JIT: Preserve RHS writes in SIMD partial stores

[Copilot]

For v.X = Helper(ref v), morphing could read the full vector before the RHS, then overwrite the callee’s writes to other fields.

• Morphing: Keep direct partial stores when the RHS has persistent or ordering side effects. Retain WithElement for side-effect-free RHS expressions.
• Coverage: Add numerics-field and SIMD-half regressions, plus assignment, pure-RHS, and volatile-read cases.

• Fixes JIT: (bug) v.X = Helper(ref v) on a Vector4 local loses the callee's write to v.Y #135392

[azure-pipelines]

Azure Pipelines:
Successfully started running 5 pipeline(s).
11 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: @dotnet/runtime-infrastructure
See info in area-owners.md if you want to be subscribed.

[Copilot]

🔵 Needs a closer look

Cross-architecture code-generation and register-allocation effects need human review and runtime validation; no builds or tests were run here.

0 open findings

What changed in this PR

Addresses #135392 by preserving right-hand-side writes when the JIT morphs partial SIMD stores.

Changes:

• Keeps direct partial stores for expressions with persistent or ordering side effects.
• Retains WithElement optimization for side-effect-free expressions.
• Adds numerics-field, SIMD-half, assignment, pure-expression, and volatile-read regressions.

File	Description
src/​tests/​JIT/​Regression_ro_2/​Runtime_135392.cs	Adds partial-store regression coverage.
src/​coreclr/​jit/​lclmorph.cpp	Prevents unsafe partial-store transformations.

🧠 Review effort: Balanced

---

Give feedback about Copilot approvals in this survey to enter a drawing for a $150 gift card.

[tannergooding]

Do we not need to consider GTF_EXCEPT? We notably now have GTF_OBS_EFFECT which is GTF_SIDE_EFFECT | GTF_ORDER_SIDEEFF (which is therefore GTF_PERSISTENT_SIDE_EFFECTS | GTF_EXCEPT | GTF_ORDER_SIDEEFF)

[EgorBo]

Yeah, let's use GTF_OBS_EFFECT here then

[tannergooding]

LGTM. Couple similar issues tracked under #135432 and then some others you logged as well (#134893 is directly related, #135393 and #135394 are indirectly related)