* [patch 06/21] mutex subsystem, add include/asm-arm/mutex.h
@ 2006-01-04 14:42 Ingo Molnar
0 siblings, 0 replies; 3+ messages in thread
From: Ingo Molnar @ 2006-01-04 14:42 UTC (permalink / raw)
To: lkml
Cc: Linus Torvalds, Andrew Morton, Arjan van de Ven, Nicolas Pitre,
Jes Sorensen, Al Viro, Oleg Nesterov, David Howells, Alan Cox,
Christoph Hellwig, Andi Kleen, Russell King
From: Nicolas Pitre <nico@cam.org>
add the ARM version of mutex.h, which is optimized in assembly for
ARMv6, and uses the xchg implementation on pre-ARMv6.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
----
include/asm-arm/mutex.h | 128 ++++++++++++++++++++++++++++++++++++++++++++++++
1 files changed, 128 insertions(+)
Index: linux/include/asm-arm/mutex.h
===================================================================
--- /dev/null
+++ linux/include/asm-arm/mutex.h
@@ -0,0 +1,128 @@
+/*
+ * include/asm-arm/mutex.h
+ *
+ * ARM optimized mutex locking primitives
+ *
+ * Please look into asm-generic/mutex-xchg.h for a formal definition.
+ */
+#ifndef _ASM_MUTEX_H
+#define _ASM_MUTEX_H
+
+#if __LINUX_ARM_ARCH__ < 6
+/* On pre-ARMv6 hardware the swp based implementation is the most efficient. */
+# include <asm-generic/mutex-xchg.h>
+#else
+
+/*
+ * Attempting to lock a mutex on ARMv6+ can be done with a bastardized
+ * atomic decrement (it is not a reliable atomic decrement but it satisfies
+ * the defined semantics for our purpose, while being smaller and faster
+ * than a real atomic decrement or atomic swap. The idea is to attempt
+ * decrementing the lock value only once. If once decremented it isn't zero,
+ * or if its store-back fails due to a dispute on the exclusive store, we
+ * simply bail out immediately through the slow path where the lock will be
+ * reattempted until it succeeds.
+ */
+#define __mutex_fastpath_lock(count, fail_fn) \
+do { \
+ int __ex_flag, __res; \
+ \
+ typecheck(atomic_t *, count); \
+ typecheck_fn(fastcall void (*)(atomic_t *), fail_fn); \
+ \
+ __asm__ ( \
+ "ldrex %0, [%2] \n" \
+ "sub %0, %0, #1 \n" \
+ "strex %1, %0, [%2] \n" \
+ \
+ : "=&r" (__res), "=&r" (__ex_flag) \
+ : "r" (&(count)->counter) \
+ : "cc","memory" ); \
+ \
+ if (unlikely(__res || __ex_flag)) \
+ fail_fn(count); \
+} while (0)
+
+#define __mutex_fastpath_lock_retval(count, fail_fn) \
+({ \
+ int __ex_flag, __res; \
+ \
+ typecheck(atomic_t *, count); \
+ typecheck_fn(fastcall int (*)(atomic_t *), fail_fn); \
+ \
+ __asm__ ( \
+ "ldrex %0, [%2] \n" \
+ "sub %0, %0, #1 \n" \
+ "strex %1, %0, [%2] \n" \
+ \
+ : "=&r" (__res), "=&r" (__ex_flag) \
+ : "r" (&(count)->counter) \
+ : "cc","memory" ); \
+ \
+ __res |= __ex_flag; \
+ if (unlikely(__res != 0)) \
+ __res = fail_fn(count); \
+ __res; \
+})
+
+/*
+ * Same trick is used for the unlock fast path. However the original value,
+ * rather than the result, is used to test for success in order to have
+ * better generated assembly.
+ */
+#define __mutex_fastpath_unlock(count, fail_fn) \
+do { \
+ int __ex_flag, __res, __orig; \
+ \
+ typecheck(atomic_t *, count); \
+ typecheck_fn(fastcall void (*)(atomic_t *), fail_fn); \
+ \
+ __asm__ ( \
+ "ldrex %0, [%3] \n" \
+ "add %1, %0, #1 \n" \
+ "strex %2, %1, [%3] \n" \
+ \
+ : "=&r" (__orig), "=&r" (__res), "=&r" (__ex_flag) \
+ : "r" (&(count)->counter) \
+ : "cc","memory" ); \
+ \
+ if (unlikely(__orig || __ex_flag)) \
+ fail_fn(count); \
+} while (0)
+
+/*
+ * If the unlock was done on a contended lock, or if the unlock simply fails
+ * then the mutex remains locked.
