Using CUDA IDE without having a NVIDIA gpu

Hello forum,

these question sounds maybe at the beginning weird but it has a logic reason behind.
My professor offered us an ssh access to the University server with a NVIDIA super extreme graphics card which name i have forgotten.
I would like now to use these CUDA 5.5 package to program my homework on my macbook air and when i think it should be fine i would like to go by ssh to the server and compile it.
Or can i compile it on my normal computer and start the binary on the server? I really know less about all that background thats why i ask such stupid questions.

If you read these and think like oh my god he should change … please tell me i am willing to continue with these topic it sound so interesting and i learned already a lot so far.

Thank you for reading.

Kind regards Alex

You should be able to compile it on a computer that doesn’t have an NVIDIA GPU. However, the latest CUDA 5.5 installer will bark at you and refuse to install if you don’t have a CUDA compatible graphics card installed. I believe the CUDA 5.0 installer does not perform any such checks, but I might be wrong.

Honestly, you’d be better off doing the development/testing on a machine that has a CUDA compatible card.

Nsight Eclipse Edition (the IDE for Linux and Mac) can be ran on a system without CUDA GPU. You can write your code and compile. It also supports remote debugging since version 5.5.

Note that you will not be able to run executables built on Mac on your university server - you need to compile them on the same OS you run them.

Dear eugeneo and vacaloca, thank your for your replies.
I will try the eclipse solution because right now my aim is it to have syntax highlighting and automatic code completion which should be both possible in eclipse. I will try now to work with it at come back to here if i have further questions to these topic.
Thank you for your fast replies. You are a great help.

Hello, i dont know why but i cant continue.
I have installed the CUDA Toolkit 5.5 on my mac. The files under /Developer/NVIDIA/CUDA-5.5 are there and i can see the samples and so on.

I did noting else only installed the Downloaded package for my MAC from there:

An tried to start he command nsight from the Terminal. How it is explained here:
The latest version of NVIDIA Nsight Eclipse Edition with support for CUDA C/C++ and support for the Kepler Architecture is available with the CUDA Toolkit 5.5 and is supported on MAC and Linux platforms. Just install the CUDA Toolkit and run ‘nsight’ on the command line.

But nothing happens it complains that these command is not known… .

I am pretty shure that i forgot some important steps but i do not find them anywhere.
Could sombuddy maybe tell me how i can start these nsight eclipse thing to write my code, copy it to the university server, gebug it and have fun?
Of course only the first element is a meter of my question. The rest is hopefully straight forward.

I have a mac book air without an NVIDIA graphics card thats maybe also a problem but like eugeneo said it should work.

Thank you for your answer.

Kind regards Alex

Please make sure you update $PATH as outlined here -

Hallo eugeneo,

I thank you so much for you great support. It helps me so much.
I started now the following two commands:
export PATH=/Developer/NVIDIA/CUDA-5.5/bin:$PATH

But nothing seems to be changed i still have to go to the directory /Developer/NVIDIA/CUDA-5.5/bin and have to type there nsight to start eclipse.

Do i have to change any parameters?

Sorry for these basic questions we learn at the university the language but not how to compile our work.

I wish you a ncie rest of the Weekend. Kind regards Alexander

Unless you type those commands in the same shell (terminal) instance or put them in your .bashrc file (so that they get executed everytime you load the bash shell) you’ll have to keep typing them every time. See: