GitHub-Mirror/Star64_linux
2
0
Fork 0
mirror of https://github.com/Fishwaldo/Star64_linux.git synced 2025-06-21 22:21:21 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

perf: Optimize throttling code

Browse source 
By pre-computing the maximum number of samples per tick we can avoid a
multiplication and a conditional since MAX_INTERRUPTS >
max_samples_per_tick.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This commit is contained in:
Peter Zijlstra 2011-02-16 11:22:34 +01:00 • committed by Ingo Molnar
parent 4979d2729a
commit 163ec4354a
3 changed files with 29 additions and 20 deletions
Download patch file Download diff file Expand all files Collapse all files

4
include/linux/perf_event.h
View file

@ -1110,6 +1110,10 @@ extern int sysctl_perf_event_paranoid;
extern int sysctl_perf_event_mlock; extern int sysctl_perf_event_mlock;
extern int sysctl_perf_event_sample_rate; extern int sysctl_perf_event_sample_rate;
extern int perf_proc_update_handler(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp,
loff_t *ppos);
static inline bool perf_paranoid_tracepoint_raw(void) static inline bool perf_paranoid_tracepoint_raw(void)
{ {
return sysctl_perf_event_paranoid > -1; return sysctl_perf_event_paranoid > -1;

43
kernel/perf_event.c
View file

@ -150,7 +150,24 @@ int sysctl_perf_event_mlock __read_mostly = 512; /* 'free' kb per user */
/* /*
* max perf event sample rate * max perf event sample rate
*/ */
int sysctl_perf_event_sample_rate __read_mostly = 100000; #define DEFAULT_MAX_SAMPLE_RATE 100000
int sysctl_perf_event_sample_rate __read_mostly = DEFAULT_MAX_SAMPLE_RATE;
static int max_samples_per_tick __read_mostly =
DIV_ROUND_UP(DEFAULT_MAX_SAMPLE_RATE, HZ);
int perf_proc_update_handler(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp,
loff_t *ppos)
{
int ret = proc_dointvec(table, write, buffer, lenp, ppos);
if (ret || !write)
return ret;
max_samples_per_tick = DIV_ROUND_UP(sysctl_perf_event_sample_rate, HZ);
return 0;
}
static atomic64_t perf_event_id; static atomic64_t perf_event_id;
@ -4941,26 +4958,14 @@ static int __perf_event_overflow(struct perf_event *event, int nmi,
if (unlikely(!is_sampling_event(event))) if (unlikely(!is_sampling_event(event)))
return 0; return 0;
if (!throttle) { if (unlikely(hwc->interrupts >= max_samples_per_tick)) {
hwc->interrupts++; if (throttle) {
} else { hwc->interrupts = MAX_INTERRUPTS;
if (hwc->interrupts != MAX_INTERRUPTS) { perf_log_throttle(event, 0);
hwc->interrupts++;
if (HZ * hwc->interrupts >
(u64)sysctl_perf_event_sample_rate) {
hwc->interrupts = MAX_INTERRUPTS;
perf_log_throttle(event, 0);
ret = 1;
}
} else {
/*
* Keep re-disabling events even though on the previous
* pass we disabled it - just in case we raced with a
* sched-in and the event got enabled again:
*/
ret = 1; ret = 1;
} }
} } else
hwc->interrupts++;
if (event->attr.freq) { if (event->attr.freq) {
u64 now = perf_clock(); u64 now = perf_clock();

2
kernel/sysctl.c
View file

@ -948,7 +948,7 @@ static struct ctl_table kern_table[] = {
.data = &sysctl_perf_event_sample_rate, .data = &sysctl_perf_event_sample_rate,
.maxlen = sizeof(sysctl_perf_event_sample_rate), .maxlen = sizeof(sysctl_perf_event_sample_rate),
.mode = 0644, .mode = 0644,
.proc_handler = proc_dointvec, .proc_handler = perf_proc_update_handler,
}, },
#endif #endif
#ifdef CONFIG_KMEMCHECK #ifdef CONFIG_KMEMCHECK