RFR: 8311125: Remove unused parameter 'phase' in AllocateNode::Ideal_allocation
Christian Hagedorn
chagedorn at openjdk.org
Fri Jun 30 06:54:55 UTC 2023
On Thu, 29 Jun 2023 22:46:57 GMT, Xin Liu <xliu at openjdk.org> wrote:
> There are 2 overloaded AllocateNode::Idea_allocation() in graphkit.cpp.
> One of them never uses 'phase' in the pattern-matching effort.
>
> C++ compiler may emit a warning for the unused parameter. We will need to take care of it if we treat
> warning as error. It also unnecessarily couple CheckCastPP with PhaseValue. In some places, we have to
> gain the instance for it.
>
> I would like to remove 'phase' as parameter. This is a pure clean-up. The other Idea_allocation() does
> use PhaseValue* phase to get constant nodes, so leave it alone.
Otherwise, the cleanup looks good.
src/hotspot/share/opto/memnode.cpp line 632:
> 630:
> 631: if (ac != nullptr && ac->is_clonebasic()) {
> 632: AllocateNode* alloc = AllocateNode::Ideal_allocation(ac->in(ArrayCopyNode::Dest));
You can also remove the `phase` parameter of this method since this was the only usage.
src/hotspot/share/opto/memnode.cpp line 1716:
> 1714: }
> 1715:
> 1716: AllocateNode* LoadNode::is_new_object_mark_load(PhaseGVN *phase) const {
You can also remove the `phase` parameter of this method since there is now no usage left.
src/hotspot/share/opto/parse1.cpp line 1014:
> 1012: // then barrier introduced by allocation node can be removed.
> 1013: if (DoEscapeAnalysis && alloc_with_final()) {
> 1014: AllocateNode *alloc = AllocateNode::Ideal_allocation(alloc_with_final());
Suggestion:
AllocateNode* alloc = AllocateNode::Ideal_allocation(alloc_with_final());
-------------
Marked as reviewed by chagedorn (Reviewer).
PR Review: https://git.openjdk.org/jdk/pull/14719#pullrequestreview-1506605427
PR Review Comment: https://git.openjdk.org/jdk/pull/14719#discussion_r1247498636
PR Review Comment: https://git.openjdk.org/jdk/pull/14719#discussion_r1247497126
PR Review Comment: https://git.openjdk.org/jdk/pull/14719#discussion_r1247500189
More information about the hotspot-compiler-dev
mailing list