Run ETH GPU backend concurrently with CPU workers

[snapss-dev]

Summary

• Added concurrent CPU worker searching to the ETH GPU backend loop.
• Updated GPU benchmark to score GPU+CPU hybrid throughput (instead of GPU-only throughput) so backend selection reflects real runtime behavior.
• Added configurable worker count support to CPU worker pool for hybrid mode sizing.

What changed

• src/ui/app.ts
  • Added getHybridCpuWorkerCount() helper.
  • In benchmark, GPU candidate now runs gpuVanity.search(...) in parallel with a smaller CPU pool.
  • In runtime GPU mode, each iteration now executes:
    • one GPU batch search
    • N CPU worker searches
      concurrently, then aggregates checked counts/results.
  • Updated labels to show GPU+CPU hybrid behavior.
• src/worker/pool.ts
  • createWorkerPool(preferredWorkerCount?) now supports capping worker count for hybrid scheduling.
• README.md
  • Added note that GPU mode can run CPU workers in parallel.

Validation

• npm run build passes.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
vanity-eth-gpu	Ready	Preview, Comment	Feb 15, 2026 11:20pm