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

[Junmo Park via llvm-commits]

flyingforyou added a comment.

Thanks Tim.

> Oops, this seems wrong on a more thorough reading. I don't think nGnRnE is actually that strong, though I still think the description of nGnRE is probably roughly right and prevents us from reordering volatiles. E.g. on the Reordering attribute:


I agree with your opinion. I learned a lot from this commit. 
Especially, I really appreciate Tim, Chad, Junbum for your time reviewing.

If we can't merge two volatile load, I think this commit is not worth for non-volatile and volatile load merge. How do you think about this? 
If you agree with this, I'll abandon this commit.

Junmo.


http://reviews.llvm.org/D14489