/**
* Conversion function. They deal with data-types in 3 ways, always making local copies.
* In order to allow performance testings, there are 3 functions:
* - one copying directly from one memory location to another one using the
* data-type copy function.
* - one which use a 2 convertors created with the same data-type
* - and one using 2 convertors created from different data-types.
*
*/
static int local_copy_ddt_count( ompi_datatype_t* pdt, int count )
{
MPI_Aint extent;
void *pdst, *psrc;
TIMER_DATA_TYPE start, end;
long total_time;
ompi_ddt_type_extent( pdt, &extent );
pdst = malloc( extent * count );
psrc = malloc( extent * count );
{
int i;
for( i = 0; i < (count * extent); i++ )
((char*)psrc)[i] = i % 128 + 32;
}
memset( pdst, 0, count * extent );
cache_trash(); /* make sure the cache is useless */
GET_TIME( start );
if( OMPI_SUCCESS != ompi_ddt_copy_content_same_ddt( pdt, count, pdst, psrc ) ) {
printf( "Unable to copy the datatype in the function local_copy_ddt_count."
" Is the datatype committed ?\n" );
}
GET_TIME( end );
total_time = ELAPSED_TIME( start, end );
printf( "direct local copy in %ld microsec\n", total_time );
free( pdst );
free( psrc );
return OMPI_SUCCESS;
}
int MPI_Scatterv(void *sendbuf, int *sendcounts, int *displs,
MPI_Datatype sendtype, void *recvbuf, int recvcount,
MPI_Datatype recvtype, int root, MPI_Comm comm)
{
int i, size, err;
MEMCHECKER(
ptrdiff_t ext;
size = ompi_comm_size(comm);
ompi_ddt_type_extent(recvtype, &ext);
memchecker_comm(comm);
if(OMPI_COMM_IS_INTRA(comm)) {
if(ompi_comm_rank(comm) == root) {
memchecker_datatype(sendtype);
/* check whether root's send buffer is defined. */
for (i = 0; i < size; i++) {
memchecker_call(&opal_memchecker_base_isdefined,
(char *)(sendbuf)+displs[i]*ext,
sendcounts[i], sendtype);
}
if(MPI_IN_PLACE != recvbuf) {
memchecker_datatype(recvtype);
/* check whether receive buffer is addressable. */
memchecker_call(&opal_memchecker_base_isaddressable, recvbuf, recvcount, recvtype);
}
} else {
memchecker_datatype(recvtype);
/* check whether receive buffer is addressable. */
memchecker_call(&opal_memchecker_base_isaddressable, recvbuf, recvcount, recvtype);
}
} else {
if(MPI_ROOT == root) {
memchecker_datatype(sendtype);
/* check whether root's send buffer is defined. */
for (i = 0; i < size; i++) {
memchecker_call(&opal_memchecker_base_isdefined,
(char *)(sendbuf)+displs[i]*ext,
sendcounts[i], sendtype);
}
} else if (MPI_PROC_NULL != root) {
/* check whether receive buffer is addressable. */
memchecker_call(&opal_memchecker_base_isaddressable, recvbuf, recvcount, recvtype);
}
}
);
int
ompi_coll_tuned_alltoallv_intra_pairwise(void *sbuf, int *scounts, int *sdisps,
struct ompi_datatype_t *sdtype,
void* rbuf, int *rcounts, int *rdisps,
struct ompi_datatype_t *rdtype,
struct ompi_communicator_t *comm,
mca_coll_base_module_t *module)
{
int line = -1, err = 0;
int rank, size, step;
int sendto, recvfrom;
void *psnd, *prcv;
ptrdiff_t sext, rext;
size = ompi_comm_size(comm);
rank = ompi_comm_rank(comm);
OPAL_OUTPUT((ompi_coll_tuned_stream,
"coll:tuned:alltoallv_intra_pairwise rank %d", rank));
ompi_ddt_type_extent(sdtype, &sext);
ompi_ddt_type_extent(rdtype, &rext);
psnd = ((char *) sbuf) + (sdisps[rank] * sext);
prcv = ((char *) rbuf) + (rdisps[rank] * rext);
if (0 != scounts[rank]) {
err = ompi_ddt_sndrcv(psnd, scounts[rank], sdtype,
prcv, rcounts[rank], rdtype);
if (MPI_SUCCESS != err) {
return err;
}
}
/* If only one process, we're done. */
if (1 == size) {
