1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236
|
/*
* Copyright (C) by Argonne National Laboratory
* See COPYRIGHT in top-level directory
*/
#include "mpi.h"
#include <stdio.h>
#include <stdlib.h>
#include "mpitest.h"
#include <assert.h>
/*
static char MTEST_Descrip[] = "Test MPI_Allreduce with non-commutative user-defined operations using matrix rotations";
*/
/* This example is similar to allred3.c, but uses only 3x3 matrics with
integer-valued entries. This is an associative but not commutative
operation.
The number of matrices is the count argument. The matrix is stored
in C order, so that
c(i,j) is cin[j+i*3]
Three different matrices are used:
I = identity matrix
A = (1 0 0 B = (0 1 0
0 0 1 1 0 0
0 1 0) 0 0 1)
The product
I^k A I^(p-2-k-j) B I^j
is
(0 1 0
0 0 1
1 0 0)
for all values of k, p, and j.
*/
void matmult(void *cinPtr, void *coutPtr, int *count, MPI_Datatype * dtype);
void matmult(void *cinPtr, void *coutPtr, int *count, MPI_Datatype * dtype)
{
const int *cin = (const int *) cinPtr;
int *cout = (int *) coutPtr;
int i, j, k, nmat;
int tempcol[3];
int offset1, offset2;
for (nmat = 0; nmat < *count; nmat++) {
for (j = 0; j < 3; j++) {
for (i = 0; i < 3; i++) {
tempcol[i] = 0;
for (k = 0; k < 3; k++) {
/* col[i] += cin(i,k) * cout(k,j) */
offset1 = k + i * 3;
offset2 = j + k * 3;
tempcol[i] += cin[offset1] * cout[offset2];
}
}
for (i = 0; i < 3; i++) {
offset1 = j + i * 3;
cout[offset1] = tempcol[i];
}
}
/* Advance to the next matrix */
cin += 9;
cout += 9;
}
}
/* Initialize the integer matrix as one of the
above matrix entries, as a function of count.
We guarantee that both the A and B matrices are included.
*/
static void initMat(int rank, int size, int nmat, int mat[])
{
int i, kind;
/* Zero the matrix */
for (i = 0; i < 9; i++) {
mat[i] = 0;
}
/* Decide which matrix to create (I, A, or B) */
if (size == 2) {
/* rank 0 is A, 1 is B */
kind = 1 + rank;
} else {
int tmpA, tmpB;
/* Most ranks are identity matrices */
kind = 0;
/* Make sure exactly one rank gets the A matrix
* and one the B matrix */
tmpA = size / 4;
tmpB = (3 * size) / 4;
if (rank == tmpA)
kind = 1;
if (rank == tmpB)
kind = 2;
}
switch (kind) {
case 0: /* Identity */
mat[0] = 1;
mat[4] = 1;
mat[8] = 1;
break;
case 1: /* A */
mat[0] = 1;
mat[5] = 1;
mat[7] = 1;
break;
case 2: /* B */
mat[1] = 1;
mat[3] = 1;
mat[8] = 1;
break;
}
}
/* Compare a matrix with the known result */
static int checkResult(int nmat, int mat[], const char *msg)
{
int n, k, errs = 0, wrank;
static int solution[9] = { 0, 1, 0,
0, 0, 1,
1, 0, 0
};
MPI_Comm_rank(MPI_COMM_WORLD, &wrank);
for (n = 0; n < nmat; n++) {
for (k = 0; k < 9; k++) {
if (mat[k] != solution[k]) {
errs++;
if (errs == 1) {
printf("Errors for communicators %s\n", MTestGetIntracommName());
fflush(stdout);
}
if (errs < 10) {
printf("[%d]matrix #%d(%s): Expected mat[%d,%d] = %d, got %d\n",
wrank, n, msg, k / 3, k % 3, solution[k], mat[k]);
fflush(stdout);
}
}
}
/* Advance to the next matrix */
mat += 9;
}
return errs;
}
int main(int argc, char *argv[])
{
int errs = 0;
int size, rank;
int minsize = 2, count;
MPI_Comm comm;
int *buf, *bufout;
MPI_Op op;
MPI_Datatype mattype;
int i;
MTest_Init(&argc, &argv);
MPI_Op_create(matmult, 0, &op);
/* A single rotation matrix (3x3, stored as 9 consecutive elements) */
MPI_Type_contiguous(9, MPI_INT, &mattype);
MPI_Type_commit(&mattype);
/* Sanity check: test that our routines work properly */
{
int one = 1;
buf = (int *) malloc(4 * 9 * sizeof(int));
initMat(0, 4, 0, &buf[0]);
initMat(1, 4, 0, &buf[9]);
initMat(2, 4, 0, &buf[18]);
initMat(3, 4, 0, &buf[27]);
matmult(&buf[0], &buf[9], &one, &mattype);
matmult(&buf[9], &buf[18], &one, &mattype);
matmult(&buf[18], &buf[27], &one, &mattype);
checkResult(1, &buf[27], "Sanity Check");
free(buf);
}
while (MTestGetIntracommGeneral(&comm, minsize, 1)) {
if (comm == MPI_COMM_NULL)
continue;
MPI_Comm_size(comm, &size);
MPI_Comm_rank(comm, &rank);
for (count = 1; count < size; count++) {
/* Allocate the matrices */
buf = (int *) malloc(count * 9 * sizeof(int));
if (!buf) {
MPI_Abort(MPI_COMM_WORLD, 1);
}
bufout = (int *) malloc(count * 9 * sizeof(int));
if (!bufout) {
MPI_Abort(MPI_COMM_WORLD, 1);
}
for (i = 0; i < count; i++) {
initMat(rank, size, i, &buf[i * 9]);
}
MPI_Allreduce(buf, bufout, count, mattype, op, comm);
errs += checkResult(count, bufout, "");
/* Try the same test, but using MPI_IN_PLACE */
for (i = 0; i < count; i++) {
initMat(rank, size, i, &bufout[i * 9]);
}
MPI_Allreduce(MPI_IN_PLACE, bufout, count, mattype, op, comm);
errs += checkResult(count, bufout, "IN_PLACE");
free(buf);
free(bufout);
}
MTestFreeComm(&comm);
}
MPI_Op_free(&op);
MPI_Type_free(&mattype);
MTest_Finalize(errs);
return MTestReturnValue(errs);
}
|