01-mm1_acc.F90

program mm
  use omp_lib
  integer(8), parameter :: N = 4096
  integer(8) :: i,j,k,tmp6,tmp2
  real(8), dimension(N,N) :: A, B, C
  real(8) :: tmp, chk, t0, t1, t2, t3

  t0 = omp_get_wtime()
  !$acc data create(A,B,C)
  !$acc kernels
  do i = 0, N-1
    do j = 0, N-1
      A(i+1,j+1) = i + j + 1d0
      B(i+1,j+1) = i + j + 1d0
    end do
  end do
  C(:,:) = 0d0
  !$acc end kernels

  t2 = omp_get_wtime()
  !$acc kernels
  do k = 1, N
    do j = 1, N
      do i = 1, N
        C(i,k) = C(i,k) + A(i,j) * B(j,k)
      end do
    end do
  end do
  !$acc end kernels
  !$acc wait
  t3 = omp_get_wtime()

  tmp6 = (N * ((2 * N) - 1) * (N - 1)) / 6
  tmp2 = (N * (N - 1)) / 2
  chk = 0
  !$acc parallel loop reduction(+:chk)
  do j = 0, N-1
    do i = 0, N-1
      tmp = tmp6                              &
          + (((i + 1d0) + (j + 1d0)) * tmp2 ) &
          + (((i + 1d0) * (j + 1d0)) * N)
      chk = chk + (C(i+1,j+1)/tmp)
    end do
  end do
  !$acc end parallel loop
  !$acc end data
  t1 = omp_get_wtime()
  write(*,'(I8," ",I8," ",F10.6," ",F10.6)') int(chk), (N*N), (t3-t2), (t1-t0)
end program

02-compile.txt

pgf90 -Minfo=all -fast -acc -ta=nvidia -I /opt/nvidia/cudatoolkit/default/include/  -o mmF90_acc mm1_acc.F90
mm:
      9, Generating create(c(:,:))
         Generating create(b(:,:))
         Generating create(a(:,:))
     10, Generating present_or_create(c(:,:))
         Generating present_or_create(b(:,:))
         Generating present_or_create(a(:,:))
         Generating compute capability 1.3 binary
         Generating compute capability 2.0 binary
     11, Loop is parallelizable
     12, Loop is parallelizable
         Accelerator kernel generated
         11, !$acc loop gang, vector(128) ! blockidx%x threadidx%x
         12, !$acc loop gang ! blockidx%y
             CC 1.3 : 13 registers; 40 shared, 8 constant, 0 local memory bytes
             CC 2.0 : 19 registers; 0 shared, 56 constant, 0 local memory bytes
     23, Generating present_or_create(c(:,:))
         Generating present_or_create(a(:,:))
         Generating present_or_create(b(:,:))
         Generating compute capability 1.3 binary
         Generating compute capability 2.0 binary
     24, Loop is parallelizable
     25, Loop carried dependence of 'c' prevents parallelization
         Loop carried backward dependence of 'c' prevents vectorization
     26, Loop is parallelizable
         Accelerator kernel generated
         24, !$acc loop gang, vector(4) ! blockidx%y threadidx%y
         26, !$acc loop gang, vector(64) ! blockidx%x threadidx%x
             CC 1.3 : 19 registers; 40 shared, 12 constant, 0 local memory bytes
             CC 2.0 : 26 registers; 0 shared, 56 constant, 0 local memory bytes
     38, Accelerator kernel generated
         38, CC 1.3 : 30 registers; 32 shared, 32 constant, 0 local memory bytes
             CC 2.0 : 30 registers; 0 shared, 80 constant, 0 local memory bytes
         39, !$acc loop gang ! blockidx%x
         40, !$acc loop vector(256) ! threadidx%x
     38, Generating present_or_create(c(:,:))
         Generating compute capability 1.3 binary
         Generating compute capability 2.0 binary
     40, Loop is parallelizable

03-run.txt

$ aprun ./mmF90_acc
16777216 16777216   2.280605   3.387837