runtime/src/ompt-specific.cpp

//******************************************************************************
// include files
//******************************************************************************

#include "kmp.h"
#include "ompt-internal.h"
#include "ompt-specific.h"

//******************************************************************************
// macros
//******************************************************************************

#define GTID_TO_OMPT_THREAD_ID(id) ((ompt_thread_id_t)(id >= 0) ? id + 1 : 0)

#define LWT_FROM_TEAM(team) (team)->t.ompt_serialized_team_info;

#define OMPT_THREAD_ID_BITS 16

// 2013 08 24 - John Mellor-Crummey
//   ideally, a thread should assign its own ids based on thread private data.
//   however, the way the intel runtime reinitializes thread data structures
//   when it creates teams makes it difficult to maintain persistent thread
//   data. using a shared variable instead is simple. I leave it to intel to
//   sort out how to implement a higher performance version in their runtime.

// when using fetch_and_add to generate the IDs, there isn't any reason to waste
// bits for thread id.
#if 0
#define NEXT_ID(id_ptr, tid)                                                   \
  ((KMP_TEST_THEN_INC64(id_ptr) << OMPT_THREAD_ID_BITS) | (tid))
#else
#define NEXT_ID(id_ptr, tid) (KMP_TEST_THEN_INC64((volatile kmp_int64 *)id_ptr))
#endif

//******************************************************************************
// private operations
//******************************************************************************

//----------------------------------------------------------
// traverse the team and task hierarchy
// note: __ompt_get_teaminfo and __ompt_get_taskinfo
//       traverse the hierarchy similarly and need to be
//       kept consistent
//----------------------------------------------------------

ompt_team_info_t *__ompt_get_teaminfo(int depth, int *size) {
  kmp_info_t *thr = ompt_get_thread();

  if (thr) {
    kmp_team *team = thr->th.th_team;
    if (team == NULL)
      return NULL;

    ompt_lw_taskteam_t *lwt = LWT_FROM_TEAM(team);

    while (depth > 0) {
      // next lightweight team (if any)
      if (lwt)
        lwt = lwt->parent;

      // next heavyweight team (if any) after
      // lightweight teams are exhausted
      if (!lwt && team) {
        team = team->t.t_parent;
        if (team) {
          lwt = LWT_FROM_TEAM(team);
        }
      }

      depth--;
    }

    if (lwt) {
      // lightweight teams have one task
      if (size)
        *size = 1;

      // return team info for lightweight team
      return &lwt->ompt_team_info;
    } else if (team) {
      // extract size from heavyweight team
      if (size)
        *size = team->t.t_nproc;

      // return team info for heavyweight team
      return &team->t.ompt_team_info;
    }
  }

  return NULL;
}

ompt_task_info_t *__ompt_get_taskinfo(int depth) {
  ompt_task_info_t *info = NULL;
  kmp_info_t *thr = ompt_get_thread();

  if (thr) {
    kmp_taskdata_t *taskdata = thr->th.th_current_task;
    ompt_lw_taskteam_t *lwt = LWT_FROM_TEAM(taskdata->td_team);

    while (depth > 0) {
      // next lightweight team (if any)
      if (lwt)
        lwt = lwt->parent;

      // next heavyweight team (if any) after
      // lightweight teams are exhausted
      if (!lwt && taskdata) {
        taskdata = taskdata->td_parent;
        if (taskdata) {
          lwt = LWT_FROM_TEAM(taskdata->td_team);
        }
      }
      depth--;
    }

    if (lwt) {
      info = &lwt->ompt_task_info;
    } else if (taskdata) {
      info = &taskdata->ompt_task_info;
    }
  }

  return info;
}

//******************************************************************************
// interface operations
//******************************************************************************

//----------------------------------------------------------
// thread support
//----------------------------------------------------------

ompt_parallel_id_t __ompt_thread_id_new() {
  static uint64_t ompt_thread_id = 1;
  return NEXT_ID(&ompt_thread_id, 0);
}

void __ompt_thread_begin(ompt_thread_type_t thread_type, int gtid) {
  ompt_callbacks.ompt_callback(ompt_event_thread_begin)(
      thread_type, GTID_TO_OMPT_THREAD_ID(gtid));
}

void __ompt_thread_end(ompt_thread_type_t thread_type, int gtid) {
  ompt_callbacks.ompt_callback(ompt_event_thread_end)(
      thread_type, GTID_TO_OMPT_THREAD_ID(gtid));
}

ompt_thread_id_t __ompt_get_thread_id_internal() {
  // FIXME: until we have a better way of assigning ids, use __kmp_get_gtid
  // since the return value might be negative, we need to test that before
  // assigning it to an ompt_thread_id_t, which is unsigned.
  int id = __kmp_get_gtid();
  assert(id >= 0);

