SUMMARY Graphical processing units have been gaining rising attention because of their high performance processing capabilities for many scientific and engineering applications. However, programming such highly parallel devices requires adequate programming tools. Many such programming tools have emerged and hold the promise for high levels of performance. Some of such tools may require specialized parallel programming skills, while others attempt to target the domain scientist. The costs versus the benefits out of such tools are often unclear. In this work we examine the use of several of these programming tools such as Compute Unified Device Architecture, Open Compute Language, Portland Group Inc., and MATLAB in developing kernels from the (NAS) NASA Advanced Supercomputing parallel benchmarking suite. The resulting performance as well as the needed programmers' efforts were quantified and used to characterize the productivity of graphical processing units using these different programming paradigms. Copyright © 2011 John Wiley & Sons, Ltd.

Despite parallelisation, scalable performance of WRF is often not achieved. Although the WRFV3 Parallel Benchmark Page provides valuable scaling information, this one-domain configuration is often not typical of the general application of WRF to nested domains of varying sizes and shapes. This study is a first step in providing a centralised WRF performance repository for a variety of typical configurations and environments that might be found in general NWP environments. Here, a real-world scenario that comes close to the ones used in regional climate studies in complex topographical areas is evaluated. The performance of WRF V3.2.1 is analysed in several multi-nest configurations varying vertical levels and parameterisations and tested for scalability on different platforms. Initial insight shows that multi-nest configurations make realization of desired scalabilities more difficult. In many real-world cases, the outer nest has substantially fewer grid points than the inner nests, and this can be problematic in that large-scale domain decomposition of the inner nests where the great majority of computations take place is limited by “over-decomposition” of the outer nest.

WRF/Chem simulations were performed using the meteorological conditions of January 2000 and alternatively the emissions of January 1990 and 2000 to examine whether increases in emissions may have caused the increasing trends in observed sulfate-aerosol concentrations at coastal Alaska sites. The analysis focused on six regions in Alaska that are exposed differently to the main emission sources. Meteorological observations at 59 sites and aerosol measurements at three sites showed that WRF/Chem captured the meteorological situation over Alaska well and simulated the aerosol concentrations acceptably. Except for the region adjacent to the Arctic Ocean that is influenced by local SO2-emissions, Alaska SO2 and SO42−-aerosol distributions are affected by long-range transport of SO2 from ship emissions and/or emissions in Canada and southern Siberia. Local changes in emissions between 1990 and 2000 are not the main cause for concentrations changes in the six regions. The increases of SO42−-aerosols and SO42−-in-cloud along the Gulf of Alaska are caused by increased ship or Canadian emissions. The study provides evidence that the increased ship and Canadian emissions during the last decades can cause increases in sulfate aerosols.

The latent semantic indexing (LSI) methodology for information retrieval applies the singular value decomposition to identify an eigensystem for a large matrix, in which cells represent the occurrence of terms (words) within documents. This methodology is used to rank text documents, such as Web pages or abstracts, based on their relevance to a topic. The LSI was introduced to address the issues of synonymy (different words with the same meaning) and polysemy (the same words with multiple meanings), thus addressing the ambiguity in human language by utilizing the statistical context of words. Rather than keeping all k possible eigenvectors and eigenvalues from the singular value decomposition which approximates the original term by document matrix, a smaller number is used - essentially allowing a fuzzy match of a topic to the original term by document matrix. In this paper, we show that the choice k impacts the resultant ranking and there is no value of k that results in stability of ranked results for similarity of the topic to documents. This is a surprising result, because prior literature indicates that eigensystems based on successively large values of k should approximate the complete (max k) eigensystems. The finding that document-query similarity rankings with larger values of k do not, in fact, maintain consistency, makes it difficult to assert that any particular value of k is optimal. This in turn renders LSI potentially untrustworthy for use in ranking text documents, even for values that differ by only 10% of the max k.

Albuminuria is associated with cardiovascular (CV) outcomes. We evaluated albuminuria, alone and in combination with estimated glomerular filtration rate (eGFR), as a predictor of mortality and CV morbidity in 12,944 patients with non–ST-segment elevation acute coronary syndromes.Baseline serum creatinine and urinary dipsticks were obtained, with albuminuria stratified into no/trace albuminuria, microalbuminuria (≥30 but <300 mg/dL), or macroalbuminuria (≥300 mg/dL). Kaplan-Meier rates and proportional Cox hazards models of CV death, overall mortality, CV death or myocardial infarction (MI), and bleeding were calculated. Incidence of acute kidney injury, identified by adverse event reporting and creatinine increase (absolute ≥0.3 mg/dL or relative ≥50%), was descriptively reported.Both dipstick albuminuria and creatinine values were available in 9473 patients (73.2%). More patients with macroalbuminuria, versus no/trace albuminuria, had diabetes (66% vs 27%) or hypertension (86% vs 68%). Rates for CV death and overall mortality per strata were 3.1% and 4.8% (no/trace albuminuria); 5.8% and 9.0% (microalbuminuria); and 7.7% and 12.6% (macroalbuminuria) at 2 years of follow-up. Corresponding rates for CV death or MI were 12.2%, 16.9%, and 23.5%, respectively. Observed acute kidney injury rates were 0.6%, 1.2%, and 2.9% (n = 79), respectively. Adjusted HRs for macroalbuminuria on CV mortality were 1.65 (95% CI 1.15-2.37), and after adjustment with eGFR, 1.37 (95% CI 0.93-2.01). Corresponding HRs for overall mortality were 1.82 (95% CI 1.37-2.42) and 1.47 (95% CI 1.08-1.98).High-risk patients with non–ST-segment elevation acute coronary syndromes and albuminuria have increased morbidity and increased overall mortality independent of eGFR.

