Difference between Go's multithreading and pthread or Java Threads

[shreyasva]

I read somewhere that Go's multithreading is implemented using
pthreads, is this true.
What is the impetus for using Go's multithreading over other options.
Taken a algorithm that can be parallelised and implemented in Go and
C, which one would run faster.

[John Asmuth]

http://golang.org/doc/go_faq.html#goroutines

:)

[Mike Ramirez]

I can't answer this definitively, but have you tried benchmarking these things yourself?

With this said, Go is still not on par with C in terms of performance yet, there is a lot of room for improvement with go.

Mike

--

Charity, n.:

A thing that begins at home and usually stays there.

[Alexander Surma]

> I read somewhere that Go's multithreading is implemented using
> pthreads, is this true.

This is not entirely true.

Afaik, pthreads are used only, when you compile your Go code with gccgo.
The 8c/6c use an own, much more lightweight version of threads (that's
why they're called go routines) with their own scheduler and cooperative
scheduling(!). You can see this, if you run a go program with multiple
go routines. Your os's process view will only show one process (in
contrast to the same program compiled with gccgo).

Regarding your performance question: This depends heavily on benchmarked
program.
Gccgo is better when it comes to optimizing. However, context switches
of go routines are much cheaper. So big number of short go routines
might be faster in 8c/6c, a few, rather long go routines might be more
efficient with gccgo.

HTH,
Surma

[El-Hassan Wanas]

IIRC, the number of threads the go runtime will spawn is undefined, but the number of goroutines is greater than or equal to the number of threads. The goroutines are scheduled by each thread that is responsible for them.

Wanas

[Paulo Pinto]

Actually how do go routines compare against using fibers in Windows,
or
Solaris LWP?

Or even against multithreading libraries like Java
util.concurrent, .Net TPL,
or C++ TBB and PPL.

--
Paulo

[Rob 'Commander' Pike]

On Feb 22, 2011, at 11:24 AM, Paulo Pinto wrote:

> Actually how do go routines compare against using fibers in Windows,
> or
> Solaris LWP?

They're supported by the language. The rest is implementation detail.

-rob

[bflm]

On Feb 22, 8:24 pm, Paulo Pinto <paulo.jpi...@gmail.com> wrote:
> Actually how do go routines compare against using fibers in Windows,
> or
> Solaris LWP?
>
> Or even against multithreading libraries like Java
> util.concurrent, .Net TPL,
> or C++ TBB and PPL.

E.g.: "Goroutines are multiplexed as needed onto system threads."
src: http://golang.org/doc/GoCourseDay3.pdf

In other words - a goroutine is not by default fixed to execute only
in the same thread in which it was created. I think this might be the
most important difference when comparing to some of the above
mentioned (eco)systems/libs.