  return GTID_TO_OMPT_THREAD_ID(id);
}

//----------------------------------------------------------
// state support
//----------------------------------------------------------

void __ompt_thread_assign_wait_id(void *variable) {
  int gtid = __kmp_gtid_get_specific();
  kmp_info_t *ti = ompt_get_thread_gtid(gtid);

  ti->th.ompt_thread_info.wait_id = (ompt_wait_id_t)variable;
}

ompt_state_t __ompt_get_state_internal(ompt_wait_id_t *ompt_wait_id) {
  kmp_info_t *ti = ompt_get_thread();

  if (ti) {
    if (ompt_wait_id)
      *ompt_wait_id = ti->th.ompt_thread_info.wait_id;
    return ti->th.ompt_thread_info.state;
  }
  return ompt_state_undefined;
}

//----------------------------------------------------------
// idle frame support
//----------------------------------------------------------

void *__ompt_get_idle_frame_internal(void) {
  kmp_info_t *ti = ompt_get_thread();
  return ti ? ti->th.ompt_thread_info.idle_frame : NULL;
}

//----------------------------------------------------------
// parallel region support
//----------------------------------------------------------

ompt_parallel_id_t __ompt_parallel_id_new(int gtid) {
  static uint64_t ompt_parallel_id = 1;
  return gtid >= 0 ? NEXT_ID(&ompt_parallel_id, gtid) : 0;
}

void *__ompt_get_parallel_function_internal(int depth) {
  ompt_team_info_t *info = __ompt_get_teaminfo(depth, NULL);
  void *function = info ? info->microtask : NULL;
  return function;
}

ompt_parallel_id_t __ompt_get_parallel_id_internal(int depth) {
  ompt_team_info_t *info = __ompt_get_teaminfo(depth, NULL);
  ompt_parallel_id_t id = info ? info->parallel_id : 0;
  return id;
}

int __ompt_get_parallel_team_size_internal(int depth) {
  // initialize the return value with the error value.
  // if there is a team at the specified depth, the default
  // value will be overwritten the size of that team.
  int size = -1;
  (void)__ompt_get_teaminfo(depth, &size);
  return size;
}

//----------------------------------------------------------
// lightweight task team support
//----------------------------------------------------------

void __ompt_lw_taskteam_init(ompt_lw_taskteam_t *lwt, kmp_info_t *thr, int gtid,
                             void *microtask, ompt_parallel_id_t ompt_pid) {
  lwt->ompt_team_info.parallel_id = ompt_pid;
  lwt->ompt_team_info.microtask = microtask;
  lwt->ompt_task_info.task_id = 0;
  lwt->ompt_task_info.frame.reenter_runtime_frame = NULL;
  lwt->ompt_task_info.frame.exit_runtime_frame = NULL;
  lwt->ompt_task_info.function = NULL;
  lwt->parent = 0;
}

void __ompt_lw_taskteam_link(ompt_lw_taskteam_t *lwt, kmp_info_t *thr) {
  ompt_lw_taskteam_t *my_parent = thr->th.th_team->t.ompt_serialized_team_info;
  lwt->parent = my_parent;
  thr->th.th_team->t.ompt_serialized_team_info = lwt;
}

ompt_lw_taskteam_t *__ompt_lw_taskteam_unlink(kmp_info_t *thr) {
  ompt_lw_taskteam_t *lwtask = thr->th.th_team->t.ompt_serialized_team_info;
  if (lwtask)
    thr->th.th_team->t.ompt_serialized_team_info = lwtask->parent;
  return lwtask;
}

//----------------------------------------------------------
// task support
//----------------------------------------------------------

ompt_task_id_t __ompt_task_id_new(int gtid) {
  static uint64_t ompt_task_id = 1;
  return NEXT_ID(&ompt_task_id, gtid);
}

ompt_task_id_t __ompt_get_task_id_internal(int depth) {
  ompt_task_info_t *info = __ompt_get_taskinfo(depth);
  ompt_task_id_t task_id = info ? info->task_id : 0;
  return task_id;
}

void *__ompt_get_task_function_internal(int depth) {
  ompt_task_info_t *info = __ompt_get_taskinfo(depth);
  void *function = info ? info->function : NULL;
  return function;
}

ompt_frame_t *__ompt_get_task_frame_internal(int depth) {
  ompt_task_info_t *info = __ompt_get_taskinfo(depth);
  ompt_frame_t *frame = info ? frame = &info->frame : NULL;
  return frame;
}

//----------------------------------------------------------
// team support
//----------------------------------------------------------

void __ompt_team_assign_id(kmp_team_t *team, ompt_parallel_id_t ompt_pid) {
  team->t.ompt_team_info.parallel_id = ompt_pid;
}