Get the latest tech news
How fast are Linux pipes anyway?
Pipes are ubiquitous in Unix --- but how fast can they go on Linux? In this post we'll iteratively improve a simple pipe-writing benchmark from 3.5GiB/s to 65GiB/s, guided by Linux `perf`.
fizzbuzz writes the output in blocks as big as the L2 cache, to strike a good balance between cheap access to memory and minimizing IO overhead. But often this compromise is worth it, and in fact it is a common pattern for high-performance server applications: we trade off possibly wasteful CPU utilization for better latency and/or throughput. Pipes and splicing in particular aren’t really hot topics when it comes to high-performance programming, but the themes we’ve touched upon are: zero-copy operations, ring buffers, paging & virtual memory, synchronization overhead.
Or read this on Hacker News
