Lee J.,Ulsan National Institute of Science and Technology |
Shrivastava A.,Arizona StateUniversity
Transactions on Embedded Computing Systems | Year: 2013
Register File (RF) is extremely vulnerable to soft errors, and traditional redundancy based schemes to protect the RF are prohibitive not only because RF is often in the timing critical path of the processor, but also since it is one of the hottest blocks on the chip. Software approaches would be ideal in this case, but previous approaches based on instruction scheduling are only moderately effective due to local scope. In this article we present a compiler approach, based on interprocedural program analysis, to reduce the vulnerability of registers by temporarily writing live variables to protected memory. We formulate the problem as an integer linear programming problem and also present a very efficient heuristic algorithm. Further we present an iterative optimization method based on Kernighan-Lin's graph partitioning algorithm. Our experiments demonstrate that our proposed techniques can reduce the vulnerability of a RF by 33 ~ 37% on average and up to 66%, with a small 2% increase in runtime. In addition, our overhead reduction optimization can effectively reduce the code size overhead, by more than 40% on average, to a mere 5 ~ 6%, compared to highly optimized binaries. Categories and Subject Descriptors: D.3.4 [Programming Languages]: Processors-Code generation, compilers, optimization; B.8.1 [Performance and Reliability]: Reliability, Testing, and Fault-Tolerance. © 2013 ACM 1539-9087/2013/11-ART38 $15.00.
Ahlstrom L.S.,University of Arizona |
Ahlstrom L.S.,University of Michigan |
Baker J.L.,University of Arizona |
Baker J.L.,James Franck Institute |
And 12 more authors.
Journal of Molecular Graphics and Modelling | Year: 2013
Effective data reduction methods are necessary for uncovering the inherent conformational relationships present in large molecular dynamics (MD) trajectories. Clustering algorithms provide a means to interpret the conformational sampling of molecules during simulation by grouping trajectory snapshots into a few subgroups, or clusters, but the relationships between the individual clusters may not be readily understood. Here we show that network analysis can be used to visualize the dominant conformational states explored during simulation as well as the connectivity between them, providing a more coherent description of conformational space than traditional clustering techniques alone. We compare the results of network visualization against 11 clustering algorithms and principal component conformer plots. Several MD simulations of proteins undergoing different conformational changes demonstrate the effectiveness of networks in reaching functional conclusions. © 2013 Elsevier Inc.
Jiang H.,China University of Petroleum - Beijing |
Chen M.,China University of Petroleum - Beijing |
Jin Y.,China University of Petroleum - Beijing |
Chen K.P.,Arizona StateUniversity
Journal of Porous Media | Year: 2015
Volume-expansion-induced gas acceleration effect in a highly compressible gas flow in porous media is studied. The importance of gas acceleration is quantified in terms of a dimensionless parameter; conditions under which gas acceleration effect need to be considered in the gas pressure profile and mass flux are identified. It is shown that when the Forchheimer drag is given, the effect of gas acceleration on the gas pressure profile and mass flux depend only on the dimensionless parameter representing the ratio between the gas acceleration and the Forchheimer drag. It is also shown that the Forchheimer drag has an insignificant effect on the outlet pressure in choked flows. The result of this analysis is successfully applied to published experimental data. © 2015 by Begell House, Inc.
Guo P.,University of Adelaide |
Saw W.,University of Adelaide |
Van Eyk P.,University of Adelaide |
Ashman P.,University of Adelaide |
And 2 more authors.
Energy Procedia | Year: 2015
A coal to liquid (CTL)polygeneration process with a solar hybrid dual fluidized bed (SDFB) gasifier (SCTL) is investigated in recently processing paper. A storage unit was integrated to store sensible heat in bed material in order to reduce the influence of solar resource transience. In this paper, a Fischer-Tropsch liquid fuel production system via solar hybrid co-gasification of coal and biomass in SDFB gasifier (SCBTL) is investigated. The energetic and environmental performanceof the SCBTL system is assessed as a function of the biomass ratio and char conversion. It is found that the performance of the SCBTL system is found to be less sensitive to char conversion in the gasification reactor (Xchar,g) than the SCTL system.As theXchar,g decreases from 100% to 57%, the annually averaged solar share of the SCTL system is reduced from 24% to 0, while the solar share of the SCBTL system with wood fraction (higher heating value basis) of 0.5 and 1only decreases to 7% and 13% respectively. It is tricky to achieve very higher char conversion (especially higher than 85%) in the gasification reactor we studied, so this reduction of impact of the char conversion is very important. To achieve a mine-to-tank (MTT) GHG emission which can match the well-to-tank (WTT) greenhouse gas (GHG) emission, a wood fraction of 0.24 and 0.37 is required respectively for the SCBTL system with a char conversion of 100% and 70%, while this required fraction is increased to 0.45 for the non-solar equivalent. However, parameters optimization and other system design options need to be studied to improve the performance of SCBTL further and adjust the ratio of FTL to net electricity in the system output. © 2015 The Authors. Published by Elsevier Ltd.
Lu J.,Arizona StateUniversity |
Bai K.,Arizona StateUniversity |
Shrivastava A.,Arizona StateUniversity
ACM Transactions on Embedded Computing Systems | Year: 2015
Scaling the memory hierarchy is a major challenge when we scale the number of cores in a multicore processor. Software Managed Multicore (SMM) architectures come up as one of the promising solutions. In an SMM architecture, there are no caches, and each core has only a local scratchpad memory [Banakar et al. 2002]. As the local memory usually is small, large applications cannot be directly executed on it. Code and data of the task mapped to each core need to be managed between global memory and local memory. This article solves the problem of efficiently managing code on an SMM architecture. The primary requirement of generating efficient code assignments is a correct management cost model. In this article, we address this problem by proposing a cost calculation graph. In addition, we develop two heuristics CMSM (Code Mapping for Software Managed multicores) and CMSM-advanced that result in efficient code management execution on the local scratchpad memory. Experimental results collected after executing applications from the MiBench suite [Guthaus et al. 2001] demonstrate that merely by adopting the correct management cost calculation, even using previous code assignment schemes, we can improve performance by an average of 12%. Combining the correct management cost model and a more optimized code mapping algorithm together, our heuristics can reduce runtime in more than 80% of the cases, and by up to 20% on our set of benchmarks, compared to the state-of-the-art code assignment approach [Jung et al. 2010]. When compared with Instruction-level Parallelism (ILP) results, CMSM-advanced performs an average of 5% worse.We also simulate the benchmarks on a cache-based system, and find that the code management overhead on SMM core with our code management is much less than memory latency of a cache-based system. © 2015 ACM.