From 8c34424b3a10eef507655b3304297d96ad71f804 Mon Sep 17 00:00:00 2001 From: Ivan Maidanski Date: Fri, 18 Nov 2016 18:22:00 +0300 Subject: [PATCH] Refine AO_nop_write comment for ARM big.LITTLE architecture * src/atomic_ops/sysdeps/gcc/arm.h [!AO_UNIPROCESSOR && AO_ARM_HAVE_DMB && !AO_PREFER_BUILTIN_ATOMICS] (AO_nop_write): Refine comment about "dmb ishst" for big.LITTLE architecture. --- src/atomic_ops/sysdeps/gcc/arm.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/atomic_ops/sysdeps/gcc/arm.h b/src/atomic_ops/sysdeps/gcc/arm.h index f548aa1..633ed3b 100644 --- a/src/atomic_ops/sysdeps/gcc/arm.h +++ b/src/atomic_ops/sysdeps/gcc/arm.h @@ -108,7 +108,8 @@ { /* AO_THUMB_GO_ARM is empty. */ /* This will target the system domain and thus be overly */ - /* conservative as the CPUs will occupy the inner shareable domain. */ + /* conservative as the CPUs (even in case of big.LITTLE SoC) will */ + /* occupy the inner shareable domain. */ /* The plain variant (dmb st) is theoretically slower, and should */ /* not be needed. That said, with limited experimentation, a CPU */ /* implementation for which it actually matters has not been found */ -- 2.40.0