When Do Goroutines Yield in Go?

At Which Points Can a Goroutine Yield Execution?

In Go programs, goroutines, lightweight threads of execution, can yield to other goroutines at various points, enabling concurrency and efficient resource utilization. However, not all goroutine execution is subject to yielding.

Traditionally, goroutines were known to yield primarily in the face of blocking system calls (syscals). However, as noted in previous discussions, function calls can also trigger goroutine switches during stack growth checks.

The behavior described in the question, where a goroutine appears to yield even without function calls or syscals, was addressed with the introduction of asynchronous preemption in Go version 1.14. This enhancement grants goroutines the ability to be preempted at almost all points in their execution, even within endless loops without function calls.

While asynchronous preemption provides a significant improvement in scheduler responsiveness, it's important to note that preemption points may vary across different Go releases. Therefore, it's generally not recommended to rely on specific preemption points for synchronization purposes.

In the example code provided in the question, the lack of synchronization around the output array and oi index complicates the analysis of yielding behavior. Using strings in the output array could potentially involve garbage collection, which can introduce scheduling points.

In essence, goroutines in Go yield execution at various points, including system calls, function calls, and asynchronous preemption points that can occur almost anywhere. Asynchronous preemption has significantly improved goroutine scheduling behavior, but it's essential to avoid relying on specific preemption points for synchronization.

The above is the detailed content of When Do Goroutines Yield in Go?. For more information, please follow other related articles on the PHP Chinese website!