On Monday, November 21, 2016 at 2:23:37 PM UTC-5, zhaoguo wang wrote: > > Hi all, > > I wrote a simple application: the main function creates two goroutines, > while these two goroutines use RPC to do simple communication continually. > > I found > *if the main function falls into a busy loop after creating the two > goroutines, * > *the whole application will hang up after sending/receiving a couple > number of RPCs* (~80 on my testbed). > > Can anyone illustrate why this happens? As each goroutine uses > independent Linux threads, why one spins can affect others? > > Threads are independent, but there's still a single runtime scheduler. IIRC the busy loop is preventing the GC from being able to stop the world, which in turn ends up blocking the scheduler.
> I use *runtime.GOMAXPROCS(4)* to make sure each goroutine uses a > Linux thread and the testbed is ubuntu 14.04 with 8 CPU cores. > I use the go1.7.3 library which is compiled from the source code. I also > attach the source code of the simple test, you should be > able to run it with "go run test.go". > Just never use a busy loop. A busy loop is always an error, and there's no reason to use 100% of a CPU just to block the main goroutine. You can use a WaitGroup, channel receive, or even an empty select statement to block efficiently if you want. -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
