Hello, Iam trying to use lu_factorize and lu_solve for solving a linear system Ax= B where matrix A has entries of type double and vector B entries of type point ( a simple class wrapping double [3] ) but I get compiler errors in in_place_solve ( all of the overrides have this problem) BOOST_UBLAS_CHECK (e1 () (n, n) != value_type (0), singular ()); apparently this expects that "double" and "point" are comparable using != operator. The problems disappear if I change the line to BOOST_UBLAS_CHECK (e1 () (n, n) != E1::value_type (0), singular ()); but I leave it for the experts to give me the correct solution. PS : I tried to switch to MTL after this, but it too, had the same problem. Thanks, Arun