You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: src/MPIOperator/mpioperator.jl
+3-1Lines changed: 3 additions & 1 deletion
Original file line number
Diff line number
Diff line change
@@ -1,4 +1,6 @@
1
1
## This shallow struct is used to indicate that each LazyMIPOperator should be evaluated on each rank and the result is to be reduced across all ranks using MPI.Allreduce
2
+
## This is the MPI-parallelized version of a linear operator
3
+
## If one added the flexibilty of choosing the reduction, one could also parallelize over products of functions etc...
2
4
struct MPIOperator{O}
3
5
parent::O
4
6
functionMPIOperator(parent::O) where {O}
@@ -15,7 +17,7 @@ end
15
17
16
18
function (Op::MPIOperator{O})(x::S) where {O,S}
17
19
y_per_rank =parent(Op)(x)
18
-
y =large_allreduce(y_per_rank, +, MPI.COMM_WORLD)
20
+
y =MPIHelper.allreduce(y_per_rank, +, MPI.COMM_WORLD)
0 commit comments