2024-04-25 10:19:23

by Uros Bizjak

Subject: [PATCH] perf/x86/amd: Use try_cmpxchg() in events/amd/{un,}core.c

Replace this pattern in events/amd/{un,}core.c:

cmpxchg(*ptr, old, new) == old

.. with the simpler and faster:

try_cmpxchg(*ptr, &old, new)

The x86 CMPXCHG instruction returns success in the ZF flag, so this change
saves a compare after the CMPXCHG.

No functional change intended.

Signed-off-by: Uros Bizjak <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
---
arch/x86/events/amd/core.c | 3 ++-
arch/x86/events/amd/uncore.c | 6 ++++--
2 files changed, 6 insertions(+), 3 deletions(-)

diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index 1fc4ce44e743..c1b8aeff7bc0 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -433,7 +433,8 @@ static void __amd_put_nb_event_constraints(struct cpu_hw_events *cpuc,
* when we come here
*/
for (i = 0; i < x86_pmu.num_counters; i++) {
- if (cmpxchg(nb->owners + i, event, NULL) == event)
+ struct perf_event *tmp = event;
+ if (try_cmpxchg(nb->owners + i, &tmp, NULL))
break;
}
}
diff --git a/arch/x86/events/amd/uncore.c b/arch/x86/events/amd/uncore.c
index 4ccb8fa483e6..cac7c20d1aee 100644
--- a/arch/x86/events/amd/uncore.c
+++ b/arch/x86/events/amd/uncore.c
@@ -162,7 +162,8 @@ static int amd_uncore_add(struct perf_event *event, int flags)
/* if not, take the first available counter */
hwc->idx = -1;
for (i = 0; i < pmu->num_counters; i++) {
- if (cmpxchg(&ctx->events[i], NULL, event) == NULL) {
+ struct perf_event *tmp = NULL;
+ if (try_cmpxchg(&ctx->events[i], &tmp, event)) {
hwc->idx = i;
break;
}
@@ -196,7 +197,8 @@ static void amd_uncore_del(struct perf_event *event, int flags)
event->pmu->stop(event, PERF_EF_UPDATE);

for (i = 0; i < pmu->num_counters; i++) {
- if (cmpxchg(&ctx->events[i], event, NULL) == event)
+ struct perf_event *tmp = event;
+ if (try_cmpxchg(&ctx->events[i], &tmp, NULL))
break;
}

--
2.44.0

2024-05-18 09:16:00

by Ingo Molnar

Subject: Re: [PATCH] perf/x86/amd: Use try_cmpxchg() in events/amd/{un,}core.c

* Uros Bizjak <[email protected]> wrote:

> Replace this pattern in events/amd/{un,}core.c:
>
> cmpxchg(*ptr, old, new) == old
>
> ... with the simpler and faster:
>
> try_cmpxchg(*ptr, &old, new)
>
> The x86 CMPXCHG instruction returns success in the ZF flag, so this change
> saves a compare after the CMPXCHG.
>
> No functional change intended.
>
> Signed-off-by: Uros Bizjak <[email protected]>
> Cc: Peter Zijlstra <[email protected]>
> Cc: Ingo Molnar <[email protected]>
> Cc: Arnaldo Carvalho de Melo <[email protected]>
> Cc: Namhyung Kim <[email protected]>
> Cc: Mark Rutland <[email protected]>
> Cc: Alexander Shishkin <[email protected]>
> Cc: Jiri Olsa <[email protected]>
> Cc: Ian Rogers <[email protected]>
> Cc: Adrian Hunter <[email protected]>
> Cc: Thomas Gleixner <[email protected]>
> Cc: Borislav Petkov <[email protected]>
> Cc: Dave Hansen <[email protected]>
> Cc: "H. Peter Anvin" <[email protected]>
> ---
> arch/x86/events/amd/core.c | 3 ++-
> arch/x86/events/amd/uncore.c | 6 ++++--
> 2 files changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
> index 1fc4ce44e743..c1b8aeff7bc0 100644
> --- a/arch/x86/events/amd/core.c
> +++ b/arch/x86/events/amd/core.c
> @@ -433,7 +433,8 @@ static void __amd_put_nb_event_constraints(struct cpu_hw_events *cpuc,
> * when we come here
> */
> for (i = 0; i < x86_pmu.num_counters; i++) {
> - if (cmpxchg(nb->owners + i, event, NULL) == event)
> + struct perf_event *tmp = event;
> + if (try_cmpxchg(nb->owners + i, &tmp, NULL))
> break;
> }
> }
> diff --git a/arch/x86/events/amd/uncore.c b/arch/x86/events/amd/uncore.c
> index 4ccb8fa483e6..cac7c20d1aee 100644
> --- a/arch/x86/events/amd/uncore.c
> +++ b/arch/x86/events/amd/uncore.c
> @@ -162,7 +162,8 @@ static int amd_uncore_add(struct perf_event *event, int flags)
> /* if not, take the first available counter */
> hwc->idx = -1;
> for (i = 0; i < pmu->num_counters; i++) {
> - if (cmpxchg(&ctx->events[i], NULL, event) == NULL) {
> + struct perf_event *tmp = NULL;
> + if (try_cmpxchg(&ctx->events[i], &tmp, event)) {
> hwc->idx = i;
> break;
> }
> @@ -196,7 +197,8 @@ static void amd_uncore_del(struct perf_event *event, int flags)
> event->pmu->stop(event, PERF_EF_UPDATE);
>
> for (i = 0; i < pmu->num_counters; i++) {
> - if (cmpxchg(&ctx->events[i], event, NULL) == event)
> + struct perf_event *tmp = event;
> + if (try_cmpxchg(&ctx->events[i], &tmp, NULL))
> break;
> }

Note that's a newline missing between local variable declaration block and
the first statement after that. I've added all 3 to this patch (see below),
but please keep an eye on this in future patches.

