask *cpus,int *idle_cpu) +{ +return __select_idle_cpu (p, core, cpus, idle_cpu); }#endif /* CONFIG_SCHED_SMT */
@@ -6124,10 +6135,11 @@ static inline int select_idle_core(struct task_struct *p, struct sched_domain *sstatic int select_idle_cpu (struct task_struct *p,struct sched_domain *sd,int target) {struct cpumask *cpus =this_cpu_cpumask_var_ptr (select_idle_mask); +int i, cpu, idle_cpu =-1 , nr =INT_MAX ; +bool smt =test_idle_cores (target, false); +int this =smp_processor_id ();struct sched_domain *this_sd;u64 time; -int this =smp_processor_id (); -int cpu, nr =INT_MAX ; this_sd =rcu_dereference (*this_cpu_ptr (&sd_llc));if (!this_sd)
@@ -6135,7 +6147,7 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tcpumask_and (cpus,sched_domain_span (sd), p->cpus_ptr); -if (sched_feat (SIS_PROP )) { +if (sched_feat (SIS_PROP ) && !smt) {u64 avg_cost, avg_idle, span_avg;/*
@@ -6159,16 +6171,29 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tfor_each_cpu_wrap (cpu, cpus, target) {if (!--nr)return -1 ; -if (available_idle_cpu (cpu) ||sched_idle_cpu (cpu)) -break ; +if (smt) { + i =select_idle_core (p, cpu, cpus, &idle_cpu); +if ((unsigned int )i < nr_cpumask_bits) +return i; + + }else { + i =__select_idle_cpu (p, cpu, cpus, &idle_cpu); +if ((unsigned int )i < nr_cpumask_bits) { + idle_cpu = i; +break ; + } + } } -if (sched_feat (SIS_PROP )) { +if (smt) +set_idle_cores (this, false); + +if (sched_feat (SIS_PROP ) && !smt) { time =cpu_clock (this) - time;update_avg (&this_sd->avg_scan_cost, time); } -return cpu; +return idle_cpu; }/*
@@ -6297,10 +6322,6 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target)if (!sd)return target; - i =select_idle_core (p, sd, target); -if ((unsigned )i < nr_cpumask_bits) -return i; - i =select_idle_cpu (p, sd, target);if ((unsigned )i < nr_cpumask_bits)return i;
Signature · 1 line
2.26.2
Vincent Guittot
maintainerRe: [PATCH 3/5] sched/fair: Make select_idle_cpu() proportional to cores
In reply to Mel GormanQuoted 21 lines · Mel Gorman
From: Peter Zijlstra (Intel) <[email protected]>
Instead of calculating how many (logical) CPUs to scan, compute how many cores to scan.
This changes behaviour for anything !SMT2.
Signed-off-by Peter Zijlstra (Intel) <[email protected]>Signed-off-by Mel Gorman <[email protected]>--- kernel/sched/core.c | 19 ++++++++++++++----- kernel/sched/fair.c | 12 ++++++++++-- 2 files changed, 24 insertions(+), 7 deletions(-)kernel/sched/core.c +0
-0@@ -7444,11 +7444,20 @@ int sched_cpu_activate(unsigned int cpu)balance_push_set (cpu, false);
kernel/sched/core.c int sched_cpu_activate(unsigned int cpu) · line 7449
Quoted 10 lines
kernel/sched/core.c +3
-5#ifdef CONFIG_SCHED_SMT -/* -* When going up, increment the number of cores with SMT present. -*/ -if (cpumask_weight (cpu_smt_mask (cpu)) ==2 ) -static_branch_inc_cpuslocked (&sched_smt_present); + do { + int weight = cpumask_weight(cpu_smt_mask(cpu)); + extern int sched_smt_weight;
coding style problem
Quoted 47 lines
kernel/sched/core.c +11
-0+ +if (weight > sched_smt_weight) + sched_smt_weight = weight; + +/* +* When going up, increment the number of cores with SMT present. +*/ +if (weight ==2 ) + static_branch_inc_cpuslocked(&sched_smt_present); + + }while (0 );#endif set_cpu_active (cpu, true);kernel/sched/fair.c +6
