Observed issues › nvhpc-managed-acc-shutdown-host-access

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

nvfortran -acc -gpu=mem:managed: host access to an allocatable after !$acc shutdown ends in SIGSEGV

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	n/a	n/a	n/a	n/a	n/a	n/a	n/a
CCE 19.0.0	correct	correct	unknown not run	unknown not run	unknown not run	n/a	n/a
CCE 20.0.2	correct	correct	unknown not run	unknown not run	unknown not run	n/a	n/a
CCE 21.0.3	correct	correct	unknown not run	unknown not run	unknown not run	n/a	n/a
CCE 21.0.4	correct	correct	unknown not run	unknown not run	unknown not run	n/a	n/a
NVHPC 25.11	n/a	n/a	n/a	n/a	n/a	correct	behaves incorrectly
NVHPC 26.9	n/a	n/a	n/a	n/a	n/a	correct	behaves incorrectly
GCC 16.2.0	fails another way	fails another way	fails another way	fails another way	fails another way	fails another way	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	NVHPC 25.11, A100 managed NVHPC 26.9, A100 managed	show after shutdown timeout: the monitored command dumped core
fails another way	GCC 16.2.0, MI300A unified GCC 16.2.0, MI300A copy GCC 16.2.0, MI250X GCC 16.2.0, MI300X GCC 16.2.0, MI350X GCC 16.2.0, A100 separate	show nvhpc-managed-acc-shutdown-host-access.f90:6:9: 6 \| !$acc shutdown \| 1 Error: Unclassifiable OpenACC directive at (1)

Reproduce

NVHPC 25.11, A100 managed

export LD_LIBRARY_PATH=$NVHPC_25_11/compilers/lib
nvfortran -acc=gpu -gpu=cc80 -gpu=mem:managed -O2 repro.f90 -o repro
./repro

NVHPC 26.9, A100 managed

export LD_LIBRARY_PATH=$NVHPC_26_9/compilers/lib
nvfortran -acc=gpu -gpu=cc80 -gpu=mem:managed -O2 repro.f90 -o repro
./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 -h acc -O2 repro.f90 -o repro
HSA_XNACK=1 ./repro

Program (38 lines)

! NVHPC 25.11, NVIDIA A100 80GB PCIe (wingtip, sm_80), managed memory
! (-gpu=mem:managed): after the OpenACC shutdown directive a HOST access to
! an allocatable array ends the program with SIGSEGV; the same program is
! correct in separate memory (the compiler's default).
! Spec: "The shutdown directive shuts down the connection to the given
! device or devices of the given device type, and frees any associated
! runtime resources. This ends all data lifetimes in device memory"
! (OpenACC 3.4 §2.14.2 p.86). It ends device data lifetimes; the host array
! is host data of the program (no data clause names it here), so reading it
! on the host afterwards is defined. Managed memory is NVHPC's mode in which
! "Dynamically allocated host data are placed in CUDA Managed Memory" (HPC
! Compilers User's Guide 25.11 §5.4): the program's own host allocation
! becomes a runtime resource that shutdown frees.
! Expected: "after shutdown" and "sum 64", exit 0.
! Observed (2026-10-24, nvfortran 25.11-0, -O2, two runs, the same):
! - -gpu=mem:separate: "after shutdown", "sum 64", exit 0 (correct);
! - -gpu=mem:managed: no output, exit status 139 (SIGSEGV; the buffered
!   output is lost).
! Found by the catalog: acc34-directive-shutdown is correct in separate
! memory and crashes in managed memory (catalog/mode_differences.yaml
! nvhpc_entries, class managed_failure); reduced by hand to this program.
! NVHPC 26.9 (26.9-0, the latest release; rerun 2026-10-24 on wingtip with
! the same flags): still present -- the same: managed memory SIGSEGV (exit 139), separate correct.
! Fix (2026-10-24): rerun on 25.11 and 26.9 in both memory modes,
! each with its own runtime libraries (LD_LIBRARY_PATH = the release's
! compilers/lib; the first 26.9 runs had loaded 25.11's): still present
! on both releases, the same output.
! NVHPC: nvfortran -acc=gpu -gpu=cc80 -gpu=mem:managed -O2
! Run:   ./a.out
program p
  implicit none
  integer, allocatable :: a(:)
  allocate(a(64))
  a = 1
  !$acc shutdown
  print *, 'after shutdown'
  print *, 'sum', sum(a)
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

Expected

sum 64

Observed where it behaves incorrectly

 after shutdown
timeout: the monitored command dumped core

What the specification requires

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

OpenACC 3.4 §2.14.2 p.86

The shutdown directive shuts down the connection to the given device or devices of the given device type, and frees any associated runtime resources. This ends all data lifetimes in device memory

Evidence

Same compiler, one change

the shutdown directive removed: CCE 19.0.0: correct; CCE 20.0.2: correct; CCE 21.0.3: correct; CCE 21.0.4: correct; NVHPC 25.11: correct; NVHPC 26.9: correct
Control program
```
program p
  implicit none
  integer, allocatable :: a(:)
  allocate(a(64))
  a = 1
  print *, 'after shutdown'
  flush(6)
  print *, 'sum', sum(a)
end program p
```

Other compilers run it correctly

CCE 19.0.0, CCE 20.0.2, CCE 21.0.3, CCE 21.0.4