linbox-team / linbox

LinBox - C++ library for exact, high-performance linear algebra
https://linbox-team.github.io/linbox
GNU Lesser General Public License v2.1
83 stars 28 forks source link

Dense dixon with high bitsize fail #197

Open Breush opened 5 years ago

Breush commented 5 years ago

Might be because of applyM (in apply.h) which handle numbers below 53 bits differently.

#include <linbox/solutions/solve.h>

using namespace LinBox;

int main(void)
{
    using Ring = Givaro::ZRing<Integer>;
    using Matrix = DenseMatrix<Ring>;
    using Vector = DenseVector<Ring>;

    Ring ZZ;

    Matrix A(ZZ, 1, 2);
    Vector x(ZZ, A.coldim());
    Vector b(ZZ, A.rowdim());
    Ring::Element d;

    A.setEntry(0, 0, "-4354606737974399");
    A.setEntry(0, 1, "1444080914074811");
    ZZ.assign(b[0], "-7270283604704528652531814423094");

    // Calling Dixon
    Method::Dixon method;
    method.singularSolutionType = SingularSolutionType::Random;
    solve(x, d, A, b, method);

    std::cout << "[ " << A.getEntry(0, 0) << " " << A.getEntry(0, 1) << " ] x = [ " << b[0] << " ]";
    std::cout << " => xNum/xDen = [ " << x[0] << " " << x[1] << " ] /" << d << std::endl;

    if ((A.getEntry(0, 0) * x[0]) + (A.getEntry(0, 1) * x[1]) != b[0] * d) {
        std::cerr << "PROBLEM! Ax != b" << std::endl;
    }

    return 0;
}
Breush commented 5 years ago

test-solve-full -m 1 -n 1 -b 1 -B 30 exhibits the same problem