We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
Hacker News·20d ago·5 min read
The Tokio/Rayon Trap and Why Async/Await Fails Concurrency
TL;DR
Async/await simplifies writing asynchronous code but introduces significant complexity in operation, shifting scheduling burdens to developers. This complexity can lead to confusion between asynchrony and concurrency.
✦ Why It Matters
Consider evaluating your project's concurrency model to determine if async/await is the best fit for your needs.
Key Takeaways
- Evaluate your current use of async/await for potential performance issues.
- Consider separating I/O and CPU-bound tasks to avoid latency spikes.
- Explore Project Tina for a more predictable concurrency model.
- Implement strict memory bounds to prevent out-of-memory crashes.
How It Works
Project Tina uses a thread-per-core model, where each core handles its own workload without migrating tasks. This eliminates the overhead of work-stealing and maintains CPU cache locality, enhancing performance.
Memory is pre-allocated, and mailboxes are bounded, ensuring that the system can handle traffic spikes predictably without crashing.