GCC OpenACC: structure and member array in one data directive stops at run time

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

GCC stops with "outer struct not mapped for attach" at a data directive naming a structure and its pointee array.

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	n/a	n/a	unknown not run	unknown not run	unknown not run	n/a	n/a
CCE 20.0.2	n/a	n/a	unknown not run	unknown not run	unknown not run	n/a	n/a
CCE 21.0.3	n/a	n/a	unknown not run	unknown not run	unknown not run	n/a	n/a
CCE 21.0.4	n/a	n/a	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	correct
NVHPC 26.9	n/a	n/a	n/a	n/a	n/a	correct	correct
GCC 16.2.0	behaves incorrectly	behaves incorrectly	behaves incorrectly	behaves incorrectly	behaves incorrectly	behaves incorrectly	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	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 libgomp: outer struct not mapped for attach

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/gcc -fopenacc -foffload=amdgcn-amdhsa -foffload-options=amdgcn-amdhsa=-march=gfx942 -O2 repro.c -o repro
HSA_XNACK=1 ./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/gcc -fopenacc -foffload=nvptx-none -foffload-options=nvptx-none=-march=sm_80 -foffload-options=nvptx-none=-latomic -O2 repro.c -o repro
./repro

NVHPC 25.11, A100 separate (runs correctly)

export LD_LIBRARY_PATH=$NVHPC_25_11/compilers/lib
nvc -acc=gpu -gpu=cc80 -gpu=mem:separate -O2 repro.c -o repro
./repro

Program (77 lines)

// # Finding (nested-data): GCC OpenACC: copyin of a structure and of the array its pointer member points to in one data
// # directive stops at run time: "outer struct not mapped for attach"
//
// - toolchains and modes that fail: gcc_g90 (GCC 16.2.0), gcc_sm80 (GCC 16.2.0); modes: copy; -O2
// - build: gcc -fopenacc -foffload=amdgcn-amdhsa -foffload-options=amdgcn-amdhsa=-march=gfx90a -O2 (gcc_g90); gcc -fopenacc
//   -foffload=nvptx-none -foffload-options=nvptx-none=-march=sm_80 -foffload-options=nvptx-none=-latomic -O2 (gcc_sm80)
// - failure: wrong_answer (check=fail)
// - observed: `libgomp: outer struct not mapped for attach`; identical signature on 2 of 2 failing cells
// - expected: compiles and prints `tot= 15` with `check=pass`
// - rule: OpenACC 3.4 §2.7.8 p.55 l.1902-1903: "If var is a pointer reference, an attach pointer action is performed
//   followed by an increment counter action"; the attach actions of a directive are scheduled after its transfers: "The
//   implementation schedules pointer updates after any data transfers due to transfer in actions that are performed for
//   the same directive." (§2.7.2 p.52 l.1785-1786), so the structure and the array its member points to may be mapped by
//   one directive
// - same-compiler control: controls/c3-ctl-nested-data.c (one change from this file; correct on gcc_g90 (GCC 16.2.0), gcc_sm80 (GCC
//   16.2.0); rows results/nested/red/red6.jsonl, results/nested/red/wgs.jsonl, results/nested/red/wnv.jsonl);
//   controls/c5-acc-serial-struct-and-array.c (one change from this file; correct on gcc_g90 (GCC 16.2.0), gcc_sm80 (GCC 16.2.0);
//   rows results/nested/red/red6.jsonl, results/nested/red/wgs.jsonl, results/nested/red/wnv.jsonl)
// - other families: nvhpc (NVHPC 25.11), nvhpc269 (NVHPC 26.9) run the file correctly (rows results/nested/red/wnv.jsonl, site
//   rows, check case site-nested--gcc-acc-c-struct-and-member-array-one-directive)
// - gate and confirmer: campaign gate of the programs below: 82 cells, 82 confirmed in the four environments (a-d; (c) on another
//   node or GPU; rows results/nested/big/gate_reruns.part.<host>.1371141.jsonl,
//   results/nested/big/gate_reruns.part.<host>.517665.jsonl,
//   results/nested/big/gate_reruns.part.wingtip.501872.jsonl,
//   results/nested/big/gate_reruns.part.wingtip.501873.jsonl,
//   results/nested/big/gate_reruns.part.wingtip.711615.jsonl,
//   results/nested/big/gate_reruns.part.wingtip.724122.jsonl, results/nested/big/rows.part.<host>.397000.jsonl,
//   results/nested/big/rows.part.wingtip.3886921.jsonl,
//   results/nested/main/gate_reruns.part.<host>.2711328.jsonl, results/nested/main/gate_reruns.part.<host>.351539.jsonl,
//   results/nested/main/gate_reruns.part.wingtip.3582436.jsonl,
//   results/nested/main/gate_reruns.part.wingtip.3582437.jsonl,
//   results/nested/main/gate_reruns.part.wingtip.3637854.jsonl,
//   results/nested/main/gate_reruns.part.wingtip.3639010.jsonl,
//   results/nested/main/rows.part.<host>.2691037.jsonl, results/nested/main/rows.part.wingtip.3363229.jsonl,
//   results/nested/main/rows.part.wingtip.3363230.jsonl)
// - programs (41): nd-c-acc-tdata-alias-d2-20ae973e, nd-c-acc-tdata-alias-d2-23c00982, nd-c-acc-tdata-alias-d2-72a1d7e8,
//   nd-c-acc-tdata-alias-d2-af2ec241, nd-c-acc-tdata-chain-d1-0ed1bf12, nd-c-acc-tdata-chain-d1-9886a523,
//   nd-c-acc-tdata-chain-d2-085c13be, nd-c-acc-tdata-chain-d2-1a584322, nd-c-acc-tdata-chain-d2-25c6c4d8,
//   nd-c-acc-tdata-chain-d2-361ca0f9, nd-c-acc-tdata-chain-d2-36b0e127, nd-c-acc-tdata-chain-d2-43985434,
//   nd-c-acc-tdata-chain-d2-462abe25, nd-c-acc-tdata-chain-d2-5e7728c0, nd-c-acc-tdata-chain-d2-85f2a830,
//   nd-c-acc-tdata-chain-d2-a3760a7a, nd-c-acc-tdata-chain-d2-af1ca1e6, nd-c-acc-tdata-chain-d2-d9e6a839,
//   nd-c-acc-tdata-chain-d2-f585dbe8, nd-c-acc-tdata-chain-d2-f85f821b, nd-c-acc-tdata-chain-d3-474c8b58,
//   nd-c-acc-tdata-chain-d3-475cfe44, nd-c-acc-tdata-chain-d3-4bf0e2c8, nd-c-acc-tdata-chain-d3-5c2aa78e,
//   nd-c-acc-tdata-chain-d3-63c6d8ed, nd-c-acc-tdata-chain-d3-8fa68caf, nd-c-acc-tdata-chain-d3-99924776,
//   nd-c-acc-tdata-chain-d3-9d855631, nd-c-acc-tdata-chain-d3-f891fd7c, nd-c-acc-tdata-diamond-d3-6974fab3,
//   nd-c-acc-tdata-diamond-d3-6fe9454b, nd-c-acc-tdata-diamond-d3-87e0b9a8, nd-c-acc-tdata-list-d2-1ab29ba9,
//   nd-c-acc-tdata-list-d2-cc8507d1, nd-c-acc-tdata-list-d3-367465a3, nd-c-acc-tdata-tree-d2-12d76b63,
//   nd-c-acc-tdata-tree-d2-87d4954e, nd-c-acc-tdata-tree-d2-aedfe6ab, nd-c-acc-tdata-tree-d2-fdcb2b5a,
//   nd-c-acc-tdata-tree-d3-335e85bb, nd-c-acc-tdata-tree-d3-98389330
// - legality: the rules (offload.engine.rules) accept it; both NVHPC releases compile and run the reduction correctly (a
//   compiler of another family; the registry has no OpenACC C clause checker, so this is acceptance and execution, not a
//   clause check)
#include <stdio.h>
#include <stdlib.h>
#include <openacc.h>
struct s0 { int v; int *a; };
int main(void)
{
  int hf = 1, i, tot = 0, ok;
  struct s0 r;
  r.v = 5; r.a = malloc(4 * sizeof(int));
  for (i = 0; i < 4; i++) r.a[i] = i + 1;
  #pragma acc serial copy(hf)
  hf = acc_on_device(acc_device_host) ? 1 : 0;
  #pragma acc data copyin(r, r.a[0:4])
  {
    #pragma acc serial copy(tot) present(r)
    {
      tot = r.v;
      for (int q = 0; q < 4; q++) tot += r.a[q];
    }
  }
  ok = tot == 15;
  printf("tot= %d\n", tot);
  printf("RESULT ran_on_host=%d check=%s\n", hf, ok ? "pass" : "fail");
  return 0;
}