Thanks,

Ingo

================>
arch/x86/events/amd/core.c | 1 +
arch/x86/events/amd/uncore.c | 2 ++
2 files changed, 3 insertions(+)

diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index c1b8aeff7bc0..18bfe3451f3a 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -434,6 +434,7 @@ static void __amd_put_nb_event_constraints(struct cpu_hw_events *cpuc,
*/
for (i = 0; i < x86_pmu.num_counters; i++) {
struct perf_event *tmp = event;
+
if (try_cmpxchg(nb->owners + i, &tmp, NULL))
break;
}
diff --git a/arch/x86/events/amd/uncore.c b/arch/x86/events/amd/uncore.c
index cac7c20d1aee..0fafe233bba4 100644
--- a/arch/x86/events/amd/uncore.c
+++ b/arch/x86/events/amd/uncore.c
@@ -163,6 +163,7 @@ static int amd_uncore_add(struct perf_event *event, int flags)
hwc->idx = -1;
for (i = 0; i < pmu->num_counters; i++) {
struct perf_event *tmp = NULL;
+
if (try_cmpxchg(&ctx->events[i], &tmp, event)) {
hwc->idx = i;
break;
@@ -198,6 +199,7 @@ static void amd_uncore_del(struct perf_event *event, int flags)

for (i = 0; i < pmu->num_counters; i++) {
struct perf_event *tmp = event;
+
if (try_cmpxchg(&ctx->events[i], &tmp, NULL))
break;
}

2024-05-18 09:43:50

by tip-bot2 for Alexey Makhalov

Subject: [tip: perf/urgent] perf/x86/amd: Use try_cmpxchg() in events/amd/{un,}core.c

The following commit has been merged into the perf/urgent branch of tip:

Commit-ID: cd84351c8c1baec86342d784feb884ace007d51c
Gitweb: https://git.kernel.org/tip/cd84351c8c1baec86342d784feb884ace007d51c
Author: Uros Bizjak <[email protected]>
AuthorDate: Thu, 25 Apr 2024 12:16:14 +02:00
Committer: Ingo Molnar <[email protected]>
CommitterDate: Sat, 18 May 2024 11:15:13 +02:00

perf/x86/amd: Use try_cmpxchg() in events/amd/{un,}core.c

Replace this pattern in events/amd/{un,}core.c:

cmpxchg(*ptr, old, new) == old

.. with the simpler and faster:

try_cmpxchg(*ptr, &old, new)

The x86 CMPXCHG instruction returns success in the ZF flag, so this change
saves a compare after the CMPXCHG.

No functional change intended.

Signed-off-by: Uros Bizjak <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
---
arch/x86/events/amd/core.c | 4 +++-
arch/x86/events/amd/uncore.c | 8 ++++++--
2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index 1fc4ce4..18bfe34 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -433,7 +433,9 @@ static void __amd_put_nb_event_constraints(struct cpu_hw_events *cpuc,
* when we come here
*/
for (i = 0; i < x86_pmu.num_counters; i++) {
- if (cmpxchg(nb->owners + i, event, NULL) == event)
+ struct perf_event *tmp = event;
+
+ if (try_cmpxchg(nb->owners + i, &tmp, NULL))
break;
}
}
diff --git a/arch/x86/events/amd/uncore.c b/arch/x86/events/amd/uncore.c
index 4ccb8fa..0fafe23 100644
--- a/arch/x86/events/amd/uncore.c
+++ b/arch/x86/events/amd/uncore.c
@@ -162,7 +162,9 @@ static int amd_uncore_add(struct perf_event *event, int flags)
/* if not, take the first available counter */
hwc->idx = -1;
for (i = 0; i < pmu->num_counters; i++) {
- if (cmpxchg(&ctx->events[i], NULL, event) == NULL) {
+ struct perf_event *tmp = NULL;
+
+ if (try_cmpxchg(&ctx->events[i], &tmp, event)) {
hwc->idx = i;
break;
}
@@ -196,7 +198,9 @@ static void amd_uncore_del(struct perf_event *event, int flags)
event->pmu->stop(event, PERF_EF_UPDATE);

for (i = 0; i < pmu->num_counters; i++) {
- if (cmpxchg(&ctx->events[i], event, NULL) == event)
+ struct perf_event *tmp = event;
+
+ if (try_cmpxchg(&ctx->events[i], &tmp, NULL))
break;
}