OffTarget

Observed issues › fuzz/gcc-target-simd-map-nontemporal-rejected

Appears not to supportRefused at compile time · Fortran · OpenMP · Open

The same array in map and nontemporal clauses of a Fortran target simd construct is rejected

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

On target simd, map applies to the target leaf and nontemporal to the simd leaf. GCC 16.2's Fortran front end rejects an array named in both ("present on multiple clauses") for both offload targets; the C form compiles and runs. Without nontemporal the program compiles and runs correctly.

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 24.3.0correctcorrectunknown
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 16.2.0not supportednot supportednot supportednot supportednot supportednot supportedn/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
not supportedGCC 16.2.0, MI300A unified
GCC 16.2.0, MI300A copy
GCC 16.2.0, MI300X
GCC 16.2.0, MI350X
show
site-fuzz--gcc-target-simd-map-nontemporal-rejected.f90:40:48:
   40 |   !$omp target simd map(tofrom: o) nontemporal(o)
      |                                                1
Error: Symbol ‘o’ present on multiple clauses at (1)
not supportedGCC 16.2.0, MI250X
GCC 16.2.0, A100 separate
show
site-fuzz--gcc-target-simd-map-nontemporal-rejected.f90:32:48:
   32 |   !$omp target simd map(tofrom: o) nontemporal(o)
      |                                                1
Error: Symbol ‘o’ present on multiple clauses at (1)
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 identifier nontemporal (site-fuzz--gcc-target-simd-map-nontemporal-rejected.f90: 32)
  0 inform,   0 warnings,   1 severes, 0 fatal for target_simd_map_nontemporal

Reproduce

GCC 16.2.0, MI300A unified

export LD_LIBRARY_PATH=$GCC_16_2/lib64:/opt/rocm-7.2.1/lib
$GCC_16_2/bin/gfortran -fopenmp -foffload=amdgcn-amdhsa -foffload-options=amdgcn-amdhsa=-march=gfx942 -O2 repro.f90 -o repro
HSA_XNACK=1 OMP_TARGET_OFFLOAD=mandatory ./repro

