Synopsis
		sfmaxwell_cpml24_backward < Fv.rsf > Fw1.rsf back=Fw2.rsf rho=Frho.rsf eta=Feta.rsf ef=Fef.rsf eb=Feb.rsf wltf=Fwltf.rsf wltb=Fwltb.rsf nb=30 nt=1000 dt=0.001 fm=20. nsnap=0 order1=y attenuating=y

It is shown that even with attenuation (moderate), reverse reconstruction
is still a feasible way to build the incident wavefield, using boundary
saving scheme. Allowing for the heavy burden of the boundary saving in 3D,
we prefer 4-th order FD in space.

Parameters
		attenuating=y add attenuation or not   file back= auxiliary output file name   dt=0.001 time sampling interval   file eb= auxiliary output file name   file ef= auxiliary output file name   file eta= auxiliary input file name   fm=20. domainant frequency for ricker wavelet   nb=30 thinkness of sponge ABC   nsnap=0 number of snapshots for re-initialization   nt=1000 number of time steps   order1=y 1st order or 2nd order accuracy   file rho= auxiliary input file name   file wltb= auxiliary output file name   file wltf= auxiliary output file name