[PATCH] Add the llvm.frameallocate and llvm.recoverframeallocation intrinsics

[Chandler Carruth]

On Mon, Dec 22, 2014 at 4:32 PM, Reid Kleckner <rnk at google.com> wrote:

> Does anyone have any high-level feedback on this idea? I think the code is
> in reasonable shape, and it would be good to get this in and iterate.


Having discussed this really extensively with you, and read through the
patch, I think this LGTM.

I think further tweaks to this design can and should happen in post-commit
review and follow-up patches. This is a reasonably invasive thing, so I
think those will come, but I don't think we need to hold this up.

If anyone ever comes up with a fundamentally better way of modeling these
concepts (accessing another function's allocations via its frame pointer)
we can always rip this stuff out. It isn't *that* invasive.

I do continue to dislike the names, but I don't have good ideas for better.
My complaints about the names are:

1) We need to make up our mind about foo_bar vs. foobar. But
llvm.frameaddress makes me not want to change course here.
2) "recoverframeallocation" is the name that just seems *weird* to me.
would just "frameallocation" be better? Unsure.

We can tweak the names right up until 3.6 branches though, so I'm not
really worried with checking this in using todays names until better ones
occur to someone.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20141222/401aad85/attachment.html>