GCC 16.2.0, A100 separate

export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=$GCC_16_2/lib64
$GCC_16_2/bin/gfortran -fopenmp -foffload=nvptx-none -foffload-options=nvptx-none=-march=sm_80 -foffload-options=nvptx-none=-latomic -O2 repro.f90 -o repro
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 (47 lines)
! # Finding (GCC 16.2, fuzz sample): the same array in map and nontemporal of target simd is rejected (Fortran)
!
! - toolchains: gcc_g90 (GCC 16.2.0, AMD Instinct MI250X gfx90a, Tioga) and gcc_sm80 (GCC 16.2.0, NVIDIA A100 sm_80, wingtip);
!   separate memory only; -O2
! - build: gfortran -fopenmp -foffload=amdgcn-amdhsa -foffload-options=amdgcn-amdhsa=-march=gfx90a -O2 (gcc_g90);
!   -foffload=nvptx-none -foffload-options=nvptx-none=-march=sm_80 -foffload-options=nvptx-none=-latomic -O2 (gcc_sm80)
! - failure: rejected: present on multiple clauses
! - observed: `Error: Symbol 'o' present on multiple clauses at (1)` at the nontemporal clause, on both targets; the C form
!   (results/parity/probes-r2/nontemporal_c.c) compiles and runs correctly on both (mandatory pass 5 of 5; unset pass 5 of 5)
! - expected: compiles; prints `bad= 0` and `RESULT ran_on_host=0 check=pass`
! - rule: on a combined construct each clause applies to the leaf constructs that allow it (5.2 §17.2 p.341): map to target,
!   nontemporal to simd; nontemporal is no data-sharing attribute clause, so the map restriction "A list item cannot appear in both
!   a map clause and a data-sharing attribute clause on the same construct" does not apply.
! - same-compiler control (compiles; correct, gcc_sm80: mandatory pass 5 of 5; unset pass 5 of 5; gcc_g90: mandatory pass 5 of 5;
!   unset pass 5 of 5 (results/parity/<machine>/<tc>/probes.jsonl)): the same construct without nontemporal
!   (controls/gcc-target-simd-map-nontemporal-control.f90).
! - GCC's documentation: no row of libgomp.texi's tables names it (undocumented; nontemporal is OpenMP 5.0's).
! - other families: AFAR and CCE 19/20/21/21.0.4 run this file correctly in both memory modes; NVHPC 25.11/26.9 fail it another way
!   (site matrix). The site's finding matrix runs this file on every compiler.
! - fuzz programs (44 cells (22 gcc_g90, 22 gcc_sm80); gate: confirmed 44; confirmer: needs_human (R12b-compile-undecided) 35,
!   compiler_bug (R11-compile-rejection) 9): b0:227, b0:294, c1:1350, c1:1688, c1:1721, c1:18, c1:209, c1:227, c1:427, c1:469,
!   s1:1336, s1:1608, s1:2321, s1:327, s1:3745, s1:4530, s1:9979, x0-alloc:191, x0-declare:70, x0-multitu:111, x0-multitu:227,
!   x0-types:220
! - legality: two witnesses that are neither GCC nor a compiler this finding is about, offload.engine.rules and AFAR flang's
!   front end,
!   accept this file and its control (offload.engine.prerun.check; results/parity/tioga/results/parity/legality.jsonl) and the fuzz
!   programs (the campaign's legality gate; its base-language check, gfortran -std=f2008 without OpenMP, is a syntax check of the
!   base language only and no witness). GCC is a compiler under test only and never a legality witness.
program target_simd_map_nontemporal
  use omp_lib, only: omp_is_initial_device
  implicit none
  integer, parameter :: n = 64
  integer :: o(n), i, bad, host_flag
  o = 0
  host_flag = 1
  !$omp target map(tofrom: host_flag)
    host_flag = merge(1, 0, omp_is_initial_device())
  !$omp end target
  ! map applies to the target leaf, nontemporal to the simd leaf
  !$omp target simd map(tofrom: o) nontemporal(o)
  do i = 1, n
    o(i) = 2 * i
  end do
  bad = count(o /= [(2 * i, i = 1, n)])
  print '(a,i0)', 'bad= ', bad
  print '(a,i0,a,a)', 'RESULT ran_on_host=', host_flag, ' check=', merge('pass', 'fail', bad == 0)
end program target_simd_map_nontemporal

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

Expected

check=pass

Observed where it behaves incorrectly

site-fuzz--gcc-target-simd-map-nontemporal-rejected.f90:40:48:
   40 |   !$omp target simd map(tofrom: o) nontemporal(o)
      |                                                1
Error: Symbol ‘o’ present on multiple clauses at (1)

What the specification requires

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

OpenMP 5.2 §17.2 p.341

Evidence

Same compiler, one change

  • the same construct without nontemporal: not run yet
    Control program
    ! Control of fuzz/gcc-target-simd-map-nontemporal-rejected.f90: the same construct without nontemporal (compiles; correct
    ! on gcc_g90 and gcc_sm80).
    program target_simd_map_nontemporal
      use omp_lib, only: omp_is_initial_device
      implicit none
      integer, parameter :: n = 64
      integer :: o(n), i, bad, host_flag
      o = 0
      host_flag = 1
      !$omp target map(tofrom: host_flag)
        host_flag = merge(1, 0, omp_is_initial_device())
      !$omp end target
      ! map applies to the target leaf, nontemporal to the simd leaf
      !$omp target simd map(tofrom: o)
      do i = 1, n
        o(i) = 2 * i
      end do
      bad = count(o /= [(2 * i, i = 1, n)])
      print '(a,i0)', 'bad= ', bad
      print '(a,i0,a,a)', 'RESULT ran_on_host=', host_flag, ' check=', merge('pass', 'fail', bad == 0)
    end program target_simd_map_nontemporal
    

Other compilers run it correctly

AFAR 24.3.0, CCE 19.0.0, CCE 20.0.2, CCE 21.0.3, CCE 21.0.4