-0@@ -6010,6 +6010,8 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *pDEFINE_STATIC_KEY_FALSE (sched_smt_present);EXPORT_SYMBOL_GPL (sched_smt_present); +int sched_smt_weight__read_mostly =1 ; +static inline void set_idle_cores (int cpu,int val) {struct sched_domain_shared *sds;@@ -6124,6 +6126,8 @@ static int select_idle_smt(struct task_struct *p, struct sched_domain *sd, int t#else /* CONFIG_SCHED_SMT */ +#define sched_smt_weight 1 +static inline int select_idle_core (struct task_struct *p,struct sched_domain *sd,int target) {return -1 ;@@ -6136,6 +6140,8 @@ static inline int select_idle_smt(struct task_struct *p, struct sched_domain *sd#endif /* CONFIG_SCHED_SMT */ +#define sis_min_cores 2 +/* * Scan the LLC domain for idle CPUs; this is dynamically regulated by * comparing the average scan cost (tracked in sd->avg_scan_cost) against the @@ -6166,10 +6172,12 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t avg_cost = this_sd->avg_scan_cost +1 ;
kernel/sched/fair.c static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t · line 6180
Quoted 10 lines
kernel/sched/fair.c +4
-2span_avg = sd->span_weight * avg_idle; -if (span_avg >4 *avg_cost) +if (span_avg > sis_min_cores*avg_cost) nr =div_u64 (span_avg, avg_cost);else - nr =4 ; + nr = sis_min_cores; + + nr *= sched_smt_weight;
Also, patch 5 will look at all CPUs of a core in select_idle_core so nr will decrement by 1 per core so i don't see the need to multiply by sched_smt_weight one patch 5 is applied
Quoted 6 lines
kernel/sched/fair.c +0
-1time =cpu_clock (this); } --2.26.2
Vincent Guittot
maintainerRe: [PATCH 5/5] sched/fair: Merge select_idle_core/cpu()
In reply to Mel GormanQuoted 138 lines · Mel Gorman
Both
select_idle_core()andselect_idle_cpu()do a loop over the same cpumask. Observe that by clearing the already visited CPUs, we can fold the iteration and iterate a core at a time.All we need to do is remember any non-idle CPU we encountered while scanning for an idle core. This way we'll only iterate every CPU once.
Signed-off-by Peter Zijlstra (Intel) <[email protected]>Signed-off-by Mel Gorman <[email protected]>--- kernel/sched/fair.c | 97 +++++++++++++++++++++++++++------------------ 1 file changed, 59 insertions(+), 38 deletions(-)kernel/sched/fair.c +42
-30@@ -6006,6 +6006,14 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *preturn new_cpu; } +static inline int __select_idle_cpu (struct task_struct *p,int core,struct cpumask *cpus,int *idle_cpu) +{ +if (available_idle_cpu (core) ||sched_idle_cpu (core)) +return core; + +return -1 ; +} +#ifdef CONFIG_SCHED_SMT DEFINE_STATIC_KEY_FALSE (sched_smt_present);EXPORT_SYMBOL_GPL (sched_smt_present);@@ -6066,40 +6074,34 @@ void __update_idle_core(struct rq *rq)* there are no idle cores left in the system; tracked through * sd_llc->shared->has_idle_cores and enabled through update_idle_core() above. */ -static int select_idle_core (struct task_struct *p,struct sched_domain *sd,int target) +static int select_idle_core (struct task_struct *p,int core,struct cpumask *cpus,int *idle_cpu) { -struct cpumask *cpus =this_cpu_cpumask_var_ptr (select_idle_mask); -int core, cpu; +bool idle = true; +int cpu;if (!static_branch_likely (&sched_smt_present)) -return -1 ; - -if (!test_idle_cores (target, false)) -return -1 ; - -cpumask_and (cpus,sched_domain_span (sd), p->cpus_ptr); +return __select_idle_cpu (p, core, cpus, idle_cpu); -for_each_cpu_wrap (core, cpus, target) { -bool idle = true; - -for_each_cpu (cpu,cpu_smt_mask (core)) { -if (!available_idle_cpu (cpu)) { - idle = false; -break ; +for_each_cpu (cpu,cpu_smt_mask (core)) { +if (!available_idle_cpu (cpu)) { + idle = false; +if (*idle_cpu ==-1 ) { +if (sched_idle_cpu (cpu) &&cpumask_test_cpu (cpu, p->cpus_ptr)) { + *idle_cpu = cpu; +break ; + } +continue ; } +break ; } - -if (idle) -return core; - -cpumask_andnot (cpus, cpus,cpu_smt_mask (core)); +if (*idle_cpu ==-1 &&cpumask_test_cpu (cpu, p->cpus_ptr)) + *idle_cpu = cpu; } -/* -* Failed to find an idle core; stop looking for one. -*/ -set_idle_cores (target,0 ); +if (idle) +return core; +cpumask_andnot (cpus, cpus,cpu_smt_mask (core));return -1 ; }@@ -6107,9 +6109,18 @@ static int select_idle_core(struct task_struct *p, struct sched_domain *sd, int#define sched_smt_weight 1 -static inline int select_idle_core (struct task_struct *p,struct sched_domain *sd,int target) +static inline void set_idle_cores (int cpu,int val) { -return -1 ; +} + +static inline bool test_idle_cores (int cpu,bool def) +{ +return def; +} + +static inline int select_idle_core (struct task_struct *p,int core,struct cpumask *cpus,int *idle_cpu) +{ +return __select_idle_cpu (p, core, cpus, idle_cpu); }#endif /* CONFIG_SCHED_SMT */ @@ -6124,10 +6135,11 @@ static inline int select_idle_core(struct task_struct *p, struct sched_domain *sstatic int select_idle_cpu (struct task_struct *p,struct sched_domain *sd,int target) {struct cpumask *cpus =this_cpu_cpumask_var_ptr (select_idle_mask); +int i, cpu, idle_cpu =-1 , nr =INT_MAX ; +bool smt =test_idle_cores (target, false); +int this =smp_processor_id ();struct sched_domain *this_sd;u64 time; -int this =smp_processor_id (); -int cpu, nr =INT_MAX ; this_sd =rcu_dereference (*this_cpu_ptr (&sd_llc));if (!this_sd)@@ -6135,7 +6147,7 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tcpumask_and (cpus,sched_domain_span (sd), p->cpus_ptr); -if (sched_feat (SIS_PROP )) { +if (sched_feat (SIS_PROP ) && !smt) {u64 avg_cost, avg_idle, span_avg;/*
kernel/sched/fair.c static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t · line 6174
Quoted 6 lines
kernel/sched/fair.c +1
-2@@ -6159,16 +6171,29 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tfor_each_cpu_wrap (cpu, cpus, target) {if (!--nr)return -1 ; -if (available_idle_cpu (cpu) ||sched_idle_cpu (cpu)) -break ; +if (smt) {
If we want to stay on something similar to the previous behavior, we want to check on all cores if test_idle_cores is true so nr should be set to number of cores
and if there is no idle core, nr should be divided by sched_smt_weight so we will only scan a limited number of core
and you can always call select_idle_core()
Quoted 41 lines
kernel/sched/fair.c +16
-7+ i =select_idle_core (p, cpu, cpus, &idle_cpu); +if ((unsigned int )i < nr_cpumask_bits) +return i; + + }else { + i =__select_idle_cpu (p, cpu, cpus, &idle_cpu); +if ((unsigned int )i < nr_cpumask_bits) { + idle_cpu = i; +break ; + } + } } -if (sched_feat (SIS_PROP )) { +if (smt) +set_idle_cores (this, false); + +if (sched_feat (SIS_PROP ) && !smt) { time =cpu_clock (this) - time;update_avg (&this_sd->avg_scan_cost, time); } -return cpu; +return idle_cpu; }/* @@ -6297,10 +6322,6 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target)if (!sd)return target; - i =select_idle_core (p, sd, target); -if ((unsigned )i < nr_cpumask_bits) -return i; - i =select_idle_cpu (p, sd, target);if ((unsigned )i < nr_cpumask_bits)return i; --2.26.2
Re: [PATCH 3/5] sched/fair: Make select_idle_cpu() proportional to cores
In reply to Vincent GuittotQuoted 15 lines · Vincent Guittot
kernel/sched/core.c int sched_cpu_activate(unsigned int cpu) · line 7449
Quoted 12 lines
kernel/sched/core.c +3
-5@@ -7444,11 +7444,20 @@ int sched_cpu_activate(unsigned int cpu)balance_push_set (cpu, false); #ifdef CONFIG_SCHED_SMT -/* -* When going up, increment the number of cores with SMT present. -*/ -if (cpumask_weight (cpu_smt_mask (cpu)) ==2 ) -static_branch_inc_cpuslocked (&sched_smt_present); + do { + int weight = cpumask_weight(cpu_smt_mask(cpu)); + extern int sched_smt_weight;coding style problem
Presumably you are referring to an extern defined in a C file. That can move to kernel/sched/sched.h in this patch.
