diff options
author | Will Deacon <will.deacon@arm.com> | 2014-02-21 17:01:48 +0100 |
---|---|---|
committer | Russell King <rmk+kernel@arm.linux.org.uk> | 2014-02-25 11:30:20 +0000 |
commit | c32ffce0f66e5d1d4856254516e24f5ef275cd00 (patch) | |
tree | 125229cdd38bfd6e7e62cff7eb8771a34cc999a7 /arch/arm/include/asm/cmpxchg.h | |
parent | 6ea41c80115f49e7d8b80312ffc99973d283471f (diff) | |
download | linux-c32ffce0f66e5d1d4856254516e24f5ef275cd00.tar.bz2 |
ARM: 7984/1: prefetch: add prefetchw invocations for barriered atomics
After a bunch of benchmarking on the interaction between dmb and pldw,
it turns out that issuing the pldw *after* the dmb instruction can
give modest performance gains (~3% atomic_add_return improvement on a
dual A15).
This patch adds prefetchw invocations to our barriered atomic operations
including cmpxchg, test_and_xxx and futexes.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Diffstat (limited to 'arch/arm/include/asm/cmpxchg.h')
-rw-r--r-- | arch/arm/include/asm/cmpxchg.h | 6 |
1 files changed, 6 insertions, 0 deletions
diff --git a/arch/arm/include/asm/cmpxchg.h b/arch/arm/include/asm/cmpxchg.h index df2fbba7efc8..abb2c3769b01 100644 --- a/arch/arm/include/asm/cmpxchg.h +++ b/arch/arm/include/asm/cmpxchg.h @@ -2,6 +2,7 @@ #define __ASM_ARM_CMPXCHG_H #include <linux/irqflags.h> +#include <linux/prefetch.h> #include <asm/barrier.h> #if defined(CONFIG_CPU_SA1100) || defined(CONFIG_CPU_SA110) @@ -35,6 +36,7 @@ static inline unsigned long __xchg(unsigned long x, volatile void *ptr, int size #endif smp_mb(); + prefetchw((const void *)ptr); switch (size) { #if __LINUX_ARM_ARCH__ >= 6 @@ -138,6 +140,8 @@ static inline unsigned long __cmpxchg(volatile void *ptr, unsigned long old, { unsigned long oldval, res; + prefetchw((const void *)ptr); + switch (size) { #ifndef CONFIG_CPU_V6 /* min ARCH >= ARMv6K */ case 1: @@ -230,6 +234,8 @@ static inline unsigned long long __cmpxchg64(unsigned long long *ptr, unsigned long long oldval; unsigned long res; + prefetchw(ptr); + __asm__ __volatile__( "1: ldrexd %1, %H1, [%3]\n" " teq %1, %4\n" |