Applications to work on Gpu

Accelerated Computing CUDA CUDA Programming and Performance
1

hi,

I’m planning to do a mini project on GPUs. I’m having Nvidia K20s and K40s available with me , and i’m familiar with cuda programming and have used Cublas and Cula acceleraters for performance. I worked mostly on image processing applications like image segmentation. I did’nt see much speedup for such applications.

I’m not not sure which project to workon. Can anyone suggest some good applications which will give high speedup by using GPUs.

Thankyou.

2

“I did’nt see much speedup for such applications.”

was any to be expected?
if so, can you explain why you did not realize any?

“which will give high speedup by using GPUs”

why is this a requirement?

3

Find it hard to believe for large data sets a GPU did not show a huge performance difference from a CPU for image processing. Can you prove this statement?

In general here is an somewhat ordered list of the types of algorithms which have large speedups over CPUs;

  1. Brute Force/Exhaustive Search
  2. Physics simulations(Monte Carlo)
  3. Image Processing/Reconstruction
  4. Sorting
  5. Linear Algebra Sub-Routines
  6. Signal Processing(FFTs)
  7. Machine Learning
  8. Graph problems (BFS based)

If you want a more detailed list look here for comparisons against high-end multi core CPUs;

http://ambermd.org/gpus/benchmarks.htm