Get Your Hand Dirty With C.U.D.A
Recently, me and my friend got our hand on research on C.U.D.A which stand for Compute Unified Device Architecture. C.U.D.A is a parallel computing engine in NVIDIA GPUs. As the lab rat we use a Nvidia GeForce 9800 GT from Digital Alliance. The main objective is to replace the main computing component from using common CPU to a GPU unit for certain function such as determine billion of prime number.
The main reason why we use GPU to replace Common CPUs is the flop of its main memory is significantly huge, than a common CPUs. Now, the fact that the common CPUs that have more complete instruction set than a GPU make it more difficult to process a huge data that have a same pattern on one cycle, but not for GPUs that meant to build to process a huge data which have a same pattern on one cycle.
For the starter we try to find as much as possible reference that can support our research, here some e-book that we find :
- Introduction to GPU Computing (PDF)
- Basic CUDA (PDF)
- CUDA Toolkit & Libraries (PDF)
- CUDA Optimization (PDF)
- Introduction to OpenCL (PDF)
- Case Study: Computational Fluid Dynamics (PDF)