Fix two latent HWIntrinsic miscompiles in gentree.cpp

[tannergooding]

Two latent correctness fixes found while auditing gentree.cpp, each with a regression test.

---

Fix GetOperForHWIntrinsicId testing the isScalar pointer instead of *isScalar (#130830)

GetOperForHWIntrinsicId refines a GT_SUB into a GT_NEG when the constant op1 is a scalar zero, but it guarded on the isScalar out-param pointer (always non-null) rather than the dereferenced value. This let a packed subtract whose constant has a zero low lane but is not all-zero (e.g. Vector128.Create(0, 1, 2, 3) - x) be reported as a negate. fgOptimizeHWIntrinsic's (-v1) + v2 => v2 - v1 transform then drops the constant entirely.

---

Fix stray block clobbering needsFixup in gtNewSimdMinMaxNode (#130831)

A stray unconditional block in the min branch of the floating constant fast path overwrote needsFixup for all four cases, making the preceding if/else dead. needsFixup signals that a signed-zero constant needs the AVX512 fixup so min(+0, -0) keeps the correct sign of zero; with it wrongly forced false, Min/MinNumber against a signed-zero constant miscompiles. The min branch now mirrors the already-correct max branch.

Existing coverage in JitBlue/Runtime_98068 exercises Min/MinNumber const-folding but always pairs an operand with NaN, so the finite opposite-signed-zero case was never tested.

---

Both are pre-existing on main, found by inspection. Tests added under JitBlue/Runtime_130830 and JitBlue/Runtime_130831.

CC. @dotnet/jit-contrib

Note

This PR description was authored with the help of GitHub Copilot.

[azure-pipelines]

Azure Pipelines:
Successfully started running 5 pipeline(s).
10 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull request overview

This PR fixes two latent correctness issues in CoreCLR JIT’s gentree.cpp HWIntrinsic handling and adds targeted JitBlue regression coverage to prevent regressions.

Changes:

• Fix GenTreeHWIntrinsic::GetOperForHWIntrinsicId(..., getEffectiveOp: true) to test *isScalar (not the out-param pointer) before treating certain SUB patterns as NEG.
• Fix gtNewSimdMinMaxNode (xarch floating const-fold fast path) so needsFixup is computed correctly for signed-zero Min/MinNumber scenarios.
• Add new JitBlue regression tests for both miscompiles.

Show a summary per file

File	Description
src/coreclr/jit/gentree.cpp	Correctness fixes for HWIntrinsic effective-op classification and signed-zero fixup gating in SIMD min/max constant folding.
src/tests/JIT/Regression/JitBlue/Runtime_130830/Runtime_130830.cs	Regression test ensuring packed integer subtract isn’t misreported as negate when only the low lane is zero.
src/tests/JIT/Regression/JitBlue/Runtime_130831/Runtime_130831.cs	Regression test intended to validate signed-zero behavior for Min/MinNumber with constant operands.

Copilot's findings

• Files reviewed: 3/3 changed files
• Comments generated: 1

[Copilot]

Copilot's findings

• Files reviewed: 3/3 changed files
• Comments generated: 2

[Copilot]

Copilot's findings

• Files reviewed: 4/4 changed files
• Comments generated: 0 new

[tannergooding]

CC. @dotnet/jit-contrib, @EgorBo. Two simple bug fixes with regression tests

[github-actions]

Workflow state for the Holistic Review Orchestrator.

{
  "version": 5,
  "last_dispatched_commit": "45c9dfe3b26431e9282bbbb6078c7973d0db59c2",
  "last_dispatched_base_ref": "main",
  "last_dispatched_base_sha": "49110af56a95223e88f9e540503c0f8a16e64e15",
  "last_reviewed_commit": "45c9dfe3b26431e9282bbbb6078c7973d0db59c2",
  "last_reviewed_base_ref": "main",
  "last_reviewed_base_sha": "49110af56a95223e88f9e540503c0f8a16e64e15",
  "last_recorded_worker_run_id": "29688130758",
  "review_attempt_commit": "",
  "review_attempt_base_ref": "",
  "review_attempt_count": 0,
  "max_review_attempts": 5,
  "review_history_format": "holistic-review-disclosure-v1",
  "review_history": [
    {
      "commit": "45c9dfe3b26431e9282bbbb6078c7973d0db59c2",
      "review_id": 4730809436
    }
  ]
}

[github-actions]

Holistic Review

Motivation: Two latent, pre-existing HWIntrinsic miscompiles in gentree.cpp, found by inspection while auditing the file. (1) GetOperForHWIntrinsicId tested the isScalar out-param pointer (always non-null) instead of the dereferenced *isScalar, so a packed GT_SUB whose constant op1 has a zero low lane but is not all-zero (e.g. Vector128.Create(0,1,2,3) - x) was wrongly refined to GT_NEG, causing fgOptimizeHWIntrinsic's (-v1)+v2 => v2-v1 transform to drop the real constant. (2) A stray unconditional block in the min branch of gtNewSimdMinMaxNode overwrote needsFixup with IsVectorZero() for all four cases, making the preceding if/else dead and dropping the required AVX512 fixup for signed-zero constants, miscompiling Min/MinNumber against a signed-zero constant.

Approach: Minimal, surgical fixes. The first dereferences the flag (*isScalar) so the negate refinement applies only to scalar ops. The second removes the stray block so the min branch mirrors the already-correct max branch: isNumber uses IsFloatPositiveZero/IsVectorZero, otherwise IsFloatNegativeZero/IsVectorNegativeZero. Two focused regression tests are added under JitBlue/Runtime_130830 and JitBlue/Runtime_130831 and wired into Regression_ro_2.csproj so they build and run in CI.

Summary: LGTM. Both changes are correct and precisely targeted. I verified the corrected min branch is a faithful inverse mirror of the max branch's fixup logic, and that *isScalar is the intended guard. The regression tests are well-constructed: Runtime_130830 asserts per-lane results that would collapse to <90,90,90,90> under the old bug, and Runtime_130831 correctly compares signed-zero results bitwise (via BitConverter), since Double/Single.Equals treat -0.0/+0.0 as equal and would otherwise mask the defect. The added <Compile> entries are necessary because sibling tests in this directory are individually listed in the merged csproj. No actionable findings.

Note

This review was generated by this repository's Holistic Review agentic workflow to complement the built-in Copilot review.

Generated by Holistic Review · 41.2 AIC · ⌖ 10.4 AIC · ⊞ 10K