Papers & Reports 2009

This paper evaluates performance of a dual-core SGI Altix 4700, a quad-core SGI Altix ICE 8200, and a dual-core IBM POWER5+ system. To measure performance, NAS computer scientists used micro-benchmarks from the High Performance Computing Challenge, NAS Parallel Benchmarks, and four real-world applications three from computational fluid dynamics and one from climate modeling.
Scientific Application-Based Performance Comparison of SGI Altix 4700, IBM Power5+, and SGI ICE 8200 Supercomputers,NAS Technical Report, NAS-09-001, February 2009.

This paper describes a methodology to quantify the effect of contention on the performance of parallel applications running on high-end computers based on commodity multicore microprocessors. In particular, the authors characterize the contention for a specific shared resource by comparing runs that use different patterns of assigning processes to cores on multicore nodes. The approach is demonstrated with a subset of the HPCC benchmarks and the NAS Parallel Benchmarks running on high-end computing platforms that use four quad-core microprocessors Intel's Clovertown, Harpertown, and Nehalem-EP, and AMD's Barcelona.
Characterizing Application Performance Sensitivity to Resource Contention in Multicore Architectures,NAS Technical Report, NAS-09-002, November 2009.

This paper studies some of the issues in accelerating the computational fluid dynamics code OVERFLOW by using a Graphics Processing Unit (GPU). OVERFLOW includes as one of its solver options an algorithm that is a fairly small piece of code but which accounts for a significant portion of the total computational time. The algorithm was modified to be suitable for a GPU, and attention was given to 64-bit and 32-bit arithmetic. Interestingly, the work done for the GPU produced ideas for accelerating the CPU code and led to significant speedup on the CPU.
Acceleration of a CFD Code with a GPU,NAS Technical Report, NAS-09-003, November 2009.

This paper presents a very large scale aerodynamic calculation run on NASA's Pleiades supercomputer using the three-dimensional Navier-Stokes code OVERFLOW. The application is a rotorcraft blade simulation, with attention focused on the resolution of the vortices, which develop off the blade tips and propagate through the flow domain. This simulation represents a level of refinement beyond any previous results presented for this problem.
Large Scale Aerodynamic Calculation on Pleiades,NAS Technical Report, NAS-09-004, November 2009.

This paper presents an early performance evaluation of a 624-core cluster based on the Intel Xeon Processor 5560 ("Nehalem-EP") the first processor from Intel with NUMA architecture managed by an on-chip integrated memory controller. Authors evaluate these features using the High Performance Computing Challenge benchmarks, NAS Parallel Benchmarks, and four full-scale scientific applications, comparing and contrasting performance results of the Xeon 5560 with Intel's previous generations of quad-core processors.
Early Performance Evaluation of a "Nehalem" Cluster Using Scientific and Engineering Applications,NAS Technical Report, NAS-09-005, November 2009.