Somebody might actually write a GPU program first, despite the primitive debug tools available. I suspect that they would feel butterflies in their stomach unless there was an easy way to verify the GPU results. As we get more experienced GPU programmers, there will be more original programs written for GPUs.

You will have an opportunity over the next few years to see how many programmers feel comfortable with GPUs. I predict a slow pickup, which is why I support writing parallel algorithms using BLAS, since they are relatively easy to convert to GPUs.

Just my prediction. So far, GPU computing has not exploded. Looking at language development, I only expect slow growth for the next three years. Let's hope I am wrong.

Lawrence

Not from GPU's. The rosetta software isn't suitable for GPU computations.

Open source was not the determinant. It was access to a much larger pool of resources. If they wrote in CUDA, they only have nVidia cards. By writing in OpenCL, they have a much easier time writing apps that cover a much larger pool of resources.
OpenCL = CPU, nVidia, and AMD/ATI. (and now even Intel at some projects)
CUDA = nVidia only


Edit: Read back through many older posts and you will find that HCC was originally trying CUDA before going OpenCL.