Dear MPI Forum, In chapter 5.9.5 (User-Defined Reduction Operations), page 185 of MPI 3.1 and page 187 of MPI 2019 Draft, it reads : Advice to implementors. We outline below a naive and inefficient implementation of MPI_REDUCE not supporting the "in place" option. Strictly speaking, the outlined implementation is only valid for intra-communicators (and does not support the "in place" option), so you might consider amending this line before MPI 2019 is out. Best regards, Gilles