vmulh.vx

{
  let SEW      = get_sew();
  let LMUL_pow = get_lmul_pow();
  let num_elem = get_num_elem(LMUL_pow, SEW);

  if illegal_normal(vd, vm) then { handle_illegal(); return RETIRE_FAIL };

  let 'n = num_elem;
  let 'm = SEW;

  let vm_val  : vector('n, dec, bool)     = read_vmask(num_elem, vm, 0b00000);
  let rs1_val : bits('m)                  = get_scalar(rs1, SEW);
  let vs2_val : vector('n, dec, bits('m)) = read_vreg(num_elem, SEW, LMUL_pow, vs2);
  let vd_val  : vector('n, dec, bits('m)) = read_vreg(num_elem, SEW, LMUL_pow, vd);
  result      : vector('n, dec, bits('m)) = undefined;
  mask        : vector('n, dec, bool)     = undefined;

  (result, mask) = init_masked_result(num_elem, SEW, LMUL_pow, vd_val, vm_val);

  foreach (i from 0 to (num_elem - 1)) {
    if mask[i] then {
      result[i] = match funct6 {
        MVX_VAADDU       => {
                              let result_add = zero_extend('m + 1, vs2_val[i]) + zero_extend('m + 1, rs1_val);
                              let rounding_incr = get_fixed_rounding_incr(result_add, 1);
                              slice(result_add >> 1, 0, 'm) + zero_extend('m, rounding_incr)
                            },
        MVX_VAADD        => {
                              let result_add = sign_extend('m + 1, vs2_val[i]) + sign_extend('m + 1, rs1_val);
                              let rounding_incr = get_fixed_rounding_incr(result_add, 1);
                              slice(result_add >> 1, 0, 'm) + zero_extend('m, rounding_incr)
                            },
        MVX_VASUBU       => {
                              let result_sub = zero_extend('m + 1, vs2_val[i]) - zero_extend('m + 1, rs1_val);
                              let rounding_incr = get_fixed_rounding_incr(result_sub, 1);
                              slice(result_sub >> 1, 0, 'm) + zero_extend('m, rounding_incr)
                            },
        MVX_VASUB        => {
                              let result_sub = sign_extend('m + 1, vs2_val[i]) - sign_extend('m + 1, rs1_val);
                              let rounding_incr = get_fixed_rounding_incr(result_sub, 1);
                              slice(result_sub >> 1, 0, 'm) + zero_extend('m, rounding_incr)
                            },
        MVX_VSLIDE1UP    => {
                              if (vs2 == vd) then { handle_illegal(); return RETIRE_FAIL };
                              if i == 0 then rs1_val else vs2_val[i - 1]
                            },
        MVX_VSLIDE1DOWN  => {
                              let last_elem = get_end_element();
                              assert(last_elem < num_elem);
                              if i < last_elem then vs2_val[i + 1] else rs1_val
                            },
        MVX_VMUL         => get_slice_int(SEW, signed(vs2_val[i]) * signed(rs1_val), 0),
        MVX_VMULH        => get_slice_int(SEW, signed(vs2_val[i]) * signed(rs1_val), SEW),
        MVX_VMULHU       => get_slice_int(SEW, unsigned(vs2_val[i]) * unsigned(rs1_val), SEW),
        MVX_VMULHSU      => get_slice_int(SEW, signed(vs2_val[i]) * unsigned(rs1_val), SEW),
        MVX_VDIVU        => {
                              let q : int = if unsigned(rs1_val) == 0 then -1 else quot_round_zero(unsigned(vs2_val[i]), unsigned(rs1_val));
                              to_bits(SEW, q)
                            },
        MVX_VDIV         => {
                              let elem_max : int = 2 ^ (SEW - 1) - 1;
                              let elem_min : int = 0 - 2 ^ (SEW - 1);
                              let q : int = if signed(rs1_val) == 0 then -1 else quot_round_zero(signed(vs2_val[i]), signed(rs1_val));
                              /* check for signed overflow */
                              let q' : int = if q > elem_max then elem_min else q;
                              to_bits(SEW, q')
                            },
        MVX_VREMU        => {
                              let r : int = if unsigned(rs1_val) == 0 then unsigned(vs2_val[i]) else rem_round_zero(unsigned(vs2_val[i]), unsigned (rs1_val));
                              /* signed overflow case returns zero naturally as required due to -1 divisor */
                              to_bits(SEW, r)
                            },
        MVX_VREM         => {
                              let r : int = if signed(rs1_val) == 0 then signed(vs2_val[i]) else rem_round_zero(signed(vs2_val[i]), signed(rs1_val));
                              /* signed overflow case returns zero naturally as required due to -1 divisor */
                              to_bits(SEW, r)
                            }
      }
    }
  };

  write_vreg(num_elem, SEW, LMUL_pow, vd, result);
  vstart = zeros();
  RETIRE_SUCCESS
}