[OpenMP] Fix dynamic schedule assertion failure on AArch64

[Nawrin]

Summary

This patch fixes assertion failure for dynamic schedule with 8 byte induction variable. It checks the steal lock of selected victim and if the lock is not initialized then pick a different victim to steal from.

The bug report: https://github.com/llvm/llvm-project/issues/54422#issuecomment-1274020428

[AndreyChurbanov]

openmp/runtime/src/kmp_dispatch.cpp

1293

This change fights with the consequence of the problem, not with the root cause. IMHO, the better fix would be to change

if (KMP_ATOMIC_LD_RLX(&v->steal_flag) != READY ||

to

if (KMP_ATOMIC_LD_ACQ(&v->steal_flag) != READY ||

Then the acquire load would synchronize with release store at buffer initialization, that should prevent the steal_lock pointer to ever be NULL here.

[AndreyChurbanov]

LGTM.

This patch fixes the issue reported in https://github.com/llvm/llvm-project/issues/54422 for me.

openmp/runtime/src/kmp_dispatch.cpp

#ifdef KMP_DEBUG
               }
 #endif
               // activate non-empty buffer and let others steal from us
               if (pr->u.p.count < (UT)pr->u.p.ub)
                 KMP_ATOMIC_ST_REL(&pr->steal_flag, READY);
               break;
             }
           }
-          if (KMP_ATOMIC_LD_RLX(&v->steal_flag) != READY ||
+          if (KMP_ATOMIC_LD_ACQ(&v->steal_flag) != READY ||
               v->u.p.count >= (UT)v->u.p.ub) {
             pr->u.p.parm4 = (victimId + 1) % nproc; // shift start victim tid
             continue; // no chunks to steal, try next victim
           }
           lckv = v->u.p.steal_lock;
           KMP_ASSERT(lckv != NULL);
           __kmp_acquire_lock(lckv, gtid);
           limit = v->u.p.ub; // keep initial ub
           if (v->u.p.count >= limit) {