Ordered Concurrency in Go: Achieving Speed Without Sacrificing Order

Go's concurrency is a powerful feature, but it can disrupt the natural order of data processing. This article explores three approaches to building a high-performance ordered concurrent map in Go. The author presents three methods: a reply-to channel approach, a sync.Cond based turn-taking approach, and a permission-passing chain approach. Benchmarks reveal the permission-passing chain, especially when combined with a channel pool to eliminate allocations, as the clear winner in terms of performance and memory efficiency. This method cleverly uses channels for efficient point-to-point signaling, avoiding the 'thundering herd' problem and achieving a balance between concurrency and order.