IEEE International Symposium on Parallel & Distributed Processing, Workshops and Phd Forum : [proceedings]. IEEE International Symposium on Parallel & Distributed Processing, Workshops and Phd Forum最新文献

The matching problem formulated as Maximum Cardinality Matching in General Graphs (MCMGG) finds the largest matching on graphs without restrictions. The Micali-Vazirani algorithm has the best asymptotic complexity for solving MCMGG when the graphs are sparse. Parallelizing matching in general graphs on the GPU is difficult for multiple reasons. First, the augmenting path procedure is highly recursive, and NVIDIA GPUs use registers to store kernel arguments, which eventually spill into cached device memory, with a performance penalty. Second, extracting parallelism from the matching process requires partitioning the graph to avoid any overlapping augmenting paths. We propose an implementation of the Micali-Vazirani algorithm which identifies bridge edges using thread-parallel breadth-first search, followed by block-parallel path augmentation and blossom contraction. Augmenting path and Union-find methods were implemented as stack-based iterative methods, with a stack allocated in shared memory. Our experimentation shows that compared to the serial implementation, our approach results in up to 15-fold speed-up for very sparse regular graphs, up to 5-fold slowdown for denser regular graphs, and finally a 50-fold slowdown for power-law distributed Kronecker graphs. This implementation has been open-sourced for further research on developing combinatorial graph algorithms on GPUs.

The Edmonds Blossom algorithm is implemented here using depth-first search, which is intrinsically serial. By streamlining the code, our serial implementation is consistently three to five times faster than the previously fastest general graph matching code. By extracting parallelism across iterations of the algorithm, with coarse-grain locking, we are able to further reduce the run time on random regular graphs four-fold and obtain a two-fold reduction of run time on real-world graphs with similar topology. Solving very sparse graphs (average degree less than four) exhibiting community structure with eight threads led to a slow down of three-fold, but this slow down is replaced by marginal speed up once the average degree is greater than four. We conclude that our parallel coarse-grain locking implementation performs well when extracting parallelism from this augmenting-path-based algorithm and may work well for similar algorithms.

Criminal justice involved (CJI) individuals with a history of opioid use disorder (OUD) are at high risk of overdose and death in the weeks following release from jail. We developed the Justice-Community Circulation Model (JCCM) to investigate OUD/CJI dynamics post-release and the effects of interventions on overdose deaths. The JCCM uses a synthetic agent-based model population of approximately 150,000 unique individuals that is generated using demographic information collected from multiple Chicago-area studies and data sets. We use a high-performance computing (HPC) workflow to implement a sequential approximate Bayesian computation algorithm for calibrating the JCCM. The calibration results in the simulated joint posterior distribution of the JCCM input parameters. The calibrated model is used to investigate the effects of a naloxone intervention for a mass jail release. The simulation results show the degree to which a targeted intervention focusing on recently released jail inmates can help reduce the risk of death from opioid overdose.

The productivity of computational biologists is limited by the speed of their workflows and subsequent overall job throughput. Because most biomedical researchers are focused on better understanding scientific phenomena rather than developing and optimizing code, a computing and data system implemented in an adventitious and/or non-optimized manner can impede the progress of scientific discovery. In our experience, most computational, life-science applications do not generally leverage the full capabilities of high-performance computing, so tuning a system for these applications is especially critical. To optimize a system effectively, systems staff must understand the effects of the applications on the system. Effective stewardship of the system includes an analysis of the impact of the applications on the compute cores, file system, resource manager and queuing policies. The resulting improved system design, and enactment of a sustainability plan, help to enable a long-term resource for productive computational and data science. We present a case study of a typical biomedical computational workload at a leading academic medical center supporting over $100 million per year in computational biology research. Over the past eight years, our high-performance computing system has enabled over 900 biomedical publications in four major areas: genetics and population analysis, gene expression, machine learning, and structural and chemical biology. We have upgraded the system several times in response to trends, actual usage, and user feedback. Major components crucial to this evolution include scheduling structure and policies, memory size, compute type and speed, parallel file system capabilities, and deployment of cloud technologies. We evolved a 70 teraflop machine to a 1.4 petaflop machine in seven years and grew our user base nearly 10-fold. For long-term stability and sustainability, we established a chargeback fee structure. Our overarching guiding principle for each progression has been to increase scientific throughput and enable enhanced scientific fidelity with minimal impact to existing user workflows or code. This highly-constrained system optimization has presented unique challenges, leading us to adopt new approaches to provide constructive pathways forward. We share our practical strategies resulting from our ongoing growth and assessments.

