// Start of type verification for Maple IR nodes.
//
// General rules:
//
// 1. For binary operations, the types of the two operands must be compatible.
//
// 2. In checking type compatibility, other than identical types, the types in
// each of the following group are compatible with each other:
//            [i32, u32, ptr, ref, a32]
//            [i64, u64, ptr, ref, a64]
//
// 3. dynany is compatiable with any dyn* type.
//
// 4. u1, i8, u8, i16, u16 must not be types of arithmetic operations, because
// many machines do not provide instructions for such types as they lack such
// register sizes.  Similarly, these types must not be used as result types for
// any read instructions: dread/iread/regread/ireadoff/ireadfpoff.
//
// 5. When an opcode only specifies one type (which is called the result type),
// it expects both operands and results to be of this same type.  Thus, the
// types among the operands and this result type must be compatible.
//
// 6. When an opcode specifies two types, the additional (second) type is
// the operand type.  The types of the operands and the operand type must be
// compatible.
//
// 7. The opcodes addrof, addroflabel, addroffunc and iaddrof form addresses.
// Thus, their result type must be in [ptr,ref,a32,a64].
//
// 8. The opcodes bnot, extractbits, sext, zext, lnot must have result type in
// [i32, u32, i64, u64].
//
// 9. The opcodes abs, neg must have result type in
// [i32, u32, i64, u64, f32, f64].
//
// 10. The opcodes recip, sqrt must have result type in [f32, f64].
//
// 11. The opcodes ceil, floor, round, trunc must have result-type in
// [i32, u32, i64, u64] and operand-type in [f32, f64].
//
// 12. The opcodes add, div, sub, mpy, max, min must have result-type in
// [i32, u32, i64, u64, f32, f64].
//
// 13. The opcodes eq, ge, gt, le, lt, ne must have result-type in
// any signed or unsigned integer type; they also specifies operand-type, and
// this operand-type and the types of their two operands must be compatible.
//
// 14. The opcodes ashr, band, bior, bxor, depositbits, land, lior, lshr, shl,
// rem must have  result-type in [i32, u32, i64, u64].
//
// 15. select's result-type and the types of its second and third operands must
// be compatible; its first operand must be of integer type.
//
// 16. array's result-type must be in [ptr,ref,a32,a64]; the type of <opnd0> must
// also be in [ptr,ref,a32,a64]; the types of the rest of the operands must be in
// [i32, u32, i64, u64].
//
// 17. The operand of brfalse, trfalse must be of integer type.
//
// 18. The operand of switch, rangegoto must be in [i32, u32, i64, u64].
code