OffTarget

Observed issues › nested/afar-cce-c-call-result-member-in-target-rejected

Appears not to supportRefused at compile time · C · OpenMP · Open

AFAR and CCE: member of a function result in a target region rejected

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

AFAR and CCE reject a member access through a function call result inside a target region.

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 24.3.0not supportednot supportedunknown
not run
not supportednot supportedn/an/a
CCE 19.0.0not supportednot supportedunknown
not run
unknown
not run
unknown
not run
n/an/a
CCE 20.0.2not supportednot supportedunknown
not run
unknown
not run
unknown
not run
n/an/a
CCE 21.0.3not supportednot supportedunknown
not run
unknown
not run
unknown
not run
n/an/a
CCE 21.0.4not supportednot supportedunknown
not run
unknown
not run
unknown
not run
n/an/a
NVHPC 25.11n/an/an/an/an/acorrectcorrect
NVHPC 26.9n/an/an/an/an/acorrectcorrect
GCC 16.2.0correctcorrectcorrectcorrectcorrectcorrectn/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
not supportedAFAR 24.3.0, MI300A unified
AFAR 24.3.0, MI300A copy
CCE 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
site-nested--afar-cce-c-call-result-member-in-target-rejected.c:48:50: error: expected addressable lvalue in 'map' clause
   48 |     for (int q = 0; q < 4; q++) tot += child(r)->a[q];
      |                                        ~~~~~~~~~~^
1 error generated.
not supportedAFAR 24.3.0, MI300X
AFAR 24.3.0, MI350X
show
site-nested--afar-cce-c-call-result-member-in-target-rejected.c:56:50: error: expected addressable lvalue in 'map' clause
   56 |     for (int q = 0; q < 4; q++) tot += child(r)->a[q];
      |                                        ~~~~~~~~~~^
1 error generated.

Reproduce

AFAR 24.3.0, MI300A unified

$AFAR/bin/amdclang -fopenmp --offload-arch=gfx942 -O2 repro.c -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
craycc -fopenmp -fopenmp-targets=amdgcn-amd-amdhsa -Xopenmp-target=amdgcn-amd-amdhsa -march=gfx942 -O2 repro.c -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
craycc -fopenmp -fopenmp-targets=amdgcn-amd-amdhsa -Xopenmp-target=amdgcn-amd-amdhsa -march=gfx942 -O2 repro.c -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
craycc -fopenmp -fopenmp-targets=amdgcn-amd-amdhsa -Xopenmp-target=amdgcn-amd-amdhsa -march=gfx942 -O2 repro.c -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
craycc -fopenmp -fopenmp-targets=amdgcn-amd-amdhsa -Xopenmp-target=amdgcn-amd-amdhsa -march=gfx942 -O2 repro.c -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
nvc -mp=gpu -gpu=cc80 -gpu=mem:separate -O2 repro.c -o repro
OMP_TARGET_OFFLOAD=mandatory ./repro
Program (62 lines)
// # Finding (nested-data): AFAR and CCE (clang front end) reject a member access through a call result inside a target region:
// # "expected addressable lvalue in 'map' clause"
//
// - toolchains and modes that fail: afar (AFAR 24.3.0), cce19 (CCE 19.0.0), cce20 (CCE 20.0.2), cce21 (CCE 21.0.3), cce21_4 (CCE
//   21.0.4); modes: unified, copy; -O2
// - build: amdclang -fopenmp --offload-arch=gfx942 -O2 (afar); craycc -fopenmp -fopenmp-targets=amdgcn-amd-amdhsa
//   -Xopenmp-target=amdgcn-amd-amdhsa -march=gfx942 -O2 (cce19, cce20, cce21, cce21_4)
// - failure: rejected
// - observed: `site-nested--afar-cce-c-call-result-member-in-target-rejected.c:48:50: error: expected addressable lvalue in 'map'
//   clause`; identical signature on 10 of 10 failing cells
// - expected: compiles and prints `tot= 10` with `check=pass`
// - rule: only variables referenced in a target construct get an implicit data-mapping attribute: "the first matching
//   rule from the following list determines the implicit data-mapping (or data-sharing) attribute for variables
//   referenced in a target construct" (5.2 §5.8.1 p.148 l.3-5); the member `child(r)->a` is a structure element reached
//   through a function result, not a variable, and r->nx->a[0:4] is mapped explicitly
// - same-compiler control: controls/p2-ctl-pointer-variable.c (one change from this file; correct on afar (AFAR 24.3.0), cce19 (CCE
//   19.0.0), cce20 (CCE 20.0.2), cce21 (CCE 21.0.3), cce21_4 (CCE 21.0.4); rows results/nested/red/red7.jsonl,
//   results/nested/red/tuo1.jsonl, results/nested/red/wgs.jsonl, results/nested/red/wnv.jsonl); controls/p3-ctl-direct-member.c
//   (one change from this file; correct on afar (AFAR 24.3.0), cce19 (CCE 19.0.0), cce20 (CCE 20.0.2), cce21 (CCE 21.0.3), cce21_4
//   (CCE 21.0.4); rows results/nested/red/red7.jsonl, results/nested/red/tuo1.jsonl, results/nested/red/wgs.jsonl,
//   results/nested/red/wnv.jsonl)
// - other families: nvhpc (NVHPC 25.11), nvhpc269 (NVHPC 26.9), gcc_g90 (GCC 16.2.0), gcc_sm80 (GCC 16.2.0) run the file correctly
//   (rows results/nested/red/red7.jsonl, results/nested/red/wgs.jsonl, results/nested/red/wnv.jsonl, site rows, check case
//   site-nested--afar-cce-c-call-result-member-in-target-rejected)
// - gate and confirmer: campaign gate of the programs below: 10 cells, 10 confirmed in the four environments (a-d; (c) on another
//   node or GPU; rows results/nested/big/gate_reruns.part.<host>.1040717.jsonl,
//   results/nested/big/gate_reruns.part.<host>.1284552.jsonl, results/nested/big/gate_reruns.part.<host>.1251749.jsonl,
//   results/nested/big/gate_reruns.part.<host>.1517583.jsonl, results/nested/big/gate_reruns.part.<host>.1434668.jsonl,
//   results/nested/big/rows.part.<host>.1956389.jsonl, results/nested/big/rows.part.<host>.1523848.jsonl,
//   results/nested/big/rows.part.<host>.1239313.jsonl, results/nested/big/rows.part.<host>.1197409.jsonl,
//   results/nested/big/rows.part.<host>.1146311.jsonl, results/nested/main/gate_reruns.part.<host>.961942.jsonl,
//   results/nested/main/gate_reruns.part.<host>.919396.jsonl,
//   results/nested/main/gate_reruns.part.<host>.3910568.jsonl,
//   results/nested/main/gate_reruns.part.<host>.4072524.jsonl, results/nested/main/rows.part.<host>.994167.jsonl)
// - programs (1): nd-fnres-c-child-pointer
// - legality: the rules (offload.engine.rules) accept it; GCC compiles and runs it; the clang front ends of AFAR and CCE are
//   the two compilers that reject it (they share a lineage, so neither witnesses the other)
#include <stdio.h>
#include <stdlib.h>
#include <omp.h>
struct n { int *a; struct n *nx; };
#pragma omp declare target
static struct n *child(struct n *r) { return r->nx; }
#pragma omp end declare target
int main(void)
{
  int hf = 1, i, tot = 0, ok;
  struct n *r = calloc(1, sizeof *r);
  r->nx = calloc(1, sizeof *r->nx);
  r->nx->a = malloc(4 * sizeof(int));
  for (i = 0; i < 4; i++) r->nx->a[i] = i + 1;
  #pragma omp target map(from: hf)
  hf = omp_is_initial_device() ? 1 : 0;
  #pragma omp target map(tofrom: tot) map(to: r[0:1]) map(to: r->nx[0:1]) map(to: r->nx->a[0:4])
  {
    for (int q = 0; q < 4; q++) tot += child(r)->a[q];
  }
  ok = tot == 10;
  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

