Writing code using CUDA already makes your program limited to NVIDIA GPU's, and hence not very portable.  (Exactly the desire of NVIDIA, of course).  Most of the GPU codes I've seen use the OpenACC set of directives (Fortran) / pragmas (C/C++) instead. These have the advantage of (relative) ease of use, and of portability to non-GPU systems, where the directives are just ignored.  The latest versions of OpenMP also have support for GPU computing. 

Hi Pejvak

NVIDIA developed CUDA to perform parallel processing under C, C++ and Fortran, all without the need for the use of assembly language. This was no mistake. These are all high performing languages, and the aim of CUDA was brute power. If you wanted to use any other software for this purpose, you would likely find you would need to use assembly language, which you would likely have to write yourself, only to discover the the whole thing would run far, far slower than what is on offer by CUDA.

I hope that helps.

Regards

Bob

You have to use one of the API (CUDA or OpenCL) on GPUs for using it General purpose computing like in your case bio-informatics or you can use matlab to parallel your code using GPUs.

Hi Pejvak,

Take a look a this:

https://developer.nvidia.com/how-to-cuda-python

There you will find the information for programming GPUs from Pyuthon.

Best.

Writing code using CUDA already makes your program limited to NVIDIA GPU's, and hence not very portable.  (Exactly the desire of NVIDIA, of course).  Most of the GPU codes I've seen use the OpenACC set of directives (Fortran) / pragmas (C/C++) instead. These have the advantage of (relative) ease of use, and of portability to non-GPU systems, where the directives are just ignored.  The latest versions of OpenMP also have support for GPU computing. 

use opencl rather!

Keep in mind that in any available, effective option for GPU programming, you don't "port" an existing code.  You either rewrite it or touch almost every part of it.  CUDA, as was mentioned before, is limited to NVIDIA GPUs while OpenCL and others attempt to generalize to other processor architectures.  The development and acceptance of a programming layer for the variety of emerging advanced architectures (Intel Knights Landing MIC, GPGPU) is a very active research area.  What is the best today may not be (will hopefully not be) the best tomorrow.

PASTA is not accelerated by any GPU technology.   If you are familiar with programming then you can use NVBIO libraries. The Matlab Bioinformatics Toolbox is also some alternative:)

I think that you should look for the parallel programming method that is compatible with your computer. For example there are problems with MAC computers where it's suitable to use OpenCL. But I prefer to use CUDA with C++ for Parallel computing with OpenCV if you are working on computer vision

Perhaps, you need to re-write the implementation. I recommend Heterogeneous Language ( like OPENCL) instead of dedicated CUDA to make a comparison between Sequential implementation, Multicores (CPUs) evaluation, Many Cores (GPU) Evaluation. Then the experiment could be done as far as the latency and throughput are concerned.

Best regards. 

If you use NVIDIA GPUs, you can use CUDA. If you use Python, you can use PyCUDA (but the code that is to be run on GPU, called "kernel", needs to be written in CUDA C, a dialect of C language).