As the demands of computational science and engineering simulations increase, the size and capabilities of High Performance Computing (HPC) clusters are also expected to grow. Consequently, the software providing the application programming abstractions for the clusters must adapt to meet these demands. Specifically, the increased cost of interprocessor synchronization and communication in larger systems must be accommodated. Non-blocking operations that allow communication latency to be hidden by overlapping it with computation have been proposed to mitigate this problem. In this work, we investigate offloading a portion of the communication processing to dedicated hardware in order to support communication/computation overlap efficiently. We work with the Message Passing Interface (MPI), the de facto standard for parallel programming in HPC environments. We investigate both point-to-point non-blocking communication and collective operations; our work with collectives focuses on the allgather operation. We develop designs for both flat and hierarchical cluster topologies and examine both eager and rendezvous communication protocols. We also develop a generalized primitive operation with the aim of simplifying further research into non-blocking collectives. We propose a new algorithm for the non-blocking allgather collective and implement it using this primitive. The algorithm has constant resource usage even when executing multiple operations simultaneously. We implemented these designs using CORE-Direct offloading support in Mellanox InfiniBand adapters. We present an evaluation of the designs using microbenchmarks and an application kernel that shows that offloaded non-blocking communication operations can provide latency that is comparable to that of their blocking counterparts while allowing most of the duration of the communication to be overlapped with computation and remaining resilient to process arrival and scheduling variations. / Thesis (Master, Electrical & Computer Engineering) -- Queen's University, 2014-05-29 11:55:53.87

Message passing interface (MPI) has been widely used for implementing parallel and distributed applications. The emergence of cloud computing offers a scalable, fault-tolerant, on-demand al-ternative to traditional on-premise clusters. In this thesis, we investigate the possibility of adopt-ing the cloud platform as an alternative to conventional MPI-based solutions. We show that cloud platform can exhibit competitive performance and benefit the users of this platform with its fault-tolerant architecture and on-demand access for a robust solution. Extensive research is done to identify the difficulties of designing and implementing an MPI-like framework for Azure cloud platform. We present the details of the key components required for implementing such a framework along with our experimental results for benchmarking multiple basic operations of MPI standard implemented in the cloud and its practical application in solving well-known large-scale algorithmic problems.

No description available.

Writing efficient parallel programs is a difficult and error-prone process. The Standard Template Adaptive Parallel Library (STAPL) is being developed to make this task easier for programmers with little experience in parallel programming. STAPL is a C++ library for writing parallel programs using a generic programming approach similar to writing sequential programs using the C++ Standard Template Library (STL). STAPL provides a collection of parallel containers (pContainers) to store data in a distributed fashion and a collection of pViews to abstract details of the data distribution. STAPL algorithms are written in terms of PARAGRAPHs which are high level descriptions of task dependence graphs. Scheduling plays a very important role in the efficient execution of parallel programs. In this thesis, we present our work to enable efficient scheduling of parallel programs written using STAPL. We abstract the scheduling activities associated with PARAGRAPHs in a software module called the scheduler which is customizable and extensible. We provide support for static scheduling of PARAGRAPHs and develop mechanisms based on migration of tasks and data to support dynamic scheduling strategies for PARAGRAPHs with arbitrary dependencies. We also provide implementations of different scheduling strategies that can be used to improve the performance of applications suffering from load imbalance. The scheduling infrastructure developed in this thesis is highly customizable and can be used to execute a variety of parallel computations. We demonstrate its usefulness by improving the performance of two applications: a widely used synthetic benchmark (UTS) and a Parallel Motion Planning application. The experiments are conducted on an Opteron cluster and a massively parallel Cray XE6 machine. Experimental results up to 6144 processors are presented.

Graduation date: 1992

Lipidomics and metabolomics are powerful tools for the examination of cellular metabolism and physiology. Methods for lipid analysis need to be developed that begin with small samples and do not overly dilute or disperse the sample in the separation process. Microchips provide a platform for interfacing lysis of small cell populations with on-chip solid phase extraction for isolating lipid samples to generate high quality mass spectra from very small samples. Chapter 1 of this dissertation presents a novel method for small scale lipidomics of bacterial cells by microchip based extraction coupled with untargeted profiling of glycerophospholipids using nanoelectrospray ionization mass spectrometry. Chapter 2 and 3 focus on the development of LC-MS/MS methods to study biological pathways. In Chapter 2, I describe a method for analysis of the phospholipids metabolite, GroPIns, in the medium of the pathogenic yeast Candida albicans. This method was applied to aid in the characterization of the GroPIns transport protein, Git1, in C. albicans. Chapter 3 extends the studies of part two and describes an efficient method based on HILIC-MS/MS for the separation and quantification of five lipid-related extracellular metabolites in yeast Saccharomyces cerevisiae. This newly developed methodology was successfully applied to determine the extracellualr levels of glycerophosphoinositol, glycerophosphocholine, glycerol 3-phosphate, inositol and choline in wild type and mutant strains. / Bayer School of Natural and Environmental Sciences / Chemistry and Biochemistry / PhD / Dissertation

