MIR Optimization: Constant Folding & Propagation

## Summary Implement constant folding and propagation on MIR to evaluate constant expressions at compile time. **Parent issue:** #687 ## Context Constant folding evaluates operations with constant operands at compile time: - `add(5, 3)` → `8` - `mul(x, 0)` → `0` (even if x is not constant) - `and(x, 0xFF)` where x is `uint8` → `x` (no-op) This reduces bytecode size and gas costs, and enables further optimizations. ## Tasks ### Basic constant folding - [ ] For arithmetic ops with constant operands, compute result - [ ] Handle 256-bit arithmetic correctly (wrapping, overflow) - [ ] Support all arithmetic ops: add, sub, mul, div, mod, exp - [ ] Support all bitwise ops: and, or, xor, not, shl, shr, sar ### Algebraic simplifications - [ ] `add(x, 0)` → `x` - [ ] `mul(x, 1)` → `x` - [ ] `mul(x, 0)` → `0` - [ ] `div(x, 1)` → `x` - [ ] `and(x, 0)` → `0` - [ ] `or(x, 0)` → `x` - [ ] `xor(x, 0)` → `x` - [ ] `shl(x, 0)` → `x` - [ ] `sub(x, x)` → `0` - [ ] `xor(x, x)` → `0` ### Comparison folding - [ ] `eq(5, 5)` → `true` - [ ] `lt(3, 5)` → `true` - [ ] `iszero(0)` → `true` - [ ] Propagate through branches (see SCCP issue) ### Integration with lowering - [ ] Fold during HIR→MIR lowering where possible - [ ] Avoid generating constant operations in the first place ## Example ``` ; Before constant folding v0 = const 10 v1 = const 5 v2 = add(v0, v1) ; → const 15 v3 = mul(v2, 2) ; → const 30 v4 = arg(0) v5 = add(v4, 0) ; → v4 return v5 ; After constant folding v4 = arg(0) return v4 ``` ## Acceptance Criteria - [ ] All 256-bit arithmetic handled correctly - [ ] Algebraic identities applied - [ ] Tests verify no incorrect simplifications - [ ] Measurable reduction in instruction counts ## Estimated Complexity **Small-Medium** - Straightforward but many cases to handle ## Dependencies - MIR structure (done)