Safepoint timeout in scimark?
vladimir.kozlov at oracle.com
Tue Jul 19 17:28:08 UTC 2016
Such If node can be folded with CLE later so you will get SP after it
again. Also if it is not folded it will prevent some loop optimizations.
I would say not converting such loop into Counted may give the same
On 7/19/16 9:51 AM, Roland Westrelin wrote:
> Hi Vladimir,
>> I am worry that moving safepoint before CLE will be correct in all cases
>> since JVMS could be different.
> In my patch, I guarded the safepoint before the CLE with a test that
> guarantees we will take the backedge of the loop. Shouldn't that be
More information about the hotspot-compiler-dev