Benchmarks for async_std, tokio and futures runtimes in comparison with synchronous versions.
Tests can be run with cargo bench.
From what can be seen in benchmarks we can say:
- Synchronous std version of IO by itself is faster than those of tokio and async_std.
- When we have blocking cases of IO and at the same time we can do a computation, it is better to use async
- Tokio, async_std and async_std + futures have roughly the same performance in the presented test cases.
If we decide to go with synchronous IO, it should be noted not to call it from async functions as it will block the thread, better to spawn a separate thread for it.
Both tokio (tracking issue) and async_std (tracking issue) seem to be currently lacking support for wasm target.