1
1
true_lb while computing the lower bound.
Этот коммит содержится в:
George Bosilca 2014-11-13 23:22:01 -05:00 коммит произвёл Gilles Gouaillardet
родитель 705147e98b
Коммит d622db783d
20 изменённых файлов: 72 добавлений и 68 удалений

Просмотреть файл

@ -2,7 +2,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2005 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -80,6 +80,7 @@ mca_coll_basic_allreduce_inter(void *sbuf, void *rbuf, int count,
{
int err, i, rank, root = 0, rsize;
ptrdiff_t lb, extent;
ptrdiff_t true_lb, true_extent;
char *tmpbuf = NULL, *pml_buffer = NULL;
ompi_request_t *req[2];
mca_coll_basic_module_t *basic_module = (mca_coll_basic_module_t*) module;
@ -102,12 +103,16 @@ mca_coll_basic_allreduce_inter(void *sbuf, void *rbuf, int count,
if (OMPI_SUCCESS != err) {
return OMPI_ERROR;
}
err = ompi_datatype_get_true_extent(dtype, &true_lb, &true_extent);
if (OMPI_SUCCESS != err) {
return OMPI_ERROR;
}
tmpbuf = (char *) malloc(count * extent);
tmpbuf = (char *) malloc(true_extent + (count - 1) * extent);
if (NULL == tmpbuf) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
pml_buffer = tmpbuf - lb;
pml_buffer = tmpbuf - true_lb;
/* Do a send-recv between the two root procs. to avoid deadlock */
err = MCA_PML_CALL(irecv(rbuf, count, dtype, 0,

Просмотреть файл

@ -2,7 +2,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -88,7 +88,7 @@ mca_coll_basic_exscan_intra(void *sbuf, void *rbuf, int count,
if (NULL == free_buffer) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
reduce_buffer = free_buffer - lb;
reduce_buffer = free_buffer - true_lb;
err = ompi_datatype_copy_content_same_ddt(dtype, count,
reduce_buffer, (char*)sbuf);

