2005-04-17 06:20:36 +08:00
|
|
|
#ifndef __ASM_SPINLOCK_H
|
|
|
|
#define __ASM_SPINLOCK_H
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Simple spin lock operations.
|
|
|
|
*
|
|
|
|
* Copyright (C) 2001-2004 Paul Mackerras <paulus@au.ibm.com>, IBM
|
|
|
|
* Copyright (C) 2001 Anton Blanchard <anton@au.ibm.com>, IBM
|
|
|
|
* Copyright (C) 2002 Dave Engebretsen <engebret@us.ibm.com>, IBM
|
|
|
|
* Rework to support virtual processors
|
|
|
|
*
|
|
|
|
* Type of int is used as a full 64b word is not necessary.
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License
|
|
|
|
* as published by the Free Software Foundation; either version
|
|
|
|
* 2 of the License, or (at your option) any later version.
|
|
|
|
*/
|
|
|
|
#include <linux/config.h>
|
|
|
|
#include <asm/paca.h>
|
|
|
|
#include <asm/hvcall.h>
|
|
|
|
#include <asm/iSeries/HvCall.h>
|
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
volatile unsigned int lock;
|
|
|
|
#ifdef CONFIG_PREEMPT
|
|
|
|
unsigned int break_lock;
|
|
|
|
#endif
|
|
|
|
} spinlock_t;
|
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
volatile signed int lock;
|
|
|
|
#ifdef CONFIG_PREEMPT
|
|
|
|
unsigned int break_lock;
|
|
|
|
#endif
|
|
|
|
} rwlock_t;
|
|
|
|
|
|
|
|
#ifdef __KERNEL__
|
|
|
|
#define SPIN_LOCK_UNLOCKED (spinlock_t) { 0 }
|
|
|
|
|
|
|
|
#define spin_is_locked(x) ((x)->lock != 0)
|
|
|
|
#define spin_lock_init(x) do { *(x) = SPIN_LOCK_UNLOCKED; } while(0)
|
|
|
|
|
|
|
|
static __inline__ void _raw_spin_unlock(spinlock_t *lock)
|
|
|
|
{
|
|
|
|
__asm__ __volatile__("lwsync # spin_unlock": : :"memory");
|
|
|
|
lock->lock = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* On a system with shared processors (that is, where a physical
|
|
|
|
* processor is multiplexed between several virtual processors),
|
|
|
|
* there is no point spinning on a lock if the holder of the lock
|
|
|
|
* isn't currently scheduled on a physical processor. Instead
|
|
|
|
* we detect this situation and ask the hypervisor to give the
|
|
|
|
* rest of our timeslice to the lock holder.
|
|
|
|
*
|
|
|
|
* So that we can tell which virtual processor is holding a lock,
|
|
|
|
* we put 0x80000000 | smp_processor_id() in the lock when it is
|
|
|
|
* held. Conveniently, we have a word in the paca that holds this
|
|
|
|
* value.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#if defined(CONFIG_PPC_SPLPAR) || defined(CONFIG_PPC_ISERIES)
|
|
|
|
/* We only yield to the hypervisor if we are in shared processor mode */
|
|
|
|
#define SHARED_PROCESSOR (get_paca()->lppaca.shared_proc)
|
|
|
|
extern void __spin_yield(spinlock_t *lock);
|
|
|
|
extern void __rw_yield(rwlock_t *lock);
|
|
|
|
#else /* SPLPAR || ISERIES */
|
|
|
|
#define __spin_yield(x) barrier()
|
|
|
|
#define __rw_yield(x) barrier()
|
|
|
|
#define SHARED_PROCESSOR 0
|
|
|
|
#endif
|
|
|
|
extern void spin_unlock_wait(spinlock_t *lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This returns the old value in the lock, so we succeeded
|
|
|
|
* in getting the lock if the return value is 0.
