PROGRAM main
      IMPLICIT NONE

      INTEGER :: n,i,j
      REAL*8, allocatable :: A(:,:)
      INTEGER :: count, count_rate, count_max, T1, T2
      REAL*8  Rand
      CHARACTER*80 :: argc1, argc2
      INTEGER :: NUM_THREADS
      
      call getarg(1,argc1)
      read(argc1,*) n
      call getarg(2,argc2)
      read(argc2,*) NUM_THREADS 
      print *, "Array size: ", n
      print *, "Number of Processors: ", NUM_THREADS

      allocate(A(n,n))
      call OMP_SET_NUM_THREADS(NUM_THREADS)
                      
      open(40, file='input')
      read(40,*) A

!     Get time
      call system_clock(count, count_rate, count_max)
      T1 = count      


!     Real work here
      call lu_factor(n,A)


!     Check time
      call system_clock(count, count_rate, count_max)
      T2 = count      
      print *, "Time = ",T2-T1


      open(60, file='par.output')
      write(60,*) A

      END PROGRAM main
      SUBROUTINE lu_factor(n,A)
      IMPLICIT NONE

      INTEGER :: n
      REAL*8 :: A(n,n)
      
      INTEGER i,j,k
      
      DO k = 1, n-1
         DO i = k+1, n          
            A(i, k) = A(i, k) / A(k, k)
         END DO
         
         DO i = k+1, n
            DO j = k+1, n
               A(i, j) = A(i, j) - A(i, k)*A(k, j)
            END DO
         END DO
      END DO

      END SUBROUTINE lu_factor