RFR: 8359602: Ideal optimizations depending on input type are missed because of missing notification mechanism from CCP [v5]
duke
duke at openjdk.org
Thu Jul 3 07:10:42 UTC 2025
On Wed, 2 Jul 2025 07:19:30 GMT, Benoît Maillard <bmaillard at openjdk.org> wrote:
>> This PR prevents some missed ideal optimizations in IGVN by notifying users of type refinements made during CCP, addressing a missed optimization that caused a verification failure with `-XX:VerifyIterativeGVN=1110`.
>>
>> ### Context
>> During the compilation of the input program (obtained from the fuzzer, then simplified and added as a test) by C2, we end up with node `591 ModI` that takes `138 Phi` as its divisor input. An existing `Ideal` optimization is to get rid of the control input of a `ModINode` when we can prove that the divisor is never `0`.
>>
>> In this specific case, the type of the `PhiNode` gets refined during CCP, but the refinement fails to propagate to its users for the IGVN phase and the ideal optimization for the `ModINode` never happens. This results in a missed optimization and hits an assert in the verification phase of IGVN (when using `-XX:VerifyIterativeGVN=1110`).
>>
>> 
>>
>> ### Detailed Analysis
>>
>> In `PhaseCCP::analyze`, we call `Value` for the `PhiNode`, which
>> results in a type refinement: the range gets restricted to `int:-13957..-1191`.
>>
>> ```c++
>> // Pull from worklist; compute new value; push changes out.
>> // This loop is the meat of CCP.
>> while (worklist.size() != 0) {
>> Node* n = fetch_next_node(worklist);
>> DEBUG_ONLY(worklist_verify.push(n);)
>> if (n->is_SafePoint()) {
>> // Make sure safepoints are processed by PhaseCCP::transform even if they are
>> // not reachable from the bottom. Otherwise, infinite loops would be removed.
>> _root_and_safepoints.push(n);
>> }
>> const Type* new_type = n->Value(this);
>> if (new_type != type(n)) {
>> DEBUG_ONLY(verify_type(n, new_type, type(n));)
>> dump_type_and_node(n, new_type);
>> set_type(n, new_type);
>> push_child_nodes_to_worklist(worklist, n);
>> }
>> if (KillPathsReachableByDeadTypeNode && n->is_Type() && new_type == Type::TOP) {
>> // Keep track of Type nodes to kill CFG paths that use Type
>> // nodes that become dead.
>> _maybe_top_type_nodes.push(n);
>> }
>> }
>> DEBUG_ONLY(verify_analyze(worklist_verify);)
>>
>>
>> At the end of `PhaseCCP::analyze`, we obtain the following types in the side table:
>> - `int` for node `591` (`ModINode`)
>> - `int:-13957..-1191` for node `138` (`PhiNode`)
>>
>> If we call `find_node(138)->bottom_type()`, we get:
>> - `int` for both nodes
>>
>> The...
>
> Benoît Maillard has updated the pull request with a new target base due to a merge or a rebase. The pull request now contains eight commits:
>
> - Fix bad test class name
> - 8359602: rename test
> - 8359602: remove requires.debug=true and add -XX:+IgnoreUnrecognizedVMOptions flag
> - 8359602: add comment
> - 8359602: add test summary and comments
> - 8359602: tag requires vm.debug == true
> - 8359602: Add test from fuzzer
> - 8359602: Add users to IGVN worklist when type is refined in CCP
@benoitmaillard
Your change (at version a66d3fb492541a17e28b3e0fe0f60080c14bdc2c) is now ready to be sponsored by a Committer.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/26017#issuecomment-3031130840
More information about the hotspot-compiler-dev
mailing list