A trend is developing in high performance computing in which commodity processors are coupled to various types of computational accelerators. Such systems are commonly called hybrid systems. In this paper, we describe our experience developing an implementation of the Linpack benchmark for a petascale hybrid system, the LANL Roadrunner cluster built by IBM for Los Alamos National Laboratory. This system combines traditional x86-64 host processors with IBM PowerXCell™ 8i accelerator processors. The implementation of Linpack we developed was the first to achieve a performance result in excess of 1.0 PFLOPS, and made Roadrunner the #1 system on the Top500 list in June 2008. We describe the design and implementation of hybrid Linpack, including the special optimizations we developed for this hybrid architecture. We then present actual results for single node and multi-node executions. From this work, we conclude that it is possible to achieve high performance for certain applications on hybrid architectures when careful attention is given to efficient use of memory bandwidth, scheduling of data movement between the host and accelerator memories, and proper distribution of work between the host and accelerator processors.

No abstract

Commercial database management systems (DBMSs) have historically seen very limited use within the scientific computing community. One reason for this absence is that previous database systems lacked support for the extensible data structures and performance features required within a high-performance computing context. However, database vendors have recently enhanced the functionality of their systems by adding object extensions to the relational engine. In principle, these extensions allow for the representation of a rich collection of scientific datatypes and common statistical operations. Utilizing these new extensions, this paper presents a study of the suitability of incorporating two popular scientific formats, NetCDF and HDF, into an object-relational system. To assess the performance of the database approach, a series of solution variables from a regional weather forecast model are used to build representative small, medium and large databases. Common statistical operations and array element queries are then performed using the object-relational database, and the execution timings are compared against native NetCDF and HDF operations.

No abstract

No abstract

No abstract

In this paper we present the design and implementation of the Linpack benchmark for the IBM BladeCenter QS22, which incorporates two IBM PowerXCell 8i1processors. The PowerXCell 8i is a new implementation of the Cell Broadband Engine™2 architecture and contains a set of special-purpose processing cores known as Synergistic Processing Elements (SPEs). The SPEs can be used as computational accelerators to augment the main PowerPC processor. The added computational capability of the SPEs results in a peak double precision floating point capability of 108.8 GFLOPS. We explain how we modified the standard open source implementation of Linpack to accelerate key computational kernels using the SPEs of the PowerXCell 8i processors. We describe in detail the implementation and performance of the computational kernels and also explain how we employed the SPEs for high-speed data movement and reformatting. The result of these modifications is a Linpack benchmark optimized for the IBM PowerXCell 8i processor that achieves 170.7 GFLOPS on a BladeCenter QS22 with 32 GB of DDR2 SDRAM memory. Our implementation of Linpack also supports clusters of QS22s, and was used to achieve a result of 11.1 TFLOPS on a cluster of 84 QS22 blades. We compare our results on a single BladeCenter QS22 with the base Linpack implementation without SPE acceleration to illustrate the benefits of our optimizations.

Current state-of-the-art in GPU networking advocates a hostcentric model that reduces performance and increases code complexity. Recently, researchers have explored several techniques for networking within a GPU kernel itself. These approaches, however, sufer from high latency, waste energy on the host, and are not scalable with larger/more GPUs on a node. In this work, we introduce Command Processor Networking (ComP-Net), which leverages the availability of scalar cores integrated on the GPU itself to provide highperformance intra-kernel networking. ComP-Net enables eicient synchronization between the Command Processors and Compute Units on the GPU through a line locking scheme implemented in the GPU's shared last-level cache. We illustrate that ComP-Net can improve application performance by up to 20% and provide up to 50% reduction in energy consumption vs. competing networking techniques across a Jacobi stencil, allreduce collective, and machine learning applications. CCS CONCEPTS • Computer systems organization → Heterogeneous (hybrid) systems;