return MPI_SUCCESS;
}
/* Perform pairwise exchange starting from 1 since local exhange is done */
for (step = 1; step < size + 1; step++) {
/* Determine sender and receiver for this step. */
sendto = (rank + step) % size;
recvfrom = (rank + size - step) % size;
/* Determine sending and receiving locations */
psnd = (char*)sbuf + sdisps[sendto] * sext;
prcv = (char*)rbuf + rdisps[recvfrom] * rext;
/* send and receive */
err = ompi_coll_tuned_sendrecv( psnd, scounts[sendto], sdtype, sendto,
MCA_COLL_BASE_TAG_ALLTOALLV,
prcv, rcounts[recvfrom], rdtype, recvfrom,
MCA_COLL_BASE_TAG_ALLTOALLV,
comm, MPI_STATUS_IGNORE, rank);
if (err != MPI_SUCCESS) { line = __LINE__; goto err_hndl; }
}
return MPI_SUCCESS;
err_hndl:
OPAL_OUTPUT((ompi_coll_tuned_stream,
"%s:%4d\tError occurred %d, rank %2d", __FILE__, line,
err, rank));
return err;
}
/*
* Linear functions are copied from the basic coll module. For
* some small number of nodes and/or small data sizes they are just as
* fast as tuned/tree based segmenting operations and as such may be
* selected by the decision functions. These are copied into this module
* due to the way we select modules in V1. i.e. in V2 we will handle this
* differently and so will not have to duplicate code.
* GEF Oct05 after asking Jeff.
*/
int
ompi_coll_tuned_alltoallv_intra_basic_linear(void *sbuf, int *scounts, int *sdisps,
struct ompi_datatype_t *sdtype,
void *rbuf, int *rcounts, int *rdisps,
struct ompi_datatype_t *rdtype,
struct ompi_communicator_t *comm,
mca_coll_base_module_t *module)
{
int i, size, rank, err;
char *psnd, *prcv;
int nreqs;
ptrdiff_t sext, rext;
MPI_Request *preq;
mca_coll_tuned_module_t *tuned_module = (mca_coll_tuned_module_t*) module;
mca_coll_tuned_comm_t *data = tuned_module->tuned_data;
size = ompi_comm_size(comm);
rank = ompi_comm_rank(comm);
OPAL_OUTPUT((ompi_coll_tuned_stream,
"coll:tuned:alltoallv_intra_basic_linear rank %d", rank));
ompi_ddt_type_extent(sdtype, &sext);
ompi_ddt_type_extent(rdtype, &rext);
/* Simple optimization - handle send to self first */
psnd = ((char *) sbuf) + (sdisps[rank] * sext);
prcv = ((char *) rbuf) + (rdisps[rank] * rext);
if (0 != scounts[rank]) {
err = ompi_ddt_sndrcv(psnd, scounts[rank], sdtype,
prcv, rcounts[rank], rdtype);
if (MPI_SUCCESS != err) {
return err;
}
}
/* If only one process, we're done. */
if (1 == size) {
return MPI_SUCCESS;
}
/* Now, initiate all send/recv to/from others. */
nreqs = 0;
preq = data->mcct_reqs;
/* Post all receives first */
for (i = 0; i < size; ++i) {
if (i == rank || 0 == rcounts[i]) {
continue;
}
prcv = ((char *) rbuf) + (rdisps[i] * rext);
err = MCA_PML_CALL(irecv_init(prcv, rcounts[i], rdtype,
i, MCA_COLL_BASE_TAG_ALLTOALLV, comm,
preq++));
++nreqs;
if (MPI_SUCCESS != err) {
ompi_coll_tuned_free_reqs(data->mcct_reqs, nreqs);
return err;
}
}
/* Now post all sends */
for (i = 0; i < size; ++i) {
if (i == rank || 0 == scounts[i]) {
continue;
}
psnd = ((char *) sbuf) + (sdisps[i] * sext);
err = MCA_PML_CALL(isend_init(psnd, scounts[i], sdtype,
i, MCA_COLL_BASE_TAG_ALLTOALLV,
MCA_PML_BASE_SEND_STANDARD, comm,
preq++));
++nreqs;
if (MPI_SUCCESS != err) {
ompi_coll_tuned_free_reqs(data->mcct_reqs, nreqs);
return err;
}
}
/* Start your engines. This will never return an error. */
MCA_PML_CALL(start(nreqs, data->mcct_reqs));
/* Wait for them all. If there's an error, note that we don't care
* what the error was -- just that there *was* an error. The PML
* will finish all requests, even if one or more of them fail.
