Eight subroutines "vinifft, vfrcos, vr2cos, vr3cos" (Gauss-Lobatto) "ginifft, gfrcos, gr2cos, gr3cos" (Gauss) are added to this package, Their usage is the same as "vinifft1, vfrcos1, vr2cos1, vr3cos1" in dvfft (vfftpack). vfrcos1.f is slower than vfrcos.f vr2cos1.f is a little slower than vr2cos.f for N>=64, but a little faster for N <=32 on SUNSpacrc-10. What happens for alpha ? An additional routine cfft2d.f is added here for 2d complex fft ! gr2cos.f and gr3cos.f are quite slow compare to vr2cos.f and vr3cos.f. They should be rewritten as vr2cos.f and vr3cos.f by using directly cosqb.f and cosqf.f in gr2cos.f and gr3cos.f !!