Preferred programing language research

[Scorcher646]

This year in school I am required to complete a year long project that results in a final product and learned skills. I have decided that I want to learn parallel/multi-core programing. The linked form is for the sole purpose of deciding which language I should use for my project based on ease of use and possible support base (popularity). I would greatly appreciate if you guys, and gals, could fill out the included Google form to assist me with this research.

Thanks,

scorcher646

https://goo.gl/forms/mrTSilLs72idd64v2

[King_of_Oz]

I use C# with the unity game engine.

[vorticalbox]

python for me i created a script download images from a webpage using all the cores the cpu. 

[Mira Yurizaki]

C, C++, C#, Python, these all work more or less for what you want.

 

However, I find just multi-threaded programming by itself vague at best. You should find a problem that can be parallelized easily and scales up with the number of cores. An easy one I think is running the basis of JPEG compression.

[mathijs727]

I have 2 favorite languages: Python and C++.

I write multi threaded applications with them regularly.

 

By the way, does GPGPU also count towards multithreading?

[Mira Yurizaki]

41 minutes ago, mathijs727 said:

By the way, does GPGPU also count towards multithreading?

I guess it depends on what the work. SIMD is multithreading, but at the same time when I think of it, I also think of multiple workers doing different, independent things that require managing resources and such multithreading (and probably the more interesting variant).

[Spigie]

Personally my favorite programming languages at the moment are: Common Lisp, Python, Lua, C and C++ (at least those are what I use most at the moment). All of them have some way of doing multicore programming (although admittingly Lua doesn't have this natively, although it is possible through something like LuaLanes). When I was programming on Windows (which I don't do anymore) I did a lot with C#, which also has support for multi-core programming. I think you will find that even if a language doesn't support multi-core/parallel programming natively there will likely be a library for the language that implements it (at least that is the case most of the time, there are exceptions to the rule of course).

 

If you don't know what language you should do your project in then you should do some research on languages which you know have support for multi-core/parallel programming, and find one that you like using.

[reniat]

6 hours ago, mathijs727 said:

I have 2 favorite languages: Python and C++.

I write multi threaded applications with them regularly.

 

By the way, does GPGPU also count towards multithreading?

As someone who only did python concurrency in college in a janky project (which ended up getting me my current job), is there a clean way to get true non-shared concurrency in python? The only way I did it was (again, janky college code from years ago), programmatically spinning up more instances of the python interpreter and having them work on sectioned segments of data, since memory isn't shared as each interpreter is its own process in windows.. It was the only way I knew how to get around the global interlock. 

Anyway, to answer the thread, C++ is my favorite hands down. 

[mathijs727]

4 hours ago, reniat said:

As someone who only did python concurrency in college in a janky project (which ended up getting me my current job), is there a clean way to get true non-shared concurrency in python? The only way I did it was (again, janky college code from years ago), programmatically spinning up more instances of the python interpreter and having them work on sectioned segments of data, since memory isn't shared as each interpreter is its own process in windows.. It was the only way I knew how to get around the global interlock. 

Anyway, to answer the thread, C++ is my favorite hands down. 

Python contains a multiprocessing module which functions to such as a parallel map and apply.

It still works by spinning up multiple processes, but the module also contains primitives to communicate between the threads/processes.

 

If you have data parallelism then you probably dont need communication and the map function will be all you need.

[Dat Guy]

12 hours ago, Scorcher646 said:

I have decided that I want to learn parallel/multi-core programing.

 

Use Common Lisp.

[AidoP]

I would recommend Haxe. It is a beautiful language that can "compile" to many different "languages" (targets) allowing it to be used for a vast array of things.

 

It doesn't see much use, but it is far from dead.

 

It is by far the cleanest and most pleasant language I've used.

[Mira Yurizaki]

9 hours ago, reniat said:

As someone who only did python concurrency in college in a janky project (which ended up getting me my current job), is there a clean way to get true non-shared concurrency in python? The only way I did it was (again, janky college code from years ago), programmatically spinning up more instances of the python interpreter and having them work on sectioned segments of data, since memory isn't shared as each interpreter is its own process in windows.. It was the only way I knew how to get around the global interlock.

There's a multithreading module you can use. I've used it in a few examples on a post I did a while ago:

 

[reniat]

2 hours ago, M.Yurizaki said:

There's a multithreading module you can use. I've used it in a few examples on a post I did a while ago:

 

does it let you get around the GIL if your goal was true simultaneous processing on different cores?

[Mira Yurizaki]

2 hours ago, reniat said:

does it let you get around the GIL if your goal was true simultaneous processing on different cores?

Probably not, but the notes about GIL on the Python wiki say it's a problem if you use CPython as your interpreter. If you use Jython or IronPython, there is no GIL

[reniat]

1 hour ago, M.Yurizaki said:

Probably not, but the notes about GIL on the Python wiki say it's a problem if you use CPython as your interpreter. If you use Jython or IronPython, there is no GIL

I wonder if they'll ever move the reference implementation from C to C++, to allow for stuff like native simultaneous concurrency.

[Mira Yurizaki]

3 minutes ago, reniat said:

I wonder if they'll ever move the reference implementation from C to C++, to allow for stuff like native simultaneous concurrency.

C++ may not be better. .NET was designed with thread safety in mind and with facilities for thread safe access. Not sure if Java or the latest revision of C++ is the same way.