[PATCH] D45576: [RFC] Allow target to handle STRICT floating-point nodes

[Ulrich Weigand via Phabricator via llvm-commits]

uweigand added a comment.

In https://reviews.llvm.org/D45576#1178328, @cameron.mcinally wrote:

> Since a and b are local to the function, the *non-strict* DIV could be hoisted out of the loop. If we assume that calls implicitly define FPC, which I think we must to model global state, wouldn't that prevent the hoist of this non-strict DIV?


Well, I was thinking that in non-strict code, calls do *not* implicitly define FPC.   Only calls in strict code (marked by the front-end when FENV_ACCESS is on at the call site) would do so.  (This is not yet implemented.)


Repository:
  rL LLVM

https://reviews.llvm.org/D45576