gathering the rank's domain to the rank master to write results - fortran

Question

Let's say that you separated your domain according to the number of ranks (with mpi_cart_create) then you have the indices of each domain. Then I want to use gatherv to put back a global array in order to write the results. This code seems to work for 2 or 4 processes but doesn't for 8. I feel like I am missing something. There should be no error message.
Edit : each rank contains its local array. Its elements are equal to the rank of the process (from 0 to n_procs). What I am trying to archieve is to put back the global array with gatherv. This is for a finite volume code. My aim is just to reproduce the domain decomposition and what I desire with gatherv. Here, there should be no -10 in the print and the block should be clearly dissociable with the decomposition done by mpi.
The code works as follow : I separate my domain, compute the delimitation/indices associated to each domain, definite my new type that will be received in the global array, then calculate the displacement and use gatherv.
program scatter
use mpi
implicit none
integer,parameter :: nrows = 8,ncols = 8
integer, dimension(8,8) :: global
integer, dimension(2) :: dims_orig,dims,coo,sizes,subsizes,coo_init
logical,dimension(2) :: periods
integer, dimension(:,:),allocatable :: local
integer, dimension(:),allocatable :: displ,counts
integer :: info,code,rank,n_procs,comm2d
integer :: ndi,ndj,nni,nnj,nni_last,nnj_last
integer :: i1,i2,j1,j2,ii1,ii2,jj1,jj2,row,col
integer :: type0,type_subarray,sizeofint,i,j
integer :: imax,jmax,loc_size
integer(kind=mpi_address_kind) :: start,extend
logical :: reorder
! dimensions of the global array
imax = nrows
jmax = ncols
! default value of the global array
global = -10
call mpi_init(code)
call mpi_comm_rank(mpi_comm_world,rank,code)
call mpi_comm_size(mpi_comm_world,n_procs,code)
dims_orig = 0
CALL MPI_DIMS_CREATE(n_procs ,2,dims_orig,code)
! number of domain in each direction
ndi = dims_orig(1)
ndj = dims_orig(2)
! nni/nnj = sizes of each domain
nni = nrows/ndi
nnj = ncols/ndj
nni_last = nrows-(ndi-1)*nrows/ndi
nnj_last = ncols-(ndj-1)*ncols/ndj
dims(1) = ndi
dims(2) = ndj
periods = .false.
reorder = .true.
call mpi_cart_create(MPI_COMM_WORLD,2,dims,periods,reorder,comm2d,code)
call mpi_comm_rank(comm2d,rank,code)
call mpi_cart_get(comm2d,2,dims,periods,coo,code)
if(coo(1)==ndi-1) then
nni = nni_last
i2 = imax
i1 = i2-nni+1
else
i1 = rank/ndj*nni+1
i2 = i1+nni-1
endif
if(coo(2)==ndj-1) then
nnj = nnj_last
j2 = jmax
j1 = j2-nnj+1
else
j1 = MOD(rank,ndj)*nnj+1
j2 = j1+nnj-1
endif
print*,rank,"|",i1,i2,j1,j2
call mpi_barrier(mpi_comm_world,code)
! create new types
sizes = [imax, jmax]
subsizes = [i2-i1+1,j2-j1+1]
coo_init = [i1-1 ,j1-1 ]
call mpi_type_create_subarray(2,sizes,subsizes,coo_init,&
mpi_order_fortran,MPI_integer,type0,code)
call mpi_type_size(MPI_INTEGER,sizeofint,code)
start = 0
extend = sizeofint*nnj
! call mpi_type_get_extent(type0,start,extend,code)
call MPI_TYPE_CREATE_RESIZED(TYPE0,start,extend,TYPE_SUBARRAY,info)
!type_subarray = type0
call MPI_TYPE_COMMIT(TYPE_SUBARRAY,code)
allocate(displ(ndi*ndj))
! forall(col=1:ndj,row=1:ndi)
! displ(1+(row-1)+(col-1)*ndi) = (row-1)+(col-1)*imax
! ! displ(1+(row-1)*ndi+(col-1)) = (row-1)*ndi+(col-1)
! endforall
! computing the displacement
do i =1,ndi
do j = 1,ndj
displ(1+(i-1)+(j-1)*ndj) = (i-1) + (j-1)*imax
enddo
enddo
allocate(local(i1:i2,j1:j2))
allocate(counts(n_procs))
counts = 1
local = rank
loc_size = (i2-i1+1)*(j2-j1+1)!(i2-i1+1)*(j2-j1+1)
call mpi_gatherv(local,loc_size,MPI_integer,&
global,counts,displ,type_subarray,0,&
MPI_COMM_WORLD,code)
call mpi_barrier(mpi_comm_world,code)
if(rank==0) then
do i = 1,nrows
print*,global(i,:)
enddo
endif
call mpi_finalize(code)
endprogram scatter