H. Sidiropoulos, G. Chatzikonstantis, D. Soudris, C. Strydis
{"title":"The VINEYARD Framework for Heterogeneous Cloud Applications: The BrainFrame Case","authors":"H. Sidiropoulos, G. Chatzikonstantis, D. Soudris, C. Strydis","doi":"10.1109/DASIP.2018.8597119","DOIUrl":"https://doi.org/10.1109/DASIP.2018.8597119","url":null,"abstract":"Emerging cloud applications like machine learning, AI, big data analytics and scientific computing require highperformance computing systems that can sustain the increased amount of data processing without consuming excessive power. To this end, many cloud operators have started deploying hardware accelerators, like GPUs and FPGAs, to increase the performance of computationally intensive tasks. However, increased performance, comes at a higher cost of increased programming complexity for utilizing these accelerators. VINEYARD has developed a versatile framework that allows the seamless deployment and utilization of heterogeneous accelerators in the cloud without increasing the programming complexity while offering the flexibility of software packages. This paper presents the main components that have been developed in the VINEYARD framework and focuses on BrainFrame, the neurocomputing case that demonstrates the new framework's value. BrainFrame not only accelerates neuronal simulations but also has an architecture that allows easy access to neuroscientists, hiding the system complexity, and enabling a modular integration of new accelerated simulators.","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121129001","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Implementation of a Real-Time Image-Based Vibration Detection and Adaptive Filtering on an FPGA","authors":"Kazuya Uetsuhara, Akane Tahara, Taito Manabe, Yuichiro Shibata","doi":"10.1109/DASIP.2018.8597053","DOIUrl":"https://doi.org/10.1109/DASIP.2018.8597053","url":null,"abstract":"In this paper, we propose and implement a field-programmable gate array (FPGA) system which extracts a vibration component of a desired frequency band from moving images in real-time, aiming at application to image-based vibration suppression such as microsurgery assistance systems. The technical challenges to this end are two-fold: fast and robust detection of vibration components in given moving images and zero-phase band-pass filtering for a desired frequency band. For the former, we employ a statistical approach using dense optical flow to derive frequency components, and design a custom optical flow computing hardware with the Lucas-Kanade (LK) method. For the latter, we implement a sort of adaptive band-pass filters called a bandlimited multiple Fourier linear combiner (BMFLC), which can recompose input signals as a mixture of sinusoidal signals with multiple frequencies in a band with no phase delay. Both designs are implemented in a deeply pipelined manner on a Xilinx Kintex-7 XC325T FPGA, without using any external memories. Empirical experiments reveal that the proposed system extracts a vibration component of high-frequency tremors in hand motions, while intentional low-frequency motions are successfully filtered out. The system processes VGA moving images at 60 FPS, with a delay of less than 1 us for the BMFLC, suggesting effectiveness of the deep pipelined architecture.","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128410999","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Efficient Task-Based Code Generation for SDF Graph Execution on Multicore Processors","authors":"Georgios Georgakarakos, J. Lilius","doi":"10.1109/DASIP.2018.8596926","DOIUrl":"https://doi.org/10.1109/DASIP.2018.8596926","url":null,"abstract":"The dataflow programming paradigm is a flexible and efficient methodology to describe parallel algorithms and optimize their scheduling and mapping in emerging multicore architectures. Transforming dataflow descriptions in executable code without scheduling overheads is however cumbersome, due to the challenge of translating effectively dataflow semantics. One of the promising techniques that have been proposed to address this problem is the correlation of dataflow and task programming models. In this paper we propose an efficient task-based code generator for PREESM dataflow framework. Our generator exploits synchronous dataflow graph information in order to improve task mapping and minimise overheads in the code's execution. We compare our task-based code against the current PREESM task generated code as well as annotated code using the popular OmpSs programming model, for the same test application. Results show that our approach achieves higher application throughput in both symmetric and asymmetric multi-core processors.","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126896097","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Design and Open Source Implementation of a Reconfigurable Hardware Model Predicitive Controller Using Online Optimization","authors":"F. Kästner, F. Werner, Erman Seydioglu, M. Hübner","doi":"10.1109/DASIP.2018.8597098","DOIUrl":"https://doi.org/10.1109/DASIP.2018.8597098","url":null,"abstract":"Advanced Process Controls (APCs) are already widely and deeply established especially for industrial plants with high but complex optimization possibilities like chemical batch processes. The most famous representer is the Model Predictive Controller (MPC). Unlike traditional controllers, an explicit process model is used to predict the future reaction of the system, given the control input and the past states. In order to find optimal control input, this prediction is used to find a sufficient solution for a dynamic optimization problem. Due to the complex algorithms needed to find a global minimum of the constrained quadratic problem using online optimization with real-time capabilities, a suitable performance of the underlying hardware is required. FPGA implementations are especially interesting due to the application specific data flow parallelization character of MPC tasks. In this research we focus on the algorithm development and reconfigurable hardware implementation of a generic MPC using High-Level-Synthesis (HLS).","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"74 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116346509","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Comparison of Different Methods Making Use of Backup Copies for Fault-Tolerant Scheduling on Embedded Multiprocessor Systems","authors":"Petr Dobiáš, E. Casseau, O. Sinnen","doi":"10.1109/DASIP.2018.8597044","DOIUrl":"https://doi.org/10.1109/DASIP.2018.8597044","url":null,"abstract":"As transistors scale down, systems are more vulnerable to faults. Their reliability consequently becomes the main concern, especially in safety-critical applications such as automotive sector, aeronautics or nuclear plants. Many methods have already been introduced to conceive fault-tolerant systems and therefore improve the reliability. Nevertheless, several of them are not suitable for real-time embedded systems since they incur significant overheads, other methods may be less intrusive but at the cost of being too specific to a dedicated system. The aim of this paper is to analyse a method making use of two task copies when on-line scheduling tasks on multiprocessor systems. This method can guarantee the system reliability without causing too much overhead and requiring any special hardware components. In addition, it remains general and thus applicable to large amount of systems. Last but not least, this paper studies two techniques of processor allocation policies: the exhaustive search and the first found solution search. It is shown that the exhaustive search is not necessary for efficient fault-tolerant scheduling and that the latter search significantly reduces the computation complexity, which is interesting for embedded systems.","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"302 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116882075","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Scott Marshall, Garrett M. Vanhoy, A. Akoglu, T. Bose, Bo Ryu
