Array expression vs explicit loop in parallel and serial - fortran

Question

I tried to OpenMP parallelize the following minimal example and compare array expressions with explicit but am surprised, that I get no speedup.
Could you explain to me the problem?
program time_routines
use iso_fortran_env, only: real64, int64
USE OMP_LIB
integer, parameter :: wp = real64
integer(kind=int64) :: start, finish, rate
integer, parameter :: problem_size = 2 * 10**4
integer :: i
real(wp), allocatable :: X(:, :), Y(:, :)
allocate(X(problem_size, problem_size), Y(problem_size, problem_size))
call random_number(X)
call system_clock(count_rate=rate)
!$omp parallel
write(6, *) "Hello from thread: ", OMP_GET_THREAD_NUM()
!$OMP END PARALLEL
! fill cache lines
Y = 2._wp * X + 3._wp
call system_clock(start)
Y = 2._wp * X + 3._wp
call system_clock(finish)
write(6, *) 'serial array expression'
write(6, *) 'Elapsed Time in seconds:', real(finish - start, kind=wp) / real(rate, kind=wp)
call system_clock(start)
!$omp workshare
Y = 2._wp * X + 3._wp
!$omp end workshare
call system_clock(finish)
write(6, *) 'parallel array expression'
write(6, *) 'Elapsed Time in seconds:', real(finish - start, kind=wp) / real(rate, kind=wp)
call system_clock(start)
do j = 1, size(X, 2)
do i = 1, size(X, 1)
Y(i, j) = 2._wp * X(i, j) + 3._wp
end do
end do
call system_clock(finish)
write(6, *) 'Explicit serial loop'
write(6, *) 'Elapsed Time in seconds:', real(finish - start, kind=wp) / real(rate, kind=wp)
call system_clock(start)
!$omp do
do j = 1, size(X, 2)
do i = 1, size(X, 1)
Y(i, j) = 2._wp * X(i, j) + 3._wp
end do
end do
!$omp end do
call system_clock(finish)
write(6, *) 'Explicit parallel loop'
write(6, *) 'Elapsed Time in seconds:', real(finish - start, kind=wp) / real(rate, kind=wp)
end program
If I compile with
gfortran -fopenmp -O2 -Wall openmp_basics.f90 -o test.exe && ./test.exe my timings are the same for all possible ways of executing the calculation.