2023-07-19 07:21:43

by Guo Ren

[permalink] [raw]
Subject: [PATCH] asm-generic: ticket-lock: Optimize arch_spin_value_unlocked

From: Guo Ren <[email protected]>

Using arch_spinlock_is_locked would cause another unnecessary memory
access to the contended value. Although it won't cause a significant
performance gap in most architectures, the arch_spin_value_unlocked
argument contains enough information. Thus, remove unnecessary
atomic_read in arch_spin_value_unlocked().

Signed-off-by: Guo Ren <[email protected]>
Signed-off-by: Guo Ren <[email protected]>
Cc: David Laight <[email protected]>
Cc: Peter Zijlstra <[email protected]>
---
Changelog:
This patch is separate from:
https://lore.kernel.org/linux-riscv/[email protected]/

Peter & David have commented on it:
https://lore.kernel.org/linux-riscv/[email protected]/
---
include/asm-generic/spinlock.h | 16 +++++++++-------
1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/include/asm-generic/spinlock.h b/include/asm-generic/spinlock.h
index fdfebcb050f4..90803a826ba0 100644
--- a/include/asm-generic/spinlock.h
+++ b/include/asm-generic/spinlock.h
@@ -68,11 +68,18 @@ static __always_inline void arch_spin_unlock(arch_spinlock_t *lock)
smp_store_release(ptr, (u16)val + 1);
}

+static __always_inline int arch_spin_value_unlocked(arch_spinlock_t lock)
+{
+ u32 val = lock.counter;
+
+ return ((val >> 16) == (val & 0xffff));
+}
+
static __always_inline int arch_spin_is_locked(arch_spinlock_t *lock)
{
- u32 val = atomic_read(lock);
+ arch_spinlock_t val = READ_ONCE(*lock);

- return ((val >> 16) != (val & 0xffff));
+ return !arch_spin_value_unlocked(val);
}

static __always_inline int arch_spin_is_contended(arch_spinlock_t *lock)
@@ -82,11 +89,6 @@ static __always_inline int arch_spin_is_contended(arch_spinlock_t *lock)
return (s16)((val >> 16) - (val & 0xffff)) > 1;
}

-static __always_inline int arch_spin_value_unlocked(arch_spinlock_t lock)
-{
- return !arch_spin_is_locked(&lock);
-}
-
#include <asm/qrwlock.h>

#endif /* __ASM_GENERIC_SPINLOCK_H */
--
2.36.1



2023-07-23 02:51:43

by Waiman Long

[permalink] [raw]
Subject: Re: [PATCH] asm-generic: ticket-lock: Optimize arch_spin_value_unlocked

On 7/19/23 03:00, [email protected] wrote:
> From: Guo Ren <[email protected]>
>
> Using arch_spinlock_is_locked would cause another unnecessary memory
> access to the contended value. Although it won't cause a significant
> performance gap in most architectures, the arch_spin_value_unlocked
> argument contains enough information. Thus, remove unnecessary
> atomic_read in arch_spin_value_unlocked().

AFAICS, only one memory access is needed for the current
arch_spinlock_is_locked(). So your description isn't quite right. OTOH,
caller of arch_spin_value_unlocked() could benefit from this change.
Currently, the only caller is lockref.

Other than that, the patch looks good to me.

Cheers,
Longman

>
> Signed-off-by: Guo Ren <[email protected]>
> Signed-off-by: Guo Ren <[email protected]>
> Cc: David Laight <[email protected]>
> Cc: Peter Zijlstra <[email protected]>
> ---
> Changelog:
> This patch is separate from:
> https://lore.kernel.org/linux-riscv/[email protected]/
>
> Peter & David have commented on it:
> https://lore.kernel.org/linux-riscv/[email protected]/
> ---
> include/asm-generic/spinlock.h | 16 +++++++++-------
> 1 file changed, 9 insertions(+), 7 deletions(-)
>
> diff --git a/include/asm-generic/spinlock.h b/include/asm-generic/spinlock.h
> index fdfebcb050f4..90803a826ba0 100644
> --- a/include/asm-generic/spinlock.h
> +++ b/include/asm-generic/spinlock.h
> @@ -68,11 +68,18 @@ static __always_inline void arch_spin_unlock(arch_spinlock_t *lock)
> smp_store_release(ptr, (u16)val + 1);
> }
>
> +static __always_inline int arch_spin_value_unlocked(arch_spinlock_t lock)
> +{
> + u32 val = lock.counter;
> +
> + return ((val >> 16) == (val & 0xffff));
> +}
> +
> static __always_inline int arch_spin_is_locked(arch_spinlock_t *lock)
> {
> - u32 val = atomic_read(lock);
> + arch_spinlock_t val = READ_ONCE(*lock);
>
> - return ((val >> 16) != (val & 0xffff));
> + return !arch_spin_value_unlocked(val);
> }
>
> static __always_inline int arch_spin_is_contended(arch_spinlock_t *lock)
> @@ -82,11 +89,6 @@ static __always_inline int arch_spin_is_contended(arch_spinlock_t *lock)
> return (s16)((val >> 16) - (val & 0xffff)) > 1;
> }
>
> -static __always_inline int arch_spin_value_unlocked(arch_spinlock_t lock)
> -{
> - return !arch_spin_is_locked(&lock);
> -}
> -
> #include <asm/qrwlock.h>
>
> #endif /* __ASM_GENERIC_SPINLOCK_H */


