review (S) for 6856025: assert(_base >= OopPtr && _base <= KlassPtr,"Not a Java pointer")

[Vladimir Kozlov]

Tom Rodriguez wrote:
> 
> On Jun 30, 2009, at 2:33 PM, Vladimir Kozlov wrote:
> 
>> After sending this I don't see that you fixed store_oop_to_unknown.
>> Is it intentional to use TypeInstPtr::BOTTOM as val_type in case of 
>> narrow field/element?
> 
> The value_type for the prebarrier is declared to be TypeOopPtr since it 
> bottoms out in LoadNode::make which always returns the wide type.  So 
> TypeInstPtr::BOTTOM is the worst case type for the store_oop_to_unknown 
> case.  Is there something wrong with that?

Nothing wrong with it. I'm just wondering why you don't use more precise
type for array's element (instance's field type is fine) by using make_oopptr().

Vladimir

> 
> tom
> 
>>
>> Thanks,
>> Vladimir
>>
>> Vladimir Kozlov wrote:
>>> Looks good.
>>> Vladimir
>>> Tom Rodriguez wrote:
>>>> http://cr.openjdk.java.net/~never/6856025/
>