Observed issues › cce21-in-reduction-taskwait-link

Appears not to supportRefused at link time · Fortran · OpenMP · Open · regression

ftn: in_reduction on a target construct fails to link (undefined _cray$mt_taskwait in the device code)

Legality: independently confirmed. 2 independent readings find the program legal and the result required. How this is checked

Affected variants

Compiler	MI300A gfx942 unified memory (XNACK 1)	MI300A gfx942 copy (XNACK 0)	MI250X gfx90a copy (XNACK 0)	MI300X gfx942 copy (XNACK 0)	MI350X gfx950 copy (XNACK 0)	A100 sm_80 separate memory	A100 sm_80 managed memory
AFAR 24.3.0	correct	correct	correct	correct	correct	n/a	n/a
CCE 19.0.0	correct	correct	correct	unknown not run	unknown not run	n/a	n/a
CCE 20.0.2	correct	correct	correct	unknown not run	unknown not run	n/a	n/a
CCE 21.0.3	not supported	not supported	not supported	unknown not run	unknown not run	n/a	n/a
CCE 21.0.4	not supported	not supported	not supported	unknown not run	unknown not run	n/a	n/a
NVHPC 25.11	n/a	n/a	n/a	n/a	n/a	fails another way	fails another way
NVHPC 26.9	n/a	n/a	n/a	n/a	n/a	fails another way	fails another way
GCC 16.2.0	fails another way	fails another way	fails another way	fails another way	fails another way	fails another way	n/a

"behaves incorrectly" / "not supported": this issue, confirmed in up to four environments (another build directory, another node, a repeat). "fails another way": a different failure, not counted as this issue. "unknown": not run there, or the runs disagree.

Messages

Result	Compiler and GPU	Message
not supported	CCE 21.0.3, MI300A unified CCE 21.0.3, MI300A copy	show Warning: Headers and libraries from cray-libsci/26.03.0 will be ignored because depends on PE_MPICH which is unavailable.. lld: error: undefined symbol: _cray$mt_taskwait >>> referenced by <dir>/pe_3067603//cooltmpdir-svH0y5/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced by <dir>/pe_3067603//cooltmpdir-svH0y5/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced by <dir>/pe_3067603//cooltmpdir-svH0y5/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced 1 more times
not supported	CCE 21.0.3, MI250X	show Warning: Headers and libraries from cray-libsci/26.03.0 will be ignored because depends on PE_MPICH which is unavailable.. lld: error: undefined symbol: _cray$mt_taskwait >>> referenced by <dir>/pe_3622278//cooltmpdir-Sd7tFy/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced by <dir>/pe_3622278//cooltmpdir-Sd7tFy/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced by <dir>/pe_3622278//cooltmpdir-Sd7tFy/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced 1 more times
not supported	CCE 21.0.4, MI300A unified CCE 21.0.4, MI300A copy	show Warning: Headers and libraries from cray-libsci/26.03.0 will be ignored because depends on PE_MPICH which is unavailable.. lld: error: undefined symbol: _cray$mt_taskwait >>> referenced by <dir>/pe_2824163//cooltmpdir-J4EMz6/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced by <dir>/pe_2824163//cooltmpdir-J4EMz6/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced by <dir>/pe_2824163//cooltmpdir-J4EMz6/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced 1 more times
not supported	CCE 21.0.4, MI250X	show Warning: Headers and libraries from cray-libsci/26.03.0 will be ignored because depends on PE_MPICH which is unavailable.. lld: error: undefined symbol: _cray$mt_taskwait >>> referenced by <dir>/pe_3695362//cooltmpdir-ZYn3xp/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced by <dir>/pe_3695362//cooltmpdir-ZYn3xp/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced by <dir>/pe_3695362//cooltmpdir-ZYn3xp/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1) >>> referenced 1 more times
fails another way	NVHPC 25.11, A100 separate NVHPC 25.11, A100 managed NVHPC 26.9, A100 separate NVHPC 26.9, A100 managed	show NVFORTRAN-S-0034-Syntax error at or near identifier task_reduction (cce21-in-reduction-taskwait-link.f90: 22) NVFORTRAN-S-0034-Syntax error at or near identifier in_reduction (cce21-in-reduction-taskwait-link.f90: 23) 0 inform, 0 warnings, 2 severes, 0 fatal for gen_prog
fails another way	GCC 16.2.0, MI300A unified GCC 16.2.0, MI300A copy GCC 16.2.0, MI250X GCC 16.2.0, MI300X GCC 16.2.0, MI350X GCC 16.2.0, A100 separate	show libgomp: OMP_TARGET_OFFLOAD is set to MANDATORY, but device not found

Reproduce

CCE 21.0.3, MI300A unified

module load cce/21.0.3 craype-accel-amd-gfx942 rocm/7.2.0
module unload cray-mpich
ftn -fopenmp -O2 repro.f90 -o repro
HSA_XNACK=1 OMP_TARGET_OFFLOAD=mandatory ./repro

CCE 21.0.4, MI300A unified

module load cce/21.0.4 craype-accel-amd-gfx942 rocm/7.2.0
module unload cray-mpich
ftn -fopenmp -O2 repro.f90 -o repro
HSA_XNACK=1 OMP_TARGET_OFFLOAD=mandatory ./repro