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

libgomp: outer struct not mapped for attach

What the specification requires

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

OpenACC 3.4 §2.7.8 p.55 l.1902-1903

If var is a pointer reference, an attach pointer action is performed followed by an increment counter action

OpenACC 3.4 §2.7.2 p.52 l.1785-1786

The implementation schedules pointer updates after any data transfers due to transfer in actions that are performed for the same directive.

Evidence

Same compiler, one change

one change from the finding; correct on gcc_g90, gcc_sm80: not run yet

Control program

// control: two nested data constructs
#include <stdio.h>
#include <stdlib.h>
#include <openacc.h>
struct s0 { int v; int *a; };
int main(void)
{
  int hf = 1, i, tot = 0, ok;
  struct s0 r;
  r.v = 5; r.a = malloc(4 * sizeof(int));
  for (i = 0; i < 4; i++) r.a[i] = i + 1;
  #pragma acc serial copy(hf)
  hf = acc_on_device(acc_device_host) ? 1 : 0;
  #pragma acc data copyin(r)
  #pragma acc data copyin(r.a[0:4])
  {
    #pragma acc serial copy(tot) present(r)
    {
      tot = r.v;
      for (int q = 0; q < 4; q++) tot += r.a[q];
    }
  }
  ok = tot == 15;
  printf("tot= %d\n", tot);
  printf("RESULT ran_on_host=%d check=%s\n", hf, ok ? "pass" : "fail");
  return 0;
}

one change from the finding; correct on gcc_g90, gcc_sm80: not run yet

Control program

// OpenACC compute construct copyin(r, r.a[0:4])
#include <stdio.h>
#include <stdlib.h>
#include <openacc.h>
struct s0 { int v; int *a; };
int main(void)
{
  int hf = 1, i, tot = 0, ok;
  struct s0 r;
  r.v = 5; r.a = malloc(4 * sizeof(int));
  for (i = 0; i < 4; i++) r.a[i] = i + 1;
  #pragma acc serial copy(hf)
  hf = acc_on_device(acc_device_host) ? 1 : 0;

  {
    #pragma acc serial copy(tot) copyin(r, r.a[0:4])
    {
      tot = r.v;
      for (int q = 0; q < 4; q++) tot += r.a[q];
    }
  }
  ok = tot == 15;
  printf("tot= %d\n", tot);
  printf("RESULT ran_on_host=%d check=%s\n", hf, ok ? "pass" : "fail");
  return 0;
}

Other compilers run it correctly

NVHPC 25.11, NVHPC 26.9