Oct. 8, 1987: insert "by the power method" in index entry for POWER,
              deleted some extraneous spaces in write state for SSIM
Nov. 4, 1987: add subroutine cg1987
Nov. 11, 1987: change dim statements on cg and cg1987 30->50.
               Also, in cg, goto 310-> goto 320
Nov. 15, 1987: modify qr so that the final column interchange is performed
               when m >= n
Nov. 16, 1987: save old cg.f in xnapack, move cg1987.f to cg.f, rm cg1987.f
Nov. 20, 1987: update overt, osolve, and omult so that ffc is replaced
               by fft, revise comment statements
Aug. 18, 1988: delete extraneous statement: na = na + 1 in cg.f, change
               armijos to armijo in write statement, delete c% lines
Sept. 3, 1988: change PFACT and PSOLVE to TFACT and TSOLVE in fasem.f
Jan. 5, 1989: corrected various problems in qr related to a row of zeros
              at the bottom of the matrix, also interchanged the threshold
              levels related to the column norms when 2 columns are switched
Jan. 19, 1989: correct the dimension of D array in fasem.f
Feb. 20, 1989: change D array to S array in comment statement of bidag
Mar. 11, 1989: change 'goto 500' to 'goto 450' in cubic part of algorithm,
               remove extraneous statement and error in cubic subroutine
               that occurs when first two arguments are the same
Mar. 22, 1989: make BOTH in cg external
Apr. 15, 1989: insert A = 1.D0 + E in the routine to estimate the machine
               epsilon in czero (sun's compiler messed up these statements)
May 27,  1989: change variable S to T in subroutine null -- this messed up
               the transition to double precision
