How to create a c++ program that can utilize multiple cpu cores.?

Question

How to create a c++ program that can utilize multiple cpu cores.?

Answer 1

Check out the Boost threading library

http://www.boost.org/doc/libs/1_37_0/doc/html/thread.html



It has the benefit of being multi platform so you can write code once and compiler it anywhere, also you don't need to investigate a series of cryptic Win API methods. Additionally, since it's C++, it's a much more streamlined, object oriented approach to doing threading which will result in more readable and shorter code.

Answer 2

On Windows:



*Include windows.h



*Find some decent win32 API doco; I reccomend using API Guide.



*The API you're looking for is called CreateThread(...)

Answer 3

Multi-threading requires a good technical analysis to be efficient, but if you're a good programmer, you should be able to get it done.



The functions to use depend on your environment (compiler, OS).



A simpler approach is often possible in batch processing. Sometimes you could simply find a way to split the task and run a few processes simultaneously. This method often performs better, if applicable. The same principle could be done in a single program using fork().

Answer 4

its beyond my ability to fully explain how to do that here. however i can point you in the right direction;

look into multi-threading and parallel programming.



just a warning, it will make your code slightly more confusing, XD

but well worth the trade off.

Answer 5

I'd say you have three basic directions you can take this:



1) redesign the program for multithreading and use std::threads (if already supported by your compiler) or boost.threads (if not): http://www.boost.org/doc/libs/1_44_0/doc/html/thread/thread_management.html



2) make use of std::packaged_task / boost::packaged_task ( http://www2.research.att.com/~bs/C++0xFAQ.html#std-future and http://www.boost.org/doc/libs/1_44_0/doc/html/thread/synchronization.html#thread.synchronization.futures.reference.packaged_task )



3) get serious and use MPI, for which boost.MPI is the most C++-friendly approach: http://www.boost.org/doc/libs/1_44_0/doc/html/mpi/tutorial.html