Was looking into some GPU CUDA samples and trying some samples out from Nvidia’s pack and also some of the demos in OpenCV. Most of Nvidia’s samples run, but OpenCV gives me mostly OpenCV Error: Gpu API call (invalid device symbol).
I have a gts 8800 card, supporting what Nvidia lists as compute capability 1.0. What are the differences and are my errors related to this? Can I run more stuff if I buy a newer card?
It seems that OpenCV supports CUDA capabilities from 1.1 and up. Sad, but true…
From OpenCV Makefile: