Message ID | 1416411982-1204-3-git-send-email-pang.xunlei@linaro.org |
---|---|
State | New |
Headers | show |
On Wed, 19 Nov 2014 23:46:21 +0800 "pang.xunlei" <pang.xunlei@linaro.org> wrote: > The call site of cpudl_find() in check_preempt_equal_dl() doesn't > use later_mask, so add this extra argument to distinquish the case. > > Signed-off-by: pang.xunlei <pang.xunlei@linaro.org> > --- > kernel/sched/cpudeadline.c | 6 ++++-- > kernel/sched/cpudeadline.h | 2 +- > kernel/sched/deadline.c | 6 +++--- > 3 files changed, 8 insertions(+), 6 deletions(-) > > diff --git a/kernel/sched/cpudeadline.c b/kernel/sched/cpudeadline.c > index c01b3aa..3047846 100644 > --- a/kernel/sched/cpudeadline.c > +++ b/kernel/sched/cpudeadline.c > @@ -98,11 +98,12 @@ static inline int cpudl_maximum(struct cpudl *cp) > * @cp: the cpudl max-heap context > * @p: the task > * @later_mask: a mask to fill in with the selected CPUs (not NULL) > + * @set_flag: indicate if later_mask should be set > * > * Returns: int - best CPU (heap maximum if suitable) > */ > int cpudl_find(struct cpudl *cp, struct task_struct *p, > - struct cpumask *later_mask) > + struct cpumask *later_mask, int set_flag) set_flag should be a bool type. > { > int best_cpu = -1; > const struct sched_dl_entity *dl_se = &p->dl; > @@ -114,7 +115,8 @@ int cpudl_find(struct cpudl *cp, struct task_struct *p, > } else if (cpumask_test_cpu(cpudl_maximum(cp), &p->cpus_allowed) && > dl_time_before(dl_se->deadline, cp->elements[0].dl)) { > best_cpu = cpudl_maximum(cp); > - cpumask_set_cpu(best_cpu, later_mask); > + if (set_flag) > + cpumask_set_cpu(best_cpu, later_mask); I'm not sure this is worth it. cpumask_set_cpu() is rather efficient. > } > > out: -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
On 20 November 2014 00:24, Steven Rostedt <rostedt@goodmis.org> wrote: > On Wed, 19 Nov 2014 23:46:21 +0800 > "pang.xunlei" <pang.xunlei@linaro.org> wrote: > >> The call site of cpudl_find() in check_preempt_equal_dl() doesn't >> use later_mask, so add this extra argument to distinquish the case. >> >> Signed-off-by: pang.xunlei <pang.xunlei@linaro.org> >> --- >> kernel/sched/cpudeadline.c | 6 ++++-- >> kernel/sched/cpudeadline.h | 2 +- >> kernel/sched/deadline.c | 6 +++--- >> 3 files changed, 8 insertions(+), 6 deletions(-) >> >> diff --git a/kernel/sched/cpudeadline.c b/kernel/sched/cpudeadline.c >> index c01b3aa..3047846 100644 >> --- a/kernel/sched/cpudeadline.c >> +++ b/kernel/sched/cpudeadline.c >> @@ -98,11 +98,12 @@ static inline int cpudl_maximum(struct cpudl *cp) >> * @cp: the cpudl max-heap context >> * @p: the task >> * @later_mask: a mask to fill in with the selected CPUs (not NULL) >> + * @set_flag: indicate if later_mask should be set >> * >> * Returns: int - best CPU (heap maximum if suitable) >> */ >> int cpudl_find(struct cpudl *cp, struct task_struct *p, >> - struct cpumask *later_mask) >> + struct cpumask *later_mask, int set_flag) > > set_flag should be a bool type. > >> { >> int best_cpu = -1; >> const struct sched_dl_entity *dl_se = &p->dl; >> @@ -114,7 +115,8 @@ int cpudl_find(struct cpudl *cp, struct task_struct *p, >> } else if (cpumask_test_cpu(cpudl_maximum(cp), &p->cpus_allowed) && >> dl_time_before(dl_se->deadline, cp->elements[0].dl)) { >> best_cpu = cpudl_maximum(cp); >> - cpumask_set_cpu(best_cpu, later_mask); >> + if (set_flag) >> + cpumask_set_cpu(best_cpu, later_mask); > > I'm not sure this is worth it. cpumask_set_cpu() is rather efficient. HI Steve, Thanks for your commenting, I've rethinked this a bit. We can do a little trick with its return value, then could avoid this extra cpumask_set_cpu() without this extra set_flag: 1) define macros for the return values of cpudl_find(), like: #define CPUDL_FIND_NONE -2 /* no available cpus */ #define CPUDL_FIND_CPUMASK -1 /* available cpus in later_mask */ then, with the return value >=0, means it returns the only one available cpu. 2) In