merge vm_version_x86_{32,64}.hpp

[Vladimir Kozlov]

Use _LP64 define

Vladimir

Vladimir Kozlov wrote:
> 
> 
> Christian Thalinger wrote:
>> On Mon, 2009-02-23 at 07:29 -0500, Paul Hohensee wrote:
>>> Imo, the #ifdef's are ok in this case because they're quite limited.
>>>
>>> Vladimir set prefetch distances based on experiment.  The difference
>>> between 320 and 384 is a single cache line.  It seems to me that 32
>>> and 64-bit would be the same because it's the same hardware, except
>>> that it may be that 64-bit code is faster than 32-bit code, thus would
>>> require a slightly greater prefetch distance.
> 
> No, it is because different average allocation object's size (different 
> oop size).
> 
>>>
>>> 32-bit doesn't use the Prefetchxxx switches because there's no common
>>> prefetch instruction for 32-bit that can be used in C++ code.  #ifdef is
>>> ok here.
> 
> That is correct.
> 
> 
>>
>> Paul, thank you for the explanation.  Last question: should I use the
>> _LP64 or AMD64 define?
>>
>> -- Christian
>>
> 
> Vladimir