Understanding GPU programming for statistical computation: Studies in massively parallel massive mixtures

This article describes advances in statistical computation for large-scale data analysis in structured Bayesian mixture models via graphics processing unit (GPU) programming. The developments are partly motivated by computational challenges arising in fitting models of increasing heterogeneity to increasingly large datasets. An example context concerns common biological studies using high-throughput technologies generating many, very large datasets and requiring increasingly high-dimensional mixture models with large numbers of mixture components.We outline important strategies and processes for GPU computation in Bayesian simulation and optimization approaches, give examples of the benefits of GPU implementations in terms of processing speed and scale-up in ability to analyze large datasets, and provide a detailed, tutorial-style exposition that will benefit readers interested in developing GPU-based approaches in other statistical models. Novel, GPU-oriented approaches to modifying existing algorithms software design can lead to vast speed-up and, critically, enable statistical analyses that presently will not be performed due to compute time limitations in traditional computational environments. Supplementalmaterials are provided with all source code, example data, and details that will enable readers to implement and explore the GPU approach in this mixture modeling context. © 2010 American Statistical Association, Institute of Mathematical Statistics, and Interface Foundation of North America.

Working at the interface of phylogenetics and population genetics: a biogeographical analysis of Triaenops spp. (Chiroptera: Hipposideridae).

New applications of genetic data to questions of historical biogeography have revolutionized our understanding of how organisms have come to occupy their present distributions. Phylogenetic methods in combination with divergence time estimation can reveal biogeographical centres of origin, differentiate between hypotheses of vicariance and dispersal, and reveal the directionality of dispersal events. Despite their power, however, phylogenetic methods can sometimes yield patterns that are compatible with multiple, equally well-supported biogeographical hypotheses. In such cases, additional approaches must be integrated to differentiate among conflicting dispersal hypotheses. Here, we use a synthetic approach that draws upon the analytical strengths of coalescent and population genetic methods to augment phylogenetic analyses in order to assess the biogeographical history of Madagascar's Triaenops bats (Chiroptera: Hipposideridae). Phylogenetic analyses of mitochondrial DNA sequence data for Malagasy and east African Triaenops reveal a pattern that equally supports two competing hypotheses. While the phylogeny cannot determine whether Africa or Madagascar was the centre of origin for the species investigated, it serves as the essential backbone for the application of coalescent and population genetic methods. From the application of these methods, we conclude that a hypothesis of two independent but unidirectional dispersal events from Africa to Madagascar is best supported by the data.

Can Cooling Technology Save Many-Core Parallel Programming from Its Programming Woes?

An abstract of this work will be presented at the Compiler, Architecture and Tools Conference (CATC), Intel Development Center, Haifa, Israel November 23, 2015.

Computational magnetohydrodynamics - Application

There are many processes, particularly in the nuclear and metals processing industries, where electromagnetic fields are used to influence the flow behaviour of a fluid. Procedures exploiting finite volume (FV) methods in both structured and unstructured meshes have recently been developed which enable this influence to be modelled in the context of conventional FV CFD codes. A range of problems have been tackled by the authors, including electromagnetic pumps and brakes, weirs and dams in steelmaking tundishes and interface effects in aluminium smelting cells. Two cases are presented here, which exemplify the application of the new procedures. The first case investigates the influence of electromagnetic fields on solidification front progression in a tin casting and the second case shows how the liquid metals free surface may be controlled through an externally imposed magnetic field in the semi-levitation casting process.

Three parallel programming paradigms: Comparisons on an archetypal PDE computation

Three paradigms for distributed-memory parallel computation that free the application programmer from the details of message passing are compared for an archetypal structured scientific computation -- a nonlinear, structured-grid partial differential equation boundary value problem -- using the same algorithm on the same hardware. All of the paradigms -- parallel languages represented by the Portland Group's HPF, (semi-)automated serial-to-parallel source-to-source translation represented by CAP-Tools from the University of Greenwich, and parallel libraries represented by Argonne's PETSc -- are found to be easy to use for this problem class, and all are reasonably effective in exploiting concurrency after a short learning curve. The level of involvement required by the application programmer under any paradigm includes specification of the data partitioning, corresponding to a geometrically simple decomposition of the domain of the PDE. Programming in SPMD style for the PETSc library requires writing only the routines that discretize the PDE and its Jacobian, managing subdomain-to-processor mappings (affine global-to-local index mappings), and interfacing to library solver routines. Programming for HPF requires a complete sequential implementation of the same algorithm as a starting point, introduction of concurrency through subdomain blocking (a task similar to the index mapping), and modest experimentation with rewriting loops to elucidate to the compiler the latent concurrency. Programming with CAPTools involves feeding the same sequential implementation to the CAPTools interactive parallelization system, and guiding the source-to-source code transformation by responding to various queries about quantities knowable only at runtime. Results representative of "the state of the practice" for a scaled sequence of structured grid problems are given on three of the most important contemporary high-performance platforms: the IBM SP, the SGI Origin 2000, and the CRAYY T3E.

The semi-automatic parallelisation of scientific application codes using a computer aided parallelisation toolkit

The shared-memory programming model can be an effective way to achieve parallelism on shared memory parallel computers. Historically however, the lack of a programming standard using directives and the limited scalability have affected its take-up. Recent advances in hardware and software technologies have resulted in improvements to both the performance of parallel programs with compiler directives and the issue of portability with the introduction of OpenMP. In this study, the Computer Aided Parallelisation Toolkit has been extended to automatically generate OpenMP-based parallel programs with nominal user assistance. We categorize the different loop types and show how efficient directives can be placed using the toolkit's in-depth interprocedural analysis. Examples are taken from the NAS parallel benchmarks and a number of real-world application codes. This demonstrates the great potential of using the toolkit to quickly parallelise serial programs as well as the good performance achievable on up to 300 processors for hybrid message passing-directive parallelisations.

