Alfonso Rodriguez

Dynamic management of multikernel multithread accelerators using Dynamic Partial Reconfiguration

Ever demanding systems with restricted resources face increasingly complex applications. Additionally, changeable environments modify working conditions over time. Therefore, a dynamic resource management is required in order to provide adaptation capabilities. By using ARTICo3, a bus-based architecture with reconfigurable slots, this adaptation is accomplished in three different but dependent areas: Consumption, Confidentiality and fault tolerance, and Computation. The proposed resource management strategies rely on an architecture and a model of computation that make execution configuration to be application-independent, but context-aware, since a CUDA-like execution model is used. The inherent and explicit application-level parallelism of multithreaded CUDA kernels is used to generate hardware accelerators that act as thread blocks. Despite other modes of operation provided by the ARTICo3 architecture, like module redundancy or dual-rail operation to mitigate Side-Channel Attacks, these thread blocks are dynamically managed and their execution is scheduled using a multiobjective optimization algorithm.

Execution modeling in self-aware FPGA-based architectures for efficient resource management

SRAM-based FPGAs have significantly improved their performance and size with the use of newer and ultra-deep-submicron technologies, even though power consumption, together with a time-consuming initial configuration process, are still major concerns when targeting energy-efficient solutions. System self-awareness enables the use of strategies to enhance system performance and power optimization taking into account run-time metrics. This is of particular importance when dealing with reconfigurable systems that may make use of such information for efficient resource management, such as in the case of the ARTICo3 architecture, which fosters dynamic execution of kernels formed by multiple blocks of threads allocated in a variable number of hardware accelerators, combined with module redundancy for fault tolerance and other dependability enhancements, e.g. side-channel-attack protection. In this paper, a model for efficient dynamic resource management focused on both power consumption and execution times in the ARTICo3 architecture is proposed. The approach enables the characterization of kernel execution by using the model, providing additional decision criteria based on energy efficiency, so that resource allocation and scheduling policies may adapt to changing conditions. Two different platforms have been used to validate the proposal and show the generalization of the model: a high-performance wireless sensor node based on a Spartan-6 and a standard off-the-shelf development board based on a Kintex-7.

Evolutionary Computing and Particle Filtering: A Hardware-Based Motion Estimation System

Particle filters constitute themselves a highly powerful estimation tool, especially when dealing with non-linear non-Gaussian systems. However, traditional approaches present several limitations, which reduce significantly their performance. Evolutionary algorithms, and more specifically their optimization capabilities, may be used in order to overcome particle-filtering weaknesses. In this paper, a novel FPGA-based particle filter that takes advantage of evolutionary computation in order to estimate motion patterns is presented. The evolutionary algorithm, which has been included inside the resampling stage, mitigates the known sample impoverishment phenomenon, very common in particle-filtering systems. In addition, a hybrid mutation technique using two different mutation operators, each of them with a specific purpose, is proposed in order to enhance estimation results and make a more robust system. Moreover, implementing the proposed Evolutionary Particle Filter as a hardware accelerator has led to faster processing times than different software implementations of the same algorithm.

A dynamically adaptable bus architecture for trading-off among performance, consumption and dependability in Cyber-Physical Systems

Cyber-Physical Systems need to handle increasingly complex tasks, which additionally, may have variable operating conditions over time. Therefore, dynamic resource management to adapt the system to different needs is required. In this paper, a new bus-based architecture, called ARTICo3, which by means of Dynamic Partial Reconfiguration, allows the replication of hardware tasks to support module redundancy, multi-thread operation or dual-rail solutions for enhanced side-channel attack protection is presented. A configuration-aware data transaction unit permits data dispatching to more than one module in parallel, or provide coalesced data dispatching among different units to maximize the advantages of burst transactions. The selection of a given configuration is application independent but context-aware, which may be achieved by the combination of a multi-thread model similar to the CUDA kernel model specification, combined with a dynamic thread/task/kernel scheduler. A multi-kernel application for face recognition is used as an application example to show one scenario of the ARTICo3 architecture.

Design of OpenCL-compatible multithreaded hardware accelerators with dynamic support for embedded FPGAs

ARTICo3 is an architecture that permits to dynamically set an arbitrary number of reconfigurable hardware accelerators, each containing a given number of threads fixed at design time according to High Level Synthesis constraints. However, the replication of these modules can be decided at runtime to accelerate kernels by increasing the overall number of threads, add modular redundancy to increase fault tolerance, or any combination of the previous. An execution scheduler is used at kernel invocation to deliver the appropriate data transfers, optimizing memory transactions, and sequencing or parallelizing execution according to the configuration specified by the resource manager of the architecture. The model of computation is compatible with the OpenCL kernel execution model, and memory transfers and architecture are arranged to match the same optimization criteria as for kernel execution in GPU architectures but, differently to other approaches, with dynamic hardware execution support. In this paper, a novel design methodology for multithreaded hardware accelerators is presented. The proposed framework provides OpenCL compatibility by implementing a memory model based on shared memory between host and compute device, which removes the overhead imposed by data transferences at global memory level, and local memories inside each accelerator, i.e. compute unit, which are connected to global memory through optimized DMA links. These local memories provide unified access, i.e. a continuous memory map, from the host side, but are divided in a configurable number of independent banks (to increase available ports) from the processing elements side to fully exploit data-level parallelism. Experimental results show OpenCL model compliance using multithreaded hardware accelerators and enhanced dynamic adaptation capabilities.

