# INITIAL GUESS HERE!
if mpi_rank == 0: print 'Reading Restart File(s)'
filename = file_address + results_dir + 'vw_array{0}.npy'.format(mpi_rank)
vw = load(filename)
assert (vw.size == x.size)
pde.matvec(vw, 1)
import sensitivities as sens
T = ((mpi_size - 1) * (n_steps-1) + n_steps)*channel.dt
# gradient plots
if mpi_rank == (mpi_size-1):
uxbar_avg = sens.velAvg(0,0,n_steps,T)
uybar_avg = sens.velAvg(1,0,n_steps,T)
uzbar_avg = sens.velAvg(2,0,n_steps,T)
ux2bar_avg = sens.vel2Avg(0,0,n_steps,T)
uy2bar_avg = sens.vel2Avg(1,0,n_steps,T)
uz2bar_avg = sens.vel2Avg(2,0,n_steps,T)
else:
uxbar_avg = sens.velAvg(0,0,n_steps-1,T)
uybar_avg = sens.velAvg(1,0,n_steps-1,T)
uzbar_avg = sens.velAvg(2,0,n_steps-1,T)
ux2bar_avg = sens.vel2Avg(0,0,n_steps-1,T)
uy2bar_avg = sens.vel2Avg(1,0,n_steps-1,T)
uz2bar_avg = sens.vel2Avg(2,0,n_steps-1,T)
len = uz2bar_avg.shape[0]
for i in range(len):
vw, info = par_minres(oper, rhs, vw, par_dot, maxiter=MR_iter, tol=1E-6,
callback=callback)
pde.matvec(vw, 1)
# save vw array to file with mpi_rank in name...
filename = file_address + results_dir + 'vw_array{0}.npy'.format(mpi_rank)
save(filename,vw)
# gradient plots
import sensitivities as sens
dirc = 0
if mpi_rank == (mpi_size-1):
uxbar_avg = sens.velAvg(dirc,0,n_steps,T)
else:
uxbar_avg = sens.velAvg(dirc,0,n_steps-1,T)
len = uxbar_avg.shape[0]
for i in range(len):
uxbar_avg[i] = mpi_comm.allreduce(uxbar_avg[i])
nb = chunk_bounds
grad = sens.uBarGrad(dirc,n_chunk,nb,T,uxbar_avg,pde.zeta)
for i in range(len):
grad[i] = mpi_comm.allreduce(grad[i])
if mpi_rank == 0:
