0

I need to calculate the determinant of a matrix (array of double) with cuda and I want use the LU decomposition to do this. I don't find the device function in cusolverSp.h but only the host function cusolverSpDcsrlsvluHost. I look in cuda 7.0 and cuda 7.5.

I don't understand why I can't find it because it is described in the official documentation: http://docs.nvidia.com/cuda/cusolver/index.html#api

Other functions like the Cholesky decomposition on the device works.

Is it implemented? If yes, where can I find it?

Thank you for your help!

TuDa
  • 77
  • 1
  • 1
  • 8

1 Answers1

1

Looking at the documentation you pointed I see:

Remark 5: only CPU (Host) path is provided

A bit of reading on the page gives this:

Next, cuSolverSP provides a new set of sparse routines based on a sparse QR factorization. Not all matrices have a good sparsity pattern for parallelism in factorization, so the cuSolverSP library also provides a CPU path to handle those sequential-like matrices. For those matrices with abundant parallelism, the GPU path will deliver higher performance. The library is designed to be called from C and C++.

Combining the two quotes, I guess the device function wasn't deemed worth providing... This is just my interpretation though.

Gilles
  • 9,269
  • 4
  • 34
  • 53
  • Thank you, I don't read the remarks and it's an error to me. Remark 6: multithreaded csrlsvlu is not avaiable yet. It's clear. Thank you. – TuDa Aug 27 '15 at 12:07