Quoted 7 lines
Quoted 7 lines
<SNIP>
/*
* Scan the LLC domain for idle CPUs; this is dynamically regulated by
* comparing the average scan cost (tracked insd->avg_scan_cost) against thekernel/sched/fair.c +0
-0@@ -6166,10 +6172,12 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t avg_cost = this_sd->avg_scan_cost +1 ;
Quoted 14 lines
kernel/sched/fair.c static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t · line 6180
Quoted 9 lines
kernel/sched/fair.c +4
-2span_avg = sd->span_weight * avg_idle; -if (span_avg >4 *avg_cost) +if (span_avg > sis_min_cores*avg_cost) nr =div_u64 (span_avg, avg_cost);else - nr =4 ; + nr = sis_min_cores; + + nr *= sched_smt_weight;Also, patch 5 will look at all CPUs of a core in
select_idle_coreso nr will decrement by 1 per core so i don't see the need to multiply bysched_smt_weightone patch 5 is applied
It makes sense in the context of this patch but can be removed again in the last patch and then I think sched_smt_weight only exists in core.c
Signature · 2 lines
Mel Gorman SUSE Labs
Re: [PATCH 5/5] sched/fair: Merge select_idle_core/cpu()
In reply to Vincent GuittotQuoted 12 lines · Vincent Guittot
kernel/sched/fair.c static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t · line 6174
Quoted 7 lines
kernel/sched/fair.c +1
-2@@ -6159,16 +6171,29 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tfor_each_cpu_wrap (cpu, cpus, target) {if (!--nr)return -1 ; -if (available_idle_cpu (cpu) ||sched_idle_cpu (cpu)) -break ; +if (smt) {If we want to stay on something similar to the previous behavior, we want to check on all cores if
test_idle_coresis true so nr should be set to number of cores
I don't think we necessarily want to do that. has_idle_cores is an effective throttling mechanism but it's not perfect. If the full domain is always scanned for a core then there can be excessive scanning in workloads like hackbench which tends to have has_idle_cores return false positives. It becomes important once average busy CPUs is over half of the domain for SMT2.
At least with the patch if that change was made, we still would not scan twice going over the same runqueues so it would still be an improvement but it would be nice to avoid excessive deep scanning when there are a lot of busy CPUs but individual tasks are rapidly idling.
However, in the event regressions are found, changing to your suggested behaviour would be an obvious starting point.
Signature · 2 lines
Mel Gorman SUSE Labs
Vincent Guittot
maintainerRe: [PATCH 5/5] sched/fair: Merge select_idle_core/cpu()
Quoted 15 lines · Mel Gorman
Quoted 12 lines · Vincent Guittot
kernel/sched/fair.c static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t · line 6174
Quoted 7 lines
kernel/sched/fair.c +1
-2@@ -6159,16 +6171,29 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tfor_each_cpu_wrap (cpu, cpus, target) {if (!--nr)return -1 ; -if (available_idle_cpu (cpu) ||sched_idle_cpu (cpu)) -break ; +if (smt) {If we want to stay on something similar to the previous behavior, we want to check on all cores if
test_idle_coresis true so nr should be set to number of cores
Quoted 3 lines
I don't think we necessarily want to do that.
has_idle_coresis an effective throttling mechanism but it's not perfect. If the full domain is always scanned for a core then there can be excessive scanning in
But that's what the code is currently doing. Can this change be done in another patch so we can check the impact of each change more easily?
This patch 5 should focus on merging select_idle_core and select_idle_cpu so we keep (almost) the same behavior but each CPU is checked only once.
Quoted 6 lines
workloads like hackbench which tends to have
has_idle_coresreturn false positives. It becomes important once average busy CPUs is over half of the domain for SMT2.At least with the patch if that change was made, we still would not scan twice going over the same runqueues so it would still be an improvement
yeah, it's for me the main goal of this patchset with the calculation of avg_can_cost being done only when SIS_PROP is true and the remove of SIS_AVG
any changes in the number of cpu/core to loop on is sensitive to regression and should be done in a separate patch IMHO
Quoted 9 lines
but it would be nice to avoid excessive deep scanning when there are a lot of busy CPUs but individual tasks are rapidly idling.
