OffTarget

Observed issues › fuzz/s1-9443-afar

Appears to behave incorrectlyCrashes or aborts at run time · Fortran · OpenMP · Open

AFAR offload runtime at HSA_XNACK=1: map(close, ...) breaks presence lookup and device access (abort, memory fault)

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

Affected variants

CompilerMI300A 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 24.3.0behaves incorrectlycorrectunknown
not run
correctcorrectn/an/a
CCE 19.0.0correctcorrectunknown
not run
unknown
not run
unknown
not run
n/an/a
CCE 20.0.2correctcorrectunknown
not run
unknown
not run
unknown
not run
n/an/a
CCE 21.0.3correctcorrectunknown
not run
unknown
not run
unknown
not run
n/an/a
CCE 21.0.4correctcorrectunknown
not run
unknown
not run
unknown
not run
n/an/a
NVHPC 25.11n/an/an/an/an/afails another wayfails another way
NVHPC 26.9n/an/an/an/an/afails another wayfails another way
GCC 16.2.0correctcorrectcorrectcorrectcorrectcorrectn/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

ResultCompiler and GPUMessage
behaves incorrectlyAFAR 24.3.0, MI300A unified
show
omptarget message: device mapping required by 'present' map type modifier does not exist for host address 0x000055555561f980 (100 bytes)
omptarget error: Call to getTargetPointer returned null pointer ('present' map type modifier).
omptarget error: Call to targetDataBegin failed, abort target.
omptarget error: Failed to process data before launching the kernel.
omptarget error: Consult https://openmp.llvm.org/design/Runtimes.html for debugging options.
omptarget error: Source location information not present. Compile with -g or -gline-tables-only.
omptarget fatal error 1: failure of target construct while offloading is mandatory
fails another wayNVHPC 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 : (s1-9443-afar.f90: 10)
  0 inform,   0 warnings,   1 severes, 0 fatal for p

Reproduce

AFAR 24.3.0, MI300A unified

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

CCE 19.0.0, MI300A unified (runs correctly)

module load cce/19.0.0 craype-accel-amd-gfx942 rocm/6.4.3
module unload cray-mpich
ftn -fopenmp -O2 repro.f90 -o repro
HSA_XNACK=1 OMP_TARGET_OFFLOAD=mandatory ./repro
Program (55 lines)
! Minimized by compliance.fuzz_min from fuzz seed 9443 (fuzz-3).
! toolchain: afar (-O2, HSA_XNACK=1, OMP_TARGET_OFFLOAD=mandatory)
! failure: run_error (crash)
! message: 'present' map type modifier
! expected: runs to completion (the host run does)
! observed: device: fail crash: SIGABRT; host reference (cce21) checksum=0
! Legal OpenMP 5.2: offload.engine.rules accepts it (offload.engine.prerun.check, rerun: with the
!   race rule and the restriction checks); NVHPC is no witness: it rejects it
!   (NVFORTRAN-S-0034-Syntax error at or near :); the check's other
!   witness, the AFAR flang front end, is the compiler under test and is not counted; no gfortran
!   or gcc is a legality witness.
! also fails on: cce19
! Finding: AFAR aborts at HSA_XNACK=1 (SIGABRT, "Call to
! getTargetPointer returned null pointer ('present' map type modifier)")
! at an empty target construct with map(close, present, to: xk1), where
! the enclosing target data construct maps xk1 (an integer(1) array); at
! HSA_XNACK=0 AFAR runs it. CCE 19 also ends with exit status 1 in both
! modes; CCE 20 and 21 run it.
! Legal OpenMP 5.2: the target data construct makes xk1 present, so the
! present modifier is satisfied (5.2 §5.8.3); close is a hint.
! Observed (tuolumne, MI300A, -O2, 2026-10-23,
! results/t28_probe/rt/9443-afar.f90.*.json): AFAR SIGABRT at HSA_XNACK=1,
! "RESULT checksum=0" at HSA_XNACK=0; CCE 19 exit status 1 (both modes);
! CCE 20, 21 "RESULT checksum=0". Reduced from s1 seed 9443 by
! tools/campaigns/fuzz.py minimize-one (146 -> 82 lines). The present-modifier aborts
! are AFAR's largest unified-only group in the copy-mode comparison.
! Related: ../afar-xnack1-present-not-enforced.f90 (AFAR at HSA_XNACK=1
! does not enforce present for an absent item: the opposite symptom of the
! same unified-memory present handling).
! Shorter form (fix; written by hand from
! the 82-line reduction, which git history keeps): an integer(1) array
! alone in target data map(to: xk1) and an empty target with
! map(close, present, to: xk1) (data/evidence/dataenv_probe/short/c9443_a.f90).
! Observed (tuolumne, MI300A, -O2, 2026-10-23, results/t28_probe/short/
! c9443_a.f90.*.json): AFAR SIGABRT at HSA_XNACK=1, "RESULT check=pass" at
! HSA_XNACK=0; CCE 19, 20, 21 pass in both modes (CCE 19's exit status 1
! needs the longer form). Without close (c9443_b) AFAR passes: the close
! modifier is the trigger. Expected: "RESULT check=pass".
! Known defect: this is the AFAR close + present abort at
! HSA_XNACK=1 of fuzz/158-afar.f90 and fuzz/1131-afar.f90 (there after
! target enter data, here inside target data); not a new finding.
program p
  implicit none
  integer, parameter :: n = 100
  integer(1) :: xk1(n)
  integer :: i
  do i = 1, n
    xk1(i) = int(mod(i * 17 + 16, 127), 1)
  end do
  !$omp target data map(to: xk1)
    !$omp target map(close, present, to: xk1)
    !$omp end target
  !$omp end target data
  print '(a)', 'RESULT check=pass'
end program p

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

omptarget message: device mapping required by 'present' map type modifier does not exist for host address 0x000055555561f980 (100 bytes)
omptarget error: Call to getTargetPointer returned null pointer ('present' map type modifier).
omptarget error: Call to targetDataBegin failed, abort target.
omptarget error: Failed to process data before launching the kernel.
omptarget error: Consult https://openmp.llvm.org/design/Runtimes.html for debugging options.
omptarget error: Source location information not present. Compile with -g or -gline-tables-only.
omptarget fatal error 1: failure of target construct while offloading is mandatory

What the specification requires

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

OpenMP 5.2 §5.8.3

Evidence

Same compiler, one change

  • map(present, to: xk1) without the close modifier: 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 p
      implicit none
      integer, parameter :: n = 100
      integer(1) :: xk1(n)
      integer :: i
      do i = 1, n
        xk1(i) = int(mod(i * 17 + 16, 127), 1)
      end do
      !$omp target data map(to: xk1)
        !$omp target map(present, to: xk1)
        !$omp end target
      !$omp end target data
      print '(a)', 'RESULT check=pass'
    end program p
    

Other compilers run it correctly

CCE 19.0.0, CCE 20.0.2, CCE 21.0.3, CCE 21.0.4, GCC 16.2.0

Reproducers of this issue