[clang] [llvm] [mlir] [IR] Make semantics of strictfp consistent v2 (PR #211769)

[Ramkumar Ramachandra via cfe-commits]

================
@@ -2800,7 +2800,8 @@ fn -> other_fn -> other_fn ; fn is norecurse
     optimizations that require assumptions about the floating-point rounding
     mode or that might alter the state of floating-point status flags that
     might otherwise be set or cleared by calling this function. LLVM will
-    not introduce any new floating-point instructions that may trap.
+    not introduce any new floating-point instructions that may trap. All
+    function definitions that contain strictfp calls must be marked strictfp.
----------------
artagnon wrote:

Interesting, and I think I understand. I think the point of confusion for me is strictfp intrinsics that are not constrained FP intrinsics: wouldn't these be sensitive to FP optimizations? Would they be selected to a platform-specific FP opcode? The original motivation for the patch was actually https://github.com/llvm/llvm-project/pull/205061 -- can you kindly give feedback on canConstantFoldIntrinsic?

https://github.com/llvm/llvm-project/pull/211769