Module vec_cmp 

Pass: broadcast scalar operand of vec/scalar comparisons.

Functions

is_arith_op 🔒

is_compare_op 🔒

rewrite_vec_scalar_compare 🔒

HLSL auto-broadcasts the scalar operand of a comparison (vec3 >= 0.1 → element-wise vec3 >= float3(0.1)); WGSL rejects vec3<f32> >= f32 with InvalidBinaryOperandTypes (a dominant validate-stage cluster, op=Ge/Le/Gt/Lt/Eq/Ne with lhs_type: Vector). Wraps the scalar operand in a vecN<f32>(…) constructor so the comparison evaluates as a vector-bool.

try_broadcast_scalar_compare 🔒

Inspect a comparison binop’s operands; if one is a known vector and the other is a known scalar f32, emit a constructor wrap around the scalar so the comparison evaluates as a vector-bool. Zero-length insertions only, so the wrap survives even when other passes have also edited the operands.

try_promote_i32_arith 🔒

Promote an i32 operand of an f32 ⟦arith⟧ i32 binop to f32 by wrapping it in float(…). WGSL refuses mixed f32 * i32 arithmetic; corpus shape (martin - pixies party 2 random mosscity.milk):

walk_block_for_vsc 🔒

walk_expr_for_vsc 🔒

walk_stmt_for_vsc 🔒