Partitioned communication was introduced in MPI 4.0 as a user-friendly interface to support pipelined communication patterns, particularly common in the context of MPI+threads. It provides the user with the ability to divide a global buffer into smaller independent chunks, called partitions, which can then be communicated independently. In this work we first model the performance gain that can be expected when using partitioned communication. Next, we describe the improvements we made to MPICH to enable those gains and provide a high-quality implementation of MPI partitioned communication. We then evaluate partitioned communication in various common use cases and assess the performance in comparison with other MPI point-to-point and one-sided approaches. Specifically, we first investigate two scenarios commonly encountered for small partition sizes in a multithreaded environment: thread contention and overhead of using many partitions. We propose two solutions to alleviate the measured penalty and demonstrate their use. We then focus on large messages and the gain obtained when exploiting the delay resulting from computations or load imbalance. We conclude with our perspectives on the benefits of partitioned communication and the various results obtained.

Massively parallel Fourier transforms are widely used in computational sciences, and specifically in computational fluid dynamics which involves unbounded Poisson problems. In practice the latter is usually the most time-consuming operation due to its inescapable all-to-all communication pattern. The originalflupslibrary tackles that issue with an implementation of the distributed Fourier transform tailor-made for successive resolutions of unbounded Poisson problems. However the proposed implementation lacks of flexibility as it only supports cell-centered data layout and features a plain communication strategy. This work extends the library along two directions. First,flups’ implementation is generalized to support a node-centered data layout. Second, three distinct approaches are provided to handle the communications: one all-to-all, and two non-blocking implementations relying on manual packing andMPI_Datatypeto communicate over the network. The proposed software is validated against analytical solutions for unbounded, semi-unbounded, and periodic domains. The performance of the approaches is then compared againstaccFFT, another distributed FFT implementation, using a periodic case. Finally the performance metrics of each implementation are analyzed and detailed on various top-tier European facilities up to 49,152 cores. This work bringsflupsup to a fully production-ready and performant distributed FFT library, featuring all the possible types of FFTs and with flexibility in the data-layout.