However, in the event regressions are found, changing to your suggested behaviour would be an obvious starting point.
--
Mel Gorman
SUSE Labs
Re: [PATCH 5/5] sched/fair: Merge select_idle_core/cpu()
Quoted 16 lines · Vincent Guittot
Quoted 15 lines · Mel Gorman
Quoted 12 lines · Vincent Guittot
kernel/sched/fair.c static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t · line 6174
Quoted 7 lines
kernel/sched/fair.c +1
-2@@ -6159,16 +6171,29 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tfor_each_cpu_wrap (cpu, cpus, target) {if (!--nr)return -1 ; -if (available_idle_cpu (cpu) ||sched_idle_cpu (cpu)) -break ; +if (smt) {If we want to stay on something similar to the previous behavior, we want to check on all cores if
test_idle_coresis true so nr should be set to number of cores
Quoted 7 lines
Quoted 3 lines
I don't think we necessarily want to do that.
has_idle_coresis an effective throttling mechanism but it's not perfect. If the full domain is always scanned for a core then there can be excessive scanning inBut that's what the code is currently doing. Can this change be done in another patch so we can check the impact of each change more easily?
Ok, when looking at this again instead of just the mail, the flow is;
int i, cpu, idle_cpu = -1, nr = INT_MAX;
...
if (sched_feat(SIS_PROP) && !smt) {
/* recalculate nr */
}The !smt check should mean that core scanning is still scanning the entire domain. There is no need to make it specific to the core account and we are already doing the full scan. Throttling that would be a separate patch.
Quoted 4 lines
This patch 5 should focus on merging
select_idle_coreandselect_idle_cpuso we keep (almost) the same behavior but each CPU is checked only once.
Which I think it's already doing. Main glitch really is that __select_idle_cpu() shouldn't be taking *idle_cpu as it does not consume the information.
Quoted 7 lines
Quoted 6 lines
workloads like hackbench which tends to have
has_idle_coresreturn false positives. It becomes important once average busy CPUs is over half of the domain for SMT2.At least with the patch if that change was made, we still would not scan twice going over the same runqueues so it would still be an improvement
Quoted 7 lines
yeah, it's for me the main goal of this patchset with the calculation of
avg_can_costbeing done only whenSIS_PROPis true and the remove ofSIS_AVGany changes in the number of cpu/core to loop on is sensitive to regression and should be done in a separate patch IMHO
Understood.
Signature · 2 lines
Mel Gorman SUSE Labs
Vincent Guittot
maintainerRe: [PATCH 5/5] sched/fair: Merge select_idle_core/cpu()
Quoted 28 lines · Mel Gorman
Quoted 16 lines · Vincent Guittot
Quoted 15 lines · Mel Gorman
Quoted 12 lines · Vincent Guittot
kernel/sched/fair.c static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t · line 6174
Quoted 7 lines
kernel/sched/fair.c +1
-2@@ -6159,16 +6171,29 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int tfor_each_cpu_wrap (cpu, cpus, target) {if (!--nr)return -1 ; -if (available_idle_cpu (cpu) ||sched_idle_cpu (cpu)) -break ; +if (smt) {If we want to stay on something similar to the previous behavior, we want to check on all cores if
test_idle_coresis true so nr should be set to number of coresQuoted 7 lines
Quoted 3 lines
I don't think we necessarily want to do that.
has_idle_coresis an effective throttling mechanism but it's not perfect. If the full domain is always scanned for a core then there can be excessive scanning inBut that's what the code is currently doing. Can this change be done in another patch so we can check the impact of each change more easily?
Ok, when looking at this again instead of just the mail, the flow is;
Quoted 7 lines
int i, cpu, idle_cpu = -1, nr = INT_MAX; ... if (sched_feat(SIS_PROP) && !smt) { /* recalculate nr */ }The !smt check should mean that core scanning is still scanning the entire
yes good point. I missed this change.
Quoted 11 lines
domain. There is no need to make it specific to the core account and we are already doing the full scan. Throttling that would be a separate patch.
