[PATCH] D14489: [AArch64] Applying load pair optimization for volatile load/store

[Tim Northover via llvm-commits]

t.p.northover added a comment.

FWIW, I think both loads would have to be volatile to prohibit that if the underlying principle was sound. We're allowed to reorder volatile accesses with non-volatile, but not 2 volatile accesses.


http://reviews.llvm.org/D14489