GPU Computing

cancel
Showing results for 
Search instead for 
Did you mean: 

Image processing using GPU-CUDA program

Hi everyone;

Anybody knows or can shared the knowledge and experienced to perform  3-d image processing using GPU(graphical processor unit) using CUDA program. Before this I used SB-RIO to perform the image processing process. For instant, i try makes comparison using FPGA and GPU. which is better; Hope everybody would shared the experience.

AbM

0 Kudos
Message 1 of 6
(11,011 Views)

Hi AbM,

I am doing image processing on FPGA on NI 1473R.

Seems like a good platform. I have good results in LabVIEW.

I don't have experience with CUDA. But while talking to companies outside of NI that doing both (matrox imaging) I received the feedback that usually both platforms are effective on very similar algorithms.

So it is depends on system configuration.

I would be very interested to find out what others think.

Amit,

Amit Shachaf
0 Kudos
Message 2 of 6
(9,413 Views)

Hi Amit,

Before this I already used SB-RIO 9632 for image processing. I tried to makes comparison, FPGA and GPU, which is get better image results.

0 Kudos
Message 3 of 6
(9,413 Views)

Hi Amit,

I am having  PCIe-1473R frame grabber &  a Dragster camera which communicate over CamLink. I am using Example Code "1-Tap 10 bit camera with frame trigger" but I am getting error of timeout whenever I am pressing Acquire button on the Host VI.

Can you share your code or can you guide me to take images from the camera

0 Kudos
Message 4 of 6
(9,413 Views)

Hi R.V.  Sorry for the late replay. I didn't get massage from this thread. Yes I can share some code if you still need it. Not the whole application. It is very large and I am using dual tap camera.  Amit,

Amit Shachaf
0 Kudos
Message 5 of 6
(9,413 Views)

Hi Amet,

Hi everyone,

I know that this thread is relative old, but I guess now its the right time to ask you guys about your experience using NI-PCIe1473 for image processing routines.

I am planning to use 1473 for image acquisition and pre-processing and then pass the images forward to a GPU for the high level image processing algos. This is my first experience with FPGA and 1473 so I'm not sure if fits into it! Maybe you can help me with a few lines of your experience

Basic ideas of using 1473 is to grab 4 images from Cameralink(full) -> store on D-Ram of 1473 -> do some basic operations on all the four images at the same time (i.e. +-/*) and then pass it to the host.

Every image would be around 10MB!

As I always read it is not a good practice to have large arrays in FPGAs, do you think it is managable?

Thank you very much in advace for sharing,

Arash

0 Kudos
Message 6 of 6
(9,413 Views)