Parallelism in Go
Writing #Go code that runs in parallel first requires your use case to support concurrent tasks. (#Concurrency vs parallelism)
You OS has a native support for threads. Most programming languages however supply some kind of abstraction on top of threads though as thread have quite the big amount of overhead. Go is no different.
In Go, to make some of your code run in parallell, you need to use so called goroutines.
Threads vs Goroutines
| Abstraction | Starting memory allocation | Memory can grow? |
|---|---|---|
| Thread | ~1 MB | No. |
| Goroutine | ~2 kB | Yes. |
Goroutines still has a little bit of overhead for each goroutine you create, but as the initial memory allocation (for their stack) is so low you as a user can approximately start up ~500 times more goroutines than threads on your computer, leading to being able to instantiate up to millions of goroutines.
There’s no way to cheat the limit of number of CPU’s, so you can’t have million of goroutines running in complete parallelism, but Go deals with the concurrency and goroutine-switching for you so all your different goroutines get all to execute.
References
- Mike Van Sickle. (2015, June 5). Concurrent Programming with Go [Course]. Pluralsight. https://www.pluralsight.com/courses/go-concurrent-programming
