llvm-project

ade0e1a4 - [cir] Refine cir::CastOp semantics for int <-> float casts (#200005)

[cir] Refine cir::CastOp semantics for int <-> float casts (#200005) Int-to-float and float-to-int casts in cir::CastOp are lowered directly to their LLVM equivalents. Update the verifier to reflect this semantics and ensure that, for vector casts, the source and destination vectors have the same length. This lets the CIR verifier reject invalid casts earlier, instead of relying on errors reported later at the LLVM IR level.