a_gpu = clrand(context, queue, (2000,), dtype=numpy.float32)

    a = a_gpu.get().astype(numpy.float64)
    a2 = a_gpu.astype(numpy.float64).get()

    assert a2.dtype == numpy.float64
    assert la.norm(a - a2) == 0, (a, a2)

    a_gpu = clrand(context, queue, (2000,), dtype=numpy.float64)

    a = a_gpu.get().astype(numpy.float32)
    a2 = a_gpu.astype(numpy.float32).get()

    assert a2.dtype == numpy.float32
    assert la.norm(a - a2)/la.norm(a) < 1e-7




if __name__ == "__main__":
    # make sure that import failures get reported, instead of skipping the tests.
    import pyopencl as cl

    import sys
    if len(sys.argv) > 1:
        exec(sys.argv[1])
    else:
        from py.test.cmdline import main
        main([__file__])

# simple test runner wrapper (useful for test debugging in IDEs)
if __name__ == "__main__":
    from py.test import cmdline
    import sys
    sys.exit(cmdline.main(['.'] + sys.argv[1:]))