Problem in splitting array with MPI_ISEND with fortran - parallel-processing

Question

I wrote the following lines of code with the aim to divide the 1D array Jp (which lives on the master process only) over different processes (included the master one).
Each process have to receive a block of non-contigouos modified data (the values are changed in the inner loop) and I created a new format newF with the MPI_TYPE_INDEXED function to select the correct portion of data to send.
I used both MPI_RECV or MPI_IRECV to receive the data.
The problem is that this part of code works fine, with any numbers of tasks (from 1 to 8), until the number of element of Jp is small, when I increase such number (i.e. n = 5000), not all the processes receive the data and the splitted array JpS show the values I used to initialized it (i.e. -10000).
The lines commented show all the changes done in order to solve this problem, Does anyone have an idea?
program test_send
use mpi
implicit none
integer :: rank, nproc, mpi_stat
integer :: n, m, k, io, i, j
integer, allocatable :: Jp(:), JpS(:), JpAux(:)
integer :: count, n_distro,newF
integer, allocatable :: sendcounts(:), displ(:), &
blocklens(:), blockdisp(:), &
request(:)
integer :: ARRAY_OF_STATUS(MPI_STATUS_SIZE), error
data count /3/
call mpi_init(mpi_stat)
call mpi_comm_size(mpi_comm_world, nproc, mpi_stat)
call mpi_comm_rank(mpi_comm_world, rank, mpi_stat)
n = 400*count
allocate(sendcounts(nproc), displ(nproc), &
blocklens(count), blockdisp(count), request(nproc))
if (rank.eq.0) then
allocate(Jp(n+1),JpAux(n+1))
Jp = 0
do i = 1,n+1
Jp(i) = i
enddo
endif
call mpi_barrier(mpi_comm_world, mpi_stat)
m = n/count
n_distro = (m+1)/nproc
k = 0
do i = 1,nproc
if (i<nproc) then
sendcounts(i) = n_distro
else
sendcounts(i) = m - (nproc-1)*n_distro
endif
displ(i) = k
k = k + sendcounts(i)
enddo
call mpi_barrier(mpi_comm_world, mpi_stat)
allocate(JpS(count*sendcounts(rank+1)+1))
call mpi_barrier(mpi_comm_world, mpi_stat)
! call mpi_irecv(JpS, (sendcounts(rank+1))*count+1,mpi_int,0,0,mpi_comm_world, request(rank+1), mpi_stat)
! call mpi_recv(JpS, (sendcounts(rank+1))*count+1,mpi_int,0,0,mpi_comm_world, MPI_STATUS_IGNORE,mpi_stat)
!call mpi_waitall(1,request,ARRAY_OF_STATUS,error)
! call mpi_barrier(mpi_comm_world, mpi_stat)
if (rank.eq.0) then
do i = 0,nproc-1
JpAux = -100000
blocklens = spread(sendcounts(i+1),1,count)
blockdisp = spread(displ(i+1),1,count) + (/ (k*m, k=0,count-1) /)
blocklens(count) = blocklens(count)+1
do j = 1,count
if (j.eq.1) then
JpAux(blockdisp(j)+1:blockdisp(j)+blocklens(j)) = Jp(blockdisp(j)+1:blockdisp(j)+blocklens(j))&
-Jp(blockdisp(j)+1)
else
JpAux(blockdisp(j)+1:blockdisp(j)+blocklens(j)) = Jp( blockdisp(j) + 1 : blockdisp(j) + blocklens(j) )&
-Jp( blockdisp(j)+1 ) + JpAux( blockdisp(j-1) + blocklens(j-1))&
+(Jp( blockdisp(j-1)+blocklens(j-1)+1 )-Jp( blockdisp(j-1)+blocklens(j-1)))
endif
enddo
call mpi_type_indexed(count, blocklens, blockdisp, mpi_int, newF, mpi_stat)
call mpi_type_commit(newF, mpi_stat)
call mpi_isend(JpAux, 1, newF, i, i, mpi_comm_world, request(i+1), mpi_stat)
call mpi_type_free(newF, mpi_stat)
enddo
endif
! call mpi_wait(request(rank+1), ARRAY_OF_STATUS, mpi_stat)
call mpi_barrier(mpi_comm_world, mpi_stat)
!call mpi_waitall(1,request,ARRAY_OF_STATUS,error)
call mpi_recv(JpS, (sendcounts(rank+1))*count+1,mpi_int,0,MPI_ANY_TAG,mpi_comm_world, MPI_STATUS_IGNORE,mpi_stat)
! print*, request
print*, 'rank: ', rank, ', size: ', size(JpS), ', Jp: ', JpS
call mpi_barrier(mpi_comm_world, mpi_stat)
call mpi_finalize(mpi_stat)
end program test_send