Solid-State Drives (SSDs) are data storage devices that use solid-state memory to store persistent data. Flash memory is the de facto nonvolatile technology used in most SSDs. It is well known that the writing performance of flash-based SSDs is much lower than the reading performance due to the fact that a flash page can be written only after it is erased. In this work, we present an SSD cache architecture designed to provide a balanced read/write performance for flash memory. An efficient automatic updating technique is proposed to provide a more responsive SSD architecture by writing back stable but dirty flash pages according to a predetermined set of policies during the SSD device idle time. Those automatic updating policies are also tested and compared. Simulation results demonstrate that both reading and writing performance are improved significantly by incorporating the proposed cache with automatic updating feature into SSDs.

The Weather Research and Forecasting (WRF) model has been used extensively in research and operational theaters on a broad range of computer architectures. There is a need for a standard suite of case studies by which to benchmark a computational platform's potential to efficiently support WRF and, concurrently, a need to push WRF into increasingly larger problem sizes in order to test the capability of WRF and its supporting software and hardware on the truly grand scale domains. The Arctic Region Supercomputing Center is implementing a suite of WRF case studies intended to support benchmarking and testing on all architectures, ranging from small, single CPU systems, to those with hundreds of thousands of cores. The benchmark cases are all based on a single weather event over a 6,075×6,075km domain centered at Fairbanks, Alaska, ranging from horizontal grid resolutions of 81km with 150,000 grid points, to 1km with over one-billion grid points. Each case is available to users in the form of a restart, boundary condition, and parameters file, so that once WRF has been installed on a system the case may be readily-tested. Implementation of the one billion grid point case has been particularly problematic, revealing issues that prevent WRF from operating on large-scale problems unless specialized software and procedures are considered. With the help of WRF and Cray experts, the giga-gridpoint barrier has been conquered, and the pursuit of even bigger computational challenges for WRF continues, with the realization that modelers will always have an insatiable appetite for higher resolution, more complicated physics and bigger domain sizes.

Alaska wildfires have strong impact on air pollution on regional and even hemispheric scales. Alaska’s climate favors annually recurring wildfires. Highly flammable material like dry tundra and leaves and needles at the floor of the boreal forests are well preserved during cold and dry seasons. As a result lightning induced fires are common during summers. Extreme fire seasons occurred in 2004 and 2005. The 2004 season was one of the warmest and driest summers on record. By the end of summer 2004 wildfires burned 6.6 millions of acres of mostly boreal forest, which broke the 1957 record for the most acres burned in a season. Particulate matter threatened human health for weeks. Local communities suffered through direct and indirect impact of wildfires. For example high fine particulate matter (PM2.5) concentration was measured in Fairbanks during the 2004 fire season, when 41 days were reported as unhealthy to hazardous; 16 (of the 41) days were classified as clearly hazardous to human health. Hazardous smoke episodes occur almost every summer in Alaska communities. Hence the impact of smoke from wildfires on community health and safety is of key concern, and a strong need for a smoke forecast

Ranking text documents based on their relevance to a topic is of great importance in information retrieval. However, giving the increasingly available avalanche of digital documents, the size of collection pool from which these documents are drawn makes this task more challenging. In addition, current computing infrastructure is unable to deal with very large corpuses directly. Thus, new algorithms are needed to seek parallel solutions and utilize more processing power to solve this problem. In this paper we propose a new algorithm that partitions a large collection of documents (a corpus) into smaller corpuses that can each be handled by a single processor for the purpose of ranking. These multiple rankings are then merged together to provide a unified listing of all selected documents from the original large corpus.

: The ARSC team made modifications to the Apache Lucene engine to accommodate go words, taken from the Google Gigaword vocabulary of n-grams. Indexing the Category B subset of the ClueWeb collection was accomplished by a divide and conquer method, working across the separate ClueWeb subsets for 1, 2 and 3-grams.