* i.e., by the end of this call, all the requests are free-able.
* So free them anyway -- even if there was an error, and return the
* error after we free everything. */
err = ompi_request_wait_all(nreqs, data->mcct_reqs,
MPI_STATUSES_IGNORE);
/* Free the requests. */
ompi_coll_tuned_free_reqs(data->mcct_reqs, nreqs);
return err;
}
int MPI_Type_create_darray(int size,
int rank,
int ndims,
int gsize_array[],
int distrib_array[],
int darg_array[],
int psize_array[],
int order,
MPI_Datatype oldtype,
MPI_Datatype *newtype)
{
ompi_datatype_t *lastType;
ptrdiff_t orig_extent, *st_offsets = NULL;
int i, start_loop, end_loop, step;
int *coords = NULL, rc = OMPI_SUCCESS;
if (MPI_PARAM_CHECK) {
int prod_psize = 1;
OMPI_ERR_INIT_FINALIZE(FUNC_NAME);
if( (rank < 0) || (size < 0) || (rank >= size) ) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_ARG, FUNC_NAME);
} else if( ndims < 0 ) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_COUNT, FUNC_NAME);
} else if( (NULL == gsize_array) || (NULL == distrib_array) || (NULL == darg_array) || (NULL == psize_array)) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_ARG, FUNC_NAME);
} else if (NULL == newtype) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_TYPE, FUNC_NAME);
} else if( !(DT_FLAG_DATA & oldtype ->flags) ) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_TYPE, FUNC_NAME);
} else if( (MPI_ORDER_C != order) && (MPI_ORDER_FORTRAN != order) ) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_ARG, FUNC_NAME);
}
for( i = 0; i < ndims; i++ ) {
if( (MPI_DISTRIBUTE_BLOCK != distrib_array[i]) &&
(MPI_DISTRIBUTE_CYCLIC != distrib_array[i]) &&
(MPI_DISTRIBUTE_NONE != distrib_array[i]) ) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_ARG, FUNC_NAME);
} else if( (gsize_array[i] < 1) || (psize_array[i] < 0) ||
((darg_array[i] < 0) && (MPI_DISTRIBUTE_DFLT_DARG != darg_array[i]) ) ) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_ARG, FUNC_NAME);
} else if( (MPI_DISTRIBUTE_DFLT_DARG != darg_array[i]) &&
(MPI_DISTRIBUTE_BLOCK == distrib_array[i]) &&
((darg_array[i] * psize_array[i]) < gsize_array[i]) ) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_ARG, FUNC_NAME);
} else if( 1 > psize_array[i] )
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_ARG, FUNC_NAME);
prod_psize *= psize_array[i];
}
if( prod_psize != size )
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_ARG, FUNC_NAME);
}
/* speedy corner case */
if (ndims < 1) {
/* Don't just return MPI_DATATYPE_NULL as that can't be
MPI_TYPE_FREE()ed, and that seems bad */
*newtype = ompi_ddt_create(0);
ompi_ddt_add(*newtype, &ompi_mpi_datatype_null, 0, 0, 0);
return MPI_SUCCESS;
}
rc = ompi_ddt_type_extent(oldtype, &orig_extent);
if (MPI_SUCCESS != rc) goto cleanup;
/* calculate position in grid using row-major ordering */
{
int tmp_rank = rank, procs = size;
coords = (int *) malloc(ndims * sizeof(int));
for (i = 0 ; i < ndims ; i++) {
procs = procs / psize_array[i];
coords[i] = tmp_rank / procs;
tmp_rank = tmp_rank % procs;
}
}
st_offsets = (ptrdiff_t *) malloc(ndims * sizeof(ptrdiff_t));
/* duplicate type to here to 1) deal with constness without
casting and 2) eliminate need to for conditional destroy below.
