! RUN: %flang_fc1 -fdebug-dump-symbols %s | FileCheck %s

! Test CUDA Fortran intrinsic can pass semantic

attributes(global) subroutine devsub()
  implicit none
  integer :: ret

  ! 3.6.4. Synchronization Functions
  call syncthreads()
  call syncwarp(1)
  call threadfence()
  call threadfence_block()
  call threadfence_system()
  ret = syncthreads_and(1)
  ret = syncthreads_count(1)
  ret = syncthreads_or(1)
end

! CHECK-LABEL: Subprogram scope: devsub
! CHECK: syncthreads (Subroutine): Use from syncthreads in cudadevice
! CHECK: syncthreads_and (Function): Use from syncthreads_and in cudadevice
! CHECK: syncthreads_count (Function): Use from syncthreads_count in cudadevice
! CHECK: syncthreads_or (Function): Use from syncthreads_or in cudadevice
! CHECK: syncwarp (Subroutine): Use from syncwarp in cudadevice
! CHECK: threadfence (Subroutine): Use from threadfence in cudadevice
! CHECK: threadfence_block (Subroutine): Use from threadfence_block in cudadevice
! CHECK: threadfence_system (Subroutine): Use from threadfence_system in cudadevice

subroutine host()
  real(4) :: af
  real(8) :: ad
  integer(4) :: ai
  integer(8) :: al
  call syncthreads()
  ai = atomicadd(ai, 1_4)
  al = atomicadd(al, 1_8)
  af = atomicadd(af, 1.0_4)
  ad = atomicadd(ad, 1.0_8)
end subroutine

! CHECK-LABEL: Subprogram scope: host
! CHECK: atomicadd, EXTERNAL: HostAssoc => atomicadd, EXTERNAL (Function, Implicit): ProcEntity REAL(4){{$}}
! CHECK: syncthreads, EXTERNAL: HostAssoc => syncthreads, EXTERNAL (Subroutine): ProcEntity{{$}}