Reconfigurable Computers (RCs) with hardware (FPGA) co-processors can achieve significant performance improvement compared with traditional microprocessor (µP)-based computers for many scientific applications. The potential amount of speedup depends on the intrinsic parallelism of the target application as well as the characteristics of the target platform. In this work, we use image processing applications as a case study to demonstrate how hardware designs are parameterized by the co-processor architecture, particularly the data I/O, i.e., the local memory of the FPGA device and the interconnect between the FPGA and the µP. The local memory has to be used by applications that access data randomly. A typical case belonging to this category is image registration. On the other hand, an application such as edge detection can directly read data through the interconnect in a sequential fashion. Two different algorithms of image registration, the exhaustive search algorithm and the Discrete Wavelet Transform (DWT)-based search algorithm, are implemented on hardware, i.e., Xilinx Vertex-IIPro 50 on the Cray XD1 reconfigurable computer. The performance improvements of hardware implementations are 10× and 2×, respectively. Regarding the category of applications that directly access the interconnect, the hardware implementation of Canny edge detection can achieve 544× speedup.

High performance computing (HPC) has often benefited from special-purpose hardware. This paper examines the potential roles for several different approaches to hardware acceleration that are currently being deployed in HPC systems. Because each technology has different performance characteristics, as well as practical considerations (such as electrical consumption, physical interface, and cost), a match of these characteristics to the desired HPC workload is desirable. Technologies discussed include multicore processors, chip multithreading, graphics processing units, field-programmable gate arrays, cell processors, and vector processors.

In the quest for higher performance and with the increasing availability of multi-core chips, many systems are currently packing more processors per node. Adopting a ccNUMA node architecture in these cases has the promise of achieving a balance between cost and performance. In this paper, a 2312 Opteron cores system based on Sun Fire servers is considered as a case study to examine the performance issues associated with such architectures. In this study, we characterize the performance behavior of the system with focus on the node level using different configurations. It will be shown that the benefits from larger nodes can be severely limited for many reasons. These reasons were isolated, the associated performance losses were assessed, and some potential solutions were proposed. With the proposed performance tunings, up to 30% application performance improvement was observed. The results revealed that such problems were mainly caused by topological imbalances, limitations of the cache coherence protocol used, operating system services distribution and the lack of intelligent management of memory affinity. In addition, provided experimental analysis can be utilized by HPC application developers in order to better understand clusters with ccNUMA nodes and also as a guideline for the use of such architectures for scientific computing.

Multi-core architectures have emerged as the dominant architecture for both desktop and high-performance systems. Multi-core systems introduce many challenges that need to be addressed to achieve the best performance. Therefore, benchmarking of these processors is necessary to identify the possible performance issues. In this paper, broad range of homogeneous multi-core architectures are investigated in terms of essential performance metrics. To measure performance, we used micro-benchmarks from High-Performance Computing Challenge (HPCC), NAS Parallel Benchmarks (NPB), LMbench, and an FFT benchmark. Performance analysis is conducted on multi-core systems from UltraSPARC and x86 architectures; including systems based on Conroe, Kentsfield, Clovertown, Santa Rosa, Barcelona, Niagara, and Victoria Falls processors. Also, the effect of multi-core architectures in cluster performance is examined using a Clovertown based cluster. Finally, cache coherence overhead is analyzed using a full-system simulator. Experimental analysis and observations in this study provide for a better understanding of the emerging homogeneous multi-core systems.

: A distributed information retrieval system with resource-selection and result-set merging capability was used to search subsets of the GOV2 document corpus for the 2008 TREC Million Query Track. The GOV2 collection was partitioned into host-name subcollections and distributed to multiple remote machines. The Multisearch demonstrations, application restricted each search to a fraction of the available sum-collections that was pre-determined by a resource-selection algorithm. Experiment results from topic-by-topic resource selection and aggregate topic resource selection are compared. The sensitivity of Multisearch retrieval performance to variations in the resource selection algorithm is discussed.

FPGA devices in Reconfigurable Computers (RCs) allow datapath, memory, and processing elements (PEs) to be customized in order to achieve very efficient algorithm implementations. However, the maximum speedup on RCs is bounded by the bandwidth available between mu Ps and FPGA hardware accelerators. In this paper, an image processing architecture is presented to fully exploit this bandwidth for achieving the maximum possible speedup. This architecture can be used to implement any convolution operation between an image and a kernel, and comprises four fully pipelined components: a line buffer, a data window, an array of PEs and a data concatenating block. Multiple image processing algorithms have been successfully implemented using this architecture, such as digital filters, edge detectors, and image transforms. In all cases, the maximum throughput is upper-bounded by the mu P-FPGA I/O bandwidth, regardless of the complexity of the algorithm. This end-to-end throughput has been measured to be 1.2 GB/s on Cray XD1 and 2.1 GB/s on SGI RC100.

With the increasing trend of putting more cores inside a single chip, more clusters adapt multicore multiprocessor nodes for high-performance computing (HPC). Cache coherent non-uniform memory access architectures (ccNUMA) are becoming an increasingly popular choice for such systems. In this paper, application performance analysis is provided using a 2312 Opteron cores system based on Sun Fire servers. Performance bottlenecks are identified and some potential solutions are proposed. With the proposed performance tunings, up to 30% application performance improvement was observed. In addition, provided experimental analysis can be utilized by HPC application developers in order to better understand clusters with ccNUMA nodes and also as a guideline for the usage of such architectures for scientific computing.