Lame, yes. But cleaner code all around. */
rc = ompi_ddt_duplicate(oldtype, &lastType);
if (OMPI_SUCCESS != rc) goto cleanup;
/* figure out ordering issues */
if (MPI_ORDER_C == order) {
start_loop = ndims - 1 ; step = -1; end_loop = -1;
} else {
start_loop = 0 ; step = 1; end_loop = ndims;
}
/* Build up array */
for (i = start_loop ; i != end_loop; i += step) {
int nprocs, rank;
switch(distrib_array[i]) {
case MPI_DISTRIBUTE_BLOCK:
rc = block(gsize_array, i, ndims, psize_array[i], coords[i],
darg_array[i], order, orig_extent,
lastType, newtype, st_offsets+i);
break;
case MPI_DISTRIBUTE_CYCLIC:
rc = cyclic(gsize_array, i, ndims, psize_array[i], coords[i],
darg_array[i], order, orig_extent,
lastType, newtype, st_offsets+i);
break;
case MPI_DISTRIBUTE_NONE:
/* treat it as a block distribution on 1 process */
if (order == MPI_ORDER_C) {
nprocs = psize_array[i]; rank = coords[i];
} else {
nprocs = 1; rank = 0;
}
rc = block(gsize_array, i, ndims, nprocs, rank,
MPI_DISTRIBUTE_DFLT_DARG, order, orig_extent,
lastType, newtype, st_offsets+i);
break;
default:
rc = MPI_ERR_ARG;
}
ompi_ddt_destroy(&lastType);
/* need to destroy the old type even in error condition, so
don't check return code from above until after cleanup. */
if (MPI_SUCCESS != rc) goto cleanup;
lastType = *newtype;
}
/* set displacement and UB correctly. Use struct instead of
resized for same reason as subarray */
{
ptrdiff_t displs[3];
ompi_datatype_t *types[3];
int tmp_size, blength[3] = { 1, 1, 1};
displs[1] = st_offsets[start_loop];
tmp_size = 1;
for (i = start_loop + step ; i != end_loop ; i += step) {
tmp_size *= gsize_array[i - step];
displs[1] += tmp_size * st_offsets[i];
}
displs[0] = 0;
displs[1] *= orig_extent;
displs[2] = orig_extent;
for (i = 0 ; i < ndims ; i++) {
displs[2] *= gsize_array[i];
}
types[0] = MPI_LB; types[1] = lastType; types[2] = MPI_UB;
rc = ompi_ddt_create_struct(3, blength, displs, types, newtype);
ompi_ddt_destroy(&lastType);
/* need to destroy the old type even in error condition, so
don't check return code from above until after cleanup. */
if (MPI_SUCCESS != rc) goto cleanup;
}
{
int* a_i[8];
a_i[0] = &size;
a_i[1] = &rank;
a_i[2] = &ndims;
a_i[3] = gsize_array;
a_i[4] = distrib_array;
a_i[5] = darg_array;
a_i[6] = psize_array;
a_i[7] = ℴ
ompi_ddt_set_args( *newtype, 4 * ndims + 4, a_i, 0, NULL, 1, &oldtype,
MPI_COMBINER_DARRAY );
}
cleanup:
if (NULL != st_offsets) free(st_offsets);
if (NULL != coords) free(coords);
OMPI_ERRHANDLER_RETURN(rc, MPI_COMM_WORLD, rc, FUNC_NAME);
}
static int local_copy_with_convertor( ompi_datatype_t* pdt, int count, int chunk )
{
MPI_Aint extent;
void *pdst = NULL, *psrc = NULL, *ptemp = NULL;
