review (S) for 6828024: verification of fixed interval usage is too weak

[Tom Rodriguez]

On Apr 15, 2009, at 6:57 PM, Vladimir Kozlov wrote:

>
> Could you put "()" around "==" expression?:
>
> 3179           check_live = move->patch_code() == lir_patch_none;

ok.

> I am not sure about next code (don't have knowledge):
>
> 3196        if (interval_at(reg_num(opr)) == interval) {
> 3197          ok = true;
> 3198          break;
> 3199        }
> 3200        int hi = reg_numHi(opr);
> 3201        if (hi != -1 && interval_at(hi) == interval) {
> 3202          ok = true;
> 3203          break;
> 3204        }
>
> should it be this?:
>
>      if (interval_at(reg_num(opr)) == interval) {
>        int hi = reg_numHi(opr);
>        if (hi != -1 && interval_at(hi) == interval) {
>          ok = true;
>          break;
>        }
>      }

It can't be equal to both.  The existing logic is just dealing with  
register pairs and a pair has two separate intervals for each half so  
you check for each one separately.  Does that make sense?

tom

>
>
> Thanks,
> Vladimir
>
> Tom Rodriguez wrote:
>> http://cr.openjdk.java.net/~never/6828024/