{"title":"GPU Based Quarter Spectral Correlation Density Function","authors":"Scott Marshall, Garrett M. Vanhoy, A. Akoglu, T. Bose, Bo Ryu","doi":"10.1109/DASIP.2018.8596977","DOIUrl":"https://doi.org/10.1109/DASIP.2018.8596977","url":null,"abstract":"In this study we investigate the parallelization of a key feature extraction method called spectral correlation density (SCD) function, which is used in signal classification systems particularly under low signal-to-noise ratio conditions for classifying numerous signals. In order to reduce the computation complexity of the SCD function, we introduce a method called Quarter SCD (QSCD) that allows extracting features of a given signal by processing only quarter of the input signal data. We then parallelize the QSCD by targeting general purpose graphics processing unit (GPU) through architecture specific optimization strategies. We present experimental evaluations on identifying the parallelization configuration for maximizing the efficiency of the program architecture in utilizing the threading power of the GPU architecture. We show that algorithmic and architecture specific optimization strategies result with improving the throughput of the state of the art GPU based Full SCD from 120 signals/second to 2719 signals/second.","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122347754","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"[Front matter]","authors":"","doi":"10.1109/dasip.2018.8597160","DOIUrl":"https://doi.org/10.1109/dasip.2018.8597160","url":null,"abstract":"","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"4 6","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114039568","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Hardware-Software Implementation of a SFM Module for Navigation an Unmanned Aerial Vehicles-A Demo","authors":"Karol Radwan, T. Kryjak, M. Gorgon","doi":"10.1109/DASIP.2018.8596937","DOIUrl":"https://doi.org/10.1109/DASIP.2018.8596937","url":null,"abstract":"In this paper the design of an embedded vision system for structure from motion (SFM) computation is presented. The solution allows for scene depth reconstruction based on two consecutive video frames registered by a moving camera. The module is developed for a Xilinx Zynq SoC (System on Chip) device and is divided into a hardware and software part. The programmable logic (PL) is used mainly for detect and match features. Fundamental matrix estimation and triangulation are executed in the processing system (PS-ARM based). The module is dedicated to navigation of unmanned aerial vehicles (UAV)-autonomous flight and landing. The system is adapted to process a 1280×720 pixel video stream @ 60 frames per second in realtime.","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114982466","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Real-Time Implementation of Contextual Image Processing Operations for 4K Video Stream in Zynq UltraScale+ MPSoC","authors":"M. Kowalczyk, Dominika Przewlocka, T. Kryjak","doi":"10.1109/DASIP.2018.8597105","DOIUrl":"https://doi.org/10.1109/DASIP.2018.8597105","url":null,"abstract":"In this paper hardware implementation of selected contextual based image pre-processing modules for a $3840times 2160$ @60 FPS video stream in a Zynq UltraScale+ MPSoC is discussed. The following operations are considered: simple averaging (box filter), Gaussian filter, edge detection using the Sobel and Canny methods, median filter and morphological erosion and dilation operations. The scheme for implementing contextual based operations for a video stream in the format of 2 and 4 pixels per clock and challenges related to the pipelined implementation of processing such data are described. Also the use of logic resources and energy efficiency of modules described in the Verilog hardware description language and using the High Level Synthesis tools (Vivado HLS, SDSoC and xfOpenCV library) are compared. All designed modules support real-time processing of a 4K@60 FPS video stream.","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"175 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116210073","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Exploring Performance Improvement Opportunities in Directive-Based GPU Programming","authors":"R. Diarra, A. Mérigot, B. Vincke","doi":"10.1109/DASIP.2018.8597015","DOIUrl":"https://doi.org/10.1109/DASIP.2018.8597015","url":null,"abstract":"GPUs offer an impressive computing power, but their effective exploitation remains an open problem. Kernel based programming models such as CUDA or OpenCL, allow a direct programming of the GPU architecture and can drive to excellent performance. However, these programming models require significant code changes, often tedious and error-prone, before getting an optimized program. Directive-based programming models (such as OpenMP and OpenACC) are now available for GPU and can offer good trade-off between performance, portability and development cost. In this paper, we do a comparative performance study between OpenACC, OpenMP 4.5 and CUDA, which is essential for facilitating parallel programming for GPUs. In order to find most significant performance issues, we port a suite of representative benchmarks and three computer vision applications to OpenACC, OpenMP and CUDA. Beyond runtime, we explore factors that influence performance, such as register counts, workload, grid and block sizes. The results of this work show that either OpenACC or OpenMP are good alternatives to kernel based programming models, provided some careful manual optimization is performed. Through the analysis of generated PTXs files, we discover that there is in general a systematic overhead in the kernel launch for OpenMP, but, for most applications, it is not a big issue provided the kernel has a sufficient workload.","PeriodicalId":328009,"journal":{"name":"2018 Conference on Design and Architectures for Signal and Image Processing (DASIP)","volume":"54 64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129521847","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}