compressed sparse column to row conversion - fortran

Question

i was trying to convert a compressed sparse column format into compressed sparse row format using fortran. here's what i have so far:
program test
implicit none
real*4,dimension(19)::csc_data=(/10.,3.,3.,9.,7.,8.,4.,8.,8.,7.,7.,9.,-2.,5.,9.,2.,3.,13.,-1./)
integer*4,dimension(19)::csc_index=(/1,2,4,2,3,5,6,3,4,3,4,5,1,4,5,6,2,5,6/)
integer*4,dimension(7)::csc_pointer=(/1,4,8,10,13,17,20/)
integer*4,dimension(7)::csr_pointer
integer*4,dimension(19)::csr_index
real*4,dimension(19)::csr_data
integer*4::global_counter,counter,i
integer*4::num_nonzero,num_cols,num_rows
integer*4::s1,s2,c,r
num_nonzero=19
num_rows=6
num_cols=6
csr_pointer(1)=1
global_counter=1
do i=1,num_rows
counter=0
do c=1,num_cols
s1=csc_pointer(c)
s2=csc_pointer(c+1)-1
do r=s1,s2
if(csc_index(r).eq.i) then
counter=counter+1
csr_data(global_counter)=csc_data(r)
csr_index(global_counter)=c
global_counter=global_counter+1
end if
end do
end do
csr_pointer(i+1)=csr_pointer(i)+counter
end do
end program test
Could anyone show more efficient approach? I would really appreciate if you could also show it with OpenMP parallelization. Thanks.

Answer 1

While I cannot comment on the correctness of the algorithm, I can make a few points on the code itself
csc_data is defined as real*4 while csr_data is defined as integer*4
csc_pointer is defined as integer*4 while csr_pointer is defined as real*4
If you output the data, aside form the integer/real difference, the data is the same, something I would not expect for converting the matrix
csr_index has an index of global_i which should be global_counter