RFR(M): 8212243: More gc interface tweaks for arraycopy and some other smaller changes in preparation for Shenandoah

[Vladimir Kozlov]

Hi Roland,

Please, split these in 3 RFEs to review and test them separately: arraycopy, predicated, small changes.

I looked on arraycopy changes.

Why you are not using InstanceKlass::_nonstatic_oop_map_size to find if class has non-static object 
fields?

PhaseMacroExpand::clone_needs_postbarrier() should have GC interface call to return false 
immediately if post barrier is not needed for current GC.

Aggressive dead nodes elimination seems fine but I want to look on separate webrev to make sure I 
did not miss something.

Thanks,
Vladimir

On 10/16/18 9:06 AM, Roland Westrelin wrote:
> 
> http://cr.openjdk.java.net/~roland/8212243/webrev.00/
> 
> This change adds 2 more entry point for arraycopy to the gc interface:
> 
> - one, when an arraycopy is turned into loads/stores and the field or
>    element being copied is a reference. In that case, barriers may have
>    to be applied.
> 
> - another, when a clone is expanded into a stub call. In that case a
>    postbarrier may be required. That one is only required when a
>    reference field or element may be encountered. That's where the
>    ciInstanceKlass changes come in.
> 
> The other complicated change is in loopPredicate.cpp. We have:
> 
> Predicate 1
> ..
> Predicate n
> 
> profile predicate 1
> ..
> profile predicate n
> 
> loop head
> 
> Shenandoah write barriers produce memory, can be hoisted out of loop but
> can also depend on some predicate so may end up with:
> 
> Predicate 1
> ..
> Predicate n
> 
> profile predicate 1
> ..
> write barrier
> ...
> profile predicate n
> 
> loop head
> 
> 
> In that case, all predicates branch to a region that dominates the
> uncommon trap call. Same is true for profile predicates. Because of the
> write barrier, the region for profile predicates has a memory phi to
> merge memory state before the barrier and after the barrier. Now when
> clone_loop_predicates() is called:
> 
> Predicate 1
> ..
> Predicate n
> 
> profile predicate 1
> ..
> write barrier
> ...
> profile predicate n
> 
> cloned predicate
> 
> cloned profile predicate
> 
> loop head
> 
> 
> the cloned predicate branches to the region shared by all
> predicates. Same goes for the cloned profile predicate and the profile
> predicates. The memory state for the cloned predicate should be the one
> following the barrier but after cloning it's the one from the predicate
> n, before the barrier. That memory state needs to be adjusted.
> 
> Other changes tweak the gc interface and make sure that when nodes are
> disconnected, nodes that need special treatment are processed.
> 
> Roland.
>