michaelh/zephyr
1
0
Fork 0
Code Releases Activity

kernel/sched: Force inlining of some routines within the scheduler guts

Browse source 
GCC 6.2.0 is making frustratingly poor inlining decisions with some of
these routines, resulting in an awful lot of runtime calls for code
that is only ever expanded once or twice within the file.

Treat with targetted ALWAYS_INLINE's to force the issue.  The
scheduler code is a hot path.

Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
This commit is contained in:
Andy Ross 2019-01-28 09:36:36 -08:00 • committed by Anas Nashif
commit b2791b0ac8
2 changed files with 9 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

2
kernel/include/ksched.h
View file

@ -215,7 +215,7 @@ static inline bool _is_valid_prio(int prio, void *entry_point)
return true;
}
static inline void _ready_thread(struct k_thread *thread)
static ALWAYS_INLINE void _ready_thread(struct k_thread *thread)
{
if (_is_thread_ready(thread)) {
_add_thread_to_ready_q(thread);

16
kernel/sched.c
View file

@ -111,7 +111,7 @@ bool _is_t1_higher_prio_than_t2(struct k_thread *t1, struct k_thread *t2)
return false;
}
static bool should_preempt(struct k_thread *th, int preempt_ok)
static ALWAYS_INLINE bool should_preempt(struct k_thread *th, int preempt_ok)
{
/* Preemption is OK if it's being explicitly allowed by
* software state (e.g. the thread called k_yield())
@ -153,7 +153,7 @@ static bool should_preempt(struct k_thread *th, int preempt_ok)
return false;
}
static struct k_thread *next_up(void)
static ALWAYS_INLINE struct k_thread *next_up(void)
{
#ifndef CONFIG_SMP
/* In uniprocessor mode, we can leave the current thread in
@ -367,8 +367,8 @@ static _wait_q_t *pended_on(struct k_thread *thread)
return thread->base.pended_on;
}
struct k_thread *_find_first_thread_to_unpend(_wait_q_t *wait_q,
struct k_thread *from)
ALWAYS_INLINE struct k_thread *_find_first_thread_to_unpend(_wait_q_t *wait_q,
struct k_thread *from)
{
ARG_UNUSED(from);
@ -381,7 +381,7 @@ struct k_thread *_find_first_thread_to_unpend(_wait_q_t *wait_q,
return ret;
}
void _unpend_thread_no_timeout(struct k_thread *thread)
ALWAYS_INLINE void _unpend_thread_no_timeout(struct k_thread *thread)
{
LOCKED(&sched_lock) {
_priq_wait_remove(&pended_on(thread)->waitq, thread);
@ -565,7 +565,7 @@ void *_get_next_switch_handle(void *interrupted)
}
#endif
void _priq_dumb_add(sys_dlist_t *pq, struct k_thread *thread)
ALWAYS_INLINE void _priq_dumb_add(sys_dlist_t *pq, struct k_thread *thread)
{
struct k_thread *t;
@ -667,7 +667,7 @@ struct k_thread *_priq_rb_best(struct _priq_rb *pq)
# endif
#endif
void _priq_mq_add(struct _priq_mq *pq, struct k_thread *thread)
ALWAYS_INLINE void _priq_mq_add(struct _priq_mq *pq, struct k_thread *thread)
{
int priority_bit = thread->base.prio - K_HIGHEST_THREAD_PRIO;
@ -675,7 +675,7 @@ void _priq_mq_add(struct _priq_mq *pq, struct k_thread *thread)
pq->bitmask |= (1 << priority_bit);
}
void _priq_mq_remove(struct _priq_mq *pq, struct k_thread *thread)
ALWAYS_INLINE void _priq_mq_remove(struct _priq_mq *pq, struct k_thread *thread)
{
int priority_bit = thread->base.prio - K_HIGHEST_THREAD_PRIO;