This page provides a way to update the community on the current status of various MPI implementation with respect to the various vesions of the MPI Standard.

MPI 4.1

Feature MPICH Open MPI
(P)MPI_WTI{ME,CK} and handle conversion macros are symbols ✅  
MPI_REQUEST_GET_STATUS makes progress ✅  
Status field get/set functions ✅  
Send with automatic buffering features ✅  
Getting multiple statuses from an array of requests ✅  
MPI_TYPE_GET_VALUE_INDEX ✅  
MPI_COMM_TYPE_RESOURCE_GUIDED and mpi_pset_name ✅  
mpi_assert_strict_persistent_collective_ordering ✅  
Memory alloc kinds ✅  
Getting names for null handles ✅  
MPI_GET_HW_RESOURCE_INFO ✅  
MPI_ERR_ERRHANDLER ✅  
Removing error class, code and string ✅  
MPI_WIN_SHARED_QUERY supports more window types ✅  
mpi_accumulate_granularity ✅  
# - Under development + - Partly done

MPI 4.0

Feature MPICH Open MPI Intel MPI
Large Counts ✅ # ✅
Partitioned Communication ✅ ✅ ✅
Sessions ✅ ✅ ✅
MPI_T Events ✅ ❌ ❌
Error Handling ✅ ✅ ✅
Non-blocking SENDRECV ✅ ✅ ✅
Persistent Collectives ✅ ✅ ✅
New Split Types ✅ ✅ ✅
MPI_COMM_DUP info ✅ ✅ ✅
Info Assertions ✅ ✅ ✅
Memory Alignment ✅ ✅ ✅
MPI_INFO_CREATE_ENV ✅ ✅ ✅
# - Under development + - Partly done

MPI 3.1

Feature MPICH MVAPICH Open MPI Cray Tianhe Intel MPI IBM (BG/Q - Legacy) IBM (PE - Legacy) IBM (Spectrum) HPE Fujitsu Microsoft MPC NEC Sunway RIKEN AMPI
Non-Blocking Collectives ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
Neighborhood Collectives ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ❌ ✅ ✅ ✅ ✅ ✅
RMA ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ +
Shared Memory ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ #
MPI_T ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ # ✅ ✅ ✅ ✅ Q1 2019
MPI_COMM_CREATE_GROUP ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ # ✅ ✅ ✅ ✅ ✅
F08 Bindings ✅ ✅ ✅ ✅ ✅ ✅ ✅ ❌ ✅ ✅ # ❌ ✅ ✅ ✅ ✅ Q2 2019
New Datatypes ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
Large Counts ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
MPI_MProbe ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
NBC I/O ✅ ✅ ✅ ✅ ❌ ✅ ❌ ❌ ✅ ✅ # ❌ # ✅ ❌ ✅ Q2 2019
# - Under development + - Partly done

Instructions for MPI Implementation Maintainers

If you would like to update the information for your implementation. Create a pull request here to make that change. If you’re unclear on how that works, contact @wesbland or @schulzm for help.