Re: [RFC PATCH v2 14/44] target/loongarch: Implement vmax/vmin

[Richard Henderson]

On 3/27/23 20:06, Song Gao wrote:
> +static void do_vminmax(unsigned vece, TCGv_vec t, TCGv_vec a, int64_t imm,
> +                       void(*gen_vminmax_vec)(unsigned,
> +                                              TCGv_vec, TCGv_vec, TCGv_vec))
> +{
> +    TCGv_vec t1;
> +
> +    t1 = tcg_temp_new_vec_matching(t);
> +    tcg_gen_dupi_vec(vece, t1, imm);

t1 = tcg_constant_vec_matching(t, vece, imm);

> +static void gen_vmini_s(unsigned vece, TCGv_vec t, TCGv_vec a, int64_t imm)
> +{
> +    do_vminmax(vece, t, a, imm, tcg_gen_smin_vec);
> +}
> +
> +static void gen_vmini_u(unsigned vece, TCGv_vec t, TCGv_vec a, int64_t imm)
> +{
> +    do_vminmax(vece, t, a, imm, tcg_gen_umin_vec);
> +}
> +
> +static void gen_vmaxi_s(unsigned vece, TCGv_vec t, TCGv_vec a, int64_t imm)
> +{
> +    do_vminmax(vece, t, a, imm, tcg_gen_smax_vec);
> +}
> +
> +static void gen_vmaxi_u(unsigned vece, TCGv_vec t, TCGv_vec a, int64_t imm)
> +{
> +    do_vminmax(vece, t, a, imm, tcg_gen_umax_vec);
> +}

Perhaps easier to expand

    tcg_gen_umax_vec(vece, t, a, tcg_constant_vec_matching(t, vece, imm));

in each instance?


r~