ompi_convertor_t *send_convertor = NULL, *recv_convertor = NULL;
struct iovec iov;
uint32_t iov_count;
size_t max_data;
int32_t length = 0, done1 = 0, done2 = 0;
TIMER_DATA_TYPE start, end, unpack_start, unpack_end;
long total_time, unpack_time = 0;
ompi_ddt_type_extent( pdt, &extent );
pdst = malloc( extent * count );
psrc = malloc( extent * count );
ptemp = malloc( chunk );
{
int i;
for( i = 0; i < (count * extent); i++ )
((char*)psrc)[i] = i % 128 + 32;
}
memset( pdst, 0, count * extent );
send_convertor = ompi_convertor_create( remote_arch, 0 );
if( OMPI_SUCCESS != ompi_convertor_prepare_for_send( send_convertor, pdt, count, psrc ) ) {
printf( "Unable to create the send convertor. Is the datatype committed ?\n" );
goto clean_and_return;
}
recv_convertor = ompi_convertor_create( remote_arch, 0 );
if( OMPI_SUCCESS != ompi_convertor_prepare_for_recv( recv_convertor, pdt, count, pdst ) ) {
printf( "Unable to create the recv convertor. Is the datatype committed ?\n" );
goto clean_and_return;
}
cache_trash(); /* make sure the cache is useless */
GET_TIME( start );
while( (done1 & done2) != 1 ) {
/* They are supposed to finish in exactly the same time. */
if( done1 | done2 ) {
printf( "WRONG !!! the send is %s but the receive is %s in local_copy_with_convertor\n",
(done1 ? "finish" : "not finish"),
(done2 ? "finish" : "not finish") );
}
max_data = chunk;
iov_count = 1;
iov.iov_base = ptemp;
iov.iov_len = chunk;
if( done1 == 0 ) {
done1 = ompi_convertor_pack( send_convertor, &iov, &iov_count, &max_data );
}
if( done2 == 0 ) {
GET_TIME( unpack_start );
done2 = ompi_convertor_unpack( recv_convertor, &iov, &iov_count, &max_data );
GET_TIME( unpack_end );
unpack_time += ELAPSED_TIME( unpack_start, unpack_end );
}
length += max_data;
}
GET_TIME( end );
total_time = ELAPSED_TIME( start, end );
printf( "copying same data-type using convertors in %ld microsec\n", total_time );
printf( "\t unpack in %ld microsec [pack in %ld microsec]\n", unpack_time,
total_time - unpack_time );
clean_and_return:
if( NULL != send_convertor ) OBJ_RELEASE( send_convertor );
if( NULL != recv_convertor ) OBJ_RELEASE( recv_convertor );
if( NULL != pdst ) free( pdst );
if( NULL != psrc ) free( psrc );
if( NULL != ptemp ) free( ptemp );
return OMPI_SUCCESS;
}
static int mpich_typeub3( void )
{
int blocklen[2], err = 0, idisp[3];
size_t sz;
MPI_Aint disp[3], lb, ub, ex;
ompi_datatype_t *types[3], *dt1, *dt2, *dt3, *dt4, *dt5;
/* Create a datatype with explicit LB and UB */
blocklen[0] = 1;
blocklen[1] = 1;
blocklen[2] = 1;
disp[0] = -3;
disp[1] = 0;
disp[2] = 6;
types[0] = &ompi_mpi_lb; /* ompi_ddt_basicDatatypes[DT_LB]; */
