Hi,
I am trying to permute large sparse matrix, program aborts with std::bad_alloc, it works for smaller dimensions, unable to figure out the problem

Code: Select all

   typedef Eigen::SparseMatrix<double, Eigen::ColMajor, std::ptrdiff_t>
    SparseMatrix;
   
    SparseMatrix G; // Very large matrix 760000 X 560000
    qrSolver.compute(G); // Using SuiteSparse QR


     SparseMatrix permutedMatrix = G *  qrSolver.colsPermutation(); // Runtime failure std::bad_alloc


Edit:
It is failing at SparseMatrix.h --> reserveInnerVectors --> m_data.reserve(totalReserveSize);
Trying to reserve 426,182,269,500 bytes! (totalReserveSize = 426182269500)!

How can I overcome this?
I am basically trying to convert rank deficient matrix to full column rank matrix

Thanks

What about G.nonZeros() ?

G.nonZeros() = 2352081.

I have debugged further and this line at

SparsePermutation.h -->
sizes[((Side==OnTheLeft) ^ Transposed) ? jp : j] = m_matrix.innerVector(((Side==OnTheRight) ^ Transposed) ? jp : j).size();

always returns the row size of matrix, so sizes[] object has same number (row size) in all cells.

Indeed, it should be nonZeros() instead of size(). Fixed.

Devel: https://bitbucket.org/eigen/eigen/commits/1d72cb9b08c1/
3.2: https://bitbucket.org/eigen/eigen/commits/1775926d6206/

Great. I have the same problem. Thank you