[OpenMP][libomp] Add support for offline CPUs in Linux

[jlpeyton]

Summary

If some CPUs are offline, then make sure they are not included in the fullMask even if norespect is given to KMP_AFFINITY.

[AndreyChurbanov]

LGTM

openmp/runtime/src/kmp.h

if (!f) {
         __kmp_fatal(KMP_MSG(CantOpenFileForReading, filename), KMP_ERR(code),
                     KMP_HNT(CheckEnvVar, env_var, filename), __kmp_msg_null);
       } else {
         __kmp_fatal(KMP_MSG(CantOpenFileForReading, filename), KMP_ERR(code),
                     __kmp_msg_null);
       }
     }
   }
+  /// Instead of erroring out, return non-zero when
+  /// unsuccessful fopen() for any reason
+  int try_open(const char *filename, const char *mode) {
+    KMP_ASSERT(!f);
+    f = fopen(filename, mode);
+    if (!f)
+      return errno;
+    return 0;
+  }
   /// Set the FILE* object to stdout and output there
   /// No open call should happen before this call.
   void set_stdout() {
     KMP_ASSERT(!f);
     f = stdout;
   }
   /// Set the FILE* object to stderr and output there
   /// No open call should happen before this call.

openmp/runtime/src/kmp_affinity.cpp

 #endif
 #if KMP_USE_HWLOC
 // Copied from hwloc
 #define HWLOC_GROUP_KIND_INTEL_MODULE 102
 #define HWLOC_GROUP_KIND_INTEL_TILE 103
 #define HWLOC_GROUP_KIND_INTEL_DIE 104
 #define HWLOC_GROUP_KIND_WINDOWS_PROCESSOR_GROUP 220
 #endif
+#include <ctype.h>
 
 // The machine topology
 kmp_topology_t *__kmp_topology = nullptr;
 // KMP_HW_SUBSET environment variable
 kmp_hw_subset_t *__kmp_hw_subset = nullptr;
 
 // Store the real or imagined machine hierarchy here
 static hierarchy_info machine_hierarchy;
while (1) {
     // Start over with new start point
     start = finish;
     if (start == mask->end())
       break;
   }
   return buf;
 }
 
-void __kmp_affinity_entire_machine_mask(kmp_affin_mask_t *mask) {
+// Return (possibly empty) affinity mask representing the offline CPUs
+// Caller must free the mask
+kmp_affin_mask_t *__kmp_affinity_get_offline_cpus() {
+  kmp_affin_mask_t *offline;
+  KMP_CPU_ALLOC(offline);
+  KMP_CPU_ZERO(offline);
+#if KMP_OS_LINUX
+  int n, begin_cpu, end_cpu;
+  kmp_safe_raii_file_t offline_file;
+  auto skip_ws = [](FILE *f) {
+    int c;
+    do {
+      c = fgetc(f);
+    } while (isspace(c));
+    if (c != EOF)
+      ungetc(c, f);
+  };
+  // File contains CSV of integer ranges representing the offline CPUs
+  // e.g., 1,2,4-7,9,11-15
+  int status = offline_file.try_open("/sys/devices/system/cpu/offline", "r");
+  if (status != 0)
+    return offline;
+  while (!feof(offline_file)) {
+    skip_ws(offline_file);
+    n = fscanf(offline_file, "%d", &begin_cpu);
+    if (n != 1)
+      break;
+    skip_ws(offline_file);
+    int c = fgetc(offline_file);
+    if (c == EOF || c == ',') {
+      // Just single CPU
+      end_cpu = begin_cpu;
+    } else if (c == '-') {
+      // Range of CPUs
+      skip_ws(offline_file);
+      n = fscanf(offline_file, "%d", &end_cpu);
+      if (n != 1)
+        break;
+      skip_ws(offline_file);
+      c = fgetc(offline_file); // skip ','
+    } else {
+      // Syntax problem
+      break;
+    }
+    // Ensure a valid range of CPUs
+    if (begin_cpu < 0 || begin_cpu >= __kmp_xproc || end_cpu < 0 ||
+        end_cpu >= __kmp_xproc || begin_cpu > end_cpu) {
+      continue;
+    }
+    // Insert [begin_cpu, end_cpu] into offline mask
+    for (int cpu = begin_cpu; cpu <= end_cpu; ++cpu) {
+      KMP_CPU_SET(cpu, offline);
+    }
+  }
+#endif
+  return offline;
+}
+
+// Return the number of available procs
+int __kmp_affinity_entire_machine_mask(kmp_affin_mask_t *mask) {
+  int avail_proc = 0;
   KMP_CPU_ZERO(mask);
 