Graduation date: 1994

Accurate quantification of phosphorus (P) fractions in manures and agricultural runoff is vital to understanding P dynamics in the environment. Phytate is an organic form of P, with 6 phosphate groups, which is found in manures, but is difficult to quantify due to its affinity for complex formation with aluminum (Al) and iron (Fe). The objective of the first study was to measure if high concentrations of aluminum (Al) and iron (Fe) hinder accurate quantification of phytate in dairy manure and broiler litter when measured by high performance ion chromatography (HPIC). In this study dairy manure and broiler litter samples were spiked with Al, Fe, and phytate. Samples were alkaline extracted, acidified,cleaned up and filtered, and then phytate was analyzed with HPIC. High concentrations of Fe did not hinder phytate recovery in manure or litter samples. While phytate recovery was close to 100% at typical manure and litter Al concentrations, high concentrations of Al inhibited phytate recovery in litter samples and in some manure samples. Overall, alkaline extraction of dairy manure and broiler litter and analysis with HPIC proved to be a relatively accurate, fast and cheap within normal Al and Fe ranges, compared to the commonly used NMR method. The developed method was then used to measure the concentration of phytate P (Pp) rather than the entire phytate molecule (6 moles Pp per mole phytate) in runoff waters and in an adsorption study. Phytate P has not been previously measured in runoff, due to the low concentrations of Pp found in runoff and the complexity Pp extraction from runoff. Runoff treatments were dairy manure with and without added Pp (DM, DM+Pp), broiler litter with and without added Pp (BL, BL+Pp), and control, with and without two levels of Pp (control, control Pp low, control Pp high). Runoff was collected under simulated rainfall and analyzed for total and dissolved reactive P (DRP). Runoff was also separated into sediment (>0.45"m) and liquid (<0.45"m and lyophilized) fractions for Pp analysis by high performance ion chromatography (HPIC). The new method for Pp analysis recovered 70% of Pp spiked into runoff samples. Phytate P was found only on the sediment fraction of runoff and was not detected in the liquid fraction, even when it was lyophilized and extracted. This agreed with the adsorption study, which showed strong adsorption of Pp. Phytate P loss from control Pp high was significantly greater than Pp loss from control. �When control Pp high was removed from analysis, Pp loss from BL, BL+Pp, DM, and DM+Pp was significantly greater than from control. Phytate P in the manures appeared to behave differently than pure Pp salts, likely to do the formation of protein-phytate and cation-phytate complexes in the manures. Phytate P had no effect on DRP in runoff for any treatment, indicating no significant release of inorganic P through competitive binding. The majority of P lost in runoff was in fractions other than DRP and Pp in the sediment fraction. Efforts to control the erosive loss of soil during rainfall events appear to be the best way to reduce total P loss, irrespective of the presence of Pp. / Master of Science

Modern high-performance digital signal processing (DSP) applications face constantly increasing performance requirements and are becoming increasingly challenging to develop and work with. In DSP paradigm, many researchers see potential in achieving algorithm speed-up by employing Field Programmable Gate Arrays (FPGAs) – reconfigurable hardware with parallelism feature. However, developing applications for FPGAs incur particular challenges on the development flow. This work proposes a scalable hybrid DSP system for performing high-performance signal processing applications. The system employs hybrid CPU + FPGA architecture of commercially available, off-the-shelf (COTS) FPGAs and central processing units (CPU) of personal computers. In this work an example implementation of a multi-channel cross-correlator is investigated and delivered using a new development paradigm. The correlator is implemented on the XD1000 development system using a high-level FPGA programming tool – Impulse CoDeveloper. Analysis of DSP application development in a hybrid CPU+FPGA system employing the high-level programming tool Impulse C is presented. Potential of the selected tool to deliver algorithm speed-ups is investigated using reference multi-channel correlator software. Particular attention is devoted to input/output (I/O) implementation, which is considered one of the most challenging problems in FPGA design development. This work delivers an I/O framework based on PCI Express interface for the proposed high-performance scalable DSP system. Using Stratix II GX PCI Express Development Board from Altera Corporation, a scalable and flexible communication approach for the multi-channel correlator is delivered. This framework can be adapted to perform other high-performance streaming DSP applications. The outcomes of this work are a multi-channel correlator developed in a reconfigurable environment with new design methodology and I/O framework with software control application. The outcomes are used to demonstrate the potential of implementing DSP applications in hybrid CPU + FPGA architecture and to discuss existing challenges and suggest possible solutions.

For pain research, the orofacial region is unique in a number of ways. The region has complex local anatomy, including substantial sensory innervation from neural pathways, and muscles of facial expression that convey important information concerning pain intensity and associated psychological traits. Although chronic orofacial pain conditions appear prevalent, useful documentation on pain intensity ratings using well established instruments is sparse. In particular, two conditions, atypical facial pain and atypical odontalgia, are poorly understood in aetiology so that definitive treatment modalities are severely limited. The region's local biofluid, saliva, has been used to diagnose various local and systemic disease states, and to quantitate drug concentrations. However, recent studies indicate that saliva also contains some of the same peptides, e.g. bradykinin, that are involved in pain mechanisms. It may be that pharmacological-pharmacokinetic studies of these peptides could shed more information on thesignificance of their presence in saliva. This thesis consists of four major sections. Section 1 comprises of three clinical studies investigating orofacial pain. Section 2 deals with clinical laboratory studies of saliva. Section 3 is concerned with the development of chromatographic methods to assay bradykinin and its pharmacokinetics in saliva. Section 4 uses chromatography for the identification of novel salivary peptides. This thesis, then, presents clinical studies of orofacial pain and pharmacological investigations of saliva as the local biofluid.