Requirements of Running Cuda or Jcuda Programs on Linux Sytems

Byte-Welt - Projekte / Projects JCuda
1

Dear all,

I want to have some practice examples up and running in JCuda.

Please let me know if there is any useful link that describes the requirements needed to run JCuda Programs and Installing & Running CUDA Packages needed for it.

However I need to run it on Hadoop Cluster.

But my first step is to run simple cuda,jcuda programs.

Thanks & best Regards,

Adarsh Sharma

2

Hello,

Sorry, I know that there should be more documentation of JCuda concerning the “Getting started” and the initial setup. So far, the first step could be to run a “trivial” program, as descibed in the FAQ Thread (at the bottom, starting at "To set up a minimum JCuda project as a basic test, you may follow these steps:…")

If this works, you may want to try some of the samples from the website. Depending on what you intend to do (I assume that you’re groing to create own kernels), the “JCudaDriverCubinSample.java” from the samples section may be a good starting point.

Figuring out specific settings or configurations for Haddop could then be the next step.

Hopefully I’ll find the time for a more extensive Tutorial soon…

bye
Marco

3

Thanx Marco for the useful links.

I just downloaded following packages :

  1. cudatoolkit_3.2.16_linux_64_rhel5.5.run

  2. devdriver_3.2_linux_64_260.19.26.run

  3. gpucomputingsdk_3.2.16_linux.run

But how to install them , is they are binary files which r executed by ./cudatoolkit_3.2.16_linux_64_rhel5.5.run command or i have to download rpm packages.

Thanks & best regards,

Adarsh Sharma

4

Hello,

Since I don’t have a CUDA capable Linux PC, I can not give specific information about the installation of the drivers. On the CUDA Downloads site there is a Getting Started Guide (PDF file, for version 3.2) which explains the installation of the driver. (This document also contains a link to a more extensive manual page, but in the best case, this is not necessary)

The installation of the Toolkit and the SDK should be straightforward, basically:

  1. Install the Toolkit by executing the .run- file as a ‘superuser’
  2. Define the environment variables.

export PATH=/usr/local/cuda/bin:$PATH 
export LD_LIBRARY_PATH=/usr/local/cuda/lib:$LD_LIBRARY_PATH
  1. Install the SDK by executing the .run-file as a ‘regular user’
    (But refer to the Guide for more detailed information!)

Hope that helps
bye
Marco