João Canas Ferreira

As applications move to the edge, efficiency in computing power and power/energy consumption is required. Heterogeneous computing promises to meet these requirements through application-specific hardware accelerators. Runtime adaptivity might be of paramount importance to realize the potential of hardware specialization, but further study is required on workload retargeting and offloading to reconfigurable hardware. This article presents our framework for the exploration of both offloading and hardware generation techniques. The framework is currently able to process instruction sequences from MicroBlaze, ARMv8, and riscv32imaf binaries, and to represent them as Control and Dataflow Graphs for transformation to implementations of hardware modules. We illustrate the framework’s capabilities for identifying binary sequences for hardware translation with a set of 13 benchmarks.

Heterogeneous platforms with FPGAs have started to be employed in the High-Performance Computing (HPC) field to improve performance and overall efficiency. These platforms allow the use of specialized hardware to accelerate software applications, but require the software to be adapted in what can be a prolonged and complex process. The main goal of this work is to describe and evaluate mechanisms that can transparently transfer the control flow between CPU and FPGA within the scope of HPC. Combining such a mechanism with transparent software profiling and accelerator configuration could lead to an automatic way of accelerating regular applications. In this work, a mechanism based on the ptrace system call is proposed, and its performance on the Intel Xeon+FPGA platform is evaluated. The feasibility of the proposed approach is demonstrated by a working prototype that performs the transparent control flow transfer of any function call to a matching hardware accelerator. This approach is more general than shared library interposition at the cost of a small time overhead in each accelerator use (about 1.3 ms in the prototype implementation).

Typical analogue-to-digital conversion (ADC) architectures, at Nyquist rate, tend to occupy a big portion of the integrated circuit die area and to consume more power than desired. Recently, with the rise of Interet-of-Things (IoT), there is a high demand for architectures that can have both reduced area and power consumption. Time encoding machines (TEM) might be a promising alternative. These types of encoders result in very simple and low-power analogue circuits, shifting most of its complexity to the decoding stage, typically stationed in a place with access to more resources. This paper focuses on a particular TEM, the integrate-and-fire neuron (IFN). The IFN modulation is based on a simplified first-order model of neural operation and it encodes the signal in a very power efficient manner. In the end, a novel hardware architecture for the reconstruction of the IFN encoded signal based on a spiking model will be presented. The method is demonstrated and implemented on FPGA, reaching an ENOB as high as 8.23.

The fifth-generation (5G) revolution represents more than a mere performance enhancement of previous generations: it will deeply transform the way humans and/or machines interact, enabling a heterogeneous expansion in the number of use cases and services. Crucial to the realization of this revolution is the design of hardware components characterized by high degrees of flexibility, versatility and resource/power efficiency. This chapter proposes a field-programmable gate array (FPGA)-oriented baseband processing architecture suitable for fast-changing communication environments such as 4G/5G waveform coexistence, noncontiguous carrier aggregation (CA) or centralized cloud radio access network (C-RAN) processing. The proposed architecture supports three 5G waveform candidates and is shown to be upgradable, resource-efficient and cost-effective. Through hardware virtualization, enabled by dynamic partial reconfiguration (DPR), the design space exploration of our architecture exceeds the hardware resources available on the Zynq xc7z020 device. Moreover, dynamic frequency scaling (DFS) enables the runtime adjustment of processing throughput and power reductions by up to 88%. The combined resource overhead for DPR and DFS is very low, and the reconfiguration latency stays two orders of magnitude below the control plane latency requirements proposed for 5G communications.

