llvm-project

9db4071a - [mlir][spirv-to-llvm] Add conversion for IsInf, IsNan, IsFinite (#232937)

[mlir][spirv-to-llvm] Add conversion for IsInf, IsNan, IsFinite (#232937) The SPIR-V floating-point classification ops are lowered to the `llvm.intr.is.fpclass` intrinsic with the matching `llvm::FPClassTest` mask: `spirv.IsInf` -> `fcInf`, `spirv.IsNan` -> `fcNan`, `spirv.IsFinite` -> `fcFinite` and `spirv.IsNormal` -> `fcNormal`.