HIR to CLIF lowering rules

Goals

• CLIF is the only executable IR.
• One lowering path for JIT and AOT.
• Lowering input is Spanned<HIRProgram> produced by the phase-indexed shared-core AST/HIR model.

Boundary requirements

• Lowering must remain backend-neutral for language semantics.
• Runtime/system operations must lower through runtime ABI entrypoints only.
• Lowering must not encode platform syscall policy.

Function lowering

• Each HIRFunctionDefinition becomes one Cranelift Function with a signature from the HIR type.
• Parameters become CLIF block params in the entry block.

Control flow

• if: entry -> then/else blocks -> merge block.
• while: loop header -> body -> backedge to header.
• match: chain of test blocks; later jump-table for enums.
• for range(...): lowers through numeric fast-path.
• for iterator_expr: lowers through Next() loop until Option::None.

Locals and variables

• Locals map to Variable in FunctionBuilder.
• def_var used at first assignment; use_var at reads.

Aggregates

• Struct allocation: alloc builtin returns pointer; field_set via stores.
• Arrays/strings use runtime builtins (array_new, str_new).

Calls

• Calls use call with signature registered via Module::declare_function.
• Runtime builtins are imported via declare_func_in_func.
• Lowering dispatches calls by typed call classification metadata (MethodDispatch, ItemCall, CallableValueCall).
• Method dispatch uses one canonical lowering path independent of source syntax shape.

Returns

• Use return terminator with explicit values.

Source locations

• Use set_srcloc for diagnostics mapping.

Non-goals

• Selecting linker strategy or artifact packaging format.
• Defining runtime ABI versioning policy.