OffTarget

Observed issues › oop/cce-select-type-guard-not-taken

Appears to behave incorrectlyWrong answer · Fortran · OpenMP · Open

SELECT TYPE in a device procedure takes the CLASS DEFAULT branch for mapped objects

Legality: independently confirmed. One independent reading was uncertain; 2 adjudicating readings that weighed both find the program legal and the result required: Both readings find the program legal. The only OpenMP polymorphism restrictions concern ALLOCATABLE polymorphic list items (map clause) or C++ types with mismatched static and dynamic types; none applies since e1, e2, b0 are non-polymorphic variables mapped tofrom. A declare target procedure with CLASS dummy and SELECT TYPE is plain Fortran, so the values are determined. How this is checked

A declare target subroutine with a CLASS(base_t) dummy argument and a SELECT TYPE construct, called in a target region with objects of two extended types, should run the matching TYPE IS branch. CCE 19, 20 and 21 run the CLASS DEFAULT branch for every object (the same with a CLASS(*) dummy). The same calls on the host take the right branches. AFAR runs the right branches.

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 19.0.0behaves incorrectlybehaves incorrectlyunknown
not run
unknown
not run
unknown
not run
n/an/a
CCE 20.0.2behaves incorrectlybehaves incorrectlyunknown
not run
unknown
not run
unknown
not run
n/an/a
CCE 21.0.3behaves incorrectlybehaves incorrectlyunknown
not run
unknown
not run
unknown
not run
n/an/a
CCE 21.0.4behaves incorrectlybehaves incorrectlyunknown
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 16.2.0fails another wayfails another wayfails another wayfails another wayfails another wayfails another wayn/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 incorrectlyCCE 19.0.0, MI300A unified
CCE 19.0.0, MI300A copy
CCE 20.0.2, MI300A unified
CCE 20.0.2, MI300A copy
CCE 21.0.3, MI300A unified
CCE 21.0.3, MI300A copy
CCE 21.0.4, MI300A unified
CCE 21.0.4, MI300A copy
show
r1= 0
r2= 0
r3= -1
RESULT ran_on_host=0 check=fail
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-1058-Call to Compiler runtime function not supported - pgf90_same_type_as_i8 (site-oop--cce-select-type-guard-not-taken.f90: 42)
  0 inform,   0 warnings,   1 severes, 0 fatal for bump
fails another wayGCC 16.2.0, MI300A unified
GCC 16.2.0, MI300A copy
show
site-oop--cce-select-type-guard-not-taken.f90:52:16: error: variable ‘__vtab_oop_m_Ext1’ has been referenced in offloaded code but hasn’t been marked to be included in the offloaded code
   52 | end module oop_m
      |                ^
site-oop--cce-select-type-guard-not-taken.f90:52:16: error: variable ‘__vtab_oop_m_Ext2’ has been referenced in offloaded code but hasn’t been marked to be included in the offloaded code
lto1: fatal error: errors during merging of translation units
compilation terminated.
gcn mkoffload: fatal error: <home>/gcc-16.2/bin/x86_64-pc-linux-gnu-accel-amdgcn-amdhsa-gcc returned 1 exit status
compilation terminated.
fails another wayGCC 16.2.0, MI250X
show
site-oop--cce-select-type-guard-not-taken.f90:53:16: error: variable ‘__vtab_oop_m_Ext1’ has been referenced in offloaded code but hasn’t been marked to be included in the offloaded code
   53 | end module oop_m
      |                ^
site-oop--cce-select-type-guard-not-taken.f90:53:16: error: variable ‘__vtab_oop_m_Ext2’ has been referenced in offloaded code but hasn’t been marked to be included in the offloaded code
lto1: fatal error: errors during merging of translation units
compilation terminated.
gcn mkoffload: fatal error: <home>/gcc-16.2/bin/x86_64-pc-linux-gnu-accel-amdgcn-amdhsa-gcc returned 1 exit status
compilation terminated.
fails another wayGCC 16.2.0, MI300X
GCC 16.2.0, MI350X
show
site-oop--cce-select-type-guard-not-taken.f90:52:16: error: variable ‘__vtab_oop_m_Ext1’ has been referenced in offloaded code but hasn’t been marked to be included in the offloaded code
   52 | end module oop_m
      |                ^
site-oop--cce-select-type-guard-not-taken.f90:52:16: error: variable ‘__vtab_oop_m_Ext2’ has been referenced in offloaded code but hasn’t been marked to be included in the offloaded code
lto1: fatal error: errors during merging of translation units
compilation terminated.
gcn mkoffload: fatal error: <home>/<user>/gcc-16.2/bin/x86_64-pc-linux-gnu-accel-amdgcn-amdhsa-gcc returned 1 exit status
compilation terminated.
fails another wayGCC 16.2.0, A100 separate
show
site-oop--cce-select-type-guard-not-taken.f90:51:16: error: variable ‘__vtab_oop_m_Ext1’ has been referenced in offloaded code but hasn’t been marked to be included in the offloaded code
   51 | end module oop_m
      |                ^
site-oop--cce-select-type-guard-not-taken.f90:51:16: error: variable ‘__vtab_oop_m_Ext2’ has been referenced in offloaded code but hasn’t been marked to be included in the offloaded code
lto1: fatal error: errors during merging of translation units
compilation terminated.
nvptx mkoffload: fatal error: <home>/gcc-16.2/bin/x86_64-pc-linux-gnu-accel-nvptx-none-gcc returned 1 exit status
compilation terminated.

