Handle GT_CCMP in GenTree::OperConsumesFlags on TARGET_AMD64

[Copilot]

GenTree::OperConsumesFlags handled GT_CCMP only on TARGET_ARM64, even though GT_CCMP is also emitted on TARGET_AMD64 (gated by #if defined(TARGET_ARM64) || defined(TARGET_AMD64) in gtlist.h).

Changes

• src/coreclr/jit/gentree.h — Add a TARGET_AMD64 branch in OperConsumesFlags. GT_JCC, GT_SETCC, GT_SELECTCC, and GT_CCMP are declared sequentially in gtlist.h (intervening lines are only comments / #if guards), so their enum values are contiguous and a range check collapses 4 comparisons to 2 — matching the existing idiom used elsewhere in the file (e.g. OperIsCompare). A static_assert(AreContiguous(...)) guards the assumption, consistent with surrounding helpers.

#if defined(TARGET_ARM64)
        if (OperIs(GT_CCMP, GT_SELECT_INCCC, GT_SELECT_INVCC, GT_SELECT_NEGCC))
        {
            return true;
        }
        return OperIs(GT_JCC, GT_SETCC, GT_SELECTCC);
#elif defined(TARGET_AMD64)
        static_assert(AreContiguous(GT_JCC, GT_SETCC, GT_SELECTCC, GT_CCMP));
        return (GT_JCC <= gtOper) && (gtOper <= GT_CCMP);
#else
        return OperIs(GT_JCC, GT_SETCC, GT_SELECTCC);
#endif

Callers (lir.cpp, liveness.cpp, lower.cpp, sideeffects.cpp) will now correctly recognize AMD64 GT_CCMP as flag-consuming when ordering / scheduling around flag-producing nodes.

Note

This pull request was created by GitHub Copilot.

[dotnet-policy-service]

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