2023-07-03 15:25:57

by Uros Bizjak

Subject: [PATCH] drm/amdgpu: Use local64_try_cmpxchg in amdgpu_perf_read

Use local64_try_cmpxchg instead of local64_cmpxchg (*ptr, old, new) == old
in amdgpu_perf_read. x86 CMPXCHG instruction returns success in ZF flag,
so this change saves a compare after cmpxchg (and related move instruction
in front of cmpxchg).

Also, try_cmpxchg implicitly assigns old *ptr value to "old" when cmpxchg
fails. There is no need to re-read the value in the loop.

No functional change intended.

Cc: Alex Deucher <[email protected]>
Cc: "Christian König" <[email protected]>
Cc: "Pan, Xinhui" <[email protected]>
Cc: David Airlie <[email protected]>
Cc: Daniel Vetter <[email protected]>
Signed-off-by: Uros Bizjak <[email protected]>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
index 71ee361d0972..6e91ea1de5aa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
@@ -276,9 +276,8 @@ static void amdgpu_perf_read(struct perf_event *event)
(!pe->adev->df.funcs->pmc_get_count))
return;

+ prev = local64_read(&hwc->prev_count);
do {
- prev = local64_read(&hwc->prev_count);
-
switch (hwc->config_base) {
case AMDGPU_PMU_EVENT_CONFIG_TYPE_DF:
case AMDGPU_PMU_EVENT_CONFIG_TYPE_XGMI:
@@ -289,7 +288,7 @@ static void amdgpu_perf_read(struct perf_event *event)
count = 0;
break;
}
- } while (local64_cmpxchg(&hwc->prev_count, prev, count) != prev);
+ } while (!local64_try_cmpxchg(&hwc->prev_count, &prev, count));

local64_add(count - prev, &event->count);
}
--
2.41.0

2023-08-09 23:24:50

by Alex Deucher

Subject: Re: [PATCH] drm/amdgpu: Use local64_try_cmpxchg in amdgpu_perf_read

Applied. Thanks!

Alex

On Mon, Jul 3, 2023 at 7:16 PM Uros Bizjak <[email protected]> wrote:
>
> Use local64_try_cmpxchg instead of local64_cmpxchg (*ptr, old, new) == old
> in amdgpu_perf_read. x86 CMPXCHG instruction returns success in ZF flag,
> so this change saves a compare after cmpxchg (and related move instruction
> in front of cmpxchg).
>
> Also, try_cmpxchg implicitly assigns old *ptr value to "old" when cmpxchg
> fails. There is no need to re-read the value in the loop.
>
> No functional change intended.
>
> Cc: Alex Deucher <[email protected]>
> Cc: "Christian König" <[email protected]>
> Cc: "Pan, Xinhui" <[email protected]>
> Cc: David Airlie <[email protected]>
> Cc: Daniel Vetter <[email protected]>
> Signed-off-by: Uros Bizjak <[email protected]>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c | 5 ++---
> 1 file changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
> index 71ee361d0972..6e91ea1de5aa 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
> @@ -276,9 +276,8 @@ static void amdgpu_perf_read(struct perf_event *event)
> (!pe->adev->df.funcs->pmc_get_count))
> return;
>
> + prev = local64_read(&hwc->prev_count);
> do {
> - prev = local64_read(&hwc->prev_count);
> -
> switch (hwc->config_base) {
> case AMDGPU_PMU_EVENT_CONFIG_TYPE_DF:
> case AMDGPU_PMU_EVENT_CONFIG_TYPE_XGMI:
> @@ -289,7 +288,7 @@ static void amdgpu_perf_read(struct perf_event *event)
> count = 0;
> break;
> }
> - } while (local64_cmpxchg(&hwc->prev_count, prev, count) != prev);
> + } while (!local64_try_cmpxchg(&hwc->prev_count, &prev, count));
>
> local64_add(count - prev, &event->count);
> }
> --
> 2.41.0
>