The paper, titled "One Communicator, Three Vendors: Extending the MoA Predictive Theory to a Real NVIDIA+AMD+CPU MPI Ensemble," describes work that builds on a companion MPI-extension paper. That earlier paper validated the first communication-bearing machine-shape ensemble, and this new paper extends the MoA predictive theory to a more realistic heterogeneous setup.

The ensemble in question consists of four NVIDIA A100 ranks and four CPU ranks on one NCSA Delta node, all running under a single Open MPI job and reassembled via MPI_Gatherv. The abstract does not give numerical results, but the stated aim is to show that the predictive theory holds for a multi-vendor mix of accelerators and CPUs in a real system.

Because the abstract is truncated, the full findings are not available from the source alone. The significance lies in moving from homogeneous or simulated cases to a concrete, three-vendor ensemble, which is a step toward practical use of MoA-style predictions in heterogeneous clusters.