Skip to content

Commit 1ebf9fd

Browse files
markallejjhursey
authored andcommitted
osc/pt2pt: Fix PSCW after Fence wrong answer.
* If the user uses PSCW synchronization after a Fence then the previous epoch is not reset which can cause the PSCW to transfer data before it is ready leading to wrong answers. * This commit resets the `eager_send_active` in the start call. Signed-off-by: Joshua Hursey <[email protected]>
1 parent d8c1a3d commit 1ebf9fd

File tree

1 file changed

+7
-1
lines changed

1 file changed

+7
-1
lines changed

ompi/mca/osc/pt2pt/osc_pt2pt_active_target.c

Lines changed: 7 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,7 @@
1010
* All rights reserved.
1111
* Copyright (c) 2007-2016 Los Alamos National Security, LLC. All rights
1212
* reserved.
13-
* Copyright (c) 2010 IBM Corporation. All rights reserved.
13+
* Copyright (c) 2010-2016 IBM Corporation. All rights reserved.
1414
* Copyright (c) 2012-2013 Sandia National Laboratories. All rights reserved.
1515
* Copyright (c) 2015 Research Organization for Information Science
1616
* and Technology (RIST). All rights reserved.
@@ -227,6 +227,12 @@ int ompi_osc_pt2pt_start (ompi_group_t *group, int assert, ompi_win_t *win)
227227
/* haven't processed any post messages yet */
228228
sync->sync_expected = sync->num_peers;
229229

230+
/* If the previous epoch was from Fence, then eager_send_active is still
231+
* set to true at this time, but it shoulnd't be true until we get our
232+
* incoming Posts. So reset to 'false' for this new epoch.
233+
*/
234+
sync->eager_send_active = false;
235+
230236
OPAL_OUTPUT_VERBOSE((50, ompi_osc_base_framework.framework_output,
231237
"ompi_osc_pt2pt_start entering with group size %d...",
232238
sync->num_peers));

0 commit comments

Comments
 (0)