Friday, 29 January 2016

Parallel Programming (OpenMP, CUDA, OpenCL)

Parallel Programming (OpenMP, CUDA, OpenCL)

One of the various useful tools which I feel will be required if we need to do some simulations, applications, etc.. to run efficiently. But  the latest computer processors from intel (corei3, corei5, corei7) have multiple cores so isn't it implied that we are getting better performance ?

The short answer is NO
They simply have the functionality however if we are writing a code which performs sequential actions then we are not utilizing the resources  efficiently.
We can see that if some computations are independent then they can be performed in parallel using threads and this reduces the computation time of the overall program drastically.
To learn openMP a YouTube_series is provided by intel and instructor is Tim Mattson. The Basics of openMP and a few parallel algorithms are taught in C/C++ and fortran.
A series called Heterogeneuous Parallel Programming is also offered in Coursera which teaches basics of CUDA and openCL .

No comments:

Post a Comment