|
|
|
|
*/
|
|
|
|
static __inline__ unsigned long __spin_trylock(spinlock_t *lock)
|
|
|
|
{
|
|
|
|
unsigned long tmp, tmp2;
|
|
|
|
|
|
|
|
__asm__ __volatile__(
|
|
|
|
" lwz %1,%3(13) # __spin_trylock\n\
|
|
|
|
1: lwarx %0,0,%2\n\
|
|
|
|
cmpwi 0,%0,0\n\
|
|
|
|
bne- 2f\n\
|
|
|
|
stwcx. %1,0,%2\n\
|
|
|
|
bne- 1b\n\
|
|
|
|
isync\n\
|
|
|
|
2:" : "=&r" (tmp), "=&r" (tmp2)
|
|
|
|
: "r" (&lock->lock), "i" (offsetof(struct paca_struct, lock_token))
|
|
|
|
: "cr0", "memory");
|
|
|
|
|
|
|
|
return tmp;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __inline__ _raw_spin_trylock(spinlock_t *lock)
|
|
|
|
{
|
|
|
|
return __spin_trylock(lock) == 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __inline__ _raw_spin_lock(spinlock_t *lock)
|
|
|
|
{
|
|
|
|
while (1) {
|
|
|
|
if (likely(__spin_trylock(lock) == 0))
|
|
|
|
break;
|
|
|
|
do {
|
|
|
|
HMT_low();
|
|
|
|
if (SHARED_PROCESSOR)
|
|
|
|
__spin_yield(lock);
|
[PATCH] ppc64: reverse prediction on spinlock busy loop code
On our raw spinlocks, we currently have an attempt at the lock, and if we do
not get it we enter a spin loop. This spinloop will likely continue for
awhile, and we pridict likely.
Shouldn't we predict that we will get out of the loop so our next instructions
are already prefetched. Even when we miss because the lock is still held, it
won't matter since we are waiting anyways.
I did a couple quick benchmarks, but the results are inconclusive.
16-way 690 running specjbb with original code
# ./specjbb 3000 16 1 1 19 30 120
...
Valid run, Score is 59282
16-way 690 running specjbb with unlikely code
# ./specjbb 3000 16 1 1 19 30 120
...
Valid run, Score is 59541
I saw a smaller increase on a JS20 (~1.6%)
JS20 specjbb w/ original code
# ./specjbb 400 2 1 1 19 30 120
...
Valid run, Score is 20460
JS20 specjbb w/ unlikely code
# ./specjbb 400 2 1 1 19 30 120
...
Valid run, Score is 20803
Anton said:
Mispredicting the spinlock busy loop also means we slow down the rate at which
we do the loads which can be good for heavily contended locks.
Note: There are some gcc issues with our default build and branch prediction,
but a CONFIG_POWER4_ONLY build should emit them correctly. I'm working with
Alan Modra on it now.
Signed-off-by: Jake Moilanen <moilanen@austin.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-01 23:58:47 +08:00
|
|
|
} while (unlikely(lock->lock != 0));
|
2005-04-17 06:20:36 +08:00
|
|
|
HMT_medium();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __inline__ _raw_spin_lock_flags(spinlock_t *lock, unsigned long flags)
|
|
|
|
{
|
|
|
|
unsigned long flags_dis;
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
if (likely(__spin_trylock(lock) == 0))
|
|
|
|
break;
|
|
|
|
local_save_flags(flags_dis);
|
|
|
|
local_irq_restore(flags);
|
|
|
|
do {
|
|
|
|
HMT_low();
|
|
|
|
if (SHARED_PROCESSOR)
|
|
|
|
__spin_yield(lock);
|
[PATCH] ppc64: reverse prediction on spinlock busy loop code
On our raw spinlocks, we currently have an attempt at the lock, and if we do
not get it we enter a spin loop. This spinloop will likely continue for
awhile, and we pridict likely.
Shouldn't we predict that we will get out of the loop so our next instructions
are already prefetched. Even when we miss because the lock is still held, it
won't matter since we are waiting anyways.
I did a couple quick benchmarks, but the results are inconclusive.
16-way 690 running specjbb with original code
# ./specjbb 3000 16 1 1 19 30 120
...
Valid run, Score is 59282
16-way 690 running specjbb with unlikely code
# ./specjbb 3000 16 1 1 19 30 120
...
Valid run, Score is 59541
I saw a smaller increase on a JS20 (~1.6%)
JS20 specjbb w/ original code
# ./specjbb 400 2 1 1 19 30 120
...
Valid run, Score is 20460
JS20 specjbb w/ unlikely code
# ./specjbb 400 2 1 1 19 30 120
...
Valid run, Score is 20803
Anton said:
Mispredicting the spinlock busy loop also means we slow down the rate at which
we do the loads which can be good for heavily contended locks.
