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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
|
/*
* test_client_obj.c: Client side of H5O routines
*/
#include <stdio.h>
#include <stdlib.h>
#include <assert.h>
#include <string.h>
#include "mpi.h"
#include "hdf5.h"
int main(int argc, char **argv) {
const char file_name[]="eff_file_obj.h5";
hid_t file_id;
hid_t gid;
hid_t did, map;
hid_t sid, dtid;
hid_t tid1, tid2, rid1, rid2;
hid_t fapl_id, dxpl_id;
hid_t e_stack;
hbool_t exists = -1;
const unsigned int nelem=60;
hsize_t dims[1];
uint64_t version;
uint64_t trans_num;
int my_rank, my_size;
int provided;
MPI_Request mpi_req;
int32_t *wdata1 = NULL, *rdata1 = NULL;
int key, value, i;
H5ES_status_t status;
size_t num_events = 0;
herr_t ret;
MPI_Init_thread(&argc, &argv, MPI_THREAD_MULTIPLE, &provided);
if(MPI_THREAD_MULTIPLE != provided) {
fprintf(stderr, "MPI does not have MPI_THREAD_MULTIPLE support\n");
exit(1);
}
/* Call EFF_init to initialize the EFF stack.
As a result of this call, the Function Shipper client is started,
and HDF5 VOL calls are registered with the function shipper.
An "IOD init" call is forwarded from the FS client to the FS server
which should already be running. */
EFF_init(MPI_COMM_WORLD, MPI_INFO_NULL);
MPI_Comm_rank(MPI_COMM_WORLD, &my_rank);
MPI_Comm_size(MPI_COMM_WORLD, &my_size);
fprintf(stderr, "APP processes = %d, my rank is %d\n", my_size, my_rank);
wdata1 = malloc (sizeof(int32_t)*nelem);
rdata1 = malloc (sizeof(int32_t)*nelem);
for(i=0;i<nelem;++i) {
wdata1[i]=i;
rdata1[i] = 0;
}
fprintf(stderr, "Create the FAPL to set the IOD VOL plugin and create the file\n");
/* Choose the IOD VOL plugin to use with this file.
First we create a file access property list. Then we call a new routine to set
the IOD plugin to use with this fapl */
fapl_id = H5Pcreate (H5P_FILE_ACCESS);
H5Pset_fapl_iod(fapl_id, MPI_COMM_WORLD, MPI_INFO_NULL);
/* create an event Queue for managing asynchronous requests.
Event Queues will releive the use from managing and completing
individual requests for every operation. Instead of passing a
request for every operation, the event queue is passed and
internally the HDF5 library creates a request and adds it to
the event queue.
Multiple Event queue can be created used by the application. */
e_stack = H5EScreate();
assert(e_stack);
/* create the file. */
file_id = H5Fcreate_ff(file_name, H5F_ACC_TRUNC, H5P_DEFAULT, fapl_id, H5_EVENT_STACK_NULL);
assert(file_id > 0);
/* create 1-D dataspace with 60 elements */
dims [0] = nelem;
sid = H5Screate_simple(1, dims, NULL);
dtid = H5Tcopy(H5T_STD_I32LE);
/* acquire container version 0 - EXACT.
This can be asynchronous, but here we need the acquired ID
right after the call to start the transaction so we make synchronous. */
version = 0;
rid1 = H5RCacquire(file_id, &version, H5P_DEFAULT, H5_EVENT_STACK_NULL);
assert(0 == version);
/* start transaction 1 with default Leader/Delegate model. Leader
which is rank 0 here starts the transaction. It can be
asynchronous, but we make it synchronous here so that the
Leader can tell its delegates that the transaction is
started. */
if(0 == my_rank) {
hid_t rid_temp;
/* create transaction object */
tid1 = H5TRcreate(file_id, rid1, (uint64_t)1);
assert(tid1);
ret = H5TRstart(tid1, H5P_DEFAULT, e_stack);
assert(0 == ret);
/* create objects */
gid = H5Gcreate_ff(file_id, "G1", H5P_DEFAULT, H5P_DEFAULT,
H5P_DEFAULT, tid1, e_stack);
assert(gid >= 0);
did = H5Dcreate_ff(gid, "D1", dtid, sid, H5P_DEFAULT, H5P_DEFAULT,
H5P_DEFAULT, tid1, e_stack);
assert(did >= 0);
ret = H5Dwrite_ff(did, dtid, H5S_ALL, H5S_ALL, dxpl_id, wdata1, tid1, e_stack);
assert(ret == 0);
ret = H5Tcommit_ff(file_id, "DT1", dtid, H5P_DEFAULT, H5P_DEFAULT,
H5P_DEFAULT, tid1, e_stack);
assert(ret == 0);
map = H5Mcreate_ff(file_id, "MAP1", H5T_STD_I32LE, H5T_STD_I32LE,
H5P_DEFAULT,H5P_DEFAULT,H5P_DEFAULT, tid1, e_stack);
assert(map >= 0);
/* write some KV pairs to each map object. */
{
key = 1;
value = 1000;
ret = H5Mset_ff(map, H5T_STD_I32LE, &key, H5T_STD_I32LE, &value,
H5P_DEFAULT, tid1, e_stack);
assert(ret == 0);
}
ret = H5TRfinish(tid1, H5P_DEFAULT, &rid_temp, e_stack);
assert(0 == ret);
/* wait on all requests and print completion status */
H5ESget_count(e_stack, &num_events);
H5ESwait_all(e_stack, &status);
H5ESclear(e_stack);
printf("%d events in event stack. Completion status = %d\n", num_events, status);
/* Close transaction object. Local op */
ret = H5TRclose(tid1);
assert(0 == ret);
/* create transaction object */
tid2 = H5TRcreate(file_id, rid_temp, (uint64_t)2);
assert(tid2);
ret = H5TRstart(tid2, H5P_DEFAULT, e_stack);
assert(0 == ret);
ret = H5Oset_comment_ff(gid, "Testing Object Comment", tid2, e_stack);
assert(ret == 0);
ret = H5TRfinish(tid2, H5P_DEFAULT, &rid2, e_stack);
assert(0 == ret);
assert(H5Gclose_ff(gid, e_stack) == 0);
assert(H5Mclose_ff(map, e_stack) == 0);
assert(H5Tclose_ff(dtid, e_stack) == 0);
assert(H5Dclose_ff(did, e_stack) == 0);
/* release container version 2. This is async. */
ret = H5RCrelease(rid_temp, e_stack);
assert(0 == ret);
ret = H5RCclose(rid_temp);
assert(0 == ret);
version = 2;
}
/* release container version 0. This is async. */
ret = H5RCrelease(rid1, e_stack);
assert(0 == ret);
/* wait on all requests and print completion status */
H5ESget_count(e_stack, &num_events);
H5ESwait_all(e_stack, &status);
H5ESclear(e_stack);
printf("%d events in event stack. Completion status = %d\n", num_events, status);
if(0 == my_rank) {
/* Close transaction object. Local op */
ret = H5TRclose(tid2);
assert(0 == ret);
}
MPI_Barrier(MPI_COMM_WORLD);
/* Process 0 tells other procs that container version 2 is acquired */
MPI_Bcast(&version, 1, MPI_UINT64_T, 0, MPI_COMM_WORLD);
assert(2 == version);
/* other processes just create a read context object; no need to
acquire it */
if(0 != my_rank) {
rid2 = H5RCcreate(file_id, version);
assert(rid2 > 0);
