RFR: 8217874: Shenandoah: AArch64: Clobbered register in ShenandoahBarrierSetAssembler::cmpxchg_oop()

[Andrew Haley]

On 1/28/19 3:59 PM, Roman Kennke wrote:
>> On 1/28/19 12:46 PM, Roman Kennke wrote:
>>> Can I get a review please?
>>
>> Isn't this a bug in the C2 pattern?
> 
> No. The declaration in C2 means that the result argument is supposed to
> be live starting from the (end of) the instruction, and input arguments
> are live up to the instruction. Overlap within an instruction need to be
> taken care of by the implementation. (Doesn't usually happen, because
> instructions are usually lowered to a single instruction.)

Oh yeah. It's all rather grim, but I guess it's not really worth
worrying about.

-- 
Andrew Haley
Java Platform Lead Engineer
Red Hat UK Ltd. <https://www.redhat.com>
EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671