mirror of
https://github.com/jemalloc/jemalloc.git
synced 2026-07-08 23:07:20 +03:00
First stage of mutex profiling.
Switched to trylock and update counters based on state.
This commit is contained in:
parent
32e7cf51cd
commit
6309df628f
5 changed files with 149 additions and 32 deletions
|
|
@ -1,9 +1,50 @@
|
|||
#ifndef JEMALLOC_INTERNAL_MUTEX_STRUCTS_H
|
||||
#define JEMALLOC_INTERNAL_MUTEX_STRUCTS_H
|
||||
|
||||
struct lock_prof_data_s {
|
||||
/*
|
||||
* Counters touched on the slow path, i.e. when there is lock
|
||||
* contention. We update them once we have the lock.
|
||||
*/
|
||||
/* Total time spent waiting on this lock. */
|
||||
nstime_t tot_wait_time;
|
||||
/* Max time spent on a single lock operation. */
|
||||
nstime_t max_wait_time;
|
||||
/* # of times have to wait for this lock (after spinning). */
|
||||
uint64_t n_wait_times;
|
||||
/* # of times acquired the lock through local spinning. */
|
||||
uint64_t n_spin_acquired;
|
||||
/* Max # of threads waiting for the lock at the same time. */
|
||||
uint32_t max_n_thds;
|
||||
/* Current # of threads waiting on the lock. Atomic synced. */
|
||||
uint32_t n_waiting_thds;
|
||||
|
||||
/*
|
||||
* Data touched on the fast path. These are modified right after we
|
||||
* grab the lock, so it's placed closest to the end (i.e. right before
|
||||
* the lock) so that we have a higher chance of them being on the same
|
||||
* cacheline.
|
||||
*/
|
||||
/* # of times the new lock holder is different from the previous one. */
|
||||
uint64_t n_owner_switches;
|
||||
/* Previous lock holder, to facilitate n_owner_switches. */
|
||||
tsdn_t *prev_owner;
|
||||
/* # of lock() operations in total. */
|
||||
uint64_t n_lock_ops;
|
||||
};
|
||||
|
||||
struct malloc_mutex_s {
|
||||
union {
|
||||
struct {
|
||||
/*
|
||||
* prof_data is defined first to reduce cacheline
|
||||
* bouncing: the data is not touched by the lock holder
|
||||
* during unlocking, while might be modified by
|
||||
* contenders. Having it before the lock itself could
|
||||
* avoid prefetching a modified cacheline (for the
|
||||
* unlocking thread).
|
||||
*/
|
||||
lock_prof_data_t prof_data;
|
||||
#ifdef _WIN32
|
||||
# if _WIN32_WINNT >= 0x0600
|
||||
SRWLOCK lock;
|
||||
|
|
@ -22,7 +63,7 @@ struct malloc_mutex_s {
|
|||
#endif
|
||||
};
|
||||
/*
|
||||
* We only touch witness when configured w/ debug. However we
|
||||
* We only touch witness when configured w/ debug. However we
|
||||
* keep the field in a union when !debug so that we don't have
|
||||
* to pollute the code base with #ifdefs, while avoid paying the
|
||||
* memory cost.
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue