Parallel Pipeline Research Articles

Trireme: Exploration of Hierarchical Multi-level Parallelism for Hardware Acceleration

The design of heterogeneous systems that include domain specific accelerators is a challenging and time-consuming process. While taking into account area constraints, designers must decide which parts of an application to accelerate in hardware and which to leave in software. Moreover, applications in domains such as Extended Reality (XR) offer opportunities for various forms of parallel execution, including loop level, task level, and pipeline parallelism. To assist the design process and expose every possible level of parallelism, we present Trireme , a fully automated tool-chain that explores multiple levels of parallelism and produces domain-specific accelerator designs and configurations that maximize performance, given an area budget. FPGA SoCs were used as target platforms, and Catapult HLS [ 7 ] was used to synthesize RTL using a commercial 12 nm FinFET technology. Experiments on demanding benchmarks from the XR domain revealed a speedup of up to 20×, as well as a speedup of up to 37× for smaller applications, compared to software-only implementations.

Image lossless compression algorithm optimization and FPGA implementation

In this thesis, a minimum redundant prefix coding with higher compression ratio and lower time complexity is proposed for lossless compression of HD images. The compression algorithm is based on Canonical Huffman coding, preprocesses the data source to be compressed according to the image data features, and then compresses the data in batches using the locally uneven features in the data, which improves the compression ratio by 1.678 times compared with the traditional canonical Huffman coding. During the implementation of the algorithm, the counting sorting method with lower time complexity and the code-length table construction method without relying on binary trees are used to reduce the complexity of the algorithm and achieve the purpose of real-time data processing by the system. Finally, the proposed compression algorithm is deployed in FPGA to improve the encoding rate by parallel hardware circuit and pipeline design.

DeHIN: A Decentralized Framework for Embedding Large-Scale Heterogeneous Information Networks

Modeling heterogeneity by extraction and exploitation of high-order information from heterogeneous information networks (HINs) has been attracting immense research attention in recent times. Such heterogeneous network embedding (HNE) methods effectively harness the heterogeneity of small-scale HINs. However, in the real world, the size of HINs grow exponentially with the continuous introduction of new nodes and different types of links, making it a billion-scale network. Learning node embeddings on such HINs creates a performance bottleneck for existing HNE methods that are commonly centralized, i.e., complete data and the model are both on a single machine. To address large-scale HNE tasks with strong efficiency and effectiveness guarantee, we present <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Decentralized Embedding Framework for Heterogeneous Information Network</i> (DeHIN) in this paper. In DeHIN, we generate a distributed parallel pipeline that utilizes hypergraphs in order to infuse parallelization into the HNE task. DeHIN presents a context preserving partition mechanism that innovatively formulates a large HIN as a hypergraph, whose hyperedges connect semantically similar nodes. Our framework then adopts a decentralized strategy to efficiently partition HINs by adopting a tree-like pipeline. Then, each resulting subnetwork is assigned to a distributed worker, which employs the deep information maximization theorem to locally learn node embeddings from the partition it receives. We further devise a novel embedding alignment scheme to precisely project independently learned node embeddings from all subnetworks onto a common vector space, thus allowing for downstream tasks like link prediction and node classification. As shown from our experimental results, DeHIN significantly improves the efficiency and accuracy of existing HNE models as well as outperforms the large-scale graph embedding frameworks by efficiently scaling up to large-scale HINs.

Consideration of Wire Transposition in Modeling the Electromagnetic Effects of Power Lines on Pipelines

In some sections, high-voltage power lines may come close to pipelines. This may give rise to problems concerned with ensuring safety of personnel and protecting the structure from corrosion. An analysis of publications in this field shows that many studies aimed at solving these problems place focus on determining the electromagnetic effects of power lines on extended metal structures. However, the authors of these publications have not set out a unified technique that would make it possible to determine the voltages induced on the structural components and the currents flowing through the pipeline. Such a technique can be implemented based on methods for modeling the electric power systems operation modes in phase coordinates. The article gives its further development and presents the results of studies aimed at solving the problem of considering the transposition of wires in modeling the electromagnetic effects of power lines on pipelines. The developed technique and digital models make it possible to adequately consider the transposition of wires in determining the voltages induced on the parts of a parallel pipeline and the currents flowing through the pipe. Based on them, it is possible to reasonably select measures to ensure the safety of personnel serving the structure, as well as to develop methods and means of protection against corrosion.

Design and Implementation of Convolutional Neural Network Accelerator Based on FPGA

Convolutional neural network, as a kind of feed-forward neural network, has been widely used in image recognition, speech processing and other fields in recent years. In this paper, an FPGA-based CNN gas pedal is designed to solve the problem of slow running and high-power consumption of CNN on resource-constrained hardware. The design invokes multi-stage pipeline parallel processing technology to accelerate convolutional operations; quantifies network parameters from 32-bit floating-point to 8-bit fixed-point while guaranteeing CNN accuracy, and uses data multiplexing to reduce resource consumption. Experimental results show that the design is 10 times faster than the intel i7-8700 and consumes only 1% of the power of the RTX 2060 at 50MHz.

A bidirectional DNN partition mechanism for efficient pipeline parallel training in cloud

Recently, deep neural networks (DNNs) have shown great promise in many fields while their parameter sizes are rapidly expanding. To break through the computation and memory limitation of a single machine, pipeline model parallelism is proposed for large-scale DNN training by fully utilizing the computation and storage power of the distributed cluster. Cloud data centers can also provide sufficient computing, storage and bandwidth resources. However, most existing approaches apply layer-wise partitioning, which is difficult to obtain an even model partition result because of the large computational overhead discrepancy between DNN layers, resulting in degraded efficiency. To tackle this issue, we propose “Bi-Partition”, a novel partitioning method based on bidirectional partitioning for forward propagation (FP) and backward propagation (BP), which improves the efficiency of the pipeline model parallelism system. By deliberated designing distinct cut positions for FP and BP of DNN training, workers in the pipeline get nearly equal computational loads, and the balanced pipeline fully utilizes the computing resources. Experiments on various DNN models and datasets validate the efficiency of our mechanism, e.g., the training efficiency achieving up to 1.9times faster than the state-of-the-art method PipeDream.

Novel multi-bit parallel pipeline-circuit design for STT-MRAM

In the paper, novel multi-bit parallel pipeline circuit design of STT-MRAM is proposed to improve the read and write efficiency. The shift register is utilized to change the series data into the parallel ones. Combined with the corresponding sense amplifier (SA) circuit and the write drive circuit, 4-bit data can be read or written in parallel in single cycle. With the 32 Kbit data test, the efficiencies of the read and the write operations of the proposed pipeline circuit are increased by 72.3% and 74.4%, separately, with the comparison of the incumbent series operations. The proposed strategy has potential application in the embedded high speed STT-MRAM.

A hybrid machine learning model for predicting crater width formed by explosions of natural gas pipelines

Despite the existence of industry models for estimating the crater width formed by the explosion of natural gas pipelines, their applicability is still limited since the complex formation mechanisms. In this work, a novel hybrid model was developed to predict crater width formed by explosions of natural gas pipelines, using artificial neural networks (ANN) as the fundamental predictor. Based on the historical accident records, the proposed hybrid model was trained by the pipeline parameter, the operating condition, the installation parameter, and the crater width. A novel nature-inspired optimization algorithm, i.e., the Lévy-Weighted Quantum particle swarm optimization (LWQPSO) algorithm, was proposed to optimize the ANN model's parameters. Three machine learning models were developed for comparative reasons to predict the crater width. The use of precision and error analysis indicators assesses prediction performance. The results show that the proposed hybrid model (LWQPSO-ANN) has high prediction accuracy and stability, which outperforms QPSO-ANN-based benchmark hybrid models and the model without an optimizer (Support Vector Machine, SVM). The parameter sensitivities of the proposed algorithm, including the maximum number of iterations, population size and contraction-expansion coefficient, were determined. The proposed hybrid model is expected to support the quantitative risk assessment (QRA), Right-of-Way (ROW) definition and the inherently safer design of the underground parallel pipelines.

Chicken Production and Human Clinical Escherichia coli Isolates Differ in Their Carriage of Antimicrobial Resistance and Virulence Factors.

Contamination of food animal products by Escherichia coli is a leading cause of foodborne disease outbreaks, hospitalizations, and deaths in humans. Chicken is the most consumed meat both in the United States and across the globe according to the U.S. Department of Agriculture. Although E. coli is a ubiquitous commensal bacterium of the guts of humans and animals, its ability to acquire antimicrobial resistance (AMR) genes and virulence factors (VFs) can lead to the emergence of pathogenic strains that are resistant to critically important antibiotics. Thus, it is important to identify the genetic factors that contribute to the virulence and AMR of E. coli. In this study, we performed in-depth genomic evaluation of AMR genes and VFs of E. coli genomes available through the National Antimicrobial Resistance Monitoring System GenomeTrackr database. Our objective was to determine the genetic relatedness of chicken production isolates and human clinical isolates. To achieve this aim, we first developed a massively parallel analytical pipeline (Reads2Resistome) to accurately characterize the resistome of each E. coli genome, including the AMR genes and VFs harbored. We used random forests and hierarchical clustering to show that AMR genes and VFs are sufficient to classify isolates into different pathogenic phylogroups and host origin. We found that the presence of key type III secretion system and AMR genes differentiated human clinical isolates from chicken production isolates. These results further improve our understanding of the interconnected role AMR genes and VFs play in shaping the evolution of pathogenic E. coli strains. IMPORTANCE Pathogenic Escherichia coli causes disease in both humans and food-producing animals. E. coli pathogenesis is dependent on a repertoire of virulence factors and antimicrobial resistance genes. Food-borne outbreaks are highly associated with the consumption of undercooked and contaminated food products. This association highlights the need to understand the genetic factors that make E. coli virulent and pathogenic in humans and poultry. This research shows that E. coli isolates originating from human clinical settings and chicken production harbor different antimicrobial resistance genes and virulence factors that can be used to classify them into phylogroups and host origins. In addition, to aid in the repeatability and reproducibility of the results presented in this study, we have made a public repository of the Reads2Resistome pipeline and have provided the accession numbers associated with the E. coli genomes analyzed.

Parallel Implementation of RC4 Data Encryption Method for Cloud Computing

Cloud computing is an on-demand availability of computing resources. The current cloud computing environment uses different algorithms for its security; RC4 is one of them. However, due to the serial implementation of RC4 over cloud computing, the process of encryption and data transmission is badly affected by the factors of poor latency. This research is proposed to cover this factor by implementing a parallel RC4 algorithm in a cloud computing environment. The parallel RC4 model has been developed in web-based architecture relevant to the cloud computing environment. The method processed data simultaneously in four parallel pipelines encrypted through the RC4 algorithm. The proposed method used RC4 algorithms’ parallel implementation to encrypt and decrypt the data. The proposed method increases the efficiency of data encryption and decryption and transmission over a cloud environment using four pipelines at once. These four pipelines receive data and encrypt them using a key and then merge those four streams into a single cipher. These four pipelines enhance the speed of communication over cloud and solve the latency issue majorly. Data sending from node devices to cloud servers is divided into four equal streams and then encrypted using RC4 distinctly but parallel. Then after the roaming, the data are again first combined and then decrypted using the RC4 algorithm. The sample data sets are evaluated on the criteria of the starting time and ending time of the encryption and decryption of data. Results show that the proposed method is able to enhance the speed of the whole process up to average 3.7% due to the implementation of parallel technique and had reduced the time notably. The future work on this research involves enhancement of security process of RC4 algorithm since it can be breached easily.

Numerical Investigation on the Migration of Leaked Pollutants after Liquid Pressurized Pipeline Leakage Regarding Oil and Gas Parallel Pipelines Situation

Numerical Investigation on the Migration of Leaked Pollutants after Liquid Pressurized Pipeline Leakage Regarding Oil and Gas Parallel Pipelines Situation

A Parallel Framework for Streaming Dimensionality Reduction.

The visualization of streaming high-dimensional data often needs to consider the speed in dimensionality reduction algorithms, the quality of visualized data patterns, and the stability of view graphs that usually change over time with new data. Existing methods of streaming high-dimensional data visualization primarily line up essential modules in a serial manner and often face challenges in satisfying all these design considerations. In this research, we propose a novel parallel framework for streaming high-dimensional data visualization to achieve high data processing speed, high quality in data patterns, and good stability in visual presentations. This framework arranges all essential modules in parallel to mitigate the delays caused by module waiting in serial setups. In addition, to facilitate the parallel pipeline, we redesign these modules with a parametric non-linear embedding method for new data embedding, an incremental learning method for online embedding function updating, and a hybrid strategy for optimized embedding updating. We also improve the coordination mechanism among these modules. Our experiments show that our method has advantages in embedding speed, quality, and stability over other existing methods to visualize streaming high-dimensional data.

