On Tue, 28 Sep 2010 15:55:26 -0700, Adam N <interf...@gmail.com> wrote: > Hello, > > So I've looked at the sparse stuff from Git, and the SparseSolve.py > examples, and I can see how to build a sparse matrix and a dense > vector, and how to use the CG stuff, solving A * x = b for x. What I > can't seem to figure out is how to do a basic > sparse-matrix-by-dense-vector multiply, and solve A * x = b for b. I'm > assuming the CG method requires this at some point, and I just can't > see where it happens? Is get_elwise_kernel doing some sort of magic in > order to return a kernel that works on sparse matrices and dense > vectors?
Assuming you're able to make a PacketedSpMV instance as in the example, the only thing you need to know is that you need to follow this procedure: xp = spmv.permute(x) yp = spmv(xp) y = spmv.unpermute(yp) I.e., I assume the permutation is what caused you grief. HTH, Andeas
pgpDodsnAxFgY.pgp
Description: PGP signature
_______________________________________________ PyCUDA mailing list PyCUDA@tiker.net http://lists.tiker.net/listinfo/pycuda
