On 07/25/2018 08:38 AM, Aleksandar Markovic wrote:
> Hi, Richard.
>
>>> + case NM_SOV:
>>> + {
>>> + TCGv t0 = tcg_temp_local_new();
>>> + TCGv t1 = tcg_temp_new();
>>> + TCGv t2 = tcg_temp_new();
>>> + TCGLabel *l1 = gen_new_label();
>>> +
>>> + gen_load_gpr(t1, rs);
>>> + gen_load_gpr(t2, rt);
>>> + tcg_gen_add_tl(t0, t1, t2);
>>> + tcg_gen_ext32s_tl(t0, t0);
>>> + tcg_gen_xor_tl(t1, t1, t2);
>>> + tcg_gen_xor_tl(t2, t0, t2);
>>> + tcg_gen_andc_tl(t1, t2, t1);
>>> +
>>> + tcg_gen_movi_tl(t0, 0);
>>> + tcg_gen_brcondi_tl(TCG_COND_GE, t1, 0, l1);
>>
>> tcg_gen_setcondi_tl.
>>
>
> Would here the correct simplification be:
>
> Replace code segment
>
> tcg_gen_movi_tl(t0, 0);
> tcg_gen_brcondi_tl(TCG_COND_GE, t1, 0, l1);
> tcg_gen_movi_tl(t0, 1);
> gen_set_label(l1);
>
> with
>
> tcg_gen_setcondi_tl(TCG_COND_GE, t0, t1, 0);
> (plus deleting the declaration of l1 of course)
Nearly. You're branching over a move of 1, and so computing the inverse of the
condition. So the correct replacement is
tcg_gen_setcondi_tl(TCG_COND_LT, t0, t1, 0);
r~