Просмотреть файл

@ -81,8 +81,7 @@ mca_coll_basic_reduce_lin_intra(void *sbuf, void *rbuf, int count,
* in the layout pattern in the datatype.
*
* For example, consider the following buffer (just talking about
* LB, extent, and true extent -- extrapolate for UB; i.e., assume
* the UB equals exactly where the data ends):
* true_lb, extent, and true extent -- extrapolate for true_ub:
*
* A B C
* --------------------------------------------------------
@ -96,7 +95,7 @@ mca_coll_basic_reduce_lin_intra(void *sbuf, void *rbuf, int count,
*
* - extent: C-A
* - true extent: C-A
* - LB: 0
* - true_lb: 0
*
* A C
* --------------------------------------------------------
@ -110,7 +109,7 @@ mca_coll_basic_reduce_lin_intra(void *sbuf, void *rbuf, int count,
*
* - extent: C-A
* - true extent: C-B
* - LB: positive
* - true_lb: positive
*
* A B C
* --------------------------------------------------------
@ -124,7 +123,7 @@ mca_coll_basic_reduce_lin_intra(void *sbuf, void *rbuf, int count,
*
* - extent: C-A
* - true extent: C-A
* - LB: negative
* - true_lb: negative
*
* A B C
* --------------------------------------------------------
@ -139,7 +138,7 @@ mca_coll_basic_reduce_lin_intra(void *sbuf, void *rbuf, int count,
*
* - extent: C-MPI_BOTTOM
* - true extent: C-B
* - LB: [potentially very large] positive
* - true_lb: [potentially very large] positive
*
* MPI_BOTTOM B C
* --------------------------------------------------------
@ -165,19 +164,19 @@ mca_coll_basic_reduce_lin_intra(void *sbuf, void *rbuf, int count,
*
* 2. If B is what we get back from malloc, but we give A to
* MPI_Send, then the buffer range [A,B) represents "dead space"
* -- no data will be put there. So it's safe to give B-LB to
* MPI_Send. More specifically, the LB is positive, so B-LB is
* -- no data will be put there. So it's safe to give B-true_lb to
* MPI_Send. More specifically, the true_lb is positive, so B-true_lb is
* actually A.
*
* 3. If A is what we get back from malloc, and B is what we give to
* MPI_Send, then the LB is negative, so A-LB will actually equal
* MPI_Send, then the true_lb is negative, so A-true_lb will actually equal
* B.
*
* 4. Although this seems like the weirdest case, it's actually
* quite similar to case #2 -- the pointer we give to MPI_Send is
* smaller than the pointer we got back from malloc().
*
* Hence, in all cases, we give (return_from_malloc - LB) to MPI_Send.
* Hence, in all cases, we give (return_from_malloc - true_lb) to MPI_Send.
*
* This works fine and dandy if we only have (count==1), which we
* rarely do. ;-) So we really need to allocate (true_extent +
@ -197,7 +196,7 @@ mca_coll_basic_reduce_lin_intra(void *sbuf, void *rbuf, int count,
if (NULL == inplace_temp) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
rbuf = inplace_temp - lb;
rbuf = inplace_temp - true_lb;
}
if (size > 1) {
@ -208,7 +207,7 @@ mca_coll_basic_reduce_lin_intra(void *sbuf, void *rbuf, int count,
}
return OMPI_ERR_OUT_OF_RESOURCE;
}
pml_buffer = free_buffer - lb;
pml_buffer = free_buffer - true_lb;
}
/* Initialize the receive buffer. */
@ -361,7 +360,7 @@ mca_coll_basic_reduce_log_intra(void *sbuf, void *rbuf, int count,
return OMPI_ERR_OUT_OF_RESOURCE;
}
pml_buffer = free_buffer - lb;
pml_buffer = free_buffer - true_lb;
/* read the comment about commutative operations (few lines down
* the page) */
if (ompi_op_is_commute(op)) {
@ -377,7 +376,7 @@ mca_coll_basic_reduce_log_intra(void *sbuf, void *rbuf, int count,
err = OMPI_ERR_OUT_OF_RESOURCE;
goto cleanup_and_return;
}
sbuf = inplace_temp - lb;
sbuf = inplace_temp - true_lb;
err = ompi_datatype_copy_content_same_ddt(dtype, count, (char*)sbuf, (char*)rbuf);
}
snd_buffer = (char*)sbuf;
@ -391,7 +390,7 @@ mca_coll_basic_reduce_log_intra(void *sbuf, void *rbuf, int count,
err = OMPI_ERR_OUT_OF_RESOURCE;
goto cleanup_and_return;
}
rbuf = free_rbuf - lb;
rbuf = free_rbuf - true_lb;
}
/* Loop over cube dimensions. High processes send to low ones in the
@ -546,7 +545,7 @@ mca_coll_basic_reduce_lin_inter(void *sbuf, void *rbuf, int count,
if (NULL == free_buffer) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
pml_buffer = free_buffer - lb;
pml_buffer = free_buffer - true_lb;
/* Initialize the receive buffer. */

Просмотреть файл

@ -111,7 +111,7 @@ mca_coll_basic_reduce_scatter_intra(void *sbuf, void *rbuf, int *rcounts,
/* temporary receive buffer. See coll_basic_reduce.c for details on sizing */
recv_buf_free = (char*) malloc(buf_size);
recv_buf = recv_buf_free - lb;
recv_buf = recv_buf_free - true_lb;
if (NULL == recv_buf_free) {
err = OMPI_ERR_OUT_OF_RESOURCE;
goto cleanup;
@ -119,7 +119,7 @@ mca_coll_basic_reduce_scatter_intra(void *sbuf, void *rbuf, int *rcounts,
/* allocate temporary buffer for results */
result_buf_free = (char*) malloc(buf_size);
result_buf = result_buf_free - lb;
result_buf = result_buf_free - true_lb;
/* copy local buffer into the temporary results */
err = ompi_datatype_sndrcv(sbuf, count, dtype, result_buf, count, dtype);
@ -323,7 +323,7 @@ mca_coll_basic_reduce_scatter_intra(void *sbuf, void *rbuf, int *rcounts,
/* temporary receive buffer. See coll_basic_reduce.c for
details on sizing */
recv_buf_free = (char*) malloc(buf_size);
recv_buf = recv_buf_free - lb;
recv_buf = recv_buf_free - true_lb;
if (NULL == recv_buf_free) {
err = OMPI_ERR_OUT_OF_RESOURCE;
goto cleanup;

Просмотреть файл

@ -2,7 +2,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -85,7 +85,7 @@ mca_coll_basic_reduce_scatter_block_intra(void *sbuf, void *rbuf, int rcount,
/* temporary receive buffer. See coll_basic_reduce.c for
details on sizing */
recv_buf_free = (char*) malloc(buf_size);
recv_buf = recv_buf_free - lb;
recv_buf = recv_buf_free - true_lb;
if (NULL == recv_buf_free) {
err = OMPI_ERR_OUT_OF_RESOURCE;
goto cleanup;

Просмотреть файл

@ -2,7 +2,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -79,7 +79,7 @@ mca_coll_basic_scan_intra(void *sbuf, void *rbuf, int count,
if (NULL == free_buffer) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
pml_buffer = free_buffer - lb;
pml_buffer = free_buffer - true_lb;
/* Copy the send buffer into the receive buffer. */

Просмотреть файл

@ -50,7 +50,7 @@ mca_coll_cuda_allreduce(void *sbuf, void *rbuf, int count,
}
opal_cuda_memcpy_sync(sbuf1, sbuf, bufsize);
sbuf2 = sbuf; /* save away original buffer */
sbuf = sbuf1 - lb;
sbuf = sbuf1 - true_lb;
}
if (opal_cuda_check_bufs(rbuf, NULL)) {
@ -61,7 +61,7 @@ mca_coll_cuda_allreduce(void *sbuf, void *rbuf, int count,
}
opal_cuda_memcpy_sync(rbuf1, rbuf, bufsize);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - lb;
rbuf = rbuf1 - true_lb;
}
rc = s->c_coll.coll_allreduce(sbuf, rbuf, count, dtype, op, comm, s->c_coll.coll_allreduce_module);
if (NULL != sbuf1) {

Просмотреть файл

@ -42,7 +42,7 @@ int mca_coll_cuda_exscan(void *sbuf, void *rbuf, int count,
}
opal_cuda_memcpy_sync(sbuf1, sbuf, bufsize);
sbuf2 = sbuf; /* save away original buffer */
sbuf = sbuf1 - lb;
sbuf = sbuf1 - true_lb;
}
if (opal_cuda_check_bufs(rbuf, NULL)) {
@ -53,7 +53,7 @@ int mca_coll_cuda_exscan(void *sbuf, void *rbuf, int count,
}
opal_cuda_memcpy_sync(rbuf1, rbuf, bufsize);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - lb;
rbuf = rbuf1 - true_lb;
}
rc = s->c_coll.coll_exscan(sbuf, rbuf, count, dtype, op, comm,

Просмотреть файл

@ -55,7 +55,7 @@ mca_coll_cuda_reduce_scatter_block(void *sbuf, void *rbuf, int rcount,
}
opal_cuda_memcpy_sync(sbuf1, sbuf, sbufsize);
sbuf2 = sbuf; /* save away original buffer */
sbuf = sbuf1 - lb;
sbuf = sbuf1 - true_lb;
}
if (opal_cuda_check_bufs(rbuf, NULL)) {
@ -66,7 +66,7 @@ mca_coll_cuda_reduce_scatter_block(void *sbuf, void *rbuf, int rcount,
}
opal_cuda_memcpy_sync(rbuf1, rbuf, rbufsize);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - lb;
rbuf = rbuf1 - true_lb;
}
rc = s->c_coll.coll_reduce_scatter_block(sbuf, rbuf, rcount, dtype, op, comm,
s->c_coll.coll_reduce_scatter_block_module);

Просмотреть файл

@ -49,7 +49,7 @@ int mca_coll_cuda_scan(void *sbuf, void *rbuf, int count,
}
opal_cuda_memcpy_sync(sbuf1, sbuf, bufsize);
sbuf2 = sbuf; /* save away original buffer */
sbuf = sbuf1 - lb;
sbuf = sbuf1 - true_lb;
}
if (opal_cuda_check_bufs(rbuf, NULL)) {
@ -60,7 +60,7 @@ int mca_coll_cuda_scan(void *sbuf, void *rbuf, int count,
}
opal_cuda_memcpy_sync(rbuf1, rbuf, bufsize);
rbuf2 = rbuf; /* save away original buffer */
rbuf = rbuf1 - lb;
rbuf = rbuf1 - true_lb;
}
rc = s->c_coll.coll_scan(sbuf, rbuf, count, dtype, op, comm,
s->c_coll.coll_scan_module);

Просмотреть файл

@ -2,7 +2,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -70,7 +70,7 @@ int mca_coll_hierarch_allreduce_intra(void *sbuf, void *rbuf, int count,
if (NULL == tbuf) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
tmpbuf = tbuf - lb;
tmpbuf = tbuf - true_lb;
if ( MPI_IN_PLACE != sbuf ) {
ret = lcomm->c_coll.coll_reduce (sbuf, tmpbuf, count, dtype,

Просмотреть файл

@ -2,7 +2,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -69,7 +69,7 @@ int mca_coll_hierarch_reduce_intra(void *sbuf, void *rbuf, int count,
if (NULL == tbuf) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
tmpbuf = tbuf - lb;
tmpbuf = tbuf - true_lb;
if ( MPI_IN_PLACE != sbuf ) {

Просмотреть файл

@ -2,7 +2,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -64,7 +64,7 @@ mca_coll_inter_reduce_inter(void *sbuf, void *rbuf, int count,
if (NULL == free_buffer) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
pml_buffer = free_buffer - lb;
pml_buffer = free_buffer - true_lb;
err = comm->c_local_comm->c_coll.coll_reduce(sbuf, pml_buffer, count,
dtype, op, 0, comm->c_local_comm,

Просмотреть файл

@ -2,7 +2,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -271,7 +271,7 @@ static int reduce_inorder(void *sbuf, void* rbuf, int count,
if (NULL == free_buffer) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
reduce_temp_buffer = free_buffer - lb;
reduce_temp_buffer = free_buffer - true_lb;
/* Trickery here: we use a potentially smaller count than
the user count -- use the largest count that is <=
@ -319,7 +319,7 @@ static int reduce_inorder(void *sbuf, void* rbuf, int count,
}
return OMPI_ERR_OUT_OF_RESOURCE;
}
sbuf = inplace_temp - lb;
sbuf = inplace_temp - true_lb;
ompi_datatype_copy_content_same_ddt(dtype, count, (char *) sbuf, (char *) rbuf);
} else {
inplace_temp = NULL;

Просмотреть файл

@ -2,7 +2,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2012 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -181,8 +181,8 @@ int ompi_coll_tuned_allgather_intra_bruck(void *sbuf, int scount,
- create temporary shift buffer,
see discussion in coll_basic_reduce.c about the size and begining
of temporary buffer.
- copy blocks [0 .. (size - rank - 1)] in rbuf to shift buffer
- move blocks [(size - rank) .. size] in rbuf to begining of rbuf
- copy blocks [0 .. (size - rank - 1)] from rbuf to shift buffer
- move blocks [(size - rank) .. size] from rbuf to begining of rbuf
- copy blocks from shift buffer starting at block [rank] in rbuf.
*/
if (0 != rank) {
@ -192,14 +192,14 @@ int ompi_coll_tuned_allgather_intra_bruck(void *sbuf, int scount,
err = ompi_datatype_get_true_extent(rdtype, &true_lb, &true_extent);
if (MPI_SUCCESS != err) { line = __LINE__; goto err_hndl; }
free_buf = (char*) calloc(((true_extent +
free_buf = (char*) calloc(((true_extent +
((ptrdiff_t)(size - rank) * (ptrdiff_t)rcount - 1) * rext)),
sizeof(char));
if (NULL == free_buf) {
line = __LINE__; err = OMPI_ERR_OUT_OF_RESOURCE; goto err_hndl;
}
shift_buf = free_buf - true_lb;
/* 1. copy blocks [0 .. (size - rank - 1)] from rbuf to shift buffer */
err = ompi_datatype_copy_content_same_ddt(rdtype, ((ptrdiff_t)(size - rank) * (ptrdiff_t)rcount),
shift_buf, rbuf);

Просмотреть файл

@ -3,7 +3,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2012 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -412,8 +412,8 @@ ompi_coll_tuned_allreduce_intra_ring(void *sbuf, void *rbuf, int count,
Early blocks are at most 1 element larger than the late ones.
*/
COLL_TUNED_COMPUTE_BLOCKCOUNT( count, size, split_rank,
early_segcount, late_segcount )
max_segcount = early_segcount;
early_segcount, late_segcount );
max_segcount = early_segcount;
max_real_segsize = true_extent + (max_segcount - 1) * extent;
@ -698,10 +698,10 @@ ompi_coll_tuned_allreduce_intra_ring_segmented(void *sbuf, void *rbuf, int count
out of the largest one will have max_segcount elements.
*/
COLL_TUNED_COMPUTE_BLOCKCOUNT( count, size, split_rank,
early_blockcount, late_blockcount )
COLL_TUNED_COMPUTE_BLOCKCOUNT( early_blockcount, num_phases, inbi,
max_segcount, k)
max_real_segsize = true_extent + (ptrdiff_t)(max_segcount - 1) * extent;
early_blockcount, late_blockcount );
COLL_TUNED_COMPUTE_BLOCKCOUNT( early_blockcount, num_phases, inbi,
max_segcount, k);
max_real_segsize = true_extent + (ptrdiff_t)(max_segcount - 1) * extent;
/* Allocate and initialize temporary buffers */
inbuf[0] = (char*)malloc(max_real_segsize);

