Search

As CMOS devices are scaled down into the nanometer regime, concerns about reliability are mounting. Instead of viewing nano-scale characteristics as an impediment, technologies such as PCMOS exploit them as a source of randomness. The technology generates random numbers that are used in probabilistic algorithms. With the PCMOS approach, different voltage levels are used to generate different probability values. If many different probability values are required, this approach becomes prohibitively expensive. In this chapter, we demonstrate a novel technique for synthesizing logic that generates new probabilities from a given set of probabilities. We focus on synthesizing combinational logic to generate arbitrary decimal probabilities from a given set of input probabilities. We demonstrate how to generate arbitrary decimal probabilities from small sets – a single probability or a pair of probabilities – through combinational logic. [ABSTRACT FROM AUTHOR]

Detecting and predicting a program's execution phases are crucial to dynamic optimizations and dynamically adaptable systems. This paper shows that a phase can be associated with dynamic code regions embedded in loops and procedures which are primary targets of compiler optimizations. This paper proposes a new phase tracking hardware, especially for dynamic optimizations, that effectively identifies and accurately predicts program phases by exploiting program control flow information. Our proposed phase tracking hardware uses a simple stack and a phase signature table to track the change of phase signature between dynamic code regions. Several design parameters of our proposed phase tracking hardware are evaluated on 10 SPEC CPU2000 benchmarks. Our proposed phase tracking hardware effectively identifies a phase at a given granularity. It correctly predicts the next program phase for 84.9% of times with a comparable small performance variance within the same phase. A longer phase length and higher phase prediction accuracy together with a reasonably small performance variance are essential to build more efficient dynamic profiling and optimization systems. [ABSTRACT FROM AUTHOR]

The spatial auto-regression (SAR) model is a popular spatial data analysis technique, which has been used in many applications with geo-spatial datasets. However, exact solutions for estimating SAR parameters are computationally expensive due to the need to compute all the eigenvalues of a very large matrix. Recently we developed a dense-exact parallel formulation of the SAR parameter estimation procedure using data parallelism and a hybrid programming technique. Though this parallel implementation showed scalability up to eight processors, the exact solution still suffers from high computational complexity and memory requirements. These limitations have led us to investigate approximate solutions for SAR model parameter estimation with the main objective of scaling the SAR model for large spatial data analysis problems. In this paper we present two candidate approximate-semi-sparse solutions of the SAR model based on Taylor series expansion and Chebyshev polynomials. Our initial experiments showed that these new techniques scale well for very large data sets, such as remote sensing images having millions of pixels. The results also show that the differences between exact and approximate SAR parameter estimates are within 0.7% and 8.2% for Chebyshev polynomials and Taylor series expansion, respectively, and have no significant effect on the prediction accuracy. [ABSTRACT FROM AUTHOR]

This paper examines the communication patterns of parallel scientific programs, including some of the NAS benchmarks and the Miami Isopycnic Coordinate Ocean Model (MICOM), that use explicit message-passing. Communication locality, including communication event locality, message destination locality, and message size locality, is proposed and studied in addition to the widely accepted metrics of message size, destination, and generation distributions. We find that the locality metrics are relatively insensitive to system and problem size variations making them robust metrics for characterizing the communication patterns of parallel applications. We observe that the communication patterns of the benchmark programs are consistent with those of the actual application. The results of this study will be useful for understanding parallel applications' communication behavior and for designing more realistic synthetic benchmarks. [ABSTRACT FROM AUTHOR]