RFR(S): 8201367: assert(current != first_mem) failed: corrupted memory graph in superword code

[Vladimir Kozlov]

Yes, this is better.

And you can initialize last_mem with first_mem without going through 
indirect loads again.

Thanks,
Vladimir

On 4/26/18 1:57 AM, Roland Westrelin wrote:
> 
> Hi Vladimir,
> 
> Thanks for looking at this.
> 
>> The logic looks fine to me. But is it optimal? Can we use
>> dependence_graph constructed before this code?
> 
> I looked at the dependence graph but couldn't find a way it would help.
> 
> What about this instead?
> 
> http://cr.openjdk.java.net/~roland/8201367/webrev.01/
> 
> Roland.
>