RFR(XS): 8075921: assert assert(allocx == alloc) fails in library_call.cpp
Roland Westrelin
roland.westrelin at oracle.com
Fri Mar 27 13:11:42 UTC 2015
Hi Vladimir,
Thanks for looking at this.
> Null checks can also make stop fast path in inline_arraycopy(). I think you should add some stopped() checks there too if needed.
All guards can stop the fast path, not only the null checks. What about this:
http://cr.openjdk.java.net/~roland/8075921/webrev.01/
Some code that generates guards already check for stopped(). I’ve added new checks only where they were missing.
Roland.
>
> Thanks,
> Vladimir
>
> On 3/26/15 6:10 AM, Roland Westrelin wrote:
>> http://cr.openjdk.java.net/~roland/8075921/webrev.00/
>>
>> During arraycopy intrinsic construction, control becomes top after arraycopy guards are added and that confuses the logic that moves the tightly coupled allocation.
>>
>> Roland.
>>
More information about the hotspot-compiler-dev
mailing list