2023-07-31 03:33:30

by Guo Ren

[permalink] [raw]
Subject: Re: [PATCH] asm-generic: ticket-lock: Optimize arch_spin_value_unlocked

On Sat, Jul 22, 2023 at 10:07:19PM -0400, Waiman Long wrote:
> On 7/19/23 03:00, [email protected] wrote:
> > From: Guo Ren <[email protected]>
> >
> > Using arch_spinlock_is_locked would cause another unnecessary memory
> > access to the contended value. Although it won't cause a significant
> > performance gap in most architectures, the arch_spin_value_unlocked
> > argument contains enough information. Thus, remove unnecessary
> > atomic_read in arch_spin_value_unlocked().
>
> AFAICS, only one memory access is needed for the current
> arch_spinlock_is_locked(). So your description isn't quite right. OTOH,
Okay, I would improve it. Here means "arch_spin_value_unlocked using
arch_spinlock_is_locked" would cause "an" unnecessary ...

> caller of arch_spin_value_unlocked() could benefit from this change.
> Currently, the only caller is lockref.
Thx for comment, I would add it in the commit msg.

New version is here:
https://lore.kernel.org/linux-riscv/[email protected]/

>
> Other than that, the patch looks good to me.
>
> Cheers,
> Longman
>
> >
> > Signed-off-by: Guo Ren <[email protected]>
> > Signed-off-by: Guo Ren <[email protected]>
> > Cc: David Laight <[email protected]>
> > Cc: Peter Zijlstra <[email protected]>
> > ---
> > Changelog:
> > This patch is separate from:
> > https://lore.kernel.org/linux-riscv/[email protected]/
> >
> > Peter & David have commented on it:
> > https://lore.kernel.org/linux-riscv/[email protected]/
> > ---
> > include/asm-generic/spinlock.h | 16 +++++++++-------
> > 1 file changed, 9 insertions(+), 7 deletions(-)
> >
> > diff --git a/include/asm-generic/spinlock.h b/include/asm-generic/spinlock.h
> > index fdfebcb050f4..90803a826ba0 100644
> > --- a/include/asm-generic/spinlock.h
> > +++ b/include/asm-generic/spinlock.h
> > @@ -68,11 +68,18 @@ static __always_inline void arch_spin_unlock(arch_spinlock_t *lock)
> > smp_store_release(ptr, (u16)val + 1);
> > }
> > +static __always_inline int arch_spin_value_unlocked(arch_spinlock_t lock)
> > +{
> > + u32 val = lock.counter;
> > +
> > + return ((val >> 16) == (val & 0xffff));
> > +}
> > +
> > static __always_inline int arch_spin_is_locked(arch_spinlock_t *lock)
> > {
> > - u32 val = atomic_read(lock);
> > + arch_spinlock_t val = READ_ONCE(*lock);
> > - return ((val >> 16) != (val & 0xffff));
> > + return !arch_spin_value_unlocked(val);
> > }
> > static __always_inline int arch_spin_is_contended(arch_spinlock_t *lock)
> > @@ -82,11 +89,6 @@ static __always_inline int arch_spin_is_contended(arch_spinlock_t *lock)
> > return (s16)((val >> 16) - (val & 0xffff)) > 1;
> > }
> > -static __always_inline int arch_spin_value_unlocked(arch_spinlock_t lock)
> > -{
> > - return !arch_spin_is_locked(&lock);
> > -}
> > -
> > #include <asm/qrwlock.h>
> > #endif /* __ASM_GENERIC_SPINLOCK_H */
>
>
> _______________________________________________
> linux-riscv mailing list
> [email protected]
> http://lists.infradead.org/mailman/listinfo/linux-riscv
>