expose llvm's stack intrinsics

[oli-obk]

while I was doing this I found loads of dead code like the ArrayAlloca function. Does it make sense to remove such functions or should we just wait until someone implements code coverage testing for rustc and does this properly?

[rust-highfive]

r? @huonw

(rust_highfive has picked a reviewer for you, use r? to override)

[nagisa]

Removing dead code is fine IMO. On the other hand, I’ve already tried adding intrinsics and not immediately using them in rustc or exposing them in std several times in the past failing each time, I suspect you might fail as well.

[oli-obk]

well... My use case is a Vla and a StackBox, so we can get DSTs on the stack, but I need compiler support for it, can't do it outside since afaik there isn't such a thing as a trans-plugin. I'll create a crate, and once it's working I'll add an rfc for std

[alexcrichton]

I think that it may be more appropriate for these to go through an RFC first. The extra morestack symbols needed by the alloca function, for example, worries me that there's some more discussion which may need to happen there.

[oli-obk]

oh, sorry, didn't know adding intrinsics needed an RFC. RFC coming up