PVF12.8 vs nvs4200M

Accelerated Computing HPC Compilers Legacy PGI Compilers
#1

hi when I compile OPENACC with GPU card NVS4200m in PVF12.8 it
returned an error call to cuMemcpyDtoH returned error 999 Unknow


program laplace
!#ifdef _OPENACC
  use openacc
!#endif
  implicit none
  integer, parameter :: fp_kind=kind(1.0)
  integer, parameter :: n=4096, m=4096, iter_max=1000
  integer :: i, j, iter
  real(fp_kind), dimension (:,:), allocatable :: A, Anew
  real(fp_kind), dimension (:),   allocatable :: y0
  real(fp_kind) :: pi=2.0_fp_kind*asin(1.0_fp_kind), tol=1.0e-5_fp_kind, error=1.0_fp_kind
  real(fp_kind) :: start_time, stop_time
  allocate ( A(0:n-1,0:m-1), Anew(0:n-1,0:m-1) )
  allocate ( y0(0:m-1) )
  A = 0.0_fp_kind
  ! Set B.C.
  y0 = sin(pi* (/ (j,j=0,m-1) /) /(m-1))
  A(0,:)   = 0.0_fp_kind
  A(n-1,:) = 0.0_fp_kind
  A(:,0)   = y0
  A(:,m-1) = y0*exp(-pi)
  
!#if _OPENACC
  call acc_init(acc_device_nvidia)
!#endif
   
  write(*,'(a,i5,a,i5,a)') 'Jacobi relaxation Calculation:', n, ' x', m, ' mesh'

  call cpu_time(start_time)
  iter=0
!$omp parallel do shared(Anew)
  do i=1,m-1
    Anew(0,i)   = 0.0_fp_kind
    Anew(n-1,i) = 0.0_fp_kind
  end do
!$omp end parallel do
!$omp parallel do shared(Anew)
  do i=1,n-1
    Anew(i,0)   = y0(i)
    Anew(i,m-1) = y0(i)*exp(-pi)
  end do
!$omp end parallel do
!$acc data copy(A), create(Anew)
  do while ( error .gt. tol .and. iter .lt. iter_max )
    error=0.0_fp_kind
!$omp parallel do shared(m, n, Anew, A) reduction( max:error )
!$acc kernels loop gang(32), vector(16)
    do j=1,m-2
!$acc loop gang(16), vector(32)
      do i=1,n-2
        Anew(i,j) = 0.25_fp_kind * ( A(i+1,j  ) + A(i-1,j  ) + &
                                     A(i  ,j-1) + A(i  ,j+1) )
        error = max( error, abs(Anew(i,j)-A(i,j)) )
      end do
!$acc end loop
    end do
!$acc end kernels
!$omp end parallel do
    if(mod(iter,100).eq.0 ) write(*,'(i5,f10.6)'), iter, error
    iter = iter +1
!$omp parallel do shared(m, n, Anew, A)
!$acc kernels loop
    do j=1,m-2
!$acc loop gang(16), vector(32)
      do i=1,n-2
        A(i,j) = Anew(i,j)
      end do
!$acc end loop
    end do
!$acc end kernels
!$omp end parallel do
  end do
!$acc end data
  call cpu_time(stop_time)
  write(*,'(a,f10.3,a)')  ' completed in ', stop_time-start_time, ' seconds'
  deallocate (A,Anew,y0)
end program laplace
#2

hi,Mat, this user told me that he update to PVF 12.9,but this question still exists

#3

hi,Mat, this user told me that he update to PVF 12.9,but this question still exists

#4

Hi Sisiy,

I do apologize, I thought I had responded to this user. This is a compiler error having to do with the default schedule for the reduction being mismatched with the user’s schedule. The work around is to use a 16x16 instead of a 16x32 vector or remove the explicit schedule.

!$acc kernels loop gang(32), vector(16)
    do j=1,m-2
!$acc loop gang(16), vector(16)
      do i=1,n-2
        Anew(i,j) = 0.25_fp_kind * ( A(i+1,j  ) + A(i-1,j  ) + &
                                     A(i  ,j-1) + A(i  ,j+1) )
        error = max( error, abs(Anew(i,j)-A(i,j)) )
      end do
    end do
!$acc end kernels




!$acc kernels loop 
    do j=1,m-2
      do i=1,n-2
        Anew(i,j) = 0.25_fp_kind * ( A(i+1,j  ) + A(i-1,j  ) + &
                                     A(i  ,j-1) + A(i  ,j+1) )
        error = max( error, abs(Anew(i,j)-A(i,j)) )
      end do
    end do
!$acc end kernels

Hope this helps,
Mat