[PATCH] D42998: [x86/retpoline] Make the external thunk names exactly match the names that happened to end up in GCC.

[Chandler Carruth via Phabricator via llvm-commits]

chandlerc added a comment.

In https://reviews.llvm.org/D42998#1000076, @jyknight wrote:

> >> Maybe it's only a theoretical issue because we only use r11 in 64bit mode though?
> > 
> > Not even a theoretical issue. LLVM has no code path that uses a register other than r11 in 64-bit mode as an explicit design decision, and we have asserts to that effect.
>
> OK, sounds good.
>
> It was not 100% clear only from looking at this patch, as there's no immediately-visible assert that we wouldn't emit `__x86_indirect_thunk_e[acd]x` with is64Bit().


Sure, I've added asserts to help document this invariant locally.


Repository:
  rL LLVM

https://reviews.llvm.org/D42998