In test-cholesky, there is a test that finds the cholesky factorization of a doubly blocked matrix.
Completing the test takes 30 or more seconds and requires 2.5 million calls to simpexpt (ouch). If it's OK
with you all, I'll change the test to a single block matrix (much faster). Other than bravado, I don't
see the point of the lengthy test (I'll comment out the old test with a note about slowness).
I'm not sure why simpexpt gets called 2.5 million times (but the factorization is a huge mess). The test is
(%i1) (matrix_element_transpose : 'transpose, matrix_element_mult : ".", 0);
(%o1) 0
(%i2) (m : matrix([matrix([matrix([8,4],[4,13]),matrix([2,4],[7,9])],[matrix([2,7],[4,9]),matrix([10,6],[6,15])]),matrix([matrix([4,4],[12,5]),matrix([0,6],[6,8])],[matrix([8,3],[13,3]),matrix([6,5],[5,6])])],[matrix([matrix([4,12],[4,5]),matrix([8,13],[3,3])],[matrix([0,6],[6,8]),matrix([6,5],[5,6])]),matrix([matrix([18,5],[5,4]),matrix([5,9],[2,6])],[matrix([5,2],[9,6]),matrix([6,3],[3,11])])]),0);
(%o2) 0
(%i3) timer(?simpexpt,?exptrl)$
(%i4) (m1 : cholesky(m.transpose(m), 'noncommutingring),0);
(%o4) 0
(%i5) timer_info();
(%o5) matrix([function,time//call,calls,runtime,gctime],[exptrl,1.016838557178039*10^-5*sec,2626769,26.71*sec,0],[simpexpt,1.7136231261319827*10^-5*sec,2532062,43.39*sec,0],[total,1.3588349763735237*10^-5*sec,5158831,70.10000000000001*sec,0])
--Barton