We explore the feasibility of a database storage engine housing up to 307 billion genetic Single Nucleotide Polymorphisms (SNP) for online access. We evaluate database storage engines and implement a solution utilizing factors such as dataset size, information gain, cost and hardware constraints. Our solution provides a full feature functional model for scalable storage and query-ability for researchers exploring the SNP's in the human genome. We address the scalability problem by building physical infrastructure and comparing final costs to a major cloud provider.

Many of the continuing scientific advances achieved through computational biology are predicated on the availability of ongoing increases in computational power required for detailed simulation and analysis of cellular processes on biologically-relevant timescales. A critical challenge facing the development of future exascale supercomputer systems is the development of new computing hardware and associated scientific applications that dramatically improve upon the energy efficiency of existing solutions, while providing increased simulation, analysis, and visualization performance. Mobile computing platforms have recently become powerful enough to support interactive molecular visualization tasks that were previously only possible on laptops and workstations, creating future opportunities for their convenient use for meetings, remote collaboration, and as head mounted displays for immersive stereoscopic viewing. We describe early experiences adapting several biomolecular simulation and analysis applications for emerging heterogeneous computing platforms that combine power-efficient system-on-chip multi-core CPUs with high-performance massively parallel GPUs. We present low-cost power monitoring instrumentation that provides sufficient temporal resolution to evaluate the power consumption of individual CPU algorithms and GPU kernels. We compare the performance and energy efficiency of scientific applications running on emerging platforms with results obtained on traditional platforms, identify hardware and algorithmic performance bottlenecks that affect the usability of these platforms, and describe avenues for improving both the hardware and applications in pursuit of the needs of molecular modeling tasks on mobile devices and future exascale computers.

We present an efficient and scalable scheme for implementing agent-based modeling (ABM) simulation with In Situ visualization of large complex systems on heterogeneous computing platforms. The scheme is designed to make optimal use of the resources available on a heterogeneous platform consisting of a multicore CPU and a GPU, resulting in minimal to no resource idle time. Furthermore, the scheme was implemented under a client-server paradigm that enables remote users to visualize and analyze simulation data as it is being generated at each time step of the model. Performance of a simulation case study of vocal fold inflammation and wound healing with 3.8 million agents shows 35× and 7× speedup in execution time over single-core and multi-core CPU respectively. Each iteration of the model took less than 200 ms to simulate, visualize and send the results to the client. This enables users to monitor the simulation in real-time and modify its course as needed.

Secondary structures of ribonucleic acid (RNA) molecules play important roles in many biological processes including gene expression and regulation. Experimental observations and computing limitations suggest that we can approach the secondary structure prediction problem for long RNA sequences by segmenting them into shorter chunks, predicting the secondary structures of each chunk individually using existing prediction programs, and then assembling the results to give the structure of the original sequence. The selection of cutting points is a crucial component of the segmenting step. Noting that stem-loops and pseudoknots always contain an inversion, i.e., a stretch of nucleotides followed closely by its inverse complementary sequence, we developed two cutting methods for segmenting long RNA sequences based on inversion excursions: the centered and optimized method. Each step of searching for inversions, chunking, and predictions can be performed in parallel. In this paper we use a MapReduce framework, i.e., Hadoop, to extensively explore meaningful inversion stem lengths and gap sizes for the segmentation and identify correlations between chunking methods and prediction accuracy. We show that for a set of long RNA sequences in the RFAM database, whose secondary structures are known to contain pseudoknots, our approach predicts secondary structures more accurately than methods that do not segment the sequence, when the latter predictions are possible computationally. We also show that, as sequences exceed certain lengths, some programs cannot computationally predict pseudoknots while our chunking methods can. Overall, our predicted structures still retain the accuracy level of the original prediction programs when compared with known experimental secondary structure.