fixes for 2026.0 compiler

[Alexandr-Solovev]

PR Description

Summary

This PR updates SYCL code to remove deprecated features and improve compatibility with recent compiler versions.

Changes

1. Replacement of deprecated address_space

   • Replaced usage of

     sycl::access::address_space::ext_intel_global_device_space

     with

     sycl::access::address_space::global_space

   • Reason: ext_intel_global_device_space has been removed in recent compiler versions.

2. Update of binary operations with is_logical flag

   • For all operations like:

     template <typename T>
     struct logical_or {
         using tag_t = reduce_binary_op_tag;
         constexpr static inline T init_value = false;
         using acc_t = T;
     };

   • Added is_logical flag and updated init_value to T(false)
   • Purpose: Ensure compatibility with

     sycl::reduce_over_group(it.get_group(), local_accum, binary.native);

     so that the binary operation matches the type of the return value during reduction.

3. Removal of old compiler version checks

   • Removed blocks like:

     #if __SYCL_COMPILER_VERSION >= 20230828

Notes

• All changes are backward compatible with current SYCL standards.
• No behavioral changes; only refactoring for compatibility and clarity.

---

Checklist:

Completeness and readability

• I have commented my code, particularly in hard-to-understand areas.
• I have updated the documentation to reflect the changes or created a separate PR with updates and provided its number in the description, if necessary.
• Git commit message contains an appropriate signed-off-by string (see CONTRIBUTING.md for details).
• I have resolved any merge conflicts that might occur with the base branch.

Testing

• I have run it locally and tested the changes extensively.
• All CI jobs are green or I have provided justification why they aren't.
• I have extended testing suite if new functionality was introduced in this PR.

Performance

• I have measured performance for affected algorithms using scikit-learn_bench and provided at least a summary table with measured data, if performance change is expected.
• I have provided justification why performance and/or quality metrics have changed or why changes are not expected.
• I have extended the benchmarking suite and provided a corresponding scikit-learn_bench PR if new measurable functionality was introduced in this PR.

[ethanglaser]

Nice work on this so far! CEV job with this branch: http://intel-ci.intel.com/f0f24793-7dcb-f1ab-af3c-a4bf010d0e2d. FYI there are still fails in oneDAL build on windows and several validation jobs on linux side

[Vika-F]

Please revert this change in logical_or, wait for the CI results and LGTM.

constexpr static inline auto native = [](const T& a, const T& b) {
    return static_cast<T>(std::logical_or<T>{}(a, b));
};

[ethanglaser]

Please share the latest CEV run from this branch. Also please address clang-format before merging

[ethanglaser]

Confused how this file compiled with an extra } in here?

[Alexandr-Solovev]

I guess it was under #else branch and was never built.

[Copilot]

Pull request overview

This PR updates SYCL code to remove deprecated compiler features and improve compatibility with the 2026.0 compiler version.

Changes:

• Replaced deprecated ext_intel_global_device_space with global_space in atomic operations
• Updated logical_or binary operations to use bool accumulator type instead of float/double
• Removed compiler version checks (#if __SYCL_COMPILER_VERSION >= ...) for obsolete fallback code
• Updated MKL SYCL library symlinks from version 5 to version 6

Reviewed changes

Copilot reviewed 30 out of 30 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
dev/bazel/deps/mkl.tpl.BUILD	Updated MKL SYCL library symlinks from .so.5 to .so.6
dev/bazel/deps/mkl.bzl	Updated MKL SYCL library symlinks from .so.5 to .so.6
cpp/oneapi/dal/backend/primitives/reduction/functors.hpp	Fixed logical_or init_value and replaced deprecated address space
cpp/oneapi/dal/backend/atomic.hpp	Replaced deprecated ext_intel_global_device_space with global_space
cpp/oneapi/dal/backend/primitives/reduction/*.cpp	Added AccT template parameter for accumulator type in reduction operations
cpp/oneapi/dal/backend/primitives/selection/*.cpp	Removed obsolete compiler version checks
cpp/oneapi/dal/algo//backend/gpu/.cpp	Replaced deprecated address space and removed compiler version checks

[Copilot]

The explicit cast static_cast<float_t>(binary_.init_value) is needed because binary_.init_value is now of type bool for logical_or operations. Consider documenting this cast requirement to clarify why the conversion is necessary for future maintainers.