perf session: Keep file mmaped instead of malloc/memcpy
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>
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index 752577f..c989583 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -418,7 +418,6 @@
os->last_flush = iter->timestamp;
list_del(&iter->list);
- free(iter->event);
free(iter);
}
@@ -531,7 +530,6 @@
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 @@
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 @@
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 @@
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 @@
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 @@
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;