types[1] = &ompi_mpi_int; /* ompi_ddt_basicDatatypes[DT_INT]; */
types[2] = &ompi_mpi_ub; /* ompi_ddt_basicDatatypes[DT_UB]; */
/* Generate samples for contiguous, hindexed, hvector, indexed, and vector (struct and contiguous tested in typeub2) */
ompi_ddt_create_struct(3,blocklen,disp, types,&dt1);
ompi_ddt_commit(&dt1);
/* This type is the same as in typeub2, and is tested there */
types[0]=dt1; types[1]=dt1;
blocklen[0]=1; blocklen[1]=1;
disp[0]=-4; disp[1]=7;
idisp[0]=-4; idisp[1]=7;
ompi_ddt_create_hindexed( 2, blocklen, disp, dt1, &dt2 );
ompi_ddt_commit( &dt2 );
ompi_ddt_type_lb( dt2, &lb ); ompi_ddt_type_ub( dt2, &ub );
ompi_ddt_type_extent( dt2, &ex ); ompi_ddt_type_size( dt2, &sz );
if (lb != -7 || ub != 13 || ex != 20) {
printf("hindexed lb %d ub %d extent %d size %d\n", (int)-7, (int)13, (int)20, (int)sz);
printf("hindexed lb %d ub %d extent %d size %d\n", (int)lb, (int)ub, (int)ex, (int)sz);
err++;
}
else
printf( "hindexed ok\n" );
ompi_ddt_create_indexed( 2, blocklen, idisp, dt1, &dt3 );
ompi_ddt_commit( &dt3 );
ompi_ddt_type_lb( dt3, &lb ); ompi_ddt_type_ub( dt3, &ub );
ompi_ddt_type_extent( dt3, &ex ); ompi_ddt_type_size( dt3, &sz );
if (lb != -39 || ub != 69 || ex != 108) {
printf("indexed lb %d ub %d extent %d size %d\n", (int)-39, (int)69, (int)108, (int)sz);
printf("indexed lb %d ub %d extent %d size %d\n", (int)lb, (int)ub, (int)ex, (int)sz);
err++;
}
else
printf( "indexed ok\n" );
ompi_ddt_create_hvector( 2, 1, 14, dt1, &dt4 );
ompi_ddt_commit( &dt4 );
ompi_ddt_type_lb( dt4, &lb ); ompi_ddt_type_ub( dt4, &ub );
ompi_ddt_type_extent( dt4, &ex ); ompi_ddt_type_size( dt4, &sz );
if (lb != -3 || ub != 20 || ex != 23) {
printf("hvector lb %d ub %d extent %d size %d\n", (int)-3, (int)20, (int)23, (int)sz);
printf("hvector lb %d ub %d extent %d size %d\n", (int)lb, (int)ub, (int)ex, (int)sz);
err++;
}
else
printf( "hvector ok\n" );
ompi_ddt_create_vector( 2, 1, 14, dt1, &dt5 );
ompi_ddt_commit( &dt5 );
ompi_ddt_type_lb( dt5, &lb ); ompi_ddt_type_ub( dt5, &ub );
ompi_ddt_type_extent( dt5, &ex ); ompi_ddt_type_size( dt5, &sz );
if (lb != -3 || ub != 132 || ex != 135) {
printf("vector lb %d ub %d extent %d size %d\n", (int)-3, (int)132, (int)135, (int)sz);
printf("vector lb %d ub %d extent %d size %d\n", (int)lb, (int)ub, (int)ex, (int)sz);
err++;
}
else
printf( "vector ok\n" );
OBJ_RELEASE( dt1 ); /*assert( dt1 == NULL );*/
OBJ_RELEASE( dt2 ); /*assert( dt2 == NULL );*/
OBJ_RELEASE( dt3 ); /*assert( dt3 == NULL );*/
OBJ_RELEASE( dt4 ); /*assert( dt4 == NULL );*/
OBJ_RELEASE( dt5 ); assert( dt5 == NULL );
return err;
}
static int mpich_typeub2( void )
{
int blocklen[3], err = 0;
size_t sz1, sz2, sz3;
