[PATCH V2] RISC-V: Using merge approach to optimize repeating sequence in vec_init

[Kito Cheng]

two minor comments:

> +  void add_ta_policy_operand (enum tail_policy vta)
> +  void add_ma_policy_operand (enum mask_policy vma)

You could just named as add_policy_operand since the arugment type is
already sufficient to distinguish.

> @@ -84,40 +92,52 @@ public:
>      add_input_operand (gen_int_mode (type, Pmode), Pmode);
>    }
>
> -  void set_dest_and_mask (rtx mask, rtx dest, machine_mode mask_mode)
> +  void set_dest_and_mask (rtx mask, rtx dest, machine_mode mask_mode,
> +                         bool scalar_move = false, bool merge_op = false)
>    {
>      dest_mode = GET_MODE (dest);
>      has_dest = true;
>
>      add_output_operand (dest, dest_mode);
>
> -    if (mask)
> -      add_input_operand (mask, GET_MODE (mask));
> -    else
> -      add_all_one_mask_operand (mask_mode);
> +    if (!merge_op)
> +      {
> +       if (mask)
> +         add_input_operand (mask, GET_MODE (mask));
> +       else
> +         {
> +           if (scalar_move)
> +             add_scalar_move_mask_operand (mask_mode);
> +           else
> +             add_all_one_mask_operand (mask_mode);
> +         }
> +      }

I would like to have set_dest_and_mask_merge_op to reduce the
complexity of set_dest_and_mask.
Accroding the code I got is:
- mask and merge_op are mutual exclusion.
- scalar_move will become meanless if merge_op is true.

So I think they mixed those togeter is not good idea.