[CIR][CIRGen][Builtin][Neon] Lower vaddh_f16, vsubh_f16, vmulh_f16 and vdivh_f16 (#1258)

Lowering:

- `vaddh_f16`
- `vsubh_f16`
- `vmulh_f16`
- `vdivh_f16`

Author: AmrDeveloper

clang/lib/CIR/CodeGen/CIRGenBuilder.h

@@ -675,6 +675,14 @@ class CIRGenBuilderTy : public cir::CIRBaseBuilderTy {
     assert(!cir::MissingFeatures::foldBinOpFMF());
     return create<cir::BinOp>(lhs.getLoc(), cir::BinOpKind::Mul, lhs, rhs);
   }
+  mlir::Value createFDiv(mlir::Value lhs, mlir::Value rhs) {
+    assert(!cir::MissingFeatures::metaDataNode());
+    if (IsFPConstrained)
+      llvm_unreachable("Constrained FP NYI");
+
+    assert(!cir::MissingFeatures::foldBinOpFMF());
+    return create<cir::BinOp>(lhs.getLoc(), cir::BinOpKind::Div, lhs, rhs);
+  }
 
   mlir::Value createDynCast(mlir::Location loc, mlir::Value src,
                             cir::PointerType destType, bool isRefCast,

clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp

@@ -3771,14 +3771,22 @@ CIRGenFunction::emitAArch64BuiltinExpr(unsigned BuiltinID, const CallExpr *E,
         Ops[0], cir::VectorType::get(&getMLIRContext(), DoubleTy, 2));
     return builder.create<cir::VecExtractOp>(getLoc(E->getExprLoc()), Ops[0],
                                              emitScalarExpr(E->getArg(1)));
-  case NEON::BI__builtin_neon_vaddh_f16:
-    llvm_unreachable("NEON::BI__builtin_neon_vaddh_f16 NYI");
-  case NEON::BI__builtin_neon_vsubh_f16:
-    llvm_unreachable("NEON::BI__builtin_neon_vsubh_f16 NYI");
-  case NEON::BI__builtin_neon_vmulh_f16:
-    llvm_unreachable("NEON::BI__builtin_neon_vmulh_f16 NYI");
-  case NEON::BI__builtin_neon_vdivh_f16:
-    llvm_unreachable("NEON::BI__builtin_neon_vdivh_f16 NYI");
+  case NEON::BI__builtin_neon_vaddh_f16: {
+    Ops.push_back(emitScalarExpr(E->getArg(1)));
+    return builder.createFAdd(Ops[0], Ops[1]);
+  }
+  case NEON::BI__builtin_neon_vsubh_f16: {
+    Ops.push_back(emitScalarExpr(E->getArg(1)));
+    return builder.createFSub(Ops[0], Ops[1]);
+  }
+  case NEON::BI__builtin_neon_vmulh_f16: {
+    Ops.push_back(emitScalarExpr(E->getArg(1)));
+    return builder.createFMul(Ops[0], Ops[1]);
+  }
+  case NEON::BI__builtin_neon_vdivh_f16: {
+    Ops.push_back(emitScalarExpr(E->getArg(1)));
+    return builder.createFDiv(Ops[0], Ops[1]);
+  }
   case NEON::BI__builtin_neon_vfmah_f16:
     // NEON intrinsic puts accumulator first, unlike the LLVM fma.
     llvm_unreachable("NEON::BI__builtin_neon_vfmah_f16 NYI");