Просмотреть файл

@ -3,7 +3,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2012 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -102,7 +102,7 @@ ompi_coll_tuned_gather_intra_binomial(void *sbuf, int scount,
err= OMPI_ERR_OUT_OF_RESOURCE; line = __LINE__; goto err_hndl;
}
ptmp = tempbuf - rlb;
ptmp = tempbuf - rtrue_lb;
if (sbuf != MPI_IN_PLACE) {
/* copy from sbuf to temp buffer */
err = ompi_datatype_sndrcv(sbuf, scount, sdtype,
@ -125,7 +125,7 @@ ompi_coll_tuned_gather_intra_binomial(void *sbuf, int scount,
err= OMPI_ERR_OUT_OF_RESOURCE; line = __LINE__; goto err_hndl;
}
ptmp = tempbuf - slb;
ptmp = tempbuf - strue_lb;
/* local copy to tempbuf */
err = ompi_datatype_sndrcv(sbuf, scount, sdtype,
ptmp, scount, sdtype);

Просмотреть файл

@ -656,7 +656,7 @@ ompi_coll_tuned_reduce_intra_basic_linear(void *sbuf, void *rbuf, int count,
if (NULL == inplace_temp) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
rbuf = inplace_temp - lb;
rbuf = inplace_temp - true_lb;
}
if (size > 1) {
@ -665,7 +665,7 @@ ompi_coll_tuned_reduce_intra_basic_linear(void *sbuf, void *rbuf, int count,
err = OMPI_ERR_OUT_OF_RESOURCE;
goto exit;
}
pml_buffer = free_buffer - lb;
pml_buffer = free_buffer - true_lb;
}
/* Initialize the receive buffer. */

Просмотреть файл

@ -3,7 +3,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2012 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -186,7 +186,7 @@ ompi_coll_tuned_reduce_scatter_intra_basic_recursivehalving(void *sbuf,
/* Allocate temporary receive buffer. */
recv_buf_free = (char*) malloc(buf_size);
recv_buf = recv_buf_free - lb;
recv_buf = recv_buf_free - true_lb;
if (NULL == recv_buf_free) {
err = OMPI_ERR_OUT_OF_RESOURCE;
goto cleanup;
@ -194,7 +194,7 @@ ompi_coll_tuned_reduce_scatter_intra_basic_recursivehalving(void *sbuf,
/* allocate temporary buffer for results */
result_buf_free = (char*) malloc(buf_size);
result_buf = result_buf_free - lb;
result_buf = result_buf_free - true_lb;
/* copy local buffer into the temporary results */
err = ompi_datatype_sndrcv(sbuf, count, dtype, result_buf, count, dtype);

Просмотреть файл

@ -3,7 +3,7 @@
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2012 The University of Tennessee and The University
* Copyright (c) 2004-2014 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
@ -100,7 +100,7 @@ ompi_coll_tuned_scatter_intra_binomial(void *sbuf, int scount,
err = OMPI_ERR_OUT_OF_RESOURCE; line = __LINE__; goto err_hndl;
}
ptmp = tempbuf - slb;
ptmp = tempbuf - strue_lb;
/* and rotate data so they will eventually in the right place */
err = ompi_datatype_copy_content_same_ddt(sdtype, (ptrdiff_t)scount * (ptrdiff_t)(size - root),
@ -128,7 +128,7 @@ ompi_coll_tuned_scatter_intra_binomial(void *sbuf, int scount,
err= OMPI_ERR_OUT_OF_RESOURCE; line = __LINE__; goto err_hndl;
}
ptmp = tempbuf - rlb;
ptmp = tempbuf - rtrue_lb;
sdtype = rdtype;
scount = rcount;