Guidance on Hinting Task Types (I/O vs. fast compute-Bound) via the api

[ymolists]

Hello maintainers!

I’m using go-taskflow to manage concurrency for a mix of tasks:

1. I/O-heavy work (e.g., network calls, file operations).

2. Fast compute-bound tasks (microsecond-scale computations that don’t inherently require goroutines).

Does the library provide a way to hint or explicitly categorize tasks as either I/O-bound or compute-bound? For example:

• Prioritizing goroutine allocation for I/O tasks ?
• Directing compute-bound tasks to avoid unnecessary goroutine overhead ?

I’d appreciate guidance on best practices for optimizing task execution based on type.

Thank you!

[noneback]

For now, this repo doesn't try to optimize executions of CPU-bound tasks.

The biggest reason is that, in golang, goroutines are arranged over GPM models, which means task scheduling is not transparent to upper-layer developers. Tasks may be executed among multiple physical threads, and we cannot control their scheduling and execution, making optimization much harder.

So for now, the framework treats both kinds of tasks in the same way. The copool indeed tries to reduce goroutine overhead in a general way, but does not optimize them separately.

A simple optimization would be setting executor's concurrency and GOMAXPROCS to CPU CORES, making task cache friendly.

[noneback]

We also want to find a good way to optimize CPU-bound tasks based on the current tasking framework, and the related issue is #26. But unfortunately, I don't have much free time thesedays.

If you have any good ideas or are interested in contributing to this topic, please let us know.