I am looking to implement a batched QR operation to solve many small Ax=b problems where A is n x m.

The geqrf and ormqr pair implement this for single matrices as seen in the C.1 cuSOLVER documentation.

There is a batched geqrf available in cuBLAS, but no accompanying batched ormqr function to complete the QR factorisation. Is there a way I can use batched gemm to take the output of geqrfBatched (tau and the lower part of the Aarray) to compute the final part and if so what do I need to do?

I appreciate that there are other ways of computing Ax=b i.e. LU factorisation, but QR seems to be the best option when A is not a square matrix.

Any help would be greatly appreciated.