HARP Cluster

FPGA devices have been proving to be good candidates to accelerate applications from different research topics. For instance, machine learning applications such as K-Means clustering usually relies on large amount of data to be processed, and, despite the performance offered by other architectures, FPGAs can offer better energy efficiency. With that in mind, Intel has launched a platform that integrates a multicore and an FPGA in the same package, enabling low latency and coherent fine-grained data offload. In this paper, we present a parallel implementation of the K-Means clustering algorithm, for this novel platform, using OpenCL language, and compared it against other platforms. We found that the CPU+FPGA platform was more energy efficient than the CPU-only approach from 70.71% to 85.92%, with Standard and Tiny input sizes respectively, and up to 68.21% of performance improvement was obtained with Tiny input size. Furthermore, it was up to 7.2×more energy efficient than an Intel® Xeon Phi ™, 21.5×than a cluster of Raspberry Pi boards, and 3.8×than the low-power MPPA-256 architecture, when the Standard input size was used.

The performance of High-Level Synthesis (HLS) applications with irregular data structures is limited by its imperative programming paradigm like C/C++. In this paper, we show that constructing concurrent data structures with channels, a programming construct derived from CSP (communicating sequential processes) paradigm, is an effective approach to improve the performance of these applications. We evaluate concurrent data structure for FPGA by synthesizing a K-means clustering algorithm on the Intel HARP2 platform. A fully pipelined KMC processing element can be synthesized from OpenCL with the help of a SPSC (single-producer-single-consumer) queue and stack built from channels, achieving 15.2x speedup over a sequential baseline. The number of processing element can be scaled up by leveraging a MPMC (multiple-producer-multiple-consumer) stack with work distribution for dynamic load balance. Evaluation shows that an additional 3.5x speedup can be achieved when 4 processing element is instantiated. These results show that the concurrent data structure built with channels has great potential for improving the parallelism of HLS applications. We hope that our study will stimulate further research into the potential of channel-based HLS.

In recent years, FPGAs have been successfully employed for the implementation of efficient, application-specific accelerators for a wide range of machine learning tasks. In this work, we consider probabilistic models, namely, (Mixed) Sum-Product Networks (SPN), a deep architecture that can provide tractable inference for multivariate distributions over mixed data-sources. We develop a fully pipelined FPGA accelerator architecture, including a pipelined interface to external memory, for the inference in (mixed) SPNs. To meet the precision constraints of SPNs, all computations are conducted using double-precision floating point arithmetic. Starting from an input description, the custom FPGA-accelerator is synthesized fully automatically by our tool flow. To the best of our knowledge, this work is the first approach to offload the SPN inference problem to FPGA-based accelerators. Our evaluation shows that the SPN inference problem benefits from offloading to our pipelined FPGA accelerator architecture.

Transactional Memory (TM) has been considered as a promising alternative to existing synchronization operations, which are often the largest stumbling block to unleashing parallelism of applications. Efficient implementations of TM, however, are challenging due to the tension between lowering performance overhead and avoiding unnecessary aborts. In this paper, we present Reachability-based Optimistic Concurrency Control for Transactional Memory (ROCoCoTM), a novel scheme which offloads concurrency control (CC) algorithms, the central building blocks of TM systems, to reconfigurable hardware. To reduce the abort rate, an innovative formalization of mainstream CC algorithms is developed to reveal a common restriction that leads to unnecessary aborts. This restriction is resolved by the ROCoCo algorithm with a centralized validation phase, which can be efficiently pipelined in hardware. Thanks to a high-performance offloading engine implemented in reconfigurable hardware, ROCoCo algorithm results in decreased abort rates and reduced performance overhead. The whole system is implemented on Intel's HARP2 platform and evaluated with the STAMP benchmark suite. Experiments show 1.55x and 8.05x geomean speedup over TinySTM and an HTM based on Intel TSX, respectively. Given the fast-growing deployment of commodity CPU-FPGA platforms, ROCoCoTM paves the way for software programmers to exploit heterogeneous computing resources with a high-level transactional abstraction to effectively extract the parallelism in modern applications.

In recent years, OpenCL has been increasingly adopted as it enables software programmers to harness the performance and power efficiency of FPGAs. Despite simplifying the FPGA programming challenge, achieving high performance