Refactor DSP Pipeline: Implement Worker Pooling and Raw Pointer API

[jLynx]

This PR significantly optimizes the DSP architecture by moving from a "one-worker-per-VFO" model to a more scalable pooled worker system. It also introduces a raw-pointer-based WASM API to bypass wasm-bindgen borrow-checking limitations, allowing multiple DSP instances to reside efficiently in a single worker.

Key Changes

1. High-Efficiency DSP Worker Pooling

• Transition from 1:1 to N:M: Instead of spawning a dedicated Web Worker for every VFO, VFOs are now assigned as "slots" to a pool of DSP workers (defaulting to a max of navigator.hardwareConcurrency).
• Dynamic Load Balancing: New VFOs are assigned to the least-loaded worker in the pool to keep per-chunk processing time low.
• Batch Processing: Workers now process all assigned slots in a single pass and return an audioBatch message, reducing the overhead of postMessage communication.

2. Raw Pointer WASM API (processor.rs)

• Bypassing RefCell: Implemented a dsp_new/dsp_process_ptr API in Rust that uses raw heap pointers (Box::into_raw).
• Parallel Coexistence: This allows multiple DspProcessor instances to exist within the same WASM module instance without triggering WasmRefCell borrow errors, which previously necessitated separate workers for each instance.

3. Decoupled Audio Mixer

• Timer-Based Mixing: The audio mixer has been moved out of the worker message handler and into a dedicated setInterval loop (50Hz).
• Complexity Reduction: This prevents the $O(N^2)$ complexity scaling where every worker message triggered a full scan of all VFO queues, leading to much smoother audio performance under high VFO counts.

4. Remote Client Enhancements

• Independent Pooling: Remote clients now utilize the same pooling logic, ensuring that multi-client usage doesn't lead to an explosion of system threads.

Performance Impact

• Reduced Thread Overhead: Significantly lower memory and CPU usage when running multiple VFOs.
• Scalability: Allows the application to handle many more concurrent VFOs (both local and remote) before hitting browser performance bottlenecks.
• Stability: Reduced "jank" in the main thread by batching messages and decoupling the mixing logic from the DSP throughput.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	browsdr	5ff6d17	Commit Preview URL Branch Preview URL	Apr 11 2026, 10:37 AM

[jLynx]

I'm still not convinced this is the correct approach to improve performance. I think more tests are needed before this PR is ready