Learn More
A low-diameter, fast interconnection network is going to be a prerequisite for building exascale machines. A two-level direct network has been proposed by several groups as a scalable design for future machines. IBM's PERCS topology and the dragonfly network discussed in the DARPA exascale hardware study are examples of this design. The presence of multiple(More)
The advent of petascale computing has introduced new challenges (e.g. heterogeneity, system failure) for programming scalable parallel applications. Increased complexity and dynamism in science and engineering applications of today have further exacerbated the situation. Addressing these challenges requires more emphasis on concepts that were previously of(More)
As machines increase in scale, many researchers have predicted that failure rates will correspondingly increase. Soft errors do not inhibit execution, but may silently generate incorrect results. Recent trends have shown that soft error rates are increasing, and hence they must be detected and handled to maintain correctness. We present a holistic(More)
We present our implementation of the HPC Challenge Class II (productivity) benchmarks in the Charm++ [1] programming paradigm 1. Our submission focuses on explaining how over-decomposed, message-driven, migratable objects enhance the clarity of expression of parallel programs and also enable the runtime system to deliver portable performance. Our submission(More)
In this paper we have modelled the behavior of the CDMA cell using clustered Poisson process. We have found that when the RF dynamics of the cell is modelled using this two dimensional description of the user arrival process, the size and the capacity of the cell are completely determined by the spatial traffic density and stability arguments. This is much(More)
Collectives are an important and frequently used component of MPI. Bucket algorithms, also known as "large vector" algorithms, were introduced in the early 90's and have since evolved as a well known paradigm for large MPI collectives. Many modern day supercomputers such as the IBM Blue Gene and Cray XT are based on torus interconnects that offer a highly(More)
Charm++ is an elegant, general-purpose parallel programming model backed by an adaptive runtime system [1]. This combination yields portable performance and a spectrum of real-world productivity benefits that have been demonstrated in production applications. Our submission to this year's HPC Challenge (Class II) comprises six benchmarks that demonstrate(More)
Task mapping on torus networks has traditionally focused on either reducing the maximum dilation or average number of hops per byte for messages in an application. These metrics make simplified assumptions about the cause of network congestion, and do not provide accurate correlation with execution time. Hence, these metrics cannot be used to reasonably(More)
Giant axonal neuropathy (GAN) is a rare disease caused by mutations in the GAN gene, which encodes gigaxonin, an E3 ligase adapter that targets intermediate filament (IF) proteins for degradation in numerous cell types, including neurons and fibroblasts. The cellular hallmark of GAN pathology is the formation of large aggregates and bundles of IFs. In this(More)