Hypre LOBPCG solver fails with 64-bit integers

I have noticed that the LOBPCG eigensolver fails when Hypre is built with 64-bit integers, but works with 32-bit integers. I have verified this by running [MFEM's example 11](https://github.com/mfem/mfem/blob/master/examples/ex11p.cpp), which implements a simple eigenvalue problem with a diffusion integrator. When MFEM is built against a 32-bit Hypre, the example runs normally and returns the correct result. If it is built against a 64-bit Hypre, I get the following output: ``` Number of unknowns: 1361 Solving generalized eigenvalue problem with preconditioning block size 5 No constraints Error in LOBPCG: Bad initial vectors: orthonormalization failed DPOTRF INFO = 129944235540480 Eigenvalue lambda nan Eigenvalue lambda nan Eigenvalue lambda nan Eigenvalue lambda nan Eigenvalue lambda nan Residual 0.00000000000000e+00 Residual 0.00000000000000e+00 Residual 0.00000000000000e+00 Residual 0.00000000000000e+00 Residual 0.00000000000000e+00 0 iterations ``` Is this issue known? And if so, is there a fix for it other than using 32-bit integer Hypre?