Note: There are some gcc issues with our default build and branch prediction,
but a CONFIG_POWER4_ONLY build should emit them correctly. I'm working with
Alan Modra on it now.
Signed-off-by: Jake Moilanen <moilanen@austin.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-01 23:58:47 +08:00
|
|
|
} while (unlikely(lock->lock != 0));
|
2005-04-17 06:20:36 +08:00
|
|
|
HMT_medium();
|
|
|
|
local_irq_restore(flags_dis);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Read-write spinlocks, allowing multiple readers
|
|
|
|
* but only one writer.
|
|
|
|
*
|
|
|
|
* NOTE! it is quite common to have readers in interrupts
|
|
|
|
* but no interrupt writers. For those circumstances we
|
|
|
|
* can "mix" irq-safe locks - any writer needs to get a
|
|
|
|
* irq-safe write-lock, but readers can get non-irqsafe
|
|
|
|
* read-locks.
|
|
|
|
*/
|
|
|
|
#define RW_LOCK_UNLOCKED (rwlock_t) { 0 }
|
|
|
|
|
|
|
|
#define rwlock_init(x) do { *(x) = RW_LOCK_UNLOCKED; } while(0)
|
|
|
|
|
|
|
|
#define read_can_lock(rw) ((rw)->lock >= 0)
|
|
|
|
#define write_can_lock(rw) (!(rw)->lock)
|
|
|
|
|
|
|
|
static __inline__ void _raw_write_unlock(rwlock_t *rw)
|
|
|
|
{
|
|
|
|
__asm__ __volatile__("lwsync # write_unlock": : :"memory");
|
|
|
|
rw->lock = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This returns the old value in the lock + 1,
|
|
|
|
* so we got a read lock if the return value is > 0.
|
|
|
|
*/
|
|
|
|
static long __inline__ __read_trylock(rwlock_t *rw)
|
|
|
|
{
|
|
|
|
long tmp;
|
|
|
|
|
|
|
|
__asm__ __volatile__(
|
|
|
|
"1: lwarx %0,0,%1 # read_trylock\n\
|
|
|
|
extsw %0,%0\n\
|
|
|
|
addic. %0,%0,1\n\
|
|
|
|
ble- 2f\n\
|
|
|
|
stwcx. %0,0,%1\n\
|
|
|
|
bne- 1b\n\
|
|
|
|
isync\n\
|
|
|
|
2:" : "=&r" (tmp)
|
|
|
|
: "r" (&rw->lock)
|
|
|
|
: "cr0", "xer", "memory");
|
|
|
|
|
|
|
|
return tmp;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __inline__ _raw_read_trylock(rwlock_t *rw)
|
|
|
|
{
|
|
|
|
return __read_trylock(rw) > 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __inline__ _raw_read_lock(rwlock_t *rw)
|
|
|
|
{
|
|
|
|
while (1) {
|
|
|
|
if (likely(__read_trylock(rw) > 0))
|
|
|
|
break;
|
|
|
|
do {
|
|
|
|
HMT_low();
|
|
|
|
if (SHARED_PROCESSOR)
|
|
|
|
__rw_yield(rw);
|
[PATCH] ppc64: reverse prediction on spinlock busy loop code
On our raw spinlocks, we currently have an attempt at the lock, and if we do
not get it we enter a spin loop. This spinloop will likely continue for
awhile, and we pridict likely.
Shouldn't we predict that we will get out of the loop so our next instructions
are already prefetched. Even when we miss because the lock is still held, it
won't matter since we are waiting anyways.
I did a couple quick benchmarks, but the results are inconclusive.
16-way 690 running specjbb with original code
# ./specjbb 3000 16 1 1 19 30 120
...
Valid run, Score is 59282
16-way 690 running specjbb with unlikely code
# ./specjbb 3000 16 1 1 19 30 120
...
Valid run, Score is 59541
I saw a smaller increase on a JS20 (~1.6%)
JS20 specjbb w/ original code
# ./specjbb 400 2 1 1 19 30 120
...
Valid run, Score is 20460
JS20 specjbb w/ unlikely code
# ./specjbb 400 2 1 1 19 30 120
...
Valid run, Score is 20803
Anton said:
Mispredicting the spinlock busy loop also means we slow down the rate at which
we do the loads which can be good for heavily contended locks.
