remove test for concurrent MPI_Waitany, this is WAI.

tests/testParallelMpiWaitany.c

-// Test MPI_Waitany in parallel.
-//
-// This test has each rank send each other rank an array of random ints with
-// asynchronous sends and receives. The program spawns as many threads
-// as it has incoming receives, which all call MPI_Waitany concurrently.
-//
-// Compile with:
-//   mpicc testParallelMpiWaitany.c -g -Wall
-//
-// Run with:
-//    mpirun --oversubscribe -n 100 ./a.out
-
-#include <mpi.h>
-#include <pthread.h>
-#include <stdio.h>
-#include <stdlib.h>
-
-#define BUFFER_SIZE 1000
-
-struct thread_func_data {
-  int num_ranks;
-  MPI_Request *requests;
-};
-
-void *thread_func(void *data_in) {
-  const struct thread_func_data *data = (struct thread_func_data *)data_in;
-
-  // Wait on any of the requests.
-  int index = MPI_UNDEFINED;
-  MPI_Status status;
-  const int res = MPI_Waitany(data->num_ranks, data->requests, &index, &status);
-
-  // Make sure all went well.
-  if (res != MPI_SUCCESS) {
-    int len = 1024;
-    char error_message[len];
-    MPI_Error_string(res, error_message, &len);
-    fprintf(stderr, "MPI_Waitany failed: %s\n", error_message);
-  }
-  if (index == MPI_UNDEFINED) {
-    fprintf(stderr, "MPI_Waitany didn't return a valid index.\n");
-  }
-
-  return NULL;
-}
-
-int main(int argc, char **argv) {
-  // Initialize the MPI environment.
-  int provided;
-  MPI_Init_thread(&argc, &argv, MPI_THREAD_MULTIPLE, &provided);
-  if (provided != MPI_THREAD_MULTIPLE) {
-    fprintf(stderr, "MPI_Init_thread does not provide MPI_THREAD_MULTIPLE.\n");
-    abort();
-  }
-
-  // Get the number of processes.
-  int num_ranks;
-  MPI_Comm_size(MPI_COMM_WORLD, &num_ranks);
-
-  // Get the rank of the process.
-  int rank;
-  MPI_Comm_rank(MPI_COMM_WORLD, &rank);
-  printf("Rank %02i is ready.\n", rank);
-
-  // Create a data buffers, fill the outgoing one with random numbers.
-  int *data_out = malloc(sizeof(int) * BUFFER_SIZE);
-  for (int k = 0; k < BUFFER_SIZE; k++) {
-    data_out[k] = rand();
-  }
-  int *data_in = malloc(sizeof(int) * BUFFER_SIZE);
-
-  // Create the request objects.
-  MPI_Request requests_in[num_ranks];
-  MPI_Request requests_out[num_ranks];
-
-  // Asynchronously send the buffer to all other ranks.
-  for (int rid = 0; rid < num_ranks; rid++) {
-    if (rid != rank) {
-      MPI_Isend(data_out, BUFFER_SIZE, MPI_INT, rid, /*tag=*/0, MPI_COMM_WORLD,
-                &requests_out[rid]);
-    } else {
-      requests_out[rid] = MPI_REQUEST_NULL;
-    }
-  }
-
-  // Launch the asynchronous receives.
-  for (int rid = 0; rid < num_ranks; rid++) {
-    if (rid != rank) {
-      MPI_Irecv(data_in, BUFFER_SIZE, MPI_INT, rid, /*tag=*/0, MPI_COMM_WORLD,
-                &requests_in[rid]);
-    } else {
-      requests_in[rid] = MPI_REQUEST_NULL;
-    }
-  }
-
-  // Launch the threads that will call MPI_Waitany.
-  struct thread_func_data thread_data = {num_ranks, requests_in};
-  pthread_t threads[num_ranks - 1];
-  for (int tid = 0; tid < num_ranks - 1; tid++) {
-    pthread_create(&threads[tid], /*attr=*/NULL, thread_func, &thread_data);
-  }
-
-  // Wait for all the threads to be done.
-  for (int tid = 0; tid < num_ranks - 1; tid++) {
-    pthread_join(threads[tid], /*value_ptr=*/NULL);
-  }
-
-  // Just to be sure, make sure the sends are all done.
-  MPI_Waitall(num_ranks, requests_out, MPI_STATUSES_IGNORE);
-
-  // Clean up allocated memory.
-  free(data_in);
-  free(data_out);
-
-  // Finalize the MPI environment.
-  MPI_Finalize();
-  printf("Rank %02i is done.\n", rank);
-}