site-nested--afar-cce-c-call-result-member-in-target-rejected.c:48:50: error: expected addressable lvalue in 'map' clause
   48 |     for (int q = 0; q < 4; q++) tot += child(r)->a[q];
      |                                        ~~~~~~~~~~^
1 error generated.

What the specification requires

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

OpenMP 5.2 §5.8.1 p.148 l.3-5

the first matching rule from the following list determines the implicit data-mapping (or data-sharing) attribute for variables referenced in a target construct

Evidence

Same compiler, one change

  • one change from the finding; correct on afar, cce19, cce20, cce21, cce21_4: not run yet
    Control program
    // control: the returned pointer kept in a local variable
    #include <stdio.h>
    #include <stdlib.h>
    #include <omp.h>
    struct n { int *a; struct n *nx; };
    #pragma omp declare target
    static struct n *child(struct n *r) { return r->nx; }
    #pragma omp end declare target
    int main(void)
    {
      int hf = 1, i, tot = 0, ok;
      struct n *r = calloc(1, sizeof *r);
      r->nx = calloc(1, sizeof *r->nx);
      r->nx->a = malloc(4 * sizeof(int));
      for (i = 0; i < 4; i++) r->nx->a[i] = i + 1;
      #pragma omp target map(from: hf)
      hf = omp_is_initial_device() ? 1 : 0;
      #pragma omp target map(tofrom: tot) map(to: r[0:1]) map(to: r->nx[0:1]) map(to: r->nx->a[0:4])
      {
        struct n *c = child(r);
        for (int q = 0; q < 4; q++) tot += c->a[q];
      }
      ok = tot == 10;
      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 afar, cce19, cce20, cce21, cce21_4: not run yet
    Control program
    // control: the member reached without the call
    #include <stdio.h>
    #include <stdlib.h>
    #include <omp.h>
    struct n { int *a; struct n *nx; };
    #pragma omp declare target
    static struct n *child(struct n *r) { return r->nx; }
    #pragma omp end declare target
    int main(void)
    {
      int hf = 1, i, tot = 0, ok;
      struct n *r = calloc(1, sizeof *r);
      r->nx = calloc(1, sizeof *r->nx);
      r->nx->a = malloc(4 * sizeof(int));
      for (i = 0; i < 4; i++) r->nx->a[i] = i + 1;
      #pragma omp target map(from: hf)
      hf = omp_is_initial_device() ? 1 : 0;
      #pragma omp target map(tofrom: tot) map(to: r[0:1]) map(to: r->nx[0:1]) map(to: r->nx->a[0:4])
      {
        for (int q = 0; q < 4; q++) tot += r->nx->a[q];
      }
      ok = tot == 10;
      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, GCC 16.2.0