Note: There are some gcc issues with our default build and branch prediction,
but a CONFIG_POWER4_ONLY build should emit them correctly. I'm working with
Alan Modra on it now.
Signed-off-by: Jake Moilanen <moilanen@austin.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-01 23:58:47 +08:00
|
|
|
} while (unlikely(rw->lock < 0));
|
2005-04-17 06:20:36 +08:00
|
|
|
HMT_medium();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __inline__ _raw_read_unlock(rwlock_t *rw)
|
|
|
|
{
|
|
|
|
long tmp;
|
|
|
|
|
|
|
|
__asm__ __volatile__(
|
|
|
|
"eieio # read_unlock\n\
|
|
|
|
1: lwarx %0,0,%1\n\
|
|
|
|
addic %0,%0,-1\n\
|
|
|
|
stwcx. %0,0,%1\n\
|
|
|
|
bne- 1b"
|
|
|
|
: "=&r"(tmp)
|
|
|
|
: "r"(&rw->lock)
|
|
|
|
: "cr0", "memory");
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This returns the old value in the lock,
|
|
|
|
* so we got the write lock if the return value is 0.
|
|
|
|
*/
|
|
|
|
static __inline__ long __write_trylock(rwlock_t *rw)
|
|
|
|
{
|
|
|
|
long tmp, tmp2;
|
|
|
|
|
|
|
|
__asm__ __volatile__(
|
|
|
|
" lwz %1,%3(13) # write_trylock\n\
|
|
|
|
1: lwarx %0,0,%2\n\
|
|
|
|
cmpwi 0,%0,0\n\
|
|
|
|
bne- 2f\n\
|
|
|
|
stwcx. %1,0,%2\n\
|
|
|
|
bne- 1b\n\
|
|
|
|
isync\n\
|
|
|
|
2:" : "=&r" (tmp), "=&r" (tmp2)
|
|
|
|
: "r" (&rw->lock), "i" (offsetof(struct paca_struct, lock_token))
|
|
|
|
: "cr0", "memory");
|
|
|
|
|
|
|
|
return tmp;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __inline__ _raw_write_trylock(rwlock_t *rw)
|
|
|
|
{
|
|
|
|
return __write_trylock(rw) == 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __inline__ _raw_write_lock(rwlock_t *rw)
|
|
|
|
{
|
|
|
|
while (1) {
|
|
|
|
if (likely(__write_trylock(rw) == 0))
|
|
|
|
break;
|
|
|
|
do {
|
|
|
|
HMT_low();
|
|
|
|
if (SHARED_PROCESSOR)
|
|
|
|
__rw_yield(rw);
|
[PATCH] ppc64: reverse prediction on spinlock busy loop code
On our raw spinlocks, we currently have an attempt at the lock, and if we do
not get it we enter a spin loop. This spinloop will likely continue for
awhile, and we pridict likely.
Shouldn't we predict that we will get out of the loop so our next instructions
are already prefetched. Even when we miss because the lock is still held, it
won't matter since we are waiting anyways.
I did a couple quick benchmarks, but the results are inconclusive.
16-way 690 running specjbb with original code
# ./specjbb 3000 16 1 1 19 30 120
...
Valid run, Score is 59282
16-way 690 running specjbb with unlikely code
# ./specjbb 3000 16 1 1 19 30 120
...
Valid run, Score is 59541
I saw a smaller increase on a JS20 (~1.6%)
JS20 specjbb w/ original code
# ./specjbb 400 2 1 1 19 30 120
...
Valid run, Score is 20460
JS20 specjbb w/ unlikely code
# ./specjbb 400 2 1 1 19 30 120
...
Valid run, Score is 20803
Anton said:
Mispredicting the spinlock busy loop also means we slow down the rate at which
we do the loads which can be good for heavily contended locks.
Note: There are some gcc issues with our default build and branch prediction,
but a CONFIG_POWER4_ONLY build should emit them correctly. I'm working with
Alan Modra on it now.
Signed-off-by: Jake Moilanen <moilanen@austin.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-01 23:58:47 +08:00
|
|
|
} while (unlikely(rw->lock != 0));
|
2005-04-17 06:20:36 +08:00
|
|
|
HMT_medium();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif /* __KERNEL__ */
|
|
|
|
#endif /* __ASM_SPINLOCK_H */
|