4 ms·
If you like algorithmic-type stuff, give opencl/cuda a shot. It's amazing to be able to really feel the kind of power locked away in a standard desktop gpu. Wri
by zalzane 13y ago
If you like algorithmic-type stuff, give opencl/cuda a shot. It's amazing to be able to really feel the kind of power locked away in a standard desktop gpu. Writing gpu code is also much different than writing standard synchronous code, especially for algorithms that aren't embarrassingly parallel.
I wouldn't call gpu computing a "hot" rising technology, but it's certainly hardcore and is used widely in research, games, and many other computationally intensive applications.
- ratsimihah 13y agoYes, it seems widespread enough that whatever it has to teach cannot be wasted. Thanks!