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
|
/*
* Copyright (C) 2010. See COPYRIGHT in top-level directory.
*/
#include <stdio.h>
#include <stdlib.h>
#include <assert.h>
#include <string.h>
#include <mpi.h>
#include <armci.h>
#ifdef MODE_SET
#include <armcix.h>
#endif
#define MAX_DATA_SIZE (1024*128*16)
#define NUM_ITERATIONS ((data_size <= 2048) ? 4096 : ((data_size <= 16384) ? 1024 : 512))
#define NUM_WARMUP_ITER 1
int main(int argc, char ** argv) {
int rank, nproc, test_iter, target_rank, data_size;
#ifdef MULTIPLE
inr thread_level;
#endif
int *buf;
void **base_ptrs;
#ifdef MODE_SET
ARMCI_Group grp_world;
#endif
#ifdef MULTIPLE
MPI_Init_thread(&argc, &argv, MPI_THREAD_MULTIPLE, &thread_level);
#else
MPI_Init(&argc, &argv);
#endif
ARMCI_Init();
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &nproc);
if (rank == 0) printf("Starting one-sided contiguous performance test with %d processes\n", nproc);
buf = ARMCI_Malloc_local(MAX_DATA_SIZE);
base_ptrs = malloc(sizeof(void*)*nproc);
ARMCI_Malloc(base_ptrs, MAX_DATA_SIZE);
memset(buf, rank, MAX_DATA_SIZE);
#ifdef MODE_SET
ARMCI_Group_get_default(&grp_world);
if (getenv("ARMCIX_MODE_SET"))
ARMCIX_Mode_set(ARMCIX_MODE_CONFLICT_FREE | ARMCIX_MODE_NO_LOAD_STORE, base_ptrs[rank], &grp_world);
else if (rank == 0)
printf("Warning: ARMCIX_MODE_SET not enabled\n");
#endif
if (rank == 0)
printf("%12s %12s %12s %12s %12s %12s %12s %12s\n", "Trg. Rank", "Xfer Size",
"Get (usec)", "Put (usec)", "Acc (usec)",
"Get (MiB/s)", "Put (MiB/s)", "Acc (MiB/s)");
for (target_rank = 1; rank == 0 && target_rank < nproc; target_rank++) {
for (data_size = sizeof(double); data_size <= MAX_DATA_SIZE; data_size *= 2) {
double t_get=0, t_put=0, t_acc=0;
for (test_iter = 0; test_iter < NUM_ITERATIONS + NUM_WARMUP_ITER; test_iter++) {
if (test_iter == NUM_WARMUP_ITER)
t_get = MPI_Wtime();
ARMCI_Get(base_ptrs[target_rank], buf, data_size, target_rank);
}
t_get = (MPI_Wtime() - t_get)/NUM_ITERATIONS;
for (test_iter = 0; test_iter < NUM_ITERATIONS + NUM_WARMUP_ITER; test_iter++) {
if (test_iter == NUM_WARMUP_ITER)
t_put = MPI_Wtime();
ARMCI_Put(buf, base_ptrs[target_rank], data_size, target_rank);
}
ARMCI_Fence(target_rank);
t_put = (MPI_Wtime() - t_put)/NUM_ITERATIONS;
for (test_iter = 0; test_iter < NUM_ITERATIONS + NUM_WARMUP_ITER; test_iter++) {
double scale = 1.0;
if (test_iter == NUM_WARMUP_ITER)
t_acc = MPI_Wtime();
#ifdef NO_ACC
int stride = 0;
ARMCI_AccS(ARMCI_ACC_DBL, &scale, buf, &stride, base_ptrs[target_rank], &stride, &data_size, 0, target_rank);
#else
ARMCI_Acc(ARMCI_ACC_DBL, &scale, buf, base_ptrs[target_rank], data_size, target_rank);
#endif
}
ARMCI_Fence(target_rank);
t_acc = (MPI_Wtime() - t_acc)/NUM_ITERATIONS;
printf("%12d %12d %12.3f %12.3f %12.3f %12.3f %12.3f %12.3f\n", target_rank, data_size,
t_get*1.0e6, t_put*1.0e6, t_acc*1.0e6, data_size/(1024.0*1024.0)/t_get, data_size/(1024.0*1024.0)/t_put, data_size/(1024.0*1024.0)/t_acc);
}
}
ARMCI_Barrier();
ARMCI_Free(base_ptrs[rank]);
ARMCI_Free_local(buf);
free(base_ptrs);
ARMCI_Finalize();
MPI_Finalize();
return 0;
}
|