Quoted 4 lines
This patch 5 should focus on merging
select_idle_coreandselect_idle_cpuso we keep (almost) the same behavior but each CPU is checked only once.Which I think it's already doing. Main glitch really is that
__select_idle_cpu()shouldn't be taking*idle_cpuas it does not consume the information.
don't really like the if (smt) else in the for_each_cpu_wrap(cpu,
cpus, target) loop because it just looks like we fail to merge idle core and idle cpu search loop at the end.
But there is probably not much we can do without changing what is accounted idle core search in the avg_scan_cost
Quoted 21 lines
Quoted 8 lines
Quoted 7 lines
workloads like hackbench which tends to have
has_idle_coresreturn false positives. It becomes important once average busy CPUs is over half of the domain for SMT2.At least with the patch if that change was made, we still would not scan twice going over the same runqueues so it would still be an improvement
Quoted 7 lines
yeah, it's for me the main goal of this patchset with the calculation of
avg_can_costbeing done only whenSIS_PROPis true and the remove ofSIS_AVGany changes in the number of cpu/core to loop on is sensitive to regression and should be done in a separate patch IMHO
Understood.
--
Mel Gorman
SUSE Labs
Re: [PATCH 5/5] sched/fair: Merge select_idle_core/cpu()
Quoted 8 lines · Vincent Guittot
Quoted 8 lines
domain. There is no need to make it specific to the core account and we are already doing the full scan. Throttling that would be a separate patch.
Quoted 4 lines
This patch 5 should focus on merging
select_idle_coreandselect_idle_cpuso we keep (almost) the same behavior but each CPU is checked only once.
Quoted 8 lines
Quoted 3 lines
Which I think it's already doing. Main glitch really is that
__select_idle_cpu()shouldn't be taking*idle_cpuas it does not consume the information.don't really like the if (smt) else in the for_each_cpu_wrap(cpu,
cpus, target) loop because it just looks like we fail to merge idle core and idle cpu search loop at the end.
While it's not the best, I did at one point have a series that fully unified this function and it wasn't pretty.
Quoted 3 lines
But there is probably not much we can do without changing what is accounted idle core search in the
avg_scan_cost
Indeed. Maybe in the future it'll make more sense to consolidate it further but between the depth search and possibly using SIS_PROP core core searches, we've bigger fish to fry.
Current delta between this series and what is being tested is simply
kernel/sched/core.c +0 -1
@@ -7446,7 +7446,6 @@ int sched_cpu_activate(unsigned int cpu) #ifdef CONFIG_SCHED_SMT do { int weight = cpumask_weight(cpu_smt_mask(cpu)); - extern int sched_smt_weight; if (weight > sched_smt_weight) sched_smt_weight = weight;
kernel/sched/fair.c +4 -4
@@ -6006,7 +6006,7 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *preturn new_cpu; } -static inline int __select_idle_cpu (struct task_struct *p,int core,struct cpumask *cpus,int *idle_cpu) +static inline int __select_idle_cpu (struct task_struct *p,int core,struct cpumask *cpus) {if (available_idle_cpu (core) ||sched_idle_cpu (core))return core;
@@ -6080,7 +6080,7 @@ static int select_idle_core(struct task_struct *p, int core, struct cpumask *cpuint cpu;if (!static_branch_likely (&sched_smt_present)) -return __select_idle_cpu (p, core, cpus, idle_cpu); +return __select_idle_cpu (p, core, cpus);for_each_cpu (cpu,cpu_smt_mask (core)) {if (!available_idle_cpu (cpu)) {
@@ -6120,7 +6120,7 @@ static inline bool test_idle_cores(int cpu, bool def)static inline int select_idle_core (struct task_struct *p,int core,struct cpumask *cpus,int *idle_cpu) { -return __select_idle_cpu (p, core, cpus, idle_cpu); +return __select_idle_cpu (p, core, cpus); }#endif /* CONFIG_SCHED_SMT */
@@ -6177,7 +6177,7 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int treturn i; }else { - i =__select_idle_cpu (p, cpu, cpus, &idle_cpu); + i =__select_idle_cpu (p, cpu, cpus);if ((unsigned int )i < nr_cpumask_bits) { idle_cpu = i;break ;
kernel/sched/sched.h +2 -0
@@ -1107,6 +1107,8 @@ static inline void update_idle_core(struct rq *rq) __update_idle_core(rq); } +extern int sched_smt_weight; +#else static inline void update_idle_core (struct rq *rq) { }#endif
Signature · 2 lines
Mel Gorman SUSE Labs