the leg of "if", it can just return CPUDL_FIND_CPUMASK, as we want to select the best_cpu in find_later_rq(). In the leg of "else if", just returns cpudl_maximum(cp), apparently there is no need to set the later_mask, since we will definitely select this cpu as the best_cpu in find_later_rq() . int cpudl_find(struct cpudl *cp, struct task_struct *p, struct cpumask *later_mask) { const struct sched_dl_entity *dl_se = &p->dl; cpumask_and(later_mask, &p->cpus_allowed, &p->cpus_allowed); if (cpumask_and(later_mask, later_mask, cp->free_cpus)) { return CPUDL_FIND_CPUMASK; } else if (cpumask_test_cpu(cpudl_maximum(cp), &p->cpus_allowed) && dl_time_before(dl_se->deadline, cp->elements[0].dl)) int cpu; cpu = cpudl_maximum(cp); WARN_ON(!cpu_present(cpu)); return cpu; } out: return CPUDL_FIND_NONE; } Thus, in find_later_rq() we can change the call site code like: best_cpu = cpudl_find(&task_rq(task)->rd->cpudl, task, later_mask); if (best_cpu == CPUDL_FIND_NONE) return -1; if (best_cpu != CPUDL_FIND_CPUMASK) return best_cpu; /* adjust the following code as that in RT find_lowest_rq(), omit here... */ What's your view about this? Thanks, Xunlei > >> } >> >> out: > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
On 20 November 2014 22:58, pang.xunlei <pang.xunlei@linaro.org> wrote: > On 20 November 2014 00:24, Steven Rostedt <rostedt@goodmis.org> wrote: >> On Wed, 19 Nov 2014 23:46:21 +0800 >> "pang.xunlei" <pang.xunlei@linaro.org> wrote: >> >>> The call site of cpudl_find() in check_preempt_equal_dl() doesn't >>> use later_mask, so add this extra argument to distinquish the case. >>> >>> Signed-off-by: pang.xunlei <pang.xunlei@linaro.org> >>> --- >>> kernel/sched/cpudeadline.c | 6 ++++-- >>> kernel/sched/cpudeadline.h | 2 +- >>> kernel/sched/deadline.c | 6 +++--- >>> 3 files changed, 8 insertions(+), 6 deletions(-) >>> >>> diff --git a/kernel/sched/cpudeadline.c b/kernel/sched/cpudeadline.c >>> index c01b3aa..3047846 100644 >>> --- a/kernel/sched/cpudeadline.c >>> +++ b/kernel/sched/cpudeadline.c >>> @@ -98,11 +98,12 @@ static inline int cpudl_maximum(struct cpudl *cp) >>> * @cp: the cpudl max-heap context >>> * @p: the task >>> * @later_mask: a mask to fill in with the selected CPUs (not NULL) >>> + * @set_flag: indicate if later_mask should be set >>> * >>> * Returns: int - best CPU (heap maximum if suitable) >>> */ >>> int cpudl_find(struct cpudl *cp, struct task_struct *p, >>> - struct cpumask *later_mask) >>> + struct cpumask *later_mask, int set_flag) >> >> set_flag should be a bool type. >> >>> { >>> int best_cpu = -1; >>> const struct sched_dl_entity *dl_se = &p->dl; >>> @@ -114,7 +115,8 @@ int cpudl_find(struct cpudl *cp, struct task_struct *p, >>> } else if (cpumask_test_cpu(cpudl_maximum(cp), &p->cpus_allowed) && >>> dl_time_before(dl_se->deadline, cp->elements[0].dl)) { >>> best_cpu = cpudl_maximum(cp); >>> - cpumask_set_cpu(best_cpu, later_mask); >>> + if (set_flag) >>> + cpumask_set_cpu(best_cpu, later_mask); >> >> I'm not sure this is worth it. cpumask_set_cpu() is rather efficient. > HI Steve, > > Thanks for your commenting, I've rethinked this a bit. > We can do a little trick with its return value, then could avoid this > extra cpumask_set_cpu() without this extra set_flag: > 1) define macros for the return values of cpudl_find(), like: > #define CPUDL_FIND_NONE -2 /* no available cpus */ > #define CPUDL_FIND_CPUMASK -1 /* available cpus in later_mask */ > > then, with the return value >=0, means it returns the only one available cpu. > > 2) In the leg of "if", it can just return CPUDL_FIND_CPUMASK, as we > want to select the best_cpu in find_later_rq(). > In the leg of "else if", just returns cpudl_maximum(cp), apparently > there is no need to set the later_mask, since we will definitely > select this cpu as the best_cpu in find_later_rq() . > > int cpudl_find(struct cpudl *cp, struct task_struct *p, > struct cpumask *later_mask) > { > const