OffTarget

Observed issues › fuzz/nvhpc-loop-character-achar-unsupported

Appears to behave incorrectlyCompiler crash · Fortran · OpenMP · Open

character value in a loop construct: unsupported

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

nvfortran -mp=gpu appears not to support assigning a character value computed in the region (achar, a concatenation) to a character array element inside a loop construct of a target region: the compile stops with "Character local variable unsupported on device". The same assignment in a distribute construct, or the copy of a mapped character element in the loop construct, 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 19.0.0fails another wayfails another wayunknown
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/abehaves incorrectlybehaves incorrectly
NVHPC 26.9n/an/an/an/an/abehaves incorrectlybehaves incorrectly
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 incorrectlyNVHPC 25.11, A100 separate
NVHPC 25.11, A100 managed
NVHPC 26.9, A100 separate
NVHPC 26.9, A100 managed
show
NVFORTRAN-F-0155-Compiler failed to translate accelerator region (see -Minfo messages): Character local variable unsupported on device (site-fuzz--nvhpc-loop-character-achar-unsupported.f90: 51)
fails another wayCCE 19.0.0, MI300A unified
CCE 19.0.0, MI300A copy
show
Warning:
 Headers and libraries from cray-libsci/25.03.0 will be ignored because depends on PE_MPICH which is unavailable..
      o2(i) = achar(64 + i)
ftn-7032 ftn: ERROR P, File = site-fuzz--nvhpc-loop-character-achar-unsupported.f90, Line = 54 
  Unsupported OpenMP construct Fortran character -- o2
Cray Fortran : 75 source lines
Cray Fortran : 1 errors, 0 warnings, 0 other messages, 0 ansi
Cray Fortran : "explain ftn-message number" gives more information about each message.

Reproduce

NVHPC 25.11, A100 separate

export LD_LIBRARY_PATH=$NVHPC_25_11/compilers/lib
nvfortran -mp=gpu -gpu=cc80 -gpu=mem:separate -O2 repro.f90 -o repro
OMP_TARGET_OFFLOAD=mandatory ./repro

NVHPC 26.9, A100 separate

export LD_LIBRARY_PATH=$NVHPC_26_9/compilers/lib
nvfortran -mp=gpu -gpu=cc80 -gpu=mem:separate -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 (64 lines)
! # Finding (parity-fuzz-s1): NVHPC: a character value computed in the region (achar, a concatenation) assigned to a character
! # array element inside a loop construct of a target region fails to translate ("Character local variable unsupported on
! # device"); the same assignment in a distribute construct compiles and runs correctly
!
! - toolchains and modes that fail: nvhpc (NVHPC 25.11), nvhpc269 (NVHPC 26.9); modes: copy, managed; -O2
! - build: nvfortran -mp=gpu -gpu=cc80 -gpu=mem:separate -O2
! - failure: ice
! - observed: `NVFORTRAN-F-0155-Compiler failed to translate accelerator region (see -Minfo messages): Character local variable
!   unsupported on device (<tmpdir>/nvhpc`; identical signature on 4 of 4 failing cells
! - expected: compiles and prints `s= 1160` with `check=pass`
! - rule: "A loop construct specifies that the logical iterations of the associated loops may execute concurrently and permits
!   the encountering threads to execute the loop accordingly." (5.2 §11.7 p.257 l.8-9); the loop body is a character assignment,
!   which no restriction of the construct excludes
! - vendor: "The private, bind, and collapse clauses are supported" on the loop construct (NVIDIA HPC Compilers User's Guide
!   26.9, Using OpenMP, item 2.9.5; data/index/claims/nvidia-hpc-26.9-user-guide-openmp.txt); the guide names no character limit
! - same-compiler control: controls/nvhpc-loop-character-distribute-control.f90 (one change from this file; correct on nvhpc
!   (NVHPC 25.11), nvhpc269 (NVHPC 26.9); rows results/parity/nvs1/tuo.cl.jsonl, results/parity/nvs1/wt.cl.jsonl);
!   controls/nvhpc-loop-character-array-value-control.f90 (the value copied from a mapped character array instead of computed:
!   four edits from this file, the array, its initialization, its map clause and the assignment; correct on nvhpc (NVHPC 25.11),
!   nvhpc269 (NVHPC 26.9); rows results/parity/nvs1/tuo.cl.jsonl, results/parity/nvs1/wt.cl.jsonl)
! - other families: afar (AFAR 24.3.0), cce20 (CCE 20.0.2), cce21 (CCE 21.0.3), cce21_4 (CCE 21.0.4) run the file correctly (rows
!   results/parity/nvs1/tuo.cl.jsonl); cce19 (CCE 19.0.0) fail it another way, e.g. `ftn-7032 ftn: ERROR P, File =
!   ../../nvhpc-loop-character-achar-unsupported.f90, Line = 10`
! - variants: files next to the reduction in results/parity/nvs1/red/, rows results/parity/nvs1/wt.cl.jsonl (both releases,
!   separate and managed memory) and results/parity/nvs1/tuo.cl.jsonl: the seed's concatenation `achar(...) // xc(i)(2:4)` in a
!   loop construct of a target parallel region (nvhpc-loop-character-concatenation-target-parallel-variant.f90) fails the same
!   way; the same concatenation in a do construct of the target parallel region compiles
!   (nvhpc-loop-character-concatenation-do-control.f90) but its run does not end on either release, the implicit barrier of a
!   worksharing loop in a combined target parallel construct (the c1 finding nvhpc-target-parallel-single-hang.f90), so it is
!   not a control of this finding. A substring copy `o2(i)(2:4) = xc(i)(2:4)` in the loop construct compiles (exploration)
! - gate and confirmer: 4 of 4 cells reproduce in all four environments (a-d; (c) on another node or GPU than (a); rows
!   results/parity/nvs1/gate_abd.cl.jsonl, results/parity/nvs1/gate_abd.cl.m2.jsonl, results/parity/nvs1/gate_c.cl.jsonl,
!   results/parity/nvs1/gate_c.cl.m2.jsonl, results/parity/nvs1/wt.cl.jsonl); campaign gate of the programs below: 2 cells, 2
!   confirmed in the four environments (a-d; (c) on another node or GPU; rows
!   results/parity/wingtip/results/parity/gate/fuzz/reruns/wingtip.abd.2465893.jsonl,
!   results/parity/wingtip/results/parity/gate/fuzz/reruns/wingtip.abd.2476695.jsonl,
!   results/parity/wingtip/results/parity/gate/fuzz/reruns/wingtip.c.3354124.jsonl,
!   results/parity/wingtip/results/parity/gate/fuzz/reruns/wingtip.c.3482833.jsonl)
! - programs (1): par-s1-nvhpc-nvhpc269:2564
! - legality: the rules (offload.engine.rules) and AFAR's flang front end accept this file, its controls and the variants above
!   (offload.engine.prerun.check on a Tuolumne compute node; results/parity/nvs1/prerun.cl.txt); AFAR 24.3.0 and CCE 20/21/21.0.4
!   compile and run this file and its controls correctly with HSA_XNACK 0 and 1 (results/parity/nvs1/tuo.cl.jsonl); CCE 19
!   rejects character data in target regions ("Unsupported OpenMP construct Fortran character"), its own documented limit
!   (afar-cce19-character-defaultmap.f90). NVHPC is the compiler under test and is not counted as a witness
program p
  implicit none
  integer, parameter :: n = 16
  character(len=4) :: o2(n)
  integer :: i, s
  o2 = ' '
  !$omp target teams map(from: o2)
    !$omp loop
    do i = 1, n
      o2(i) = achar(64 + i)
    end do
    !$omp end loop
  !$omp end target teams
  s = 0
  do i = 1, n
    s = s + ichar(o2(i)(1:1))
  end do
  print '(a,i0)', 's= ', s
  print '(a,a)', 'RESULT ran_on_host=0 check=', merge('pass', 'fail', s == 16 * 64 + 136)
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

