c49460bae7
Summary: This enables primarily `stop.cpp` and `descriptor.cpp`. Requires a little bit of wrangling to get it to compile. Unlike the CUDA build, this build uses an in-tree libc++ configured for the GPU. This is configured without thread support, environment, or filesystem, and it is not POSIX at all. So, no mutexes, pthreads, or get/setenv. I tested stop, but i don't know if it's actually legal to exit from OpenMP offloading.
51 lines
1.1 KiB
Fortran
51 lines
1.1 KiB
Fortran
! REQUIRES: flang, amdgpu
|
|
|
|
! RUN: %libomptarget-compile-fortran-run-and-check-generic
|
|
program main
|
|
implicit none
|
|
integer :: result
|
|
|
|
! CHECK: 100
|
|
result = 0
|
|
!$omp target map(from: result)
|
|
block
|
|
integer, allocatable :: arr(:)
|
|
integer :: i
|
|
allocate(arr(4))
|
|
do i = 1, 4
|
|
arr(i) = i * 10
|
|
end do
|
|
result = arr(1) + arr(2) + arr(3) + arr(4)
|
|
deallocate(arr)
|
|
end block
|
|
!$omp end target
|
|
print *, result
|
|
|
|
! CHECK: 21
|
|
result = 0
|
|
!$omp target map(from: result)
|
|
block
|
|
integer, allocatable :: mat(:,:)
|
|
allocate(mat(2, 3))
|
|
mat(1,1) = 1; mat(2,1) = 2
|
|
mat(1,2) = 3; mat(2,2) = 4
|
|
mat(1,3) = 5; mat(2,3) = 6
|
|
result = mat(1,1) + mat(2,1) + mat(1,2) + mat(2,2) + mat(1,3) + mat(2,3)
|
|
deallocate(mat)
|
|
end block
|
|
!$omp end target
|
|
print *, result
|
|
|
|
! CHECK: 17
|
|
result = 0
|
|
!$omp target map(from: result)
|
|
block
|
|
integer, allocatable :: arr(:)
|
|
allocate(arr(8))
|
|
result = size(arr) + lbound(arr, 1) + ubound(arr, 1)
|
|
deallocate(arr)
|
|
end block
|
|
!$omp end target
|
|
print *, result
|
|
end program main
|