I’ve seen software being labeled as “GPU accelerated”, but I wonder what that means in practice.
I read that graphics(duh) and number crunching are good use cases, but by what criteria?
In the end, graphics are just arrays of bits visualized. So could you use the GPU for linear algebra, or maybe even MapReduce?
Number crunching is a broad term. Can I use the GPU to calculate prime numbers or fibonacci numbers?
Bonus points for sketching out the broader picture(cell processors, quantum computing, etc.).
Recent GPUs tend to have hundreds of core, many more than CPU. This would boost program performance when used efficiently (read: correctly parallelized).
As long as it involves pure computation (no I/O, sockets, or anything else), yes. I’ve made some programs (including prime number generations) using CUDA when I took parallel computing class at college.