+ */
+#define __mutex_slowpath_needs_to_unlock() 1
+
+/*
+ * For __mutex_fastpath_trylock we use another construct which could be
+ * described as a "single value cmpxchg".
+ *
+ * This provides the needed trylock semantics like cmpxchg would, but it is
+ * lighter and less generic than a true cmpxchg implementation.
+ */
+static inline int
+__mutex_fastpath_trylock(atomic_t *count, int (*fail_fn)(atomic_t *))
+{
+ int __ex_flag, __res, __orig;
+
+ __asm__ (
+
+ "1: ldrex %0, [%3] \n"
+ "subs %1, %0, #1 \n"
+ "strexeq %2, %1, [%3] \n"
+ "movlt %0, #0 \n"
+ "cmpeq %2, #0 \n"
+ "bgt 1b \n"
+
+ : "=&r" (__orig), "=&r" (__res), "=&r" (__ex_flag)
+ : "r" (&count->counter)
+ : "cc", "memory" );
+
+ return __orig;
+}
+
+#endif
+#endif
^ permalink raw reply [flat|nested] 3+ messages in thread
* [patch 06/21] mutex subsystem, add include/asm-arm/mutex.h
@ 2006-01-05 15:37 Ingo Molnar
2006-01-07 3:04 ` Nicolas Pitre
0 siblings, 1 reply; 3+ messages in thread
From: Ingo Molnar @ 2006-01-05 15:37 UTC (permalink / raw)
To: lkml
Cc: Linus Torvalds, Andrew Morton, Arjan van de Ven, Nicolas Pitre,
Jes Sorensen, Al Viro, Oleg Nesterov, David Howells, Alan Cox,
Christoph Hellwig, Andi Kleen, Russell King
From: Nicolas Pitre <nico@cam.org>
add the ARM version of mutex.h, which is optimized in assembly for
ARMv6, and uses the xchg implementation on pre-ARMv6.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
----
include/asm-arm/mutex.h | 128 ++++++++++++++++++++++++++++++++++++++++++++++++
1 files changed, 128 insertions(+)
Index: linux/include/asm-arm/mutex.h
===================================================================
--- /dev/null
+++ linux/include/asm-arm/mutex.h
@@ -0,0 +1,128 @@
+/*
+ * include/asm-arm/mutex.h
+ *
+ * ARM optimized mutex locking primitives
+ *
+ * Please look into asm-generic/mutex-xchg.h for a formal definition.
+ */
+#ifndef _ASM_MUTEX_H
+#define _ASM_MUTEX_H
+
+#if __LINUX_ARM_ARCH__ < 6
+/* On pre-ARMv6 hardware the swp based implementation is the most efficient. */
+# include <asm-generic/mutex-xchg.h>
+#else
+
+/*
+ * Attempting to lock a mutex on ARMv6+ can be done with a bastardized
+ * atomic decrement (it is not a reliable atomic decrement but it satisfies
+ * the defined semantics for our purpose, while being smaller and faster
+ * than a real atomic decrement or atomic swap. The idea is to attempt
+ * decrementing the lock value only once. If once decremented it isn't zero,
+ * or if its store-back fails due to a dispute on the exclusive store, we
+ * simply bail out immediately through the slow path where the lock will be
+ * reattempted until it succeeds.
+ */
+#define __mutex_fastpath_lock(count, fail_fn) \
+do { \
+ int __ex_flag, __res; \
+ \
+ typecheck(atomic_t *, count); \
+ typecheck_fn(fastcall void (*)(atomic_t *), fail_fn); \
+ \
+ __asm__ ( \
+ "ldrex %0, [%2] \n" \
+ "sub %0, %0, #1 \n" \
+ "strex %1, %0, [%2] \n" \
+ \
+ : "=&r" (__res), "=&r" (__ex_flag) \
+ : "r" (&(count)->counter) \
+ : "cc","memory" ); \
+ \
+ if (unlikely(__res || __ex_flag)) \
+ fail_fn(count); \
+} while (0)
+
+#define __mutex_fastpath_lock_retval(count, fail_fn) \
+({ \
+ int __ex_flag, __res; \
+ \
+ typecheck(atomic_t *, count); \
+ typecheck_fn(fastcall int (*)(atomic_t *), fail_fn); \
+ \
+ __asm__ ( \
+ "ldrex %0, [%2] \n" \
+ "sub %0, %0, #1 \n" \
+ "strex %1, %0, [%2] \n" \
+ \
+ : "=&r" (__res), "=&r" (__ex_flag) \
+ : "r" (&(count)->counter) \
+ : "cc","memory" ); \
+ \
+ __res |= __ex_flag; \
+ if (unlikely(__res != 0)) \
+ __res = fail_fn(count); \
+ __res; \
+})
+
+/*
+ * Same trick is used for the unlock fast path. However the original value,
+ * rather than the result, is used to test for success in order to have
+ * better generated assembly.
