2013-09-24 22:22:35

by Tim Chen

Subject: [PATCH v5 1/6] rwsem: check the lock before cpmxchg in down_write_trylock

Cmpxchg will cause the cacheline bouning when do the value checking,
that cause scalability issue in a large machine (like a 80 core box).

So a lock pre-read can relief this contention.

Signed-off-by: Alex Shi <[email protected]>
---
include/asm-generic/rwsem.h | 8 ++++----
1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/include/asm-generic/rwsem.h b/include/asm-generic/rwsem.h
index bb1e2cd..5ba80e7 100644
--- a/include/asm-generic/rwsem.h
+++ b/include/asm-generic/rwsem.h
@@ -70,11 +70,11 @@ static inline void __down_write(struct rw_semaphore *sem)

static inline int __down_write_trylock(struct rw_semaphore *sem)
{
- long tmp;
+ if (unlikely(sem->count != RWSEM_UNLOCKED_VALUE))
+ return 0;

- tmp = cmpxchg(&sem->count, RWSEM_UNLOCKED_VALUE,
- RWSEM_ACTIVE_WRITE_BIAS);
- return tmp == RWSEM_UNLOCKED_VALUE;
+ return cmpxchg(&sem->count, RWSEM_UNLOCKED_VALUE,
+ RWSEM_ACTIVE_WRITE_BIAS) == RWSEM_UNLOCKED_VALUE;
}

/*
--
1.7.4.4

2013-09-24 23:22:56

by Jason Low

Subject: Re: [PATCH v5 1/6] rwsem: check the lock before cpmxchg in down_write_trylock

Should we do something similar with __down_read_trylock, such as
the following?


Signed-off-by: Jason Low <[email protected]>
---
include/asm-generic/rwsem.h | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/include/asm-generic/rwsem.h b/include/asm-generic/rwsem.h
index bb1e2cd..47990dc 100644
--- a/include/asm-generic/rwsem.h
+++ b/include/asm-generic/rwsem.h
@@ -42,6 +42,9 @@ static inline int __down_read_trylock(struct
rw_semaphore *sem)
long tmp;

while ((tmp = sem->count) >= 0) {
+ if (sem->count != tmp)
+ continue;
+
if (tmp == cmpxchg(&sem->count, tmp,
tmp + RWSEM_ACTIVE_READ_BIAS)) {
return 1;
--
1.7.1

On Tue, Sep 24, 2013 at 3:22 PM, Tim Chen <[email protected]> wrote:
> Cmpxchg will cause the cacheline bouning when do the value checking,
> that cause scalability issue in a large machine (like a 80 core box).
>
> So a lock pre-read can relief this contention.
>
> Signed-off-by: Alex Shi <[email protected]>
> ---
> include/asm-generic/rwsem.h | 8 ++++----
> 1 files changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/include/asm-generic/rwsem.h b/include/asm-generic/rwsem.h
> index bb1e2cd..5ba80e7 100644
> --- a/include/asm-generic/rwsem.h
> +++ b/include/asm-generic/rwsem.h
> @@ -70,11 +70,11 @@ static inline void __down_write(struct rw_semaphore *sem)
>
> static inline int __down_write_trylock(struct rw_semaphore *sem)
> {
> - long tmp;
> + if (unlikely(sem->count != RWSEM_UNLOCKED_VALUE))
> + return 0;
>
> - tmp = cmpxchg(&sem->count, RWSEM_UNLOCKED_VALUE,
> - RWSEM_ACTIVE_WRITE_BIAS);
> - return tmp == RWSEM_UNLOCKED_VALUE;
> + return cmpxchg(&sem->count, RWSEM_UNLOCKED_VALUE,
> + RWSEM_ACTIVE_WRITE_BIAS) == RWSEM_UNLOCKED_VALUE;
> }
>
> /*
> --
> 1.7.4.4
>
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/

2013-09-24 23:30:27

by Tim Chen

Subject: Re: [PATCH v5 1/6] rwsem: check the lock before cpmxchg in down_write_trylock

On Tue, 2013-09-24 at 16:22 -0700, Jason Low wrote:
> Should we do something similar with __down_read_trylock, such as
> the following?
>
>
> Signed-off-by: Jason Low <[email protected]>
> ---
> include/asm-generic/rwsem.h | 3 +++
> 1 files changed, 3 insertions(+), 0 deletions(-)
>
> diff --git a/include/asm-generic/rwsem.h b/include/asm-generic/rwsem.h
> index bb1e2cd..47990dc 100644
> --- a/include/asm-generic/rwsem.h
> +++ b/include/asm-generic/rwsem.h
> @@ -42,6 +42,9 @@ static inline int __down_read_trylock(struct
> rw_semaphore *sem)
> long tmp;
>
> while ((tmp = sem->count) >= 0) {
> + if (sem->count != tmp)
> + continue;
> +

Considering that tmp has just been assigned the value of sem->count, the
added if check failure is unlikely and probably not needed. We should
proceed to cmpxchg below.

> if (tmp == cmpxchg(&sem->count, tmp,
> tmp + RWSEM_ACTIVE_READ_BIAS)) {
> return 1;

Tim