cannot increase dimensionality of MPI-distributed Data

[bening-gawitsa]

I try to apply FWI for marmousi 2D model and run it using MPI. This is the setting of my acquisition geometry

nshots = 60  
nreceivers = 80  
fwi_iterations = 3
shape = (1601, 401)  # Number of grid point (nx, nz)
spacing = (5, 5)  # Grid spacing in m. The domain size is now 1km by 1km
origin = (0., 0.)  # Need origin to define relative source and receiver locations
filter_sigma = (25,25)
model = demo_model('marmousi2D-isotropic', data_path= '/home/bening2019/data/', 
                   origin=origin, shape=shape, spacing=spacing, nbl=20, space_order=2)

model0 = demo_model('marmousi2D-isotropic', data_path= '/home/bening2019/data/',
                    origin=origin, shape=shape, spacing=spacing, nbl=20, grid = model.grid, space_order=2)
gaussian_smooth(model0.vp, sigma=filter_sigma)

Then when executing this code below, I got the error message that said "cannot increase dimensionality of MPI-distributed Data"

from examples.seismic import AcquisitionGeometry

t0 = 0.
tn = 5000. 
f0 = 0.015
# First, position source centrally in all dimensions, then set depth
src_coordinates = np.empty((nshots, 2), dtype=np.float32)
src_coordinates[:,0] = np.linspace(0, model.domain_size[0], num=nshots) * 1
src_coordinates[:, 1] = 1.  # Depth is 

rec_coordinates = np.empty((nreceivers, 2))
rec_coordinates[:,0] = np.linspace(0, model.domain_size[0], num=nreceivers) 
rec_coordinates[:,1] = 200.

# Geometry
geometry = AcquisitionGeometry(model, rec_coordinates, src_coordinates, t0, tn, f0=f0, src_type='Ricker')
# We can plot the time signature to see the wavelet
geometry.src.show()

and I haven't found the solution yet

[FabioLuporini]

Hi, could you turn this into an MFE?

See here for instructions: https://github.com/devitocodes/devito/blob/master/FAQ.md#how-do-i-find-the-source-of-my-bug-quickly-and-get-support

[mloubout]

geometry.src.show() will error with MPI because only one rank will have a non-empty source. Display/data debug should only be done in serial or gathered then displayed on a single rank.