+ */
+#define __mutex_fastpath_unlock(count, fail_fn) \
+do { \
+ int __ex_flag, __res, __orig; \
+ \
+ typecheck(atomic_t *, count); \
+ typecheck_fn(fastcall void (*)(atomic_t *), fail_fn); \
+ \
+ __asm__ ( \
+ "ldrex %0, [%3] \n" \
+ "add %1, %0, #1 \n" \
+ "strex %2, %1, [%3] \n" \
+ \
+ : "=&r" (__orig), "=&r" (__res), "=&r" (__ex_flag) \
+ : "r" (&(count)->counter) \
+ : "cc","memory" ); \
+ \
+ if (unlikely(__orig || __ex_flag)) \
+ fail_fn(count); \
+} while (0)
+
+/*
+ * If the unlock was done on a contended lock, or if the unlock simply fails
+ * then the mutex remains locked.
+ */
+#define __mutex_slowpath_needs_to_unlock() 1
+
+/*
+ * For __mutex_fastpath_trylock we use another construct which could be
+ * described as a "single value cmpxchg".
+ *
+ * This provides the needed trylock semantics like cmpxchg would, but it is
+ * lighter and less generic than a true cmpxchg implementation.
+ */
+static inline int
+__mutex_fastpath_trylock(atomic_t *count, int (*fail_fn)(atomic_t *))
+{
+ int __ex_flag, __res, __orig;
+
+ __asm__ (
+
+ "1: ldrex %0, [%3] \n"
+ "subs %1, %0, #1 \n"
+ "strexeq %2, %1, [%3] \n"
+ "movlt %0, #0 \n"
+ "cmpeq %2, #0 \n"
+ "bgt 1b \n"
+
+ : "=&r" (__orig), "=&r" (__res), "=&r" (__ex_flag)
+ : "r" (&count->counter)
+ : "cc", "memory" );
+
+ return __orig;
+}
+
+#endif
+#endif
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [patch 06/21] mutex subsystem, add include/asm-arm/mutex.h
2006-01-05 15:37 Ingo Molnar
@ 2006-01-07 3:04 ` Nicolas Pitre
0 siblings, 0 replies; 3+ messages in thread
From: Nicolas Pitre @ 2006-01-07 3:04 UTC (permalink / raw)
To: Ingo Molnar
Cc: lkml, Linus Torvalds, Andrew Morton, Arjan van de Ven,
Jes Sorensen, Al Viro, Oleg Nesterov, David Howells, Alan Cox,
Christoph Hellwig, Andi Kleen, Russell King
On Thu, 5 Jan 2006, Ingo Molnar wrote:
> From: Nicolas Pitre <nico@cam.org>
>
> add the ARM version of mutex.h, which is optimized in assembly for
> ARMv6, and uses the xchg implementation on pre-ARMv6.
>
> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Please consider the ARM cleanups below. I was reviewing the generated
code and found out that, from my original version, you changed:
if (unlikely(__res | __ex_flag))
for:
if (unlikely(__res || __ex_flag))
which produces worse code on ARM. I therefore made it more explicit:
__res |= __ex_flag;
if (unlikely(__res != 0))
that the single | (bitwise or) was not a mistake.
I also made everything static inline rather than macros for better
readability (both produce the same code after all).
And finally added missing \t from multi-line assembly code.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Index: linux-2.6/include/asm-arm/mutex.h
===================================================================
--- linux-2.6.orig/include/asm-arm/mutex.h
+++ linux-2.6/include/asm-arm/mutex.h
@@ -23,72 +23,71 @@
* simply bail out immediately through the slow path where the lock will be
* reattempted until it succeeds.
*/
-#define __mutex_fastpath_lock(count, fail_fn) \
-do { \
- int __ex_flag, __res; \
- \
- typecheck(atomic_t *, count); \
- typecheck_fn(fastcall void (*)(atomic_t *), fail_fn); \
- \
- __asm__ ( \
- "ldrex %0, [%2] \n" \
- "sub %0, %0, #1 \n" \
- "strex %1, %0, [%2] \n" \
- \
- : "=&r" (__res), "=&r" (__ex_flag) \
- : "r" (&(count)->counter) \
- : "cc","memory" ); \
- \
- if (unlikely(__res || __ex_flag)) \
- fail_fn(count); \
-} while (0)
-
-#define __mutex_fastpath_lock_retval(count, fail_fn) \
-({ \
- int __ex_flag, __res; \
- \
- typecheck(atomic_t *, count); \
- typecheck_fn(fastcall int (*)(atomic_t *), fail_fn); \
- \
- __asm__ ( \
- "ldrex %0, [%2] \n" \
- "sub %0, %0, #1 \n" \
- "strex %1, %0, [%2] \n" \
- \
- : "=&r" (__res), "=&r" (__ex_flag) \
- : "r" (&(count)->counter) \
- : "cc","memory" ); \
- \
- __res |= __ex_flag; \
- if (unlikely(__res != 0)) \
- __res = fail_fn(count); \
- __res; \
-})
+static inline void
+__mutex_fastpath_lock(atomic_t *count, fastcall void (*fail_fn)(atomic_t *))
+{
+ int __ex_flag, __res;
+
+ __asm__ (
+
+ "ldrex %0, [%2] \n\t"
+ "sub %0, %0, #1 \n\t"
+ "strex %1, %0, [%2] "
+
+ : "=&r" (__res), "=&r" (__ex_flag)
+ : "r" (&(count)->counter)
+ : "cc","memory" );
+
+ __res |= __ex_flag;
+ if (unlikely(__res != 0))
+ fail_fn(count);
+}
+
+static inline int
+__mutex_fastpath_lock_retval(atomic_t *count, fastcall int (*fail_fn)(atomic_t *))
+{
+ int __ex_flag, __res;
+
+ __asm__ (
+
+ "ldrex %0, [%2] \n\t"
+ "sub %0, %0, #1 \n\t"
+ "strex %1, %0, [%2] "
+
+ : "=&r" (__res), "=&r" (__ex_flag)
+ : "r" (&(count)->counter)
+ : "cc","memory" );
+
+ __res |= __ex_flag;
+ if (unlikely(__res != 0))
+ __res = fail_fn(count);
+ return __res;
+}
/*
* Same trick is used for the unlock fast path. However the original value,
* rather than the result, is used to test for success in order to have
* better generated assembly.
*/
-#define __mutex_fastpath_unlock(count, fail_fn) \
-do { \
- int __ex_flag, __res, __orig; \
- \
- typecheck(atomic_t *, count); \
- typecheck_fn(fastcall void (*)(atomic_t *), fail_fn); \
- \
- __asm__ ( \
- "ldrex %0, [%3] \n" \
- "add %1, %0, #1 \n" \
- "strex %2, %1, [%3] \n" \
- \
- : "=&r" (__orig), "=&r" (__res), "=&r" (__ex_flag) \
- : "r" (&(count)->counter) \
- : "cc","memory" ); \
- \
- if (unlikely(__orig || __ex_flag)) \
- fail_fn(count); \
-} while (0)
+static inline void
+__mutex_fastpath_unlock(atomic_t *count, fastcall void (*fail_fn)(atomic_t *))
+{
+ int __ex_flag, __res, __orig;
+
+ __asm__ (
+
+ "ldrex %0, [%3] \n\t"
+ "add %1, %0, #1 \n\t"
+ "strex %2, %1, [%3] "
+
+ : "=&r" (__orig), "=&r" (__res), "=&r" (__ex_flag)
+ : "r" (&(count)->counter)
+ : "cc","memory" );
+
+ __orig |= __ex_flag;
+ if (unlikely(__orig != 0))
+ fail_fn(count);
+}
/*
* If the unlock was done on a contended lock, or if the unlock simply fails
@@ -110,12 +109,12 @@ __mutex_fastpath_trylock(atomic_t *count
__asm__ (
- "1: ldrex %0, [%3] \n"
- "subs %1, %0, #1 \n"
- "strexeq %2, %1, [%3] \n"
- "movlt %0, #0 \n"
- "cmpeq %2, #0 \n"
- "bgt 1b \n"
+ "1: ldrex %0, [%3] \n\t"
+ "subs %1, %0, #1 \n\t"
+ "strexeq %2, %1, [%3] \n\t"
+ "movlt %0, #0 \n\t"
+ "cmpeq %2, #0 \n\t"
+ "bgt 1b "
: "=&r" (__orig), "=&r" (__res), "=&r" (__ex_flag)
: "r" (&count->counter)
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2006-01-07 3:05 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-01-04 14:42 [patch 06/21] mutex subsystem, add include/asm-arm/mutex.h Ingo Molnar
-- strict thread matches above, loose matches on Subject: below --
2006-01-05 15:37 Ingo Molnar
2006-01-07 3:04 ` Nicolas Pitre
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox