summaryrefslogtreecommitdiffstats
path: root/examples
diff options
context:
space:
mode:
authorMohamad Chaarawi <chaarawi@hdfgroup.org>2014-06-27 17:26:27 (GMT)
committerMohamad Chaarawi <chaarawi@hdfgroup.org>2014-06-27 17:26:27 (GMT)
commit41864628c953fad25685767a00ace0269c3a61f4 (patch)
treeb4783349b8041c652fcb2b566d506728928cb0d8 /examples
parentd2313249ca4245b702d1bf7ccae28f758afcc5fe (diff)
downloadhdf5-41864628c953fad25685767a00ace0269c3a61f4.zip
hdf5-41864628c953fad25685767a00ace0269c3a61f4.tar.gz
hdf5-41864628c953fad25685767a00ace0269c3a61f4.tar.bz2
[svn-r25368] add test that does incremental updates for datasets to test incremental index building.
Diffstat (limited to 'examples')
-rw-r--r--examples/h5ff_client_analysis.c2
-rw-r--r--examples/h5ff_client_index.c25
-rw-r--r--examples/h5ff_client_index_inr.c281
3 files changed, 301 insertions, 7 deletions
diff --git a/examples/h5ff_client_analysis.c b/examples/h5ff_client_analysis.c
index 83fa699..710deac 100644
--- a/examples/h5ff_client_analysis.c
+++ b/examples/h5ff_client_analysis.c
@@ -276,7 +276,7 @@ main(int argc, char **argv)
MPI_Barrier(MPI_COMM_WORLD);
write_dataset(file_name, dataset_name, total, ncomponents, H5T_NATIVE_INT,
- ntuples, start, data);
+ ntuples, start, data);
MPI_Barrier(MPI_COMM_WORLD);
diff --git a/examples/h5ff_client_index.c b/examples/h5ff_client_index.c
index 923e02c..2a8c746 100644
--- a/examples/h5ff_client_index.c
+++ b/examples/h5ff_client_index.c
@@ -96,7 +96,8 @@ write_dataset(hid_t file_id, const char *dataset_name,
}
static void
-create_index(hid_t file_id, const char *dataset_name, hid_t estack_id)
+create_index(hid_t file_id, const char *dataset_name, unsigned plugin_id,
+ hid_t estack_id)
{
hid_t dataset_id, trans_id, rcxt_id;
hid_t trspl_id;
@@ -127,7 +128,7 @@ create_index(hid_t file_id, const char *dataset_name, hid_t estack_id)
estack_id);
/* Add indexing information */
- ret = H5Xcreate_ff(file_id, H5X_PLUGIN_ALACRITY, dataset_id, H5P_DEFAULT,
+ ret = H5Xcreate_ff(file_id, plugin_id, dataset_id, H5P_DEFAULT,
trans_id, estack_id);
assert(0 == ret);
@@ -162,10 +163,11 @@ query_and_view(hid_t file_id, const char *dataset_name, hid_t estack_id)
hid_t rcxt_id;
uint64_t version = 3;
herr_t ret;
+ double t1, t2;
/* Create a simple query */
/* query = (39.1 < x < 42.1) || (295 < x < 298) */
- query_lb = (my_rank == 0) ? 39.1f : 295.f;
+ query_lb = (my_rank == 0) ? 38.8f : 295.f;
query_ub = (my_rank == 0) ? 42.1f : 298.f;
query_id1 = H5Qcreate(H5Q_TYPE_DATA_ELEM, H5Q_MATCH_GREATER_THAN,
@@ -193,8 +195,12 @@ query_and_view(hid_t file_id, const char *dataset_name, hid_t estack_id)
// view_id = H5Vcreate_ff(dataset_id, query_id, H5P_DEFAULT, rid2,
// estack_id);
// assert(view_id > 0);
+
+ t1 = MPI_Wtime();
H5Dquery_ff(dataset_id, query_id, -1, rcxt_id);
+ t2 = MPI_Wtime();
+ printf("Query time: %lf ms\n", (t2 - t1) * 1000);
/* TODO use view_id for analysis shipping etc */
// H5Vclose(view_id);
@@ -217,6 +223,7 @@ query_and_view(hid_t file_id, const char *dataset_name, hid_t estack_id)
int
main(int argc, char **argv)
{
+ unsigned plugin_id;
char file_name[50];
char dataset_name[64];
hsize_t ntuples = NTUPLES;
@@ -231,12 +238,18 @@ main(int argc, char **argv)
sprintf(file_name, "%s_%s", getenv("USER"), "eff_file_index.h5");
MPI_Init(&argc, &argv);
+ MPI_Comm_rank(MPI_COMM_WORLD, &my_rank);
+ MPI_Comm_size(MPI_COMM_WORLD, &my_size);
+
+ if (argc < 2) {
+ if (my_rank ==0) printf("Usage: %s <plugin_id>\n", argv[0]);
+ exit(0);
+ }
+ plugin_id = atoi(argv[1]);
/* Call EFF_init to initialize the EFF stack. */
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);
/* In this example write one dataset per process */
@@ -274,7 +287,7 @@ main(int argc, char **argv)
MPI_Barrier(MPI_COMM_WORLD);
- create_index(file_id, dataset_name, estack_id);
+ create_index(file_id, dataset_name, plugin_id, estack_id);
MPI_Barrier(MPI_COMM_WORLD);
diff --git a/examples/h5ff_client_index_inr.c b/examples/h5ff_client_index_inr.c
new file mode 100644
index 0000000..a521d5f
--- /dev/null
+++ b/examples/h5ff_client_index_inr.c
@@ -0,0 +1,281 @@
+/*
+ * h5ff_client_index.c: Client side test for index routines.
+ */
+
+#include <hdf5.h>
+
+#include <mpi.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <assert.h>
+#include <string.h>
+
+#define NTUPLES 256
+static int my_rank = 0, my_size = 1;
+
+static void
+query_and_view(hid_t file_id, hid_t dataset_id, hid_t rcxt_id, hid_t estack_id)
+{
+ float query_lb, query_ub;
+ hid_t query_id1, query_id2;
+ hid_t query_id;
+ herr_t ret;
+ double t1, t2;
+
+ /* Create a simple query */
+ /* query = (39.1 < x < 42.1) || (295 < x < 298) */
+ query_lb = (my_rank == 0) ? 38.8f : 295.f;
+ query_ub = (my_rank == 0) ? 42.1f : 298.f;
+
+ query_id1 = H5Qcreate(H5Q_TYPE_DATA_ELEM, H5Q_MATCH_GREATER_THAN,
+ H5T_NATIVE_FLOAT, &query_lb);
+ assert(query_id1);
+
+ query_id2 = H5Qcreate(H5Q_TYPE_DATA_ELEM, H5Q_MATCH_LESS_THAN,
+ H5T_NATIVE_FLOAT, &query_ub);
+ assert(query_id2);
+
+ query_id = H5Qcombine(query_id1, H5Q_COMBINE_AND, query_id2);
+ assert(query_id);
+
+// view_id = H5Vcreate_ff(dataset_id, query_id, H5P_DEFAULT, rid2,
+// estack_id);
+// assert(view_id > 0);
+
+ t1 = MPI_Wtime();
+ H5Dquery_ff(dataset_id, query_id, -1, rcxt_id);
+ t2 = MPI_Wtime();
+
+ printf("Query time: %lf ms\n", (t2 - t1) * 1000);
+ /* TODO use view_id for analysis shipping etc */
+
+// H5Vclose(view_id);
+
+ H5Qclose(query_id);
+ H5Qclose(query_id2);
+ H5Qclose(query_id1);
+}
+
+int
+main(int argc, char **argv)
+{
+ unsigned plugin_id;
+ char file_name[50];
+ char dataset_name[64];
+ hsize_t ntuples = NTUPLES;
+ hsize_t ncomponents = 3;
+ hsize_t start, total;
+ float *data;
+ hid_t file_id, fapl_id, dataset_id;
+ hid_t file_space_id;
+ hid_t tid2, rid1, trspl_id, rc_id;
+ uint64_t version;
+ void *dset_token;
+ size_t token_size;
+ hid_t estack_id = H5_EVENT_STACK_NULL;
+ herr_t ret;
+ hsize_t dims[2];
+ int rank;
+ hsize_t i, j;
+ int n;
+
+ sprintf(file_name, "%s_%s", getenv("USER"), "eff_file_index.h5");
+
+ MPI_Init(&argc, &argv);
+ MPI_Comm_rank(MPI_COMM_WORLD, &my_rank);
+ MPI_Comm_size(MPI_COMM_WORLD, &my_size);
+
+ if (argc < 2) {
+ if (my_rank ==0) printf("Usage: %s <plugin_id>\n", argv[0]);
+ exit(0);
+ }
+ plugin_id = atoi(argv[1]);
+
+ /* Call EFF_init to initialize the EFF stack. */
+ EFF_init(MPI_COMM_WORLD, MPI_INFO_NULL);
+
+ fprintf(stderr, "APP processes = %d, my rank is %d\n", my_size, my_rank);
+
+ /* In this example write one dataset per process */
+ memset(dataset_name, '\0', 64);
+ sprintf(dataset_name, "indexed_dataset");
+
+ start = ntuples * (hsize_t) my_rank;
+ total = ntuples * (hsize_t) my_size;
+ dims[0] = total;
+ dims[1] = ncomponents;
+ rank = (ncomponents == 1) ? 1 : 2;
+
+ data = (float *) malloc(sizeof(float) * ncomponents * ntuples);
+ for (i = 0; i < ntuples; i++) {
+ for (j = 0; j < ncomponents; j++) {
+ data[ncomponents * i + j] = (float) (((hsize_t) my_rank) * ntuples + i);
+ }
+ }
+
+ MPI_Barrier(MPI_COMM_WORLD);
+
+ /* Choose the IOD VOL plugin to use with this file. */
+ fapl_id = H5Pcreate(H5P_FILE_ACCESS);
+ H5Pset_fapl_iod(fapl_id, MPI_COMM_WORLD, MPI_INFO_NULL);
+
+ /* Open an existing file. */
+ file_id = H5Fcreate_ff(file_name, H5F_ACC_TRUNC, H5P_DEFAULT, fapl_id,
+ estack_id);
+
+ ret = H5Pclose(fapl_id);
+ assert(0 == ret);
+
+ /* acquire container version 1 - EXACT. */
+ if(0 == my_rank) {
+ version = 1;
+ rid1 = H5RCacquire(file_id, &version, H5P_DEFAULT, estack_id);
+ assert(1 == version);
+ }
+ MPI_Bcast( &version, 1, MPI_UINT64_T, 0, MPI_COMM_WORLD );
+ if(0 != my_rank) {
+ assert(version == 1);
+ rid1 = H5RCcreate(file_id, version);
+ }
+
+ /* create transaction object */
+ tid2 = H5TRcreate(file_id, rid1, (uint64_t)2);
+ assert(tid2);
+
+ trspl_id = H5Pcreate(H5P_TR_START);
+ ret = H5Pset_trspl_num_peers(trspl_id, (unsigned int) my_size);
+ assert(0 == ret);
+ ret = H5TRstart(tid2, trspl_id, estack_id);
+ assert(0 == ret);
+ ret = H5Pclose(trspl_id);
+ assert(0 == ret);
+
+ /* Create the data space for the first dataset. */
+ file_space_id = H5Screate_simple(rank, dims, NULL);
+ assert(file_space_id);
+
+ if(0 == my_rank) {
+ /* Create a dataset. */
+ dataset_id = H5Dcreate_ff(file_id, dataset_name, H5T_NATIVE_FLOAT, file_space_id,
+ H5P_DEFAULT, H5P_DEFAULT, H5P_DEFAULT, tid2, estack_id);
+ assert(dataset_id);
+
+ /* get the token size of each dset */
+ ret = H5Oget_token(dataset_id, NULL, &token_size);
+ assert(0 == ret);
+ /* allocate buffers for each token */
+ dset_token = malloc(token_size);
+ /* get the token buffer */
+ ret = H5Oget_token(dataset_id, dset_token, &token_size);
+ assert(0 == ret);
+ }
+
+ /* bcast the token sizes and the tokens */
+ MPI_Bcast(&token_size, sizeof(size_t), MPI_BYTE, 0, MPI_COMM_WORLD);
+
+ if(0 != my_rank)
+ dset_token = malloc(token_size);
+
+ MPI_Bcast(dset_token, token_size, MPI_BYTE, 0, MPI_COMM_WORLD);
+
+ if(0 != my_rank)
+ dataset_id = H5Oopen_by_token(dset_token, tid2, estack_id);
+
+ /* Add indexing information */
+ //ret = H5Xcreate_ff(file_id, plugin_id, dataset_id, H5P_DEFAULT, tid2, estack_id);
+ //assert(0 == ret);
+
+ ret = H5TRfinish(tid2, H5P_DEFAULT, NULL, estack_id);
+ assert(0 == ret);
+ H5TRclose(tid2);
+
+ if(0 == my_rank) {
+ ret = H5RCrelease(rid1, H5_EVENT_STACK_NULL);
+ assert(0 == ret);
+ }
+ H5RCclose(rid1);
+
+ MPI_Barrier(MPI_COMM_WORLD);
+
+ /* do incremental updates */
+ for(n=0; n<my_size ; n++) {
+ if(my_rank == n) {
+ hid_t tid, rid, mem_space_id;
+ hsize_t count[2] = {ntuples, ncomponents};
+ hsize_t offset[2] = {start, 0};
+
+ version = (uint64_t)(2+n);
+ rid = H5RCacquire(file_id, &version, H5P_DEFAULT, estack_id);
+ assert((uint64_t)(2+n) == version);
+
+ /* create transaction object */
+ tid = H5TRcreate(file_id, rid, (uint64_t)(3+n));
+ assert(tid);
+ ret = H5TRstart(tid, H5P_DEFAULT, estack_id);
+ assert(0 == ret);
+
+ fprintf(stderr, "Rank %d Doing its Updates with TR %d RC %d\n",
+ my_rank, 3+n, 2+n);
+
+ mem_space_id = H5Screate_simple(rank, count, NULL);
+ assert(mem_space_id);
+ /* write data to datasets */
+ ret = H5Sselect_hyperslab(file_space_id, H5S_SELECT_SET, offset,
+ NULL, count, NULL);
+ assert(0 == ret);
+ ret = H5Dwrite_ff(dataset_id, H5T_NATIVE_FLOAT, mem_space_id, file_space_id,
+ H5P_DEFAULT, data, tid, estack_id);
+ assert(0 == ret);
+ ret = H5Sclose(mem_space_id);
+ assert(0 == ret);
+
+ ret = H5TRfinish(tid, H5P_DEFAULT, NULL, estack_id);
+ assert(0 == ret);
+ H5TRclose(tid);
+ ret = H5RCrelease(rid, H5_EVENT_STACK_NULL);
+ assert(0 == ret);
+ ret = H5RCclose(rid);
+ }
+ MPI_Barrier(MPI_COMM_WORLD);
+ }
+
+ /* acquire last container version - EXACT. */
+ if(0 == my_rank) {
+ fprintf(stderr, "Rank 0 acquiring latest context: %d\n", 2+my_size);
+ version = (uint64_t)(2+my_size);
+ rc_id = H5RCacquire(file_id, &version, H5P_DEFAULT, estack_id);
+ assert((uint64_t)(2+my_size) == version);
+ }
+ MPI_Bcast( &version, 1, MPI_UINT64_T, 0, MPI_COMM_WORLD );
+ if(0 != my_rank) {
+ assert(version == (uint64_t)(2+my_size));
+ rc_id = H5RCcreate(file_id, version);
+ }
+
+ query_and_view(file_id, dataset_id, rc_id, estack_id);
+
+ MPI_Barrier(MPI_COMM_WORLD);
+
+ if(0 == my_rank) {
+ ret = H5RCrelease(rc_id, H5_EVENT_STACK_NULL);
+ assert(0 == ret);
+ }
+ H5RCclose(rc_id);
+
+ /* Close. */
+ H5Sclose(file_space_id);
+ ret = H5Dclose_ff(dataset_id, estack_id);
+ assert(0 == ret);
+ ret = H5Fclose_ff(file_id, 1, estack_id);
+ assert(0 == ret);
+
+ free(data);
+
+ MPI_Barrier(MPI_COMM_WORLD);
+
+ EFF_finalize();
+
+ MPI_Finalize();
+
+ return 0;
+}