FPGA-Based High-Performance Embedded Systems for Adaptive Edge Computing in Cyber-Physical Systems: The ARTICo 3 Framework.

Cyber-Physical Systems are experiencing a paradigm shift in which processing has been relocated to the distributed sensing layer and is no longer performed in a centralized manner. This approach, usually referred to as Edge Computing, demands the use of hardware platforms that are able to manage the steadily increasing requirements in computing performance, while keeping energy efficiency and the adaptability imposed by the interaction with the physical world. In this context, SRAM-based FPGAs and their inherent run-time reconfigurability, when coupled with smart power management strategies, are a suitable solution. However, they usually fail in user accessibility and ease of development. In this paper, an integrated framework to develop FPGA-based high-performance embedded systems for Edge Computing in Cyber-Physical Systems is presented. This framework provides a hardware-based processing architecture, an automated toolchain, and a runtime to transparently generate and manage reconfigurable systems from high-level system descriptions without additional user intervention. Moreover, it provides users with support for dynamically adapting the available computing resources to switch the working point of the architecture in a solution space defined by computing performance, energy consumption and fault tolerance. Results show that it is indeed possible to explore this solution space at run time and prove that the proposed framework is a competitive alternative to software-based edge computing platforms, being able to provide not only faster solutions, but also higher energy efficiency for computing-intensive algorithms with significant levels of data-level parallelism.

Analysis of a heterogeneous multi-core, multi-hw-accelerator-based system designed using PREESM and SDSoC

Nowadays, new heterogeneous system technologies are flooding the market: through the past years, it is possible to observe the move from single CPUs to multi-core devices featuring CPUs, GPUs and large FPGAs, such as Xilinx Zynq-7000 or Zynq UltraScale+ MPSoC architectures. In this context, providing developers with transparent deployment capabilities to efficiently execute different applications on such complex devices is important. In this paper, a design flow that combines, on one side, PREESM, a dataflow-based prototyping framework and, on the other side, Xilinx SDSoC, an HLS-based framework to automatically generate and manage hardware accelerators, is presented. This integration leverages the automatic, static task scheduling obtained from PREESM with asynchronous invocations that trigger the parallel execution of multiple hardware accelerators from some of their associated sequential software threads. An image processing application is used as a proof of concept, showing the interoperability possibilities of both tools, the level of design automation achieved and, for the resulting computing architecture, the good performance scalability according to the number of accelerators and sw threads.

Teaching hybrid HW/SW embedded system design using FPGA-based devices

Complex computing platforms involving pipelined processors, memory hierarchies, multi-core and many-core architectures are very common nowadays. These approaches require a deep understanding of the underlying hardware and the corresponding programing model to be able to decide which alternative is more suitable, i.e. obtain the best performance at the minimum cost, for a given application. Hence, it is important to cover all these aspects in academic curricula in order to provide engineers with competitive advantages and increase industrial productivity. In this paper, the methodology followed in a subject on Advanced Processing Architectures from a MSc program is presented. The theoretical content is complemented using hands-on exercises to further analyze the concepts discussed in class. As an example, the practical lessons on single-core Systems on Programmable Chip are reviewed in detail, showing the key ideas that are to be acquired by the students enrolled in the subject. Moreover, the proposed strategy has been evaluated using a voluntary and anonymous questionnaire to detect the strong and weak points of the proposed approach. This feedback is essential to provide students with valuable knowledge and meet quality criteria in academic education.

Live demonstration: A dynamically adaptable image processing application running in an FPGA-based WSN platform

This 1-Page Demonstration paper is included in the track “Multimedia Systems and Applications”. The work has been already published in [1] and [2]. The main idea of the demonstration is to show how the Virtual Architecture ARTICo3 works within a high performance wireless sensor node called HiReCookie. The selected demo includes an image processing application with several filters running as different kernels within the architecture ARTICo3. The virtual architecture works in a Spartan-6 FPGA included in the HiReCookie Node, [3] and [4]. During the demonstration, an image taken from a video camera attached to the node will be processed in real time by several dynamically reconfigurable kernels (median filters and edge detectors) under different working conditions. The solution scope includes solutions trading off among Low Power, Dependability and High Performance Computing.