AFAR 24.3.0, MI300A unified (runs correctly)

$AFAR/bin/amdflang -fopenmp --offload-arch=gfx942 -O2 repro.f90 -o repro
HSA_XNACK=1 OMP_TARGET_OFFLOAD=mandatory ./repro

Program (53 lines)

! Generated by compliance.gen (OpenMP 5.2, free-form Fortran). Do not edit.
! id: omp52-clause-in_reduction-and
! primary atom: omp.clause.in_reduction.and
! legal: yes
! observable: yes
! ran_on_host: inside the target region of interest
! evidence: construct
! needs: none
! std: f2008
program gen_prog
  use omp_lib, only: omp_is_initial_device
  implicit none
  integer, parameter :: n = 64
  integer, allocatable, target :: a(:), b(:)
  integer :: ref(n)
  integer :: i, host_flag
  logical :: ok
  logical :: ls, lsref
  allocate(a(n), b(n))
  do i = 1, n
    a(i) = i
    b(i) = 0
    ref(i) = 2*i + 1
  end do
  host_flag = 1
  ls = .true.
  lsref = .true.
  do i = 1, n
    lsref = lsref .and. (a(i) /= 17)
  end do
  !$omp taskgroup task_reduction(.and.: ls)
    !$omp target map(to: a) map(from: b) map(tofrom: host_flag) in_reduction(.and.: ls)
      host_flag = merge(1, 0, omp_is_initial_device())
      do i = 1, n
        b(i) = 2*a(i) + 1
        ls = ls .and. (a(i) /= 17)
      end do
    !$omp end target
  !$omp end taskgroup
  ok = all(b == ref) .and. (ls .eqv. lsref)
  print '(a,i0,a,a)', 'RESULT ran_on_host=', host_flag, ' check=', merge('pass', 'fail', ok)
  deallocate(a, b)
end program gen_prog
! spec (requirement): An in_reduction clause on a target construct makes the target task a participant in the task
!   reduction defined by an enclosing taskgroup with a matching task_reduction clause.
! "For the target construct, the target task becomes the participating task." (5.2 §5.5.10 p.138 l.21-22)
! spec (requirement): The private copy of an in_reduction list item on a target construct is implicitly mapped into
!   the device data environment and combined at the end of the task region.
! "This private copy will be implicitly mapped into the device data environment of the target device, if the target
!   device is not the parent device." (5.2 §5.5.10 p.138 l.23-24)
! spec (this program): The target construct lists in_reduction(.and.: ls) inside a taskgroup with
!   task_reduction(.and.: ls), which matches in operator and list item; the compiler fails to link the device code
!   (undefined _cray$mt_taskwait) instead of performing the reduction.

Source in the repository

Something to add? Dispute its legality · Report it fixed in a newer release · Report a mistake on this page

Expected and observed

Observed where it behaves incorrectly

Warning:
 Headers and libraries from cray-libsci/26.03.0 will be ignored because depends on PE_MPICH which is unavailable..
lld: error: undefined symbol: _cray$mt_taskwait
>>> referenced by <dir>/pe_3067603//cooltmpdir-svH0y5/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1)
>>> referenced by <dir>/pe_3067603//cooltmpdir-svH0y5/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1)
>>> referenced by <dir>/pe_3067603//cooltmpdir-svH0y5/repro-cce-openmppost_lld.amdgpu.lto.o:(gen_prog_$ck_L23_1)
>>> referenced 1 more times

What the specification requires

The specification text this program relies on and the compiler is expected to support.

OpenMP 5.2 §5.5.10 p.138 l.21-22

For the target construct, the target task becomes the participating task.

OpenMP 5.2 §5.5.10 p.138 l.23-24

This private copy will be implicitly mapped into the device data environment of the target device, if the target device is not the parent device.

Evidence

Same compiler, one change

in_reduction(.and.: ls) on the target replaced by map(tofrom: ls), without the enclosing taskgroup: AFAR 24.3.0: correct; CCE 19.0.0: correct; CCE 20.0.2: correct; CCE 21.0.3: correct; CCE 21.0.4: correct

Control program

program gen_prog
  use omp_lib, only: omp_is_initial_device
  implicit none
  integer, parameter :: n = 64
  integer, allocatable, target :: a(:), b(:)
  integer :: ref(n)
  integer :: i, host_flag
  logical :: ok
  logical :: ls, lsref
  allocate(a(n), b(n))
  do i = 1, n
    a(i) = i
    b(i) = 0
    ref(i) = 2*i + 1
  end do
  host_flag = 1
  ls = .true.
  lsref = .true.
  do i = 1, n
    lsref = lsref .and. (a(i) /= 17)
  end do
    !$omp target map(to: a) map(from: b) map(tofrom: host_flag) map(tofrom: ls)
      host_flag = merge(1, 0, omp_is_initial_device())
      do i = 1, n
        b(i) = 2*a(i) + 1
        ls = ls .and. (a(i) /= 17)
      end do
    !$omp end target
  ok = all(b == ref) .and. (ls .eqv. lsref)
  print '(a,i0,a,a)', 'RESULT ran_on_host=', host_flag, ' check=', merge('pass', 'fail', ok)
  deallocate(a, b)
end program gen_prog

Other compilers run it correctly

AFAR 24.3.0