Cuda Least Squares

I’m a relative Cuda novice (using a RTX4070 mobile GPU on a laptop) and am interested in solving least squares problems on a GPU. In forum searches I’ve found several references to a C-Cuda library routine cusolverSpDcsrlsqvqr (sparce only, no equivalent routine for dense matrix) as a host only routine (no GPU implementation even in Cuda 12.2). So, what are people using? Magma? Python? Something else?

1 Like