Skip to content

Commit 88aad4d

Browse files
committed
coll/libnbc: correctly handle MPI_BOTTOM
if both send and receive buffers are MPI_BOTTOM, this is not equivalent to MPI_IN_PLACE Thanks Lisandro Dalcin for reporting this issue. Refs. #9650 Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
1 parent ed05ee1 commit 88aad4d

File tree

1 file changed

+2
-2
lines changed

1 file changed

+2
-2
lines changed

ompi/mca/coll/libnbc/nbc_internal.h

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,7 @@
1010
*
1111
* Copyright (c) 2012 Oracle and/or its affiliates. All rights reserved.
1212
* Copyright (c) 2014 NVIDIA Corporation. All rights reserved.
13-
* Copyright (c) 2015-2018 Research Organization for Information Science
13+
* Copyright (c) 2015-2021 Research Organization for Information Science
1414
* and Technology (RIST). All rights reserved.
1515
* Copyright (c) 2015 Los Alamos National Security, LLC. All rights
1616
* reserved.
@@ -565,7 +565,7 @@ static inline void NBC_SchedCache_dictwipe(hb_tree *dict_in, int *size) {
565565
#define NBC_IN_PLACE(sendbuf, recvbuf, inplace) \
566566
{ \
567567
inplace = 0; \
568-
if(recvbuf == sendbuf) { \
568+
if(recvbuf == sendbuf && MPI_BOTTOM != sendbuf) { \
569569
inplace = 1; \
570570
} else \
571571
if(sendbuf == MPI_IN_PLACE) { \

0 commit comments

Comments
 (0)