redonkable/alistair23-linux
redonkable/alistair23-linux
1
0
Fork 0
Code Releases Activity

perf session: Keep file mmaped instead of malloc/memcpy

Browse source 
Profiling perf with perf revealed that a large part of the processing time is
spent in malloc/memcpy/free in the sample ordering code. That code copies the
data from the mmap into malloc'ed memory. That's silly. We can keep the mmap
and just store the pointer in the queuing data structure. For 64 bit this is
not a problem as we map the whole file anyway. On 32bit we keep 8 maps around
and unmap the oldest before mmaping the next chunk of the file.

Performance gain: 2.95s -> 1.23s (Faktor 2.4)

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163820.278787719@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This commit is contained in:
Thomas Gleixner 2010-11-30 17:49:49 +00:00 committed by Arnaldo Carvalho de Melo
parent 55b44629f5
commit fe17420784
1 changed files with 11 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

27
tools/perf/util/session.c
View file

@ -418,7 +418,6 @@ static void flush_sample_queue(struct perf_session *s,
os->last_flush = iter->timestamp;
list_del(&iter->list);
free(iter->event);
free(iter);
}
@ -531,7 +530,6 @@ static int queue_sample_event(event_t *event, struct sample_data *data,
u64 timestamp = data->time;
struct sample_queue *new;
if (timestamp < s->ordered_samples.last_flush) {
printf("Warning: Timestamp below last timeslice flush\n");
return -EINVAL;
@ -542,14 +540,7 @@ static int queue_sample_event(event_t *event, struct sample_data *data,
return -ENOMEM;
new->timestamp = timestamp;
new->event = malloc(event->header.size);
if (!new->event) {
free(new);
return -ENOMEM;
}
memcpy(new->event, event, event->header.size);
new->event = event;
__queue_sample_event(new, s);
@ -747,12 +738,12 @@ int __perf_session__process_events(struct perf_session *session,
u64 file_size, struct perf_event_ops *ops)
{
u64 head, page_offset, file_offset, file_pos, progress_next;
int err, mmap_prot, mmap_flags;
int err, mmap_prot, mmap_flags, map_idx = 0;
struct ui_progress *progress;
size_t page_size, mmap_size;
char *buf, *mmaps[8];
event_t *event;
uint32_t size;
char *buf;
perf_event_ops__fill_defaults(ops);
@ -774,6 +765,8 @@ int __perf_session__process_events(struct perf_session *session,
if (mmap_size > file_size)
mmap_size = file_size;
memset(mmaps, 0, sizeof(mmaps));
mmap_prot = PROT_READ;
mmap_flags = MAP_SHARED;
@ -789,6 +782,8 @@ remap:
err = -errno;
goto out_err;
}
mmaps[map_idx] = buf;
map_idx = (map_idx + 1) & (ARRAY_SIZE(mmaps) - 1);
file_pos = file_offset + head;
more:
@ -801,10 +796,10 @@ more:
size = 8;
if (head + event->header.size >= mmap_size) {
int munmap_ret;
munmap_ret = munmap(buf, mmap_size);
assert(munmap_ret == 0);
if (mmaps[map_idx]) {
munmap(mmaps[map_idx], mmap_size);
mmaps[map_idx] = NULL;
}
page_offset = page_size * (head / page_size);
file_offset += page_offset;