[PATCH V2 3/3] riscv: atomic: Optimize memory barrier semantics of LRSC-pairs

[guoren at kernel.org]

From: Guo Ren <guoren at linux.alibaba.com>

The current implementation is the same with 8e86f0b409a4 ("arm64:
atomics: fix use of acquire + release for full barrier semantics").
RISC-V could combine acquire and release into the AMO instructions
and it could reduce the cost of instruction in performance. Here
are the reasons for optimization:
 - Reduce one extra fence instruction
 - The "LR/SC" instruction with "acquire and release" operation is
   less cost than ACQUIRE_BARRIER/RELEASE_BARRIER which used
   precedes-loads/subsequent-stores prohibit to protect only LR/SC
   self-instruction.
 - Putting acquire/release barrier into the loop shouldn't cost
   extra performance problems from the micro-arch design view.
   Because LR and SC are sequential in the loop by RVWMO rules.

Signed-off-by: Guo Ren <guoren at linux.alibaba.com>
Signed-off-by: Guo Ren <guoren at kernel.org>
Cc: Palmer Dabbelt <palmer at dabbelt.com>
Cc: Mark Rutland <mark.rutland at arm.com>
---
 arch/riscv/include/asm/atomic.h  |  6 ++----
 arch/riscv/include/asm/cmpxchg.h | 18 ++++++------------
 2 files changed, 8 insertions(+), 16 deletions(-)

diff --git a/arch/riscv/include/asm/atomic.h b/arch/riscv/include/asm/atomic.h
index 20ce8b83bc18..4aaf5b01e7c6 100644
--- a/arch/riscv/include/asm/atomic.h
+++ b/arch/riscv/include/asm/atomic.h
@@ -382,9 +382,8 @@ static __always_inline int arch_atomic_sub_if_positive(atomic_t *v, int offset)
 		"0:	lr.w     %[p],  %[c]\n"
 		"	sub      %[rc], %[p], %[o]\n"
 		"	bltz     %[rc], 1f\n"
-		"	sc.w.rl  %[rc], %[rc], %[c]\n"
+		"	sc.w.aqrl %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		: [o]"r" (offset)
@@ -404,9 +403,8 @@ static __always_inline s64 arch_atomic64_sub_if_positive(atomic64_t *v, s64 offs
 		"0:	lr.d     %[p],  %[c]\n"
 		"	sub      %[rc], %[p], %[o]\n"
 		"	bltz     %[rc], 1f\n"
-		"	sc.d.rl  %[rc], %[rc], %[c]\n"
+		"	sc.d.aqrl %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		: [o]"r" (offset)
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
index 1af8db92250b..dfb51c98324d 100644
--- a/arch/riscv/include/asm/cmpxchg.h
+++ b/arch/riscv/include/asm/cmpxchg.h
@@ -215,9 +215,8 @@
 		__asm__ __volatile__ (					\
 			"0:	lr.w %0, %2\n"				\
 			"	bne  %0, %z3, 1f\n"			\
-			"	sc.w %1, %z4, %2\n"			\
+			"	sc.w.aq %1, %z4, %2\n"			\
 			"	bnez %1, 0b\n"				\
-			RISCV_ACQUIRE_BARRIER				\
 			"1:\n"						\
 			: "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)	\
 			: "rJ" ((long)__old), "rJ" (__new)		\
@@ -227,9 +226,8 @@
 		__asm__ __volatile__ (					\
 			"0:	lr.d %0, %2\n"				\
 			"	bne %0, %z3, 1f\n"			\
-			"	sc.d %1, %z4, %2\n"			\
+			"	sc.d.aq %1, %z4, %2\n"			\
 			"	bnez %1, 0b\n"				\
-			RISCV_ACQUIRE_BARRIER				\
 			"1:\n"						\
 			: "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)	\
 			: "rJ" (__old), "rJ" (__new)			\
@@ -259,8 +257,7 @@
 	switch (size) {							\
 	case 4:								\
 		__asm__ __volatile__ (					\
-			RISCV_RELEASE_BARRIER				\
-			"0:	lr.w %0, %2\n"				\
+			"0:	lr.w.rl %0, %2\n"			\
 			"	bne  %0, %z3, 1f\n"			\
 			"	sc.w %1, %z4, %2\n"			\
 			"	bnez %1, 0b\n"				\
@@ -271,8 +268,7 @@
 		break;							\
 	case 8:								\
 		__asm__ __volatile__ (					\
-			RISCV_RELEASE_BARRIER				\
-			"0:	lr.d %0, %2\n"				\
+			"0:	lr.d.rl %0, %2\n"			\
 			"	bne %0, %z3, 1f\n"			\
 			"	sc.d %1, %z4, %2\n"			\
 			"	bnez %1, 0b\n"				\
@@ -307,9 +303,8 @@
 		__asm__ __volatile__ (					\
 			"0:	lr.w %0, %2\n"				\
 			"	bne  %0, %z3, 1f\n"			\
-			"	sc.w.rl %1, %z4, %2\n"			\
+			"	sc.w.aqrl %1, %z4, %2\n"		\
 			"	bnez %1, 0b\n"				\
-			"	fence rw, rw\n"				\
 			"1:\n"						\
 			: "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)	\
 			: "rJ" ((long)__old), "rJ" (__new)		\
@@ -319,9 +314,8 @@
 		__asm__ __volatile__ (					\
 			"0:	lr.d %0, %2\n"				\
 			"	bne %0, %z3, 1f\n"			\
-			"	sc.d.rl %1, %z4, %2\n"			\
+			"	sc.d.aqrl %1, %z4, %2\n"		\
 			"	bnez %1, 0b\n"				\
-			"	fence rw, rw\n"				\
 			"1:\n"						\
 			: "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)	\
 			: "rJ" (__old), "rJ" (__new)			\
-- 
2.25.1