High Level Synthesis (HLS) tools targeting Field Programmable Gate Arrays (FPGAs) aim to provide a method for programming these devices via high-level abstractions. Initially, HLS support for FPGAs focused on compiling C/C++ to hardware circuits. This raised the issue of determining the programming practices which resulted in the best performing circuits. Recently, to further increase the applicability of HLS approaches, renewed effort was placed on support for HLS of OpenCL code for FPGA, raising the same issues of coding practices and performance portability. This paper explores the performance of OpenCL code compiled for FPGAs for different coding techniques. We evaluate the use of task-kernels versus NDRange kernels, data vectorization, the use of on-chip local memories, and data transfer optimizations by exploiting burst access inference. We present this exploration via a case study of the k-means algorithm, and produce a total of 10 OpenCL implementations of the kernel. To determine the effects of different data set characteristics, and to determine the gains from specialization based on number of attributes, we generated a total of 12 integer data sets. The data sets vary regarding the number of instances, number of attributes (i.e., features), and number of clusters. We also vary the number of processing cores, and present the resulting required resources and operating frequencies. Finally, we execute the same OpenCL code on a 4 GHz Intel i7-6700K CPU, showing that the FPGA achieves speedups up to $1.54 {\times } $ for four cases, and energy savings up to 80% in all cases.

Performance and power efficiency in edge and embedded systems can benefit from specialized hardware. To avoid the effort of manual hardware design, we explore the generation of accelerator circuits from binary instruction traces for several Instruction Set Architectures.

A multifunctional router IC to be included in the nodes of a wearable body sensor network is described and evaluated. The router targets different application scenarios, especially those including tens of sensors, embedded into textile materials and with high data-rate communication demands. The router IC supports two different functionality sets, one for sensor nodes and another for the base node, both based on the same circuit module. The nodes are connected to each other by means of woven thick conductive yarns forming a mesh topology with the base node at the center. From the standpoint of the network, each sensor node is a four port router capable of handling packets from destination nodes to the base node, with sufficient redundant paths. The adopted hybrid circuit and packet switching scheme significantly improve network performance in terms of end-to-end delay, throughput and power consumption. The IC also implements a highly precise, sub-microsecond one-way time synchronization protocol which is used for time stamping the acquired data. The communication module was implemented in a 4-metal, $0.35~\mu \text{m}$ CMOS technology. The maximum data rate of the system is 35 Mbps while supporting up to 250 sensors, which exceeds current BAN applications scenarios.

The K-means algorithm is widely used to find correlations between data in different application domains. However, given the massive amount of data stored, known as Big Data, the need for high-speed processing to analyze data has become even more critical, especially for real-time applications. A solution that has been adopted to increase the processing speed is the use of parallel implementations on FPGA, which has proved to be more efficient than sequential systems. Hence, this paper proposes a fully parallel implementation of the K-means algorithm on FPGA to optimize the system’s processing time, thus enabling real-time applications. This proposal, unlike most implementations proposed in the literature, even parallel ones, do not have sequential steps, a limiting factor of processing speed. Results related to processing time (or throughput) and FPGA area occupancy (or hardware resources) were analyzed for different parameters, reaching performances higher than 53 millions of data points processed per second. Comparisons to the state of the art are also presented, showing speedups of more than $15573\times $ over a partially serial implementation.

The breakdown of Dennard scaling has resulted in a decade-long stall of the maximum operating clock frequencies of processors. To mitigate this issue, computing shifted to multi-core devices. This introduced the need for programming flows and tools that facilitate the expression of workload parallelism at high abstraction levels. However, not all workloads are easily parallelizable, and the minor improvements to processor cores have not significantly increased single-threaded performance. Simultaneously, Instruction Level Parallelism in applications is considerably underexplored. This article reviews notable approaches that focus on exploiting this potential parallelism via automatic generation of specialized hardware from binary code. Although research on this topic spans over more than 20 years, automatic acceleration of software via translation to hardware has gained new importance with the recent trend toward reconfigurable heterogeneous platforms. We characterize this kind of binary acceleration approach and the accelerator architectures on which it relies. We summarize notable state-of-the-art approaches individually and present a taxonomy and comparison. Performance gains from 2.6× to 5.6× are reported, mostly considering bare-metal embedded applications, along with power consumption reductions between 1.3× and 3.9×. We believe the methodologies and results achievable by automatic hardware generation approaches are promising in the context of emergent reconfigurable devices.