Exercise E4:

• Implement the program all_to_all using send=MPI_Ssend, receive=MPI_Recv. Explain the result and correct it.
• Does the same problem occur if you work with send=MPI_Isend, receive=MPI_Recv? Explain the result.
• Implement all_to_all again with send=MPI_Send, receive=MPI_Recv. For which message size do you find a deadlock?
• Implement all_to_all using MPI_Alltoall

---

 Back