RFR: 8367483: C2 crash in PhaseValues::type: assert(t != nullptr) failed: must set before get - missing notification for CastX2P(SubL(x, y))
Christian Hagedorn
chagedorn at openjdk.org
Fri Sep 12 08:44:26 UTC 2025
On Fri, 12 Sep 2025 08:18:40 GMT, Emanuel Peter <epeter at openjdk.org> wrote:
> `CastX2PNode::Ideal` optimizes cases:
>
> CastX2P(AddX(x, y)) -> AddP(CastX2P(x), y)
> CastX2P(SubL(x, y)) -> AddP(CastX2P(x), SubL(0, y))
>
>
> But the notification code `PhaseIterGVN::add_users_of_use_to_worklist` only adds `CastX2P` to the worklist for the `AddX` and not the `SubX` cases.
>
> ---------------------------------------
>
> A little brag: this is the second (unrelated, i.e. non aliasing) bug that `TestAliasingFuzzer.java` found. Fuzzing access to native MemorySegment seems to trigger new/rare patterns.
Otherwise, looks good!
test/hotspot/jtreg/compiler/c2/gvn/MissedOptimizationWithCastX2PSubX.java line 1:
> 1: /*
There is a `compiler/igvn` test folder. I think this suits better than `c2/gvn`.
test/hotspot/jtreg/compiler/c2/gvn/MissedOptimizationWithCastX2PSubX.java line 34:
> 32: * -XX:-TieredCompilation
> 33: * -XX:+IgnoreUnrecognizedVMOptions
> 34: * -XX:+UnlockDiagnosticVMOptions
These are not required:
Suggestion:
* -XX:CompileCommand=compileonly,compiler.c2.gvn.MissedOptimizationWithCastX2PSubX::test
* -XX:-TieredCompilation
* -XX:+IgnoreUnrecognizedVMOptions
test/hotspot/jtreg/compiler/c2/gvn/MissedOptimizationWithCastX2PSubX.java line 37:
> 35: * -XX:VerifyIterativeGVN=1110
> 36: * compiler.c2.gvn.MissedOptimizationWithCastX2PSubX
> 37: * @run driver compiler.c2.gvn.MissedOptimizationWithCastX2PSubX
Should be `main` to allow additional flags to be passed in.
Suggestion:
* @run main compiler.c2.gvn.MissedOptimizationWithCastX2PSubX
-------------
Marked as reviewed by chagedorn (Reviewer).
PR Review: https://git.openjdk.org/jdk/pull/27249#pullrequestreview-3215370623
PR Review Comment: https://git.openjdk.org/jdk/pull/27249#discussion_r2343421069
PR Review Comment: https://git.openjdk.org/jdk/pull/27249#discussion_r2343415661
PR Review Comment: https://git.openjdk.org/jdk/pull/27249#discussion_r2343416742
More information about the hotspot-compiler-dev
mailing list