[Flang] Assignments for integer arrays with FORALL/WHERE are slower than GFortran

[yus3710-fj]

The execution of the following codes is slower than GFortran.

• test.f90

subroutine dianfn008(lhs,rhs,n)
  integer :: rhs(:)
  integer :: lhs(:)
  integer i,n
  forall(i=1:n)
    lhs(i) = rhs(i)
  end forall
end subroutine

• main.f90

implicit none

integer, parameter :: count = 10000
integer :: i,j,k,a1(100),a2(100)
real :: start_t, end_t
interface
  subroutine dianfn008(lhs,rhs)
    integer :: rhs(:)
    integer :: lhs(:)
  end subroutine
end interface

call cpu_time(start_t)
do i=1,count
  do j=1,count
    call dianfn008(a1,a2)
  end do
end do
call cpu_time(end_t)

print *, end_t - start_t
end

• commands

$ flang -Ofast -mcpu=native test.f90 main.f90 && ./a.out # for A64FX/Grace
$ flang -Ofast -march=native test.f90 main.f90 && ./a.out # for Xeon

	Flang [s]	GFortran [s]
A64FX	3.5503642	2.57492518
Grace	0.7966712	0.821618974
Xeon	0.7875499	0.659658015

[yus3710-fj]

Flang vectorizes and unrolls the loop, while GFortran replaces it with memcpy. This means the performance gap depends on the performance of memcpy.
I believe this is not a matter of the compilers themselves.