[PATCH] [Power] Use AtomicExpandPass for fence insertion, and use lwsync where appropriate
wschmidt at linux.vnet.ibm.com
Fri Sep 5 08:38:46 PDT 2014
I've reviewed this against what gcc does, and your emitLeadingFence matches exactly. As you've noted, what you're doing for emitTrailingFence is conservative. GCC generates an isync in rs6000.c:rs6000_post_atomic_barrier() for these cases.
In short, the patch LGTM, and you can improve it later if desired.
More information about the llvm-commits