Reproduce

CCE 19.0.0, MI300A unified

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

CCE 20.0.2, MI300A unified

module load cce/20.0.2 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

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 (84 lines)
! # Finding (fortran-oop): SELECT TYPE in a declare target procedure takes CLASS DEFAULT for mapped objects of extended types
!
! - toolchains and modes that fail: cce19 (CCE 19.0.0: copy, unified), cce20 (CCE 20.0.2: copy, unified), cce21 (CCE 21.0.3: copy,
!   unified), cce21_4 (CCE 21.0.4: copy, unified); -O2
! - build: ftn -fopenmp -O2 (module load cce/<v> craype-accel-amd-gfx942 rocm/<r>; OpenACC: ftn -h acc)
! - failure: wrong_answer (check=fail); OMP_TARGET_OFFLOAD=mandatory
! - observed: `r1= 0 r2= 0 r3= -1`: the calls with an ext1 and an ext2 actual take the CLASS DEFAULT branch (oop-poly-select-type,
!   every CCE release, both modes); a CLASS(*) dummy classifies every derived-type actual as CLASS DEFAULT (oop-unlimited-classify
!   `r1= 0`, every CCE, both modes); with intrinsic and derived actuals of a CLASS(*) dummy CCE 19 misses only the derived-type
!   actual (oop-unlimited-select-type `r1= 42 r2= 20 r3= 5`, expected r3= 7; CCE 20 and 21 correct). All three programs map
!   non-polymorphic objects from the host and dispatch on them in the device procedure
! - expected: `r1= 10 r2= 100 r3= -1` (the TYPE IS (ext1) and TYPE IS (ext2) branches run for those actuals)
! - rule: Fortran 2018 11.1.11 (a TYPE IS guard is chosen when the dynamic type of the selector is that type); a CLASS dummy
!   argument of a declare target procedure is not restricted by OpenMP (5.2 §1.2.6 p.19 l.1-3: the type-bound procedures accessed
!   must be declare target; this program has none)
! - same-compiler control: the same three calls on the host (controls/host-select-type.f90 and
!   controls/host-class-star-select-type.f90: correct on every CCE release in both modes, results/oop/probe/controls.jsonl)
! - other families: AFAR correct on all three programs; GCC 16.2 rejects them at compile time (a separate finding:
!   gcc-vtab-not-marked-for-offload.f90)
! - gate and confirmer: 18 cells, 18 confirmed in the gate's four environments (a-d; results/oop/main/gate_reruns*.jsonl);
!   confirmer (results/oop/main/confirm.jsonl): compiler_bug (R20-runtime-bug) 18
! - programs (3): oop-poly-select-type, oop-unlimited-classify, oop-unlimited-select-type (offload/engine/oop_gen.py; this file
!   is the
!   first one)
! - legality: for the cce cells: the rules (offload.engine.rules, with the fortran-oop restrictions) and AFAR's flang front end
!   (-fsyntax-only) (tools/campaigns/oop.py legality, results/oop/main/legality.json)
module oop_m
  implicit none
  integer, parameter :: n = 64
  type base_t
    integer :: a = 0
  end type base_t
  type, extends(base_t) :: ext1
    integer :: u = 0
  end type ext1
  type, extends(base_t) :: ext2
    integer :: v = 0
  end type ext2
contains
  subroutine bump(s)
    !$omp declare target
    class(base_t), intent(inout) :: s
    select type (s)
      type is (ext1)
        s%u = s%u + 10
      type is (ext2)
        s%v = s%v + 100
      class default
        s%a = -1
    end select
  end subroutine bump
end module oop_m

