Observed issues › gcc/gcc-acc-c-link-in-routine-rejected
OpenACC: a declare link variable used in a routine is rejected (C)
Legality: not independently confirmed. no independent review yet How this is checked
Same issue as gcc/gcc-acc-link-in-routine-rejected; this program is another reproducer of it.
A variable in a declare link clause may be used in an accelerator routine (OpenACC 3.4 §2.13.3). GCC 16.2 rejects a routine seq procedure that reads such a variable: "'lv' with 'link' clause used in 'routine' function", in Fortran and C. CCE compiles and runs the Fortran program.
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 | fails another way | fails another way |
| NVHPC 26.9 | n/a | n/a | n/a | n/a | n/a | fails another way | fails another way |
| GCC 16.2.0 | not supported | not supported | not supported | not supported | not supported | not supported | 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 |
|---|---|---|
| not supported | 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 | showsite-gcc--gcc-acc-c-link-in-routine-rejected.c: In function ‘get_lv’:
site-gcc--gcc-acc-c-link-in-routine-rejected.c:40:5: error: ‘lv’ with ‘link’ clause used in ‘routine’ function
40 | int lv[64];
| ^~
site-gcc--gcc-acc-c-link-in-routine-rejected.c:40:5: error: ‘lv’ with ‘link’ clause used in ‘routine’ function |
| fails another way | NVHPC 25.11, A100 separate NVHPC 25.11, A100 managed NVHPC 26.9, A100 separate NVHPC 26.9, A100 managed | showRESULT ran_on_host=0 check=fail |
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
Program (75 lines)
// # Finding (GCC 16.2): a declare link variable used in an OpenACC routine is rejected (C); the reproducer is the catalog program
// acc34-c-clause-link
//
// - toolchains: gcc_g90 (GCC 16.2.0, AMD Instinct MI250X gfx90a, Tioga) and gcc_sm80 (GCC 16.2.0,
// NVIDIA A100 sm_80, wingtip); separate memory only; -O2; OMP_TARGET_OFFLOAD=mandatory
// - build: gcc -fopenacc -foffload=amdgcn-amdhsa -foffload-options=amdgcn-amdhsa=-march=gfx90a -O2 (gcc_g90); -foffload=nvptx-none
// -foffload-options=nvptx-none=-march=sm_80 -foffload-options=nvptx-none=-latomic -O2 (gcc_sm80)
// - failure: rejected: with .link. clause used in .routine. function
// - observed: `error: 'lv' with 'link' clause used in 'routine' function` on both GPUs.
// - expected: compiles; `RESULT ran_on_host=0 check=pass`
// - spec: "The global variable or common block variables may be used in accelerator routines" (OpenACC 3.4 §2.13.3 l.3083-3084),
// for a variable in a declare link clause; the device copy exists while a data clause holds it (l.3079-3081), here the parallel
// loop's copyin(lv).
// - GCC 16.2 documentation (libgomp.texi "Enabling OpenACC"): "A complete description of all OpenACC directives accepted may be
// found in the OpenACC Application Programming Interface manual, version 2.6"; _OPENACC is 201711 (2.6). No list of OpenACC gaps
// is documented.
// - other families: NVHPC 25.11/26.9 compile it and give a wrong answer (not studied here).
// - gate and confirmer: 2 cells (1 gcc_g90, 1 gcc_sm80), each confirmed in the gate's four environments (a-d;
// results/gcc/gate-tioga/ and gate-wingtip/ verdicts.jsonl); confirmer verdicts (results/gcc/confirm/batch.gcc_*.jsonl):
// compiler_bug (R11-compile-rejection) 2
// - catalog programs (1): acc34-c-clause-link
// - legality: the catalog programs are legal by construction (offload.engine.gen, c_gen and acc_gen
// check every program against the catalog rules and the offload scope as it is built; no
// compiler). GCC is a compiler under test only: never a legality witness, here or for another
// family's case (catalog/machines.yaml under_test_only).
// Generated by compliance.acc_c_gen (OpenACC 3.4, C11): the C port (offload.engine.f2c) of acc34-clause-link. Do not edit.
// id: acc34-c-clause-link
// primary atom: acc.clause.link
// legal: yes
// observable: yes
// ran_on_host: inside the compute region of interest (acc_on_device)
// evidence: construct
// needs: none
// std: c11
#include <stdio.h>
#include <stdbool.h>
#include <openacc.h>
enum { n = 64 };
int lv[64];
#pragma acc declare link(lv)
#pragma acc routine seq
int get_lv(int k) {
int r_;
r_ = lv[k - 1];
return r_;
}
int main(void) {
int a[n];
int b[n];
int ref[n];
int host_flag;
bool ok;
for (int i = 1; i <= n; i++) {
a[i - 1] = i;
b[i - 1] = 0;
ref[i - 1] = 2 * i + 1;
}
host_flag = 1;
for (int i = 1; i <= n; i++) {
lv[i - 1] = 2 * i + 1;
}
#pragma acc parallel loop copyin(a) copyout(b) copy(host_flag) copyin(lv)
for (int i = 1; i <= n; i++) {
if (i == 1) host_flag = (acc_on_device(acc_device_host) ? 1 : 0);
b[i - 1] = get_lv(a[i - 1]);
}
bool t1_ = true;
for (int k2_ = 0; k2_ < n; k2_++)
t1_ = t1_ && (b[k2_] == ref[k2_]);
ok = t1_;
printf("RESULT ran_on_host=%d check=%s\n", host_flag, ok ? "pass" : "fail");
return 0;
}
Something to add? Dispute its legality · Report it fixed in a newer release · Report a mistake on this page
Expected and observed
Expected
check=pass
Observed where it behaves incorrectly
site-gcc--gcc-acc-c-link-in-routine-rejected.c: In function ‘get_lv’:
site-gcc--gcc-acc-c-link-in-routine-rejected.c:40:5: error: ‘lv’ with ‘link’ clause used in ‘routine’ function
40 | int lv[64];
| ^~
site-gcc--gcc-acc-c-link-in-routine-rejected.c:40:5: error: ‘lv’ with ‘link’ clause used in ‘routine’ functionWhat the specification requires
The specification text this program relies on and the compiler is expected to support.
OpenACC 3.4 §2.13.3
The global variable or common block variables may be used in accelerator routines
OpenACC 3.4 §2.13.3