[PATCH] D74651: Add IR constructs for inalloca replacement llvm.call.setup

[Eli Friedman via Phabricator via llvm-commits]

efriedma accepted this revision.
efriedma added a comment.
This revision is now accepted and ready to land.

LGTM.  Please give Reid a chance to comment before you merge, though.



================
Comment at: llvm/docs/LangRef.rst:1069
+    The preallocated attribute requires a type argument, which must be
+    the same as the pointee type of the argument.
 .. _attr_inalloca:
----------------
aeubanks wrote:
> efriedma wrote:
> > aeubanks wrote:
> > > efriedma wrote:
> > > > Are there any rules for what value has to be passed to a "preallocated" argument?  Or is the value ignored?
> > > It's ignored, clarified.
> > For the purpose of actually generating code, saying it's ignored is fine.  It might be inconvenient for IPO transforms/analysis, though: any transform that examines a preallocated argument would have to explicitly blacklist optimizations on the argument.  It might make sense to require that the argument has the "correct" value, even if code generation won't use it, for the sake of optimizations.
> Done, PTAL at new wording.
New wording seems fine.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D74651/new/

https://reviews.llvm.org/D74651