From: Augustin Degomme <adegomme@gmail.com>
Date: Wed, 23 Feb 2022 08:52:36 +0000 (+0000)
Subject: pending implementation of new algorithms (see fixme), avoid using _lr as replacement... 
X-Git-Tag: v3.31~341
X-Git-Url: http://bilbo.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/commitdiff_plain/6f125e09100261d6272a81b794a81ab01baf0df7

pending implementation of new algorithms (see fixme), avoid using _lr as replacement in these cases.
_lr is not valid for some sizes, and may default to calling _ompi again, which causes an infinite loop.
todo: find theses cases and eliminate them
---

diff --git a/src/smpi/colls/smpi_openmpi_selector.cpp b/src/smpi/colls/smpi_openmpi_selector.cpp
index 2809babd76..d1607bdff3 100644
--- a/src/smpi/colls/smpi_openmpi_selector.cpp
+++ b/src/smpi/colls/smpi_openmpi_selector.cpp
@@ -30,8 +30,8 @@ int allreduce__ompi(const void *sbuf, void *rbuf, int count,
     int communicator_size = comm->size();
     int alg = 1;
     int(*funcs[]) (const void*, void*, int, MPI_Datatype, MPI_Op, MPI_Comm)={
-        &allreduce__lr,
-        &allreduce__lr,
+        &allreduce__redbcast,
+        &allreduce__redbcast,
         &allreduce__rdb,
         &allreduce__lr,
         &allreduce__ompi_ring_segmented,