[LLVMdev] Performance regression on ARM

[Tim Northover]

> Some patch which shows this approach (and the whole problem) is attached :)

I wonder if it wouldn't be better to add an intrinsic. LLVM already
deals quite happily with emitting compiler-rt calls having different
calling conventions (& names -- it's entirely possible ARM will add an
__aeabi_mul*c3 for example). It seems a bit nasty to make Clang
understand all that too, unless we have to.

Cheers.

Tim.