struct sched_dl_entity *dl_se = &p->dl; > > cpumask_and(later_mask, &p->cpus_allowed, &p->cpus_allowed); Apologies for this typo, it should be: cpumask_and(later_mask, cpu_active_mask, &p->cpus_allowed); > if (cpumask_and(later_mask, later_mask, cp->free_cpus)) { > return CPUDL_FIND_CPUMASK; > } else if (cpumask_test_cpu(cpudl_maximum(cp), &p->cpus_allowed) && > dl_time_before(dl_se->deadline, cp->elements[0].dl)) > int cpu; > > cpu = cpudl_maximum(cp); > WARN_ON(!cpu_present(cpu)); > return cpu; > } > > out: also delete this lable. > > return CPUDL_FIND_NONE; > } > > Thus, in find_later_rq() we can change the call site code like: > best_cpu = cpudl_find(&task_rq(task)->rd->cpudl, task, > later_mask); > if (best_cpu == CPUDL_FIND_NONE) > return -1; > if (best_cpu != CPUDL_FIND_CPUMASK) > return best_cpu; > > /* adjust the following code as that in RT find_lowest_rq(), omit here... */ > > What's your view about this? > > Thanks, > Xunlei >> >>> } >>> >>> out: >> -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
diff --git a/kernel/sched/cpudeadline.c b/kernel/sched/cpudeadline.c index c01b3aa..3047846 100644 --- a/kernel/sched/cpudeadline.c +++ b/kernel/sched/cpudeadline.c @@ -98,11 +98,12 @@ static inline int cpudl_maximum(struct cpudl *cp) * @cp: the cpudl max-heap context * @p: the task * @later_mask: a mask to fill in with the selected CPUs (not NULL) + * @set_flag: indicate if later_mask should be set * * Returns: int - best CPU (heap maximum if suitable) */ int cpudl_find(struct cpudl *cp, struct task_struct *p, - struct cpumask *later_mask) + struct cpumask *later_mask, int set_flag) { int best_cpu = -1; const struct sched_dl_entity *dl_se = &p->dl; @@ -114,7 +115,8 @@ int cpudl_find(struct cpudl *cp, struct task_struct *p, } else if (cpumask_test_cpu(cpudl_maximum(cp), &p->cpus_allowed) && dl_time_before(dl_se->deadline, cp->elements[0].dl)) { best_cpu = cpudl_maximum(cp); - cpumask_set_cpu(best_cpu, later_mask); + if (set_flag) + cpumask_set_cpu(best_cpu, later_mask); } out: diff --git a/kernel/sched/cpudeadline.h b/kernel/sched/cpudeadline.h index 4a10a65..a3a85e8 100644 --- a/kernel/sched/cpudeadline.h +++ b/kernel/sched/cpudeadline.h @@ -22,7 +22,7 @@ struct cpudl { #ifdef CONFIG_SMP int cpudl_find(struct cpudl *cp, struct task_struct *p, struct cpumask *later_mask); -void cpudl_set(struct cpudl *cp, int cpu, u64 dl, int is_valid); +void cpudl_set(struct cpudl *cp, int cpu, u64 dl, int is_valid, int set_flag); int cpudl_init(struct cpudl *cp); void cpudl_set_freecpu(struct cpudl *cp, int cpu); void cpudl_cleanup(struct cpudl *cp); diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index 3ecf838..e8208d0 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -976,7 +976,7 @@ static void check_preempt_equal_dl(struct rq *rq, struct task_struct *p) * let's hope p can move out. */ if (rq->curr->nr_cpus_allowed == 1 || - cpudl_find(&rq->rd->cpudl, rq->curr, later_mask) == -1) + cpudl_find(&rq->rd->cpudl, rq->curr, later_mask, 0) == -1) return; /* @@ -984,7 +984,7 @@ static void check_preempt_equal_dl(struct rq *rq, struct task_struct *p) * see if it is pushed or pulled somewhere else. */ if (p->nr_cpus_allowed != 1 && - cpudl_find(&rq->rd->cpudl, p, later_mask) != -1) + cpudl_find(&rq->rd->cpudl, p, later_mask, 0) != -1) return; resched_curr(rq); @@ -1190,7 +1190,7 @@ static int find_later_rq(struct task_struct *task) * first, then we can look for a suitable cpu. */ best_cpu = cpudl_find(&task_rq(task)->rd->cpudl, - task, later_mask); + task, later_mask, 1); if (best_cpu == -1) return -1;
The call site of cpudl_find() in check_preempt_equal_dl() doesn't use later_mask, so add this extra argument to distinquish the case. Signed-off-by: pang.xunlei <pang.xunlei@linaro.org> --- kernel/sched/cpudeadline.c | 6 ++++-- kernel/sched/cpudeadline.h | 2 +- kernel/sched/deadline.c | 6 +++--- 3 files changed, 8 insertions(+), 6 deletions(-)