Schedule(static, 4) and schedule(static) on a parallel do in a target region deal iterations out one at a time

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	behaves incorrectly	behaves incorrectly	behaves incorrectly	behaves incorrectly	behaves incorrectly	n/a	n/a
CCE 19.0.0	behaves incorrectly	behaves incorrectly	behaves incorrectly	unknown not run	unknown not run	n/a	n/a
CCE 20.0.2	behaves incorrectly	behaves incorrectly	behaves incorrectly	unknown not run	unknown not run	n/a	n/a
CCE 21.0.3	behaves incorrectly	behaves incorrectly	behaves incorrectly	unknown not run	unknown not run	n/a	n/a
CCE 21.0.4	behaves incorrectly	behaves incorrectly	behaves incorrectly	unknown not run	unknown not run	n/a	n/a
NVHPC 25.11	n/a	n/a	n/a	n/a	n/a	correct	correct
NVHPC 26.9	n/a	n/a	n/a	n/a	n/a	correct	correct
GCC 16.2.0	correct	correct	correct	correct	correct	correct	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
behaves incorrectly	AFAR 24.3.0, MI300A unified AFAR 24.3.0, MI300A copy AFAR 24.3.0, MI250X AFAR 24.3.0, MI300X AFAR 24.3.0, MI350X CCE 19.0.0, MI300A unified CCE 19.0.0, MI300A copy CCE 19.0.0, MI250X CCE 20.0.2, MI300A unified CCE 20.0.2, MI300A copy CCE 20.0.2, MI250X CCE 21.0.3, MI300A unified CCE 21.0.3, MI300A copy CCE 21.0.3, MI250X CCE 21.0.4, MI300A unified CCE 21.0.4, MI300A copy CCE 21.0.4, MI250X	show threads=256 static,4 tid(1:12)= 0 1 2 3 4 5 6 7 8 9 10 11 static tid(1:12)= 0 1 2 3 4 5 6 7 8 9 10 11 static,4 iterations on the wrong thread: 997 of 1000 (expected 0) static: iterations starting a second chunk of a thread: 744 (expected 0)

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

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

NVHPC 25.11, A100 separate (runs correctly)

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

Program (48 lines)

! Schedule(static, 4) on a parallel do in a target region:
! "iterations are divided into chunks of size chunk_size, and the chunks
! are assigned to the threads in the team in a round-robin fashion in the
! order of the thread number" (5.2 §11.5.3 p.252), so iteration i runs on
! thread mod((i - 1)/4, T). schedule(static) without a chunk size: "at
! most one chunk is distributed to each thread" (same page), so each
! thread's iterations are contiguous.
! Expected: 0 wrong, 0 second chunks. Observed (MI300A, HSA_XNACK 0 and 1,
! 3 runs each): AFAR flang, CCE 19, 20 and 21 ftn: 256 threads,
! tid(1:12) = 0 1 2 ... 11 for both loops (one iteration at a time,
! round-robin), 997 of 1000 iterations on the wrong thread, 744 second
! chunks. The C form (sched_repro.c in results/t27a/diag/d2, AMD clang)
! is correct: tid = 0 0 0 0 1 1 1 1 ...
program sched_repro
  use omp_lib
  implicit none
  integer, parameter :: m = 1000
  integer :: tid(m), tid2(m), nthr, i, bad, bad2
  tid = -1
  tid2 = -1
  nthr = -1
  !$omp target map(from: tid, tid2, nthr)
    !$omp parallel do schedule(static, 4)
      do i = 1, m
        if (i == 1) nthr = omp_get_num_threads()
        tid(i) = omp_get_thread_num()
      end do
    !$omp end parallel do
    !$omp parallel do schedule(static)
      do i = 1, m
        tid2(i) = omp_get_thread_num()
      end do
    !$omp end parallel do
  !$omp end target
  bad = 0
  do i = 1, m
    if (tid(i) /= mod((i - 1)/4, nthr)) bad = bad + 1
  end do
  ! static without chunk: at most one chunk (contiguous) per thread
  bad2 = 0
  do i = 2, m
    if (tid2(i) /= tid2(i - 1) .and. any(tid2(1:i - 1) == tid2(i))) bad2 = bad2 + 1
  end do
  print '(a,i0,a,12(1x,i0))', 'threads=', nthr, ' static,4 tid(1:12)=', tid(1:12)
  print '(a,12(1x,i0))', 'static tid(1:12)=', tid2(1:12)
  print '(a,i0,a,i0,a)', 'static,4 iterations on the wrong thread: ', bad, ' of ', m, ' (expected 0)'
  print '(a,i0,a)', 'static: iterations starting a second chunk of a thread: ', bad2, ' (expected 0)'
end program sched_repro

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

threads=256 static,4 tid(1:12)= 0 1 2 3 4 5 6 7 8 9 10 11
static tid(1:12)= 0 1 2 3 4 5 6 7 8 9 10 11
static,4 iterations on the wrong thread: 997 of 1000 (expected 0)
static: iterations starting a second chunk of a thread: 744 (expected 0)

What the specification requires

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

OpenMP 5.2 §11.5.3 p.252

iterations are divided into chunks of size chunk_size, and the chunks are assigned to the threads in the team in a round-robin fashion in the order of the thread number

Evidence

Same compiler, one change

schedule(static, 1) and its check (iteration i on thread mod(i - 1, T)) instead of schedule(static, 4): 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 sched_repro
  use omp_lib
  implicit none
  integer, parameter :: m = 1000
  integer :: tid(m), tid2(m), nthr, i, bad, bad2
  tid = -1
  tid2 = -1
  nthr = -1
  !$omp target map(from: tid, tid2, nthr)
    !$omp parallel do schedule(static, 1)
      do i = 1, m
        if (i == 1) nthr = omp_get_num_threads()
        tid(i) = omp_get_thread_num()
      end do
    !$omp end parallel do
    !$omp parallel do schedule(static)
      do i = 1, m
        tid2(i) = omp_get_thread_num()
      end do
    !$omp end parallel do
  !$omp end target
  bad = 0
  do i = 1, m
    if (tid(i) /= mod((i - 1)/1, nthr)) bad = bad + 1
  end do
  bad2 = 0
  do i = 2, m
    if (tid2(i) /= tid2(i - 1) .and. any(tid2(1:i - 1) == tid2(i))) bad2 = bad2 + 1
  end do
  print '(a,i0,a,12(1x,i0))', 'threads=', nthr, ' static,4 tid(1:12)=', tid(1:12)
  print '(a,12(1x,i0))', 'static tid(1:12)=', tid2(1:12)
  print '(a,i0,a,i0,a)', 'static,4 iterations on the wrong thread: ', bad, ' of ', m, ' (expected 0)'
  print '(a,i0,a)', 'static: iterations starting a second chunk of a thread: ', bad2, ' (expected 0)'
end program sched_repro

Other compilers run it correctly

NVHPC 25.11, NVHPC 26.9, GCC 16.2.0