}
gid = H5Oopen_ff(file_id, "G1", H5P_DEFAULT, rid2);
assert(gid);
dtid = H5Oopen_ff(file_id, "DT1", H5P_DEFAULT, rid2);
assert(dtid);
did = H5Oopen_ff(gid,"D1", H5P_DEFAULT, rid2);
assert(did);
map = H5Oopen_ff(file_id,"MAP1", H5P_DEFAULT, rid2);
assert(did);
MPI_Barrier(MPI_COMM_WORLD);
if((my_size > 1 && 1 == my_rank) ||
(my_size == 1 && 0 == my_rank)) {
//if(0 == my_rank) {
hid_t mapl_id, tapl_id, dapl_id;
hrpl_t map_replica, dt_replica, dset_replica;
ret = H5Mprefetch_ff(map, rid2, &map_replica, H5P_DEFAULT, H5_EVENT_STACK_NULL);
assert(0 == ret);
printf("prefetched map with replica id %"PRIx64"\n", map_replica);
ret = H5Tprefetch_ff(dtid, rid2, &dt_replica, H5P_DEFAULT, H5_EVENT_STACK_NULL);
assert(0 == ret);
printf("prefetched datatype with replica id %"PRIx64"\n", dt_replica);
ret = H5Dprefetch_ff(did, rid2, &dset_replica, H5P_DEFAULT, H5_EVENT_STACK_NULL);
assert(0 == ret);
printf("prefetched dataset with replica id %"PRIx64"\n", dset_replica);
/* set dxpl to read from replica in BB */
dxpl_id = H5Pcreate (H5P_DATASET_XFER);
ret = H5Pset_read_replica(dxpl_id, dset_replica);
assert(0 == ret);
ret = H5Dread_ff(did, dtid, H5S_ALL, H5S_ALL, dxpl_id, rdata1, rid2, H5_EVENT_STACK_NULL);
assert(ret == 0);
printf("Read Data1: ");
for(i=0;i<nelem;++i)
printf("%d ",rdata1[i]);
printf("\n");
key = 1;
value = -1;
ret = H5Mget_ff(map, H5T_STD_I32LE, &key, H5T_STD_I32LE, &value,
dxpl_id, rid2, H5_EVENT_STACK_NULL);
printf("Value recieved = %d\n", value);
H5Pclose(dxpl_id);
tapl_id = H5Pcreate (H5P_DATATYPE_ACCESS);
ret = H5Pset_evict_replica(tapl_id, dt_replica);
assert(0 == ret);
ret = H5Tevict_ff(dtid, 2, tapl_id, H5_EVENT_STACK_NULL);
assert(0 == ret);
H5Pclose(tapl_id);
#if 0
dapl_id = H5Pcreate (H5P_DATASET_ACCESS);
ret = H5Pset_evict_replica(dapl_id, dset_replica);
assert(0 == ret);
ret = H5Devict_ff(did, 2, dapl_id, H5_EVENT_STACK_NULL);
assert(0 == ret);
H5Pclose(dapl_id);
mapl_id = H5Pcreate (H5P_MAP_ACCESS);
ret = H5Pset_evict_replica(mapl_id, map_replica);
assert(0 == ret);
ret = H5Mevict_ff(map, 2, mapl_id, H5_EVENT_STACK_NULL);
assert(0 == ret);
H5Pclose(mapl_id);
#endif
}
MPI_Barrier(MPI_COMM_WORLD);
if(my_rank == 0) {
/* release container version 2. This is async. */
ret = H5RCrelease(rid2, e_stack);
assert(0 == ret);
}
assert(H5Oclose_ff(gid, e_stack) == 0);
assert(H5Oclose_ff(did, e_stack) == 0);
assert(H5Oclose_ff(dtid, e_stack) == 0);
assert(H5Oclose_ff(map, e_stack) == 0);
ret = H5RCclose(rid1);
assert(0 == ret);
ret = H5RCclose(rid2);
assert(0 == ret);
/* wait on all requests and print completion status */
H5ESget_count(e_stack, &num_events);
H5ESwait_all(e_stack, &status);
H5ESclear(e_stack);
printf("%d events in event stack. Completion status = %d\n", num_events, status);
/* closing the container also acts as a wait all on all pending requests
on the container. */
assert(H5Fclose_ff(file_id, H5_EVENT_STACK_NULL) == 0);
H5Sclose(sid);
H5Pclose(fapl_id);
H5ESclose(e_stack);
free(wdata1);
free(rdata1);
/* This finalizes the EFF stack. ships a terminate and IOD finalize to the server
and shutsdown the FS server (when all clients send the terminate request)
and client */
MPI_Barrier(MPI_COMM_WORLD);
ret = EFF_finalize();
assert(ret >= 0);
MPI_Finalize();
return 0;
}
|