[Hardware][Mac] Fix the installation fail for Apple Silicon (CPU)

[OYE93]

Purpose

To fix the Installation for Apple Silicon (CPU), when following the doc, I met the error like

/vllm/csrc/cpu/dnnl_helper.cpp:89:8: error: no template
      named 'optional' in namespace 'std'
         89 |   std::optional<value_iterator_t> get_value(const KT& key) {
            |   ~~~~~^

After some checking, the cause could be that std::optional is a C++17 feature. So in this PR, just changed to force the C++17 compilation in the .cmake file.

Test Plan

uv pip install -e .

Test Result

After

Resolved 137 packages in 6.54s
      Built vllm @ file:///XXXXX/vllm
Prepared 1 package in 1m 42s
Installed 1 package in 2ms
 + vllm==0.1.dev8866+g92687fbe1.cpu (from file:///XXXXXX/vllm)

(Optional) Documentation Update

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

[gemini-code-assist]

Code Review

This pull request correctly identifies and fixes a build failure on Apple Silicon by enforcing the C++17 standard for CPU builds. The use of std::optional in the C++ source code makes this change necessary. The proposed fix is effective. I have one suggestion regarding CMake best practices to improve the long-term maintainability of the build configuration.

[gemini-code-assist]

While setting CMAKE_CXX_STANDARD is a valid way to enable C++17 and fixes the build, it's a directory-level setting. A more modern and maintainable approach in CMake is to specify C++ standard requirements on a per-target basis using target_compile_features. This improves modularity and makes the build configuration easier to maintain, as target requirements are co-located with the target definition.

For example:

# After the target is defined around line 300
target_compile_features(_C PRIVATE cxx_std_17)

This would make the target's requirements explicit. Since this change works and its scope is limited to CPU builds, this is a suggestion for better long-term maintenance of the build system.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors.

You ask your reviewers to trigger select CI tests on top of fastcheck CI.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

If you have any questions, please reach out to us on Slack at https://slack.vllm.ai.

🚀

[bigPYJ1151]

Thanks for the fix!