Optimize generic pooled SML indexed dispatch and benchmark

[gabewillen]

Summary

• rewrites benchmark/connection/tensor_cache_locality.cpp as a generic SML dispatch benchmark (no link/unlink-specific logic)
• compares direct flat dispatch, per-actor SML dispatch, pooled indexed SML dispatch, and pooled indexed batch dispatch
• optimizes sm_pool::process_indexed_batch to reuse a single indexed-event wrapper in the hot loop
• removes an extra wrapper call in sm_pool::process_event_batch

Validation

• ran scripts/quality_gates.sh
• result: All quality gates passed (format, clang-tidy, regression matrix, sanitizer matrix, coverage gate, benchmark smoke build)

[Copilot]

Pull request overview

This PR refactors an existing cache-locality benchmark into a generic “10k actor” dispatch benchmark to compare several dispatch strategies, and applies small hot-loop optimizations to sml::utility::sm_pool batch processing to reduce wrapper overhead.

Changes:

• Refactors benchmark/connection/tensor_cache_locality.cpp into a generic dispatch-overhead benchmark (direct, per-actor SML, pooled indexed SML, pooled indexed batch).
• Optimizes sm_pool::process_indexed_batch by reusing an indexed_event wrapper inside the loop.
• Avoids an extra wrapper function call in sm_pool::process_event_batch.

Reviewed changes

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

File	Description
include/boost/sml/utility/sm_pool.hpp	Reduces per-iteration overhead in indexed/batch dispatch by reusing wrapper objects and calling sm_.process_event directly in batch loops.
benchmark/connection/tensor_cache_locality.cpp	Replaces tensor/link-unlink-specific benchmark with a generic dispatch benchmark comparing multiple dispatch approaches.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.