Fix bitmask upper bounds check

[jlpeyton]

Summary

Rather than checking KMP_CPU_SETSIZE, which doesn't exist when using Hwloc, we use the get_max_proc() function which can vary based on the operating system. For example on Windows with multiple processor groups, it might be the case that the highest bit possible in the bitmask is not equal to the number of hardware threads on the machine but something higher than that.

[AndreyChurbanov]

LGTM

openmp/trunk/runtime/src/kmp.h

 extern void __kmp_affinity_uninitialize(void);
 extern void __kmp_affinity_set_init_mask(int gtid, int isa_root); /* set affinity according to KMP_AFFINITY */
 #if OMP_40_ENABLED
 extern void __kmp_affinity_set_place(int gtid);
 #endif
 extern void __kmp_affinity_determine_capable( const char *env_var );
 extern int __kmp_aux_set_affinity(void **mask);
 extern int __kmp_aux_get_affinity(void **mask);
+extern int __kmp_aux_get_affinity_max_proc();
 extern int __kmp_aux_set_affinity_mask_proc(int proc, void **mask);
 extern int __kmp_aux_unset_affinity_mask_proc(int proc, void **mask);
 extern int __kmp_aux_get_affinity_mask_proc(int proc, void **mask);
 extern void __kmp_balanced_affinity( int tid, int team_size );
 #endif /* KMP_AFFINITY_SUPPORTED */
 
 extern void __kmp_cleanup_hierarchy();
 extern void __kmp_get_hierarchy(kmp_uint32 nproc, kmp_bstate_t *thr_bar);

openmp/trunk/runtime/src/kmp_affinity.cpp

# else
     KMP_CPU_COPY((kmp_affin_mask_t *)(*mask), th->th.th_affin_mask);
     return 0;
 
 # endif /* KMP_OS_WINDOWS */
 
 }
 
 int
+__kmp_aux_get_affinity_max_proc() {
+    if (!  KMP_AFFINITY_CAPABLE()) {
+        return 0;
+    }
+#if KMP_GROUP_AFFINITY
+    if ( __kmp_num_proc_groups > 1 ) {
+        return (int)(__kmp_num_proc_groups*sizeof(DWORD_PTR)*CHAR_BIT);
+    }
+#endif
+    return __kmp_xproc;
+}
+
+int
 __kmp_aux_set_affinity_mask_proc(int proc, void **mask)
 {
     int retval;
 
     if (! KMP_AFFINITY_CAPABLE()) {
         return -1;
     }
 
     KA_TRACE(1000, ;{
         int gtid = __kmp_entry_gtid();
         char buf[KMP_AFFIN_MASK_PRINT_LEN];
         __kmp_affinity_print_mask(buf, KMP_AFFIN_MASK_PRINT_LEN,
           (kmp_affin_mask_t *)(*mask));
         __kmp_debug_printf("kmp_set_affinity_mask_proc: setting proc %d in affinity mask for thread %d = %s\n",
           proc, gtid, buf);
     });
 
     if (__kmp_env_consistency_check) {
         if ((mask == NULL) || (*mask == NULL)) {
             KMP_FATAL(AffinityInvalidMask, "kmp_set_affinity_mask_proc");
         }
     }
 
-    if ((proc < 0)
-# if !KMP_USE_HWLOC
-         || ((unsigned)proc >= KMP_CPU_SETSIZE)
-# endif
-       ) {
+    if ((proc < 0) || (proc >= __kmp_aux_get_affinity_max_proc())) {
         return -1;
     }
     if (! KMP_CPU_ISSET(proc, __kmp_affin_fullMask)) {
         return -2;
     }
 
     KMP_CPU_SET(proc, (kmp_affin_mask_t *)(*mask));
     return 0;
__kmp_aux_unset_affinity_mask_proc(int proc, void **mask)
     });
 
     if (__kmp_env_consistency_check) {
         if ((mask == NULL) || (*mask == NULL)) {
             KMP_FATAL(AffinityInvalidMask, "kmp_unset_affinity_mask_proc");
         }
     }
 
-    if ((proc < 0)
-# if !KMP_USE_HWLOC
-         || ((unsigned)proc >= KMP_CPU_SETSIZE)
-# endif
-       ) {
+    if ((proc < 0) || (proc >= __kmp_aux_get_affinity_max_proc())) {
         return -1;
     }
     if (! KMP_CPU_ISSET(proc, __kmp_affin_fullMask)) {
         return -2;
     }
 
     KMP_CPU_CLR(proc, (kmp_affin_mask_t *)(*mask));
     return 0;
__kmp_aux_get_affinity_mask_proc(int proc, void **mask)
     });
 
     if (__kmp_env_consistency_check) {
         if ((mask == NULL) || (*mask == NULL)) {
             KMP_FATAL(AffinityInvalidMask, "kmp_get_affinity_mask_proc");
         }
     }
 
-    if ((proc < 0)
-# if !KMP_USE_HWLOC
-         || ((unsigned)proc >= KMP_CPU_SETSIZE)
-# endif
-       ) {
+    if ((proc < 0) || (proc >= __kmp_aux_get_affinity_max_proc())) {
         return -1;
     }
     if (! KMP_CPU_ISSET(proc, __kmp_affin_fullMask)) {
         return 0;
     }
 
     return KMP_CPU_ISSET(proc, (kmp_affin_mask_t *)(*mask));
 }

openmp/trunk/runtime/src/kmp_ftn_entry.h

#if defined(KMP_STUB) || !KMP_AFFINITY_SUPPORTED
         return 0;
     #else
         //
         // We really only NEED serial initialization here.
         //
         if ( ! TCR_4(__kmp_init_middle) ) {
             __kmp_middle_initialize();
         }
-        if ( ! ( KMP_AFFINITY_CAPABLE() ) ) {
-            return 0;
-        }
-
-    #if KMP_GROUP_AFFINITY
-        if ( __kmp_num_proc_groups > 1 ) {
-            return (int)(__kmp_num_proc_groups*sizeof(DWORD_PTR)*CHAR_BIT);
-        }
-    #endif /* KMP_GROUP_AFFINITY */
-        return __kmp_xproc;
+        return __kmp_aux_get_affinity_max_proc();
     #endif
 }
 
 void FTN_STDCALL
 FTN_CREATE_AFFINITY_MASK( void **mask )
 {
     #if defined(KMP_STUB) || !KMP_AFFINITY_SUPPORTED
         *mask = NULL;