program oop
  use omp_lib
  use oop_m
  implicit none
  integer :: host_flag
  logical :: ok
  integer :: r1, r2, r3
  type(ext1) :: e1
  type(ext2) :: e2
  type(base_t) :: b0
  host_flag = 1
  !$omp target map(from: host_flag)
  host_flag = merge(1, 0, omp_is_initial_device())
  !$omp end target
  !$omp target map(tofrom: e1, e2, b0)
  call bump(e1)
  call bump(e2)
  call bump(b0)
  !$omp end target
  r1 = e1%u
  r2 = e2%v
  r3 = b0%a
  ok = .true.
  ok = ok .and. (r1 == 10)
  ok = ok .and. (r2 == 100)
  ok = ok .and. (r3 == -1)
  print '(a,i0)', 'r1= ', r1
  print '(a,i0)', 'r2= ', r2
  print '(a,i0)', 'r3= ', r3
  print '(a,i0,a,a)', 'RESULT ran_on_host=', host_flag, ' check=', merge('pass', 'fail', ok)
end program oop

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

r1= 10

Observed where it behaves incorrectly

r1= 0
r2= 0
r3= -1
RESULT ran_on_host=0 check=fail

What the specification requires

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

OpenMP 5.2 §1.2.6 p.19 l.1-3

Evidence

Same compiler, one change

  • the same calls of the same procedure on the host: not run yet
    Control program
    ! control for oop-poly-select-type: the same calls on the host (one change: no target construct)
    ! fortran-oop poly/select-type (OpenMP 5.2, free-form Fortran):
    ! a declare target subroutine with a CLASS(base_t) dummy and SELECT TYPE, called on the device
    !   with mapped non-polymorphic objects of three types
    ! spec: OpenMP 5.2 §1.2.6 p.19 l.1-3 (type-bound procedures accessed in a target region appear in a
    !   declare target directive); OpenMP 5.2 §7.8.1 p.205-207 (declare target, procedures); OpenMP 5.2
    !   §5.3 p.107 l.17-19 (SELECT TYPE in OpenMP regions); Fortran 2018 11.1.11 (SELECT TYPE)
    ! legality: defined
    ! expected (computed by compliance/oop_gen.py): r1= 10, r2= 100, r3= -1
    module oop_m
      implicit none
      integer, parameter :: n = 64
      type base_t
        integer :: a = 0
      end type base_t
      type, extends(base_t) :: ext1
        integer :: u = 0
      end type ext1
      type, extends(base_t) :: ext2
        integer :: v = 0
      end type ext2
    contains
      subroutine bump(s)
        !$omp declare target
        class(base_t), intent(inout) :: s
        select type (s)
          type is (ext1)
            s%u = s%u + 10
          type is (ext2)
            s%v = s%v + 100
          class default
            s%a = -1
        end select
      end subroutine bump
    end module oop_m
    
    program oop
      use omp_lib
      use oop_m
      implicit none
      integer :: host_flag
      logical :: ok
      integer :: r1, r2, r3
      type(ext1) :: e1
      type(ext2) :: e2
      type(base_t) :: b0
      host_flag = 1
      !$omp target map(from: host_flag)
      host_flag = merge(1, 0, omp_is_initial_device())
      !$omp end target
      call bump(e1)
      call bump(e2)
      call bump(b0)
      r1 = e1%u
      r2 = e2%v
      r3 = b0%a
      ok = .true.
      ok = ok .and. (r1 == 10)
      ok = ok .and. (r2 == 100)
      ok = ok .and. (r3 == -1)
      print '(a,i0)', 'r1= ', r1
      print '(a,i0)', 'r2= ', r2
      print '(a,i0)', 'r3= ', r3
      print '(a,i0,a,a)', 'RESULT ran_on_host=', host_flag, ' check=', merge('pass', 'fail', ok)
    end program oop
    

Other compilers run it correctly

AFAR 24.3.0