Segmentation Fault (Core Dumped) when Using Intel Sequential MKL DGEMM Inside OpenMP Parallel Loop

[AkhilAkkapelli]

I'm trying to perform matrix multiplication using Intel Fortran Compiler (ifort) with OpenMP and MKL's DGEMM routine. The code divides matrices into blocks and uses OpenMP for parallel execution. However, I encounter a "Segmentation fault (core dumped)" error when running the program.

compilation command:  ifort -qopenmp -qmkl=sequential matmul.f90

Here’s a simplified version of my code:

integer, parameter :: n = 9 ! Number of rows and columns of matrix
integer, parameter :: nt = 3 ! Sq root of number of threads

real(8), dimension(n,n) :: A, B, C

integer :: nb ! Block size
integer :: bi(nt,2) ! Start and End indices of block matrices
integer :: i, j, k
integer :: starttime, endtime, countrate
call omp_set_num_threads(nt*nt)

print*, "Number of OMP Threads", nt*nt

call random_number(A)
call random_number(B)

if (mod(n,nt) == 0) then
    nb = n/nt
else
    nb = n/nt + 1
end if

!print*, "block size", nb

do i = 1, nt
    bi(i,1) = 1 + nb*(i-1)
    bi(i,2) = min(nb*i,n)
end do

call system_clock(starttime, countrate)
!$OMP PARALLEL DO COLLAPSE(2)
do i  = 1, nt
    do j = 1, nt
        call DGEMM('N', 'N', n, n, n, 1.0d0, A(bi(i,1):bi(i,2), :), n, &
                   B(:, bi(j,1):bi(j,2)), n, 0.0d0, &
                   C(bi(i,1):bi(i,2), bi(j,1):bi(j,2)), n)
    end do
end do
!$OMP END PARALLEL DO

call system_clock(endtime)