GH-49835: [C++] A constexpr dynamic dispatch with static dispatch when possible

[AntoinePrv]

Rationale for this change

• Reduce verbosity when using DynamicDispatch
• Optimizations: when all targets can be statically dispatched, simply pick the best one at compile time, avoiding jump and CpuInfo creation.

What changes are included in this PR?

• Large refactor of dispatch_internal.h
• Cleanup of call sites

Are these changes tested?

Via existing tests

Are there any user-facing changes?

No

• GitHub Issue: [C++] Constexpr implementation for dynamic dispatch #49835

[github-actions]

⚠️ GitHub issue #49835has been automatically assigned in GitHub to PR creator.

[pitrou]

@github-actions crossbow submit -g cpp

[pitrou]

+1, quite a nice improvement. I'll wait for CI and benchmark results though.

[pitrou]

@ursabot please benchmark lang=C++

[github-actions]

Revision: 8398f97

Submitted crossbow builds: ursacomputing/crossbow @ actions-784cd4e111

Task	Status
example-cpp-minimal-build-static
example-cpp-minimal-build-static-system-dependency
example-cpp-tutorial
test-build-cpp-fuzz
test-conda-cpp
test-conda-cpp-valgrind
test-debian-13-cpp-amd64
test-debian-13-cpp-i386
test-debian-experimental-cpp-gcc-15
test-fedora-42-cpp
test-ubuntu-22.04-cpp
test-ubuntu-22.04-cpp-bundled
test-ubuntu-22.04-cpp-emscripten
test-ubuntu-22.04-cpp-no-threading
test-ubuntu-24.04-cpp
test-ubuntu-24.04-cpp-bundled-offline
test-ubuntu-24.04-cpp-gcc-13-bundled
test-ubuntu-24.04-cpp-gcc-14
test-ubuntu-24.04-cpp-minimal-with-formats
test-ubuntu-24.04-cpp-thread-sanitizer

[rok]

Benchmark runs are scheduled for commit 8398f97. Watch https://buildkite.com/apache-arrow and https://conbench.arrow-dev.org for updates. A comment will be posted here when the runs are complete.

[pitrou]

This PR gives surprisingly massive improvements on our ARM64 benchmarking machine (with Neon):
https://conbench.arrow-dev.org/compare/runs/596185620b9943f48060d1da301bc537...789520580536479fa686c5ad63873abb/

Up to 500% faster in some Parquet benchmarks. 😲

@cyb70289

[pitrou]

I'll merge now, thanks a lot for this @AntoinePrv !

[AntoinePrv]

Perhaps from skipping the dynamic dispatch in level_comparison.cc? I think the other two already skipped it.

[conbench-apache-arrow]

Thanks for your patience. Conbench analyzed the 3 benchmarking runs that have been run so far on PR commit 8398f97.

There were 6 benchmark results indicating a performance regression:

• Pull Request Run on amd64-c6a-4xlarge-linux at 2026-04-28 14:16:36Z
  • BuildNonInlineBinaryViewArray (C++) with source=cpp-micro, suite=arrow-builder-benchmark
  • ExecuteScalarExpressionOverhead (C++) with params=ref_only_expression/rows_per_batch:10000/real_time/threads:1, source=cpp-micro, suite=arrow-acero-expression-benchmark
• and 4 more (see the report linked below)

The full Conbench report has more details.

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 3 benchmarking runs that have been run so far on merge-commit 656e8e6.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 9 possible false positives for unstable benchmarks that are known to sometimes produce them.