[aarch64-port-dev ] Query regarding backport of memory order changes to JDK7

[Andrew Haley]

On 11/14/2014 04:15 PM, Andrew Dinn wrote:
> 
> 
> On 14/11/14 15:56, Andrew Haley wrote:
>> On 11/14/2014 03:39 PM, Andrew Dinn wrote:
>>> Is there a reason why the call to OrderAccess::release() in
>>> set_last_Java_fp comes before the assignment of _last_Java_fp?
>>
>> See the comment: // clearing _last_Java_sp must be first
>>
>> clearing _last_Java_sp is, as I understand it, always supposed to be
>> visible before clearing _last_Java_fp.  That is because _last_Java_sp
>> is used as a flag: it indicates that the frame anchor is valid.
> 
> Ok, so why isn't that already catered for by the calls to
> OrderAccess::release() which follow the assignment of _last_Java_sp in
> set_last_Java_sp() and clear()? Is this call not redundant?

Please show me the code you're describing.

Andrew.