perf-record: Share per-cpu buffers
authorPeter Zijlstra <a.p.zijlstra@chello.nl>
Thu, 20 May 2010 12:45:26 +0000 (14:45 +0200)
committerIngo Molnar <mingo@elte.hu>
Fri, 21 May 2010 09:37:58 +0000 (11:37 +0200)
It seems a waste of space to create a buffer per
event, share it per-cpu.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <20100521090710.634824884@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
tools/perf/builtin-record.c

index 94e210f..9bc8905 100644 (file)
@@ -82,7 +82,7 @@ struct mmap_data {
        unsigned int            prev;
 };
 
-static struct mmap_data                *mmap_array[MAX_NR_CPUS][MAX_COUNTERS];
+static struct mmap_data                mmap_array[MAX_NR_CPUS];
 
 static unsigned long mmap_read_head(struct mmap_data *md)
 {
@@ -365,18 +365,29 @@ try_again:
                if (group && group_fd == -1)
                        group_fd = fd[nr_cpu][counter][thread_index];
 
-               event_array[nr_poll].fd = fd[nr_cpu][counter][thread_index];
-               event_array[nr_poll].events = POLLIN;
-               nr_poll++;
-
-               mmap_array[nr_cpu][counter][thread_index].counter = counter;
-               mmap_array[nr_cpu][counter][thread_index].prev = 0;
-               mmap_array[nr_cpu][counter][thread_index].mask = mmap_pages*page_size - 1;
-               mmap_array[nr_cpu][counter][thread_index].base = mmap(NULL, (mmap_pages+1)*page_size,
-                       PROT_READ|PROT_WRITE, MAP_SHARED, fd[nr_cpu][counter][thread_index], 0);
-               if (mmap_array[nr_cpu][counter][thread_index].base == MAP_FAILED) {
-                       error("failed to mmap with %d (%s)\n", errno, strerror(errno));
-                       exit(-1);
+               if (counter || thread_index) {
+                       ret = ioctl(fd[nr_cpu][counter][thread_index],
+                                       PERF_EVENT_IOC_SET_OUTPUT,
+                                       fd[nr_cpu][0][0]);
+                       if (ret) {
+                               error("failed to set output: %d (%s)\n", errno,
+                                               strerror(errno));
+                               exit(-1);
+                       }
+               } else {
+                       mmap_array[nr_cpu].counter = counter;
+                       mmap_array[nr_cpu].prev = 0;
+                       mmap_array[nr_cpu].mask = mmap_pages*page_size - 1;
+                       mmap_array[nr_cpu].base = mmap(NULL, (mmap_pages+1)*page_size,
+                               PROT_READ|PROT_WRITE, MAP_SHARED, fd[nr_cpu][counter][thread_index], 0);
+                       if (mmap_array[nr_cpu].base == MAP_FAILED) {
+                               error("failed to mmap with %d (%s)\n", errno, strerror(errno));
+                               exit(-1);
+                       }
+
+                       event_array[nr_poll].fd = fd[nr_cpu][counter][thread_index];
+                       event_array[nr_poll].events = POLLIN;
+                       nr_poll++;
                }
 
                if (filter != NULL) {
@@ -477,16 +488,11 @@ static struct perf_event_header finished_round_event = {
 
 static void mmap_read_all(void)
 {
-       int i, counter, thread;
+       int i;
 
        for (i = 0; i < nr_cpu; i++) {
-               for (counter = 0; counter < nr_counters; counter++) {
-                       for (thread = 0; thread < thread_num; thread++) {
-                               if (mmap_array[i][counter][thread].base)
-                                       mmap_read(&mmap_array[i][counter][thread]);
-                       }
-
-               }
+               if (mmap_array[i].base)
+                       mmap_read(&mmap_array[i]);
        }
 
        if (perf_header__has_feat(&session->header, HEADER_TRACE_INFO))
@@ -861,9 +867,7 @@ int cmd_record(int argc, const char **argv, const char *prefix __used)
        for (i = 0; i < MAX_NR_CPUS; i++) {
                for (j = 0; j < MAX_COUNTERS; j++) {
                        fd[i][j] = malloc(sizeof(int)*thread_num);
-                       mmap_array[i][j] = zalloc(
-                               sizeof(struct mmap_data)*thread_num);
-                       if (!fd[i][j] || !mmap_array[i][j])
+                       if (!fd[i][j])
                                return -ENOMEM;
                }
        }