Feb 12
20

Debugging CUDA Device in Visual Studio – Installing NVIDIA Nsight

4 Comments

This post assumes you already have Visual Studio and CUDA installed.

It is quite a process to get Nsight installed if you are not yet registered as an NVIDIA developer (registration required). Below is a list of all the steps you need to follow to get Nsight up and running in Visual Studio assuming you are not yet a registered NVIDIA developer.

  1. Register as a Parallel Nsight Developer here: http://developer.nvidia.com/user/register
  2. Login here: http://developer.nvidia.com/user/login
  3. Click on the link to complete the “Basic Registered Developer Program Profile”. See pic below:
    debugging_cuda_visual_studio_with_nsight_01
  4. Once you have saved this information. Navigate to My Account and click “Apply” for the Parallel Nsight Registered Developer Program. Once you have submitted this form you have to wait for their support team to get back to you they say it takes an hour for this process but it took about 5 hours for me.
  5. Log back into your profile page via http://developer.nvidia.com/user/login
  6. Now apply for the “Parallel Nsight Registered Developer Program” click on the associated “Apply” link. See pic below:
    debugging_cuda_visual_studio_with_nsight_02
  7. Fill out the form and submit. You then need to wait for the approval email. They say this takes a few minutes, it
    took 5 minutes for me.
  8. Ensure you are still logged in and navigate to http://developer.nvidia.com/rdp/parallel-nsight-downloads
  9. On this page you will see the minimum required driver for this version. If you do not meet the requirements (can check this via the NVIDIA control panel Help->System Information: Driver Version) you can download the lastest drivers from http://www.nvidia.com/. Alternatively if your gfx card does cannot support the latest driver then download an older version of Nsight by going to http://developer.nvidia.com/rdp/parallel-nsight-archive (Version 2.0 of Nsight requires a driver of atleast version 270.81)
  10. You can then download and install Nsight from here.
Was my post helpful? :)

Posted in CUDA Programming by stevenmarkford 4 Comments

4 Responses to "Debugging CUDA Device in Visual Studio – Installing NVIDIA Nsight"

  • Qadha'a says:
    March 3, 2012 at 9:06 am
    Reply

    Hello..
    I read your previous posts about CUDA they where detailed and helpfull..
    I need your help, I’m new for CUDA … I know the CUDA architecture very well
    but programming by CUDA not yet .. I feel strange
    .. all tutorials I read not
    talking about how to have a .cu file and do I need to have both .cpp and CUDA
    togather or only .cu .. I’m really confused would you please recommend to me
    a good book or tutorial that will help me from scrach…

  • Qadha'a says:
    March 3, 2012 at 9:06 am
    Reply

    the other point …
    I tried to work with Parallel Nsight 2.1 and I have some
    problems could you please help me??

    I have Qosmio X775, windows7 x64

    1.Update drive device to be 285.86

    2.install VS 2010

    3. install CUDAToolkit 4.1

    4. install CUDA SDK 4.1

    5. test a sample code by executing an exe file and its work fine

    4. install Parallel Nsight 2.1
    5. make the connection between Nsight monitor and VS .. and it is work.

    when I debug and excuted a project from Parallel Nsight sample that came with with the package …
    it worked fine no errors.

    but when I created new project I got many kind of errors

    1. create my new project.
    2. check CUDA 4.1 box in the build customization of the project.

    3. in the project properties > configuration > VC++ directories I add CUDA macro directories (INC, BIN, LIB)

    4. I added item from NVIDIA >> CUDA C/C++ file.

    5. put a sample code from the book “CUDA by Examples” with the headers

    #include
    #include
    #include “device_launch_parameters.h”
    #include
    #include
    #include
    #include “cuda_runtime.h”

    __global__ void kernel( void ) {
    }

    int main( void ) {

    kernel<<>>();

    printf( “Hello, World!\n” );

    return 0;

    }

    it gave me LNK errors I google it and I found that because of
    main() in the wrong place it should be in .cpp file

    I moved the code to be in .cpp file but the problem was
    with the “<<<" kernel call … I tried every thing to solve
    this proble but no use … I think there are some
    problems
    in my project configration since the sample project are
    worknig perfectly … please if you know some that will help tell me

  • Qadha'a says:
    March 4, 2012 at 9:26 am
    Reply

    Thanks alot you are a great person >>> it just work after I followed your instructions :))))
    I’m Praying for you :))

    Good Luck

Leave a Comment

Notify via Email Only if someone replies to My Comment