[ltt-dev] [URCU PATCH] Optimize caa_get_cycles() for PowerPC64

[Tulio Magno Quites Machado Filho]

Make caa_get_cycles() read from the Time Base register with only 1 instruction.

Signed-off-by: Tulio Magno Quites Machado Filho <tuliom at linux.vnet.ibm.com>
---
 urcu/arch/ppc.h |   11 +++++++++++
 1 files changed, 11 insertions(+), 0 deletions(-)

diff --git a/urcu/arch/ppc.h b/urcu/arch/ppc.h
index d7317bb..2b606d0 100644
--- a/urcu/arch/ppc.h
+++ b/urcu/arch/ppc.h
@@ -48,10 +48,20 @@ extern "C" {
 		rval;					\
 	})
 
+#define mftb()						\
+	({						\
+		unsigned long long rval;		\
+		asm volatile("mftb %0" : "=r" (rval));	\
+		rval;					\
+	})
+
 typedef unsigned long long cycles_t;
 
 static inline cycles_t caa_get_cycles (void)
 {
+#ifdef __powerpc64__
+	return (cycles_t) mftb();
+#else
 	long h, l;
 
 	for (;;) {
@@ -62,6 +72,7 @@ static inline cycles_t caa_get_cycles (void)
 		if (mftbu() == h)
 			return (((cycles_t) h) << 32) + l;
 	}
+#endif
 }
 
 #ifdef __cplusplus 
-- 
1.7.4.4