Exploring Deep Ensemble Model for Insect and Pest Detection from Images

Pests and insects are the main hazards to crops, and they have a significant negative impact on both human health and wealth. Identification and classification of different types of crop insects accurately is an important task to avoid these losses. A manual examination is a quite sluggish and less efficient way of accomplishing this task while computer vision technology can significantly assist in this issue. A CNN-based deep learning model can classify insects and pests efficiently. In this paper, an ensemble-based model has been proposed using transfer learning. The experimental setup comprises pre-trained models like VGG16, VGG19, and ResNetv50 with a voting classifier ensemble technique. These pre-trained models are used to train the training dataset in a parallel pipeline model which is united with Ensemble Voting Classifier to generate the final prediction over the input sample. The benchmark IP102 dataset, having more than 75000 samples over 102 classes, is used to train and evaluate the model's performance. Results show that the proposed ensemble model achieves an accuracy of 82.5%, which is significant improvement over the recent state-of-the-art models proposed over the same selected dataset, hence strongly supporting the efficacy of the model.

A Gate Multiplexing Architecture‐Based Artificial Visual Sensor and Memory System

The human visual system plays the most important role in information perception. Inspired by its structure and processing mechanism, novel optoelectronic memristor‐based artificial visual systems have emerged, thanks to the advantages of time latency and energy efficiency. However, the established systems lack a complete hardware supporting system to demonstrate the potential of large‐scale integration. Herein for the first time, a gate multiplexing architecture for optical signal sensor and memory integration is proposed, which can achieve multivalue storage of 3‐bit optical signal in the analog domain. Based on this novel structure, an artificial visual sensor and memory system (AVSMS) is implemented with a 5 × 6 photodiode array, an I/V conversion circuit, and a 32 × 32 1T1R memristor array for image sensor and memory integration. Furthermore, a parallel pipeline storage strategy is proposed to enable AVSMS to store 32 images with a size of 5 × 6 pixels and 3‐bit grayscale without erasing operation and achieve an ideal frame rate of more than 60 000 frames per second for high‐speed storage. Experimental results demonstrate that AVSMS can directly eliminate repeated background information in sequential images through peripheral hardware circuitry. The AVSMS provides a promising hardware solution for large‐scale image sensor and memory integration in artificial visual systems.

Assessment of permafrost disturbances caused by two parallel buried warm-oil pipelines: A case study at a high-latitude wetland site in Northeast China

Multiple studies demonstrate that narrow-linear strong disturbances triggered by pipeline engineering activities have resulted in rapid permafrost degradation. However, the extent, duration, and severity of permafrost disturbances induced by two parallel buried warm-oil pipelines remain unclear. Here, we examine the thermal disturbances of the China-Russia crude oil pipelines (CRCOPs I and II) on the surrounding permafrost. Ground temperature measurements on and off the right-of-way of the CRCOPs and an electrical resistivity tomography survey were conducted at a selected high-latitude wetland site in northeastern China. The observations showed that warm oil flow in pipelines dissipated heat to the surrounding soil, resulting in the thawing of underlying permafrost accompanied by the formation and development of thaw bulbs around the pipes. Currently, the thaw bulb of CRCOP-I was about four times as large as that of CRCOP-II, mainly due to the longer 7-year working duration of CRCOP-I. To quantify the permafrost disturbances around the pipeline, a conductive heat transfer model was established. A numerical simulation of the soil thermal regime around the pipeline suggested the initial growth of the thaw bulb was quite fast but gradually slowed with time, of which the shape was modified to ellipse from a semicircle. The existence of CRCOP-II would accelerate the warming of permafrost on the CRCOP-I right-of-way due to the small separation distance between them. We expect that the thaw bulbs of the two pipelines will expand larger than previously estimated due to the adverse effects of pipeline settlement and surface water flow, sparking a wider permafrost disturbance.

UROSCAN and UROSCANSEQ: a large-scale multicenter effort towards translation of molecular bladder cancer subtypes into clinical practice – from biobank to RNA-sequencing in real time

Background Bladder cancer is molecularly one of the most heterogenous malignancies characterized by equally heterogenous clinical outcomes. Standard morphological assessment with pathology and added immunohistochemical analyses is unable to fully address the heterogeneity, but up to now treatment decisions have been made based on such information only. Bladder cancer molecular subtypes will likely provide means for a more personalized bladder cancer care. Methods To facilitate further development of bladder cancer molecular subtypes and clinical translation, the UROSCAN-biobank was initiated in 2013 to achieve systematic biobanking of preoperative blood and fresh frozen tumor tissue in a population-based setting. In a second phase, we established in 2018 a parallel logistic pipeline for molecular profiling by RNA-sequencing, to develop and validate clinical implementation of molecular subtyping and actionable molecular target identification in real-time. Results Until June 2021, 1825 individuals were included in the UROSCAN-biobank, of which 1650 (90%) had primary bladder cancer, 127 (7%) recurrent tumors, and 48 (3%) unknown tumor status. In 159 patients, multiple tumors were sampled, and metachronous tumors were collected in 83 patients. Between 2016 and 2020 the UROSCAN-biobanking included 1122/2999 (37%) of all primary bladder cancer patients in the Southern Healthcare Region. Until June 2021, the corresponding numbers subjected to RNA-sequencing and molecular subtyping was 605 (UROSCANSEQ), of which 52 (9%) samples were not sequenced due to inadequate RNA-quality (n = 47) or technical failure/lost sample (n = 5). Conclusions The UROSCAN-biobanking and UROSCANSEQ-infrastructure for molecular subtyping by real-time RNA-sequencing represents, to our knowledge, the largest effort of evaluating population-wide molecular classification of bladder cancer.

Memory-Throughput Trade-off for CNN-Based Applications at the Edge

Many modern applications require execution of Convolutional Neural Networks (CNNs) on edge devices, such as mobile phones or embedded platforms. This can be challenging, as the state-of-the art CNNs are memory costly, whereas the memory budget of edge devices is highly limited. To address this challenge, a variety of CNN memory reduction methodologies have been proposed. Typically, the memory of a CNN is reduced using methodologies such as pruning and quantization. These methodologies reduce the number or precision of CNN parameters, thereby reducing the CNN memory cost. When more aggressive CNN memory reduction is required, the pruning and quantization methodologies can be combined with CNN memory reuse methodologies. The latter methodologies reuse device memory allocated for storage of CNN intermediate computational results, thereby further reducing the CNN memory cost. However, the existing memory reuse methodologies are unfit for CNN-based applications that exploit pipeline parallelism available within the CNNs or use multiple CNNs to perform their functionality. In this article, we therefore propose a novel CNN memory reuse methodology. In our methodology, we significantly extend and combine two existing CNN memory reuse methodologies to offer efficient memory reuse for a wide range of CNN-based applications.

Energy Consumption Optimization Model of Large Parallel Natural Gas Pipeline Network: Using Compressors with Multiple Operating Modes

In long-distance parallel pipelines, compressor stations are built together, and connecting pipes and cut-off valves are arranged between pipelines, which makes it possible for different stations to operate jointly or independently. In order to reduce pipeline energy consumption, this paper considers the valve switch state, deeply analyzes the station operation mode, and establishes the optimal model of minimum energy consumption of a parallel pipeline. Taking the second and third parallel pipelines of China’s west-east gas transmission as an example, the phasor particle swarm optimization algorithm is used to solve the model. The results show that after optimization, the internal flow was redistributed and energy consumption was reduced by 15.54%.

Fast Algorithms for Denial Constraint Discovery

Denial constraints (DCs) are an integrity constraint formalism widely used to detect inconsistencies in data. Several algorithms have been devised to discover DCs from data, as manually specifying them is burdensome and, worse yet, error-prone. The existing algorithms follow two basic steps: building an intermediate data structure from records, then enumerating the DCs from that intermediate. However, current algorithms are often inefficient in computing these intermediates. Also, it is still unclear which enumeration algorithm performs best since some of the available algorithms have not yet been compared to each other. In response, we present a set of new algorithms with improved design choices. We introduce a parallel pipeline for rapidly computing the intermediate using custom data representations, algorithms, and indexes. For DC enumeration, we propose an inverted index, pruning, and parallel search strategies. We present hybrid approaches that integrate our techniques with previous enumeration algorithms, improving their performance in many scenarios. Our experimental study shows that the proposed DC discovery algorithms are consistently much faster (up to an order of magnitude) than the current state-of-the-art.

Research on low carbon emission optimization operation technology of natural gas pipeline under multi-energy structure

In order to reduce the carbon emissions of natural gas pipelines, based on the background of different energy structures, this paper proposes a general low carbon and low consumption operation model of natural gas pipelines, which is used to fine calculate the carbon emissions and energy consumption of natural gas pipeline. In this paper, an improved particle swarm optimization (NHPSO-JTVAC) algorithm is used to solve the model and the optimal scheduling scheme is given. Taking a parallel pipeline located in western China as an example, the case is analyzed. The results show that after optimization, under the existing energy types, the pipeline system can reduce 31.14% of carbon emissions, and after introducing part of new energy, the pipeline system can reduce 34.02% of carbon emissions, but the energy consumption has increased.