which IDE should I use to develop CUDA application

[cudars]

I am quite new to CUDA.
anyone could help to give some advices?
Thanks

[avidday]

There presently isn’t a “preferred” IDE. People posting here seem to have had varying amounts of success with either NetBeans, Eclipse, and Kdevelop, but I couldn’t say any of them sound like they really works all that well. Amongst editors, Vim (which is what I use) has very complete syntax highlighting and built-in interfaces to the Gnu toolchain and Cmake, so building and error parsing works well. On linux, cuda-gdb and the profiler are stand alone applications that I don’t believe work wrapped inside another application anyway.

The Cuda build process is so straightforward and minimal (given nvcc takes care of most of the hard stuff), that Makefiles to build even multi language, multi-source projects can be very simple.

[cudars]

There presently isn’t a “preferred” IDE. People posting here seem to have had varying amounts of success with either NetBeans, Eclipse, and Kdevelop, but I couldn’t say any of them sound like they really works all that well. Amongst editors, Vim (which is what I use) has very complete syntax highlighting and built-in interfaces to the Gnu toolchain and Cmake, so building and error parsing works well. On linux, cuda-gdb and the profiler are stand alone applications that I don’t believe work wrapped inside another application anyway.

The Cuda build process is so straightforward and minimal (given nvcc takes care of most of the hard stuff), that Makefiles to build even multi language, multi-source projects can be very simple.

Thanks. The Cuda build process is straightforward.

I want to an automatic IDE to do some routines for me.

But it seems that I should do it myself.

[avidday]

If you are looking for some degree of automation of code generation and hiding of a lot of the CUDA APIs, then something like PyCUDA might be worth investigating. You get full access to all the features of CUDA, in a native python environment that plays perfectly with shells like ipython and all of the Python IDEs.

[cudars]

If you are looking for some degree of automation of code generation and hiding of a lot of the CUDA APIs, then something like PyCUDA might be worth investigating. You get full access to all the features of CUDA, in a native python environment that plays perfectly with shells like ipython and all of the Python IDEs.

thank you. avidday. You are so kind.

I want to deal with a lot tiff image file by cuda. however ,I could not find the tiff format support in cuda.

by goolging, I knew libgeotiff could do the job.

but i do not know how to use libgeotiff in cuda.

[avidday]

CUDA doesn’t have any image file support at all. RIght now you get a C compiler and a standard math library for the GPU, a BLAS implementation for the GPU with a host API, and a fast Fourier transform implementation for the GPU with a host API.

If you want to process images, then you will need to write your own host code and/or use libraries to read the image files to host memory, then transfer the bitmaps (or parts of them) into GPU memory and then process them with whatever algorithms you code in CUDA for the GPU.

[cudars]

CUDA doesn’t have any image file support at all. RIght now you get a C compiler and a standard math library for the GPU, a BLAS implementation for the GPU with a host API, and a fast Fourier transform implementation for the GPU with a host API.

If you want to process images, then you will need to write your own host code and/or use libraries to read the image files to host memory, then transfer the bitmaps (or parts of them) into GPU memory and then process them with whatever algorithms you code in CUDA for the GPU.

Thanks. I got it.

By the way, Where can I find the manuls of the BLAS implementation for the GPU?

[avidday]

There is a manual in pdf form shipped in the Cuda Toolkit.

[highman]

I am quite new to CUDA.

anyone could help to give some advices?

Thanks

CB, code::block, several posts in this forum talk about cuda and codeblock