The beamforming [1] implementation has the following requirements:
- cuda toolkit 5.5 compatible
- GTX480 GPU compatible
- kernel code implementation emphasis
- caller function implementation
- must use provided existing skeleton code for reading the input and output; validation tests included
- developer must be able to test and develop locally (i.e. access to CUDA capable device on which to compile and run code)
- documentation in code (including design decisions: i.e. why this is used)
- positive speedup
- preferably without using additional libraries
[1] [login to view URL]
I have enough experience in CUDA. My present Ph.D. thesis is related to calculations on GPU in astrophysics. I think that my skills could be useful for your task and our possible cooperation would be productive.
Thank you!