perf: Optimize the hotpath by converting the perf output buffer to local_t
authorPeter Zijlstra <a.p.zijlstra@chello.nl>
Tue, 18 May 2010 08:54:20 +0000 (10:54 +0200)
committerIngo Molnar <mingo@elte.hu>
Tue, 18 May 2010 16:35:49 +0000 (18:35 +0200)
Since there is now only a single writer, we can use
local_t instead and avoid all these pesky LOCK insn.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
include/linux/perf_event.h
kernel/perf_event.c

index f1f853a9d5eb0cbd0fbca519b9a4904c94b27e07..ce7667616fcbe8bf40cd85c6ffedb3cbd2e3fe02 100644 (file)
@@ -485,6 +485,7 @@ struct perf_guest_info_callbacks {
 #include <linux/ftrace.h>
 #include <linux/cpu.h>
 #include <asm/atomic.h>
+#include <asm/local.h>
 
 #define PERF_MAX_STACK_DEPTH           255
 
@@ -588,20 +589,18 @@ struct perf_mmap_data {
 #ifdef CONFIG_PERF_USE_VMALLOC
        struct work_struct              work;
 #endif
-       int                             data_order;
+       int                             data_order;     /* allocation order  */
        int                             nr_pages;       /* nr of data pages  */
        int                             writable;       /* are we writable   */
        int                             nr_locked;      /* nr pages mlocked  */
 
        atomic_t                        poll;           /* POLL_ for wakeups */
-       atomic_t                        events;         /* event_id limit       */
 
-       atomic_long_t                   head;           /* write position    */
-
-       atomic_t                        wakeup;         /* needs a wakeup    */
-       atomic_t                        lost;           /* nr records lost   */
-
-       atomic_t                        nest;           /* nested writers    */
+       local_t                         head;           /* write position    */
+       local_t                         nest;           /* nested writers    */
+       local_t                         events;         /* event limit       */
+       local_t                         wakeup;         /* needs a wakeup    */
+       local_t                         lost;           /* nr records lost   */
 
        long                            watermark;      /* wakeup watermark  */
 
index 8cf737da3ec4926c23e0a8f58cd137a94c482950..1f98c78c33437d2fba494e38c75e11efca7a2232 100644 (file)
@@ -2916,7 +2916,7 @@ static void perf_output_get_handle(struct perf_output_handle *handle)
        struct perf_mmap_data *data = handle->data;
 
        preempt_disable();
-       atomic_inc(&data->nest);
+       local_inc(&data->nest);
 }
 
 static void perf_output_put_handle(struct perf_output_handle *handle)
@@ -2925,13 +2925,13 @@ static void perf_output_put_handle(struct perf_output_handle *handle)
        unsigned long head;
 
 again:
-       head = atomic_long_read(&data->head);
+       head = local_read(&data->head);
 
        /*
         * IRQ/NMI can happen here, which means we can miss a head update.
         */
 
-       if (!atomic_dec_and_test(&data->nest))
+       if (!local_dec_and_test(&data->nest))
                return;
 
        /*
@@ -2945,12 +2945,12 @@ again:
         * Now check if we missed an update, rely on the (compiler)
         * barrier in atomic_dec_and_test() to re-read data->head.
         */
-       if (unlikely(head != atomic_long_read(&data->head))) {
-               atomic_inc(&data->nest);
+       if (unlikely(head != local_read(&data->head))) {
+               local_inc(&data->nest);
                goto again;
        }
 
-       if (atomic_xchg(&data->wakeup, 0))
+       if (local_xchg(&data->wakeup, 0))
                perf_output_wakeup(handle);
 
        preempt_enable();
@@ -3031,7 +3031,7 @@ int perf_output_begin(struct perf_output_handle *handle,
        if (!data->nr_pages)
                goto out;
 
-       have_lost = atomic_read(&data->lost);
+       have_lost = local_read(&data->lost);
        if (have_lost)
                size += sizeof(lost_event);
 
@@ -3045,24 +3045,24 @@ int perf_output_begin(struct perf_output_handle *handle,
                 */
                tail = ACCESS_ONCE(data->user_page->data_tail);
                smp_rmb();
-               offset = head = atomic_long_read(&data->head);
+               offset = head = local_read(&data->head);
                head += size;
                if (unlikely(!perf_output_space(data, tail, offset, head)))
                        goto fail;
-       } while (atomic_long_cmpxchg(&data->head, offset, head) != offset);
+       } while (local_cmpxchg(&data->head, offset, head) != offset);
 
        handle->offset  = offset;
        handle->head    = head;
 
        if (head - tail > data->watermark)
-               atomic_inc(&data->wakeup);
+               local_inc(&data->wakeup);
 
        if (have_lost) {
                lost_event.header.type = PERF_RECORD_LOST;
                lost_event.header.misc = 0;
                lost_event.header.size = sizeof(lost_event);
                lost_event.id          = event->id;
-               lost_event.lost        = atomic_xchg(&data->lost, 0);
+               lost_event.lost        = local_xchg(&data->lost, 0);
 
                perf_output_put(handle, lost_event);
        }
@@ -3070,7 +3070,7 @@ int perf_output_begin(struct perf_output_handle *handle,
        return 0;
 
 fail:
-       atomic_inc(&data->lost);
+       local_inc(&data->lost);
        perf_output_put_handle(handle);
 out:
        rcu_read_unlock();
@@ -3086,10 +3086,10 @@ void perf_output_end(struct perf_output_handle *handle)
        int wakeup_events = event->attr.wakeup_events;
 
        if (handle->sample && wakeup_events) {
-               int events = atomic_inc_return(&data->events);
+               int events = local_inc_return(&data->events);
                if (events >= wakeup_events) {
-                       atomic_sub(wakeup_events, &data->events);
-                       atomic_inc(&data->wakeup);
+                       local_sub(wakeup_events, &data->events);
+                       local_inc(&data->wakeup);
                }
        }