[llvm] [AMDGPU] expand-fp: Change frem expansion criterion (PR #158285)

[Matt Arsenault via llvm-commits]

================
@@ -1000,8 +1008,8 @@ static bool runImpl(Function &F, const TargetLowering &TLI,
   if (ExpandFpConvertBits != llvm::IntegerType::MAX_INT_BITS)
     MaxLegalFpConvertBitWidth = ExpandFpConvertBits;
 
-  if (MaxLegalFpConvertBitWidth >= llvm::IntegerType::MAX_INT_BITS)
----------------
arsenm wrote:

We wouldn't have to solve this problem if we could demote frem to an intrinsic. 

Given the pass has to explicitly handle certain types, you might as well just handle those ones. It gets uglier when you start to consider vectors.

https://github.com/llvm/llvm-project/pull/158285