MPI_Aint disp[3], lb, ub, ex1, ex2, ex3;
ompi_datatype_t *types[3], *dt1, *dt2, *dt3;
blocklen[0] = 1;
blocklen[1] = 1;
blocklen[2] = 1;
disp[0] = -3;
disp[1] = 0;
disp[2] = 6;
types[0] = &ompi_mpi_lb; /* ompi_ddt_basicDatatypes[DT_LB]; */
types[1] = &ompi_mpi_int; /* ompi_ddt_basicDatatypes[DT_INT]; */
types[2] = &ompi_mpi_ub; /* ompi_ddt_basicDatatypes[DT_UB]; */
ompi_ddt_create_struct(3,blocklen,disp, types,&dt1);
ompi_ddt_commit(&dt1);
ompi_ddt_type_lb(dt1, &lb); ompi_ddt_type_ub(dt1, &ub);
ompi_ddt_type_extent(dt1,&ex1); ompi_ddt_type_size(dt1,&sz1);
/* Values should be lb = -3, ub = 6 extent 9; size depends on implementation */
if (lb != -3 || ub != 6 || ex1 != 9) {
printf("Example 3.26 type1 lb %d ub %d extent %d size %d\n", (int)lb, (int)ub, (int)ex1, (int)sz1);
err++;
}
else
printf("Example 3.26 type1 correct\n" );
ompi_ddt_create_contiguous(2,dt1,&dt2);
ompi_ddt_type_lb(dt2, &lb); ompi_ddt_type_ub(dt2, &ub);
ompi_ddt_type_extent(dt2,&ex2); ompi_ddt_type_size(dt2,&sz2);
/* Values should be lb = -3, ub = 15, extent = 18, size depends on implementation */
if (lb != -3 || ub != 15 || ex2 != 18) {
printf("Example 3.26 type2 lb %d ub %d extent %d size %d\n", (int)-3, (int)15, (int)18, 8);
printf("Example 3.26 type2 lb %d ub %d extent %d size %d\n", (int)lb, (int)ub, (int)ex2, (int)sz2);
err++;
}
else
printf("Example 3.26 type1 correct\n" );
OBJ_RELEASE( dt2 ); assert( dt2 == NULL );
ompi_ddt_create_contiguous(2,dt1,&dt2);
ompi_ddt_type_lb(dt2, &lb); ompi_ddt_type_ub(dt2, &ub);
ompi_ddt_type_extent(dt2,&ex2); ompi_ddt_type_size(dt2,&sz2);
/* Values should be lb = -3, ub = 15, extent = 18, size depends on implementation */
if (lb != -3 || ub != 15 || ex2 != 18) {
printf("Example 3.26 type2 lb %d ub %d extent %d size %d\n", (int)-3, (int)15, (int)18, 8);
printf("Example 3.26 type2 lb %d ub %d extent %d size %d\n", (int)lb, (int)ub, (int)ex2, (int)sz2);
err++;
}
else
printf( "Example 3.26 type2 correct\n" );
types[0]=dt1; types[1]=dt1;
blocklen[0]=1; blocklen[1]=1;
disp[0]=0; disp[1]=ex1;
ompi_ddt_create_struct(2, blocklen, disp, types, &dt3);
ompi_ddt_commit(&dt3);
ompi_ddt_type_lb(dt3, &lb); ompi_ddt_type_ub(dt3, &ub);
ompi_ddt_type_extent(dt3,&ex3); ompi_ddt_type_size(dt3,&sz3);
/* Another way to express type2 */
if (lb != -3 || ub != 15 || ex3 != 18) {
printf("type3 lb %d ub %d extent %d size %d\n", (int)-3, (int)15, (int)18, 8);
printf("type3 lb %d ub %d extent %d size %d\n", (int)lb, (int)ub, (int)ex3, (int)sz2);
err++;
}
else
printf( "type3 correct\n" );
OBJ_RELEASE( dt1 ); /*assert( dt1 == NULL );*/
OBJ_RELEASE( dt2 ); /*assert( dt2 == NULL );*/
OBJ_RELEASE( dt3 ); assert( dt3 == NULL );
return err;
}