A design pattern for integration of business process management systems

A cross-domain workflow application may be constructed using a standard reference model such as the one by the Workflow Management Coalition (WfMC) [7] but the requirements for this type of application are inherently different from one organization to another. The existing models and systems built around them meet some but not all the requirements from all the organizations involved in a collaborative process. Furthermore the requirements change over time. This makes the applications difficult to develop and distribute. Service Oriented Architecture (SOA) based approaches such as the BPET (Business Process Execution Language) intend to provide a solution but fail to address the problems sufficiently, especially in the situations where the expectations and level of skills of the users (e.g. the participants of the processes) in different organisations are likely to be different. In this paper, we discuss a design pattern that provides a novel approach towards a solution. In the solution, business users can design the applications at a high level of abstraction: the use cases and user interactions; the designs are documented and used, together with the data and events captured later that represents the user interactions with the systems, to feed an intermediate component local to the users -the IFM (InterFace Mapper) -which bridges the gaps between the users and the systems. We discuss the main issues faced in the design and prototyping. The approach alleviates the need for re-programming with the APIs to any back-end service thus easing the development and distribution of the applications

GeneGrid: Architecture, Implementation and Application

The emergence of Grid computing technology has opened up an unprecedented opportunity for biologists to share and access data, resources and tools in an integrated environment leading to a greater chance of knowledge discovery. GeneGrid is a Grid computing framework that seamlessly integrates a myriad of heterogeneous resources spanning multiple administrative domains and locations. It provides scientists an integrated environment for the streamlined access of a number of bioinformatics programs and databases through a simple and intuitive interface. It acts as a virtual bioinformatics laboratory by allowing scientists to create, execute and manage workflows that represent bioinformatics experiments. A number of cooperating Grid services interact in an orchestrated manner to provide this functionality. This paper gives insight into the details of the architecture, components and implementation of GeneGrid.

A Novel Multimodal Interface for Improving Visually Impaired People's Web Accessibility

This paper introduces a novel interface designed to help blind and visually impaired people to explore and navigate on the Web. In contrast to traditionally used assistive tools, such as screen readers and magnifiers, the new interface employs a combination of both audio and haptic features to provide spatial and navigational information to users. The haptic features are presented via a low-cost force feedback mouse allowing blind people to interact with the Web, in a similar fashion to their sighted counterparts. The audio provides navigational and textual information through the use of non-speech sounds and synthesised speech. Interacting with the multimodal interface offers a novel experience to target users, especially to those with total blindness. A series of experiments have been conducted to ascertain the usability of the interface and compare its performance to that of a traditional screen reader. Results have shown the advantages that the new multimodal interface offers blind and visually impaired people. This includes the enhanced perception of the spatial layout of Web pages, and navigation towards elements on a page. Certain issues regarding the design of the haptic and audio features raised in the evaluation are discussed and presented in terms of recommendations for future work.

Application of semiquantitative proteomics techniques to the Maillard reaction

Proteomic tools-in particular, mass spectrometry (MS)-have advanced significantly in recent years, and the identification of proteins within complex mixtures is now a routine procedure. Quantitative methods of analysis are less well advanced and continue to develop. These include the use of stable isotope ratio approaches, isotopically labeled peptide standards, and nonlabeling methods. This paper summarizes the use of MS as a proteomics tool to identify and semiquantify proteins and their modified forms by using examples of relevance to the Maillard reaction. Finally, some challenges for the future are presented.

A role of cellular prion protein in programming T-cell cytokine responses in disease.

The cellular prion protein (PrPC) is widely expressed in neural and non-neural tissues, but its function is unknown. Elucidation of the part played by PrPC in adaptive immunity has been a particular conundrum: increased expression of cell surface PrPC has been documented during T-cell activation, yet the functional significance of this activation remains unclear, with conflicting data on the effects of Prnp gene knockout on various parameters of T-cell immunity. We show here that Prnp mRNA is highly inducible within 8–24 h of T-cell activation, with surface protein levels rising from 24 h. When measured in parallel with CD69 and CD25, PrPC is a late activation antigen. Consistent with its up-regulation being a late activation event, PrP deletion did not alter T-cell-antigen presenting cell conjugate formation. Most important, activated PrP0/0 T cells demonstrated much reduced induction of several T helper (Th) 1, Th2, and Th17 cytokines, whereas others, such as TNF- and IL-9, were unaffected. These changes were investigated in the context of an autoimmune model and a bacterial challenge model. In experimental autoimmune encephalomyelitis, PrP-knockout mice showed enhanced disease in the face of reduced IL-17 responses. In a streptococcal sepsis model, this constrained cytokine program was associated with poorer local control of infection, although with reduced bacteremia. The findings indicate that PrPC is a potentially important molecule influencing T-cell activation and effector function.

Semi-formal Models to Support Program Development: Autonomic Management within Component Based Parallel and Distributed Programming

Functional and non-functional concerns require different programming effort, different techniques and different methodologies when attempting to program efficient parallel/distributed applications. In this work we present a "programmer oriented" methodology based on formal tools that permits reasoning about parallel/distributed program development and refinement. The proposed methodology is semi-formal in that it does not require the exploitation of highly formal tools and techniques, while providing a palatable and effective support to programmers developing parallel/distributed applications, in particular when handling non-functional concerns.