NVFORTRAN-F-0155-Compiler failed to translate accelerator region (see -Minfo messages): Character local variable unsupported on device (site-fuzz--nvhpc-loop-character-achar-unsupported.f90: 51)

What the specification requires

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

OpenMP 5.2 §11.7 p.257 l.8-9

A loop construct specifies that the logical iterations of the associated loops may execute concurrently and permits the encountering threads to execute the loop accordingly.

Evidence

Same compiler, one change

  • one change from the finding; correct on nvhpc, nvhpc269: not run yet
    Control program
    program p
      implicit none
      integer, parameter :: n = 16
      character(len=4) :: o2(n)
      integer :: i, s
      o2 = ' '
      !$omp target teams map(from: o2)
        !$omp distribute
        do i = 1, n
          o2(i) = achar(64 + i)
        end do
        !$omp end distribute
      !$omp end target teams
      s = 0
      do i = 1, n
        s = s + ichar(o2(i)(1:1))
      end do
      print '(a,i0)', 's= ', s
      print '(a,a)', 'RESULT ran_on_host=0 check=', merge('pass', 'fail', s == 16 * 64 + 136)
    end program p
    
  • the value copied from a mapped character array (four edits from the finding); correct on nvhpc, nvhpc269: not run yet
    Control program
    program p
      implicit none
      integer, parameter :: n = 16
      character(len=4) :: o2(n), xc(n)
      integer :: i, s
      do i = 1, n
        xc(i) = achar(64 + i)
      end do
      o2 = ' '
      !$omp target teams map(from: o2) map(to: xc)
        !$omp loop
        do i = 1, n
          o2(i) = xc(i)
        end do
        !$omp end loop
      !$omp end target teams
      s = 0
      do i = 1, n
        s = s + ichar(o2(i)(1:1))
      end do
      print '(a,i0)', 's= ', s
      print '(a,a)', 'RESULT ran_on_host=0 check=', merge('pass', 'fail', s == 16 * 64 + 136)
    end program p
    

Other compilers run it correctly

AFAR 24.3.0, CCE 20.0.2, CCE 21.0.3, CCE 21.0.4, GCC 16.2.0