Feasibility study of MPI implementation on the heterogeneous multi-core cell BE™ architecture

Abstract

The Cell Broadband Engine TM is a new heterogeneous multi-core processor from IBM, Sony, and Toshiba. It contains eight coprocessors, called Synergistic Processing Elements (SPEs), which operate directly on distinct 256 KB local stores, and also have access to a shared 512 MB to 2 GB main memory. The combined peak speed of the SPEs is 204.8 Gflop/s in single precision and 14.64 Gflop/s in double precision. There is, therefore, much interest in using the Cell BE TM for high performance computing applications. However, the unconventional architecture of the SPEs, in particular their local stores, creates some programming challenges. We describe our implementation of certain core features of MPI, such as blocking point-to-point calls and collective communication calls, which can help meet these challenges, by enabling a large class of MPI applications to be ported to the Cell BE TM processor. This implementation views each SPE as a node for an MPI process. We store the application data in main memory in order to avoid being limited by the local store size. The local store is abstracted in the library and thus hidden from the application with respect to MPI calls. We have achieved bandwidth up to 6.01 GB/s and latency as low as 0.41 μs on the ping-pong test. The contribution of this work lies in (i) demonstrating that the Cell BE TM has good potential for running intra-Cell BE TM MPI applications, (ii) enabling such applications to be ported to the Cell BE TM with minimal effort, and (iii) evaluating the performance impact of different design choices.

Extended characterization of DMA transfers on the Cell BE processor

“…Therefore, it is vital to implement MPI efficiently on the Cell BE processor to leverage its tremendous computational power. For this, there has been a feasibility study of MPI implementation on the Cell BE [6,7]. In this study, a minimal set of a synchronous mode MPI on the Cell BE has been implemented, and the results show the potential of the Cell BE to run MPI applications efficiently.…”

Study on Explicit Memory Management for CBEA Green Computing Architecture

“…A recent publication by Bellens et al [4] describes CellSs, a new programming model for Cell-like architectures. Arun Kumar et al [5] provide a MPI implementation on Cell. Its memory access relies on software cache technology.…”

A Profile-based Memory Access Optimizing Technology on CBE Architecture

“…Eichenberger et al [13] describe a series of compiler optimizations, while the proposal of software cache deals with memory access and considers the efficiency of DMA transfers. Arun Kumar et al [14] provide a MPI implementation on Cell with memory access relying on software-managed cache. But performance degradation is a major issue of current software-managed cache technologies.…”