 #if KMP_GROUP_AFFINITY
 
   if (__kmp_num_proc_groups > 1) {
     int group;
     KMP_DEBUG_ASSERT(__kmp_GetActiveProcessorCount != NULL);
     for (group = 0; group < __kmp_num_proc_groups; group++) {
       int i;
       int num = __kmp_GetActiveProcessorCount(group);
       for (i = 0; i < num; i++) {
         KMP_CPU_SET(i + group * (CHAR_BIT * sizeof(DWORD_PTR)), mask);
+        avail_proc++;
       }
     }
   } else
 
 #endif /* KMP_GROUP_AFFINITY */
 
   {
     int proc;
+    kmp_affin_mask_t *offline_cpus = __kmp_affinity_get_offline_cpus();
     for (proc = 0; proc < __kmp_xproc; proc++) {
+      // Skip offline CPUs
+      if (KMP_CPU_ISSET(proc, offline_cpus))
+        continue;
       KMP_CPU_SET(proc, mask);
+      avail_proc++;
     }
+    KMP_CPU_FREE(offline_cpus);
   }
+
+  return avail_proc;
 }
 
 // All of the __kmp_affinity_create_*_map() routines should allocate the
 // internal topology object and set the layer ids for it.  Each routine
 // returns a boolean on whether it was successful at doing so.
 kmp_affin_mask_t *__kmp_affin_fullMask = NULL;
 
 #if KMP_USE_HWLOC
if (__kmp_affinity_respect_mask) {
       }
     } else {
       if (__kmp_affinity_verbose) {
         char buf[KMP_AFFIN_MASK_PRINT_LEN];
         __kmp_affinity_print_mask(buf, KMP_AFFIN_MASK_PRINT_LEN,
                                   __kmp_affin_fullMask);
         KMP_INFORM(InitOSProcSetNotRespect, "KMP_AFFINITY", buf);
       }
+      __kmp_avail_proc =
           __kmp_affinity_entire_machine_mask(__kmp_affin_fullMask);
-      __kmp_avail_proc = __kmp_xproc;
 #if KMP_OS_WINDOWS
       // Set the process affinity mask since threads' affinity
       // masks must be subset of process mask in Windows* OS
       __kmp_affin_fullMask->set_process_affinity(true);
 #endif
     }
   }
 

openmp/runtime/src/z_Linux_util.cpp

void __kmp_clear_system_time(void) {
   KMP_CHECK_SYSFAIL_ERRNO("gettimeofday", status);
   TIMEVAL_TO_TIMESPEC(&tval, &__kmp_sys_timer_data.start);
 }
 
 static int __kmp_get_xproc(void) {
 
   int r = 0;
 
-#if KMP_OS_LINUX || KMP_OS_DRAGONFLY || KMP_OS_FREEBSD || KMP_OS_NETBSD ||     \
-    KMP_OS_OPENBSD || KMP_OS_HURD
+#if KMP_OS_LINUX
+
+  __kmp_type_convert(sysconf(_SC_NPROCESSORS_CONF), &(r));
+
+#elif KMP_OS_DRAGONFLY || KMP_OS_FREEBSD || KMP_OS_NETBSD || KMP_OS_OPENBSD || \
+    KMP_OS_HURD
 
   __kmp_type_convert(sysconf(_SC_NPROCESSORS_ONLN), &(r));
 
 #elif KMP_OS_DARWIN
 
   // Bug C77011 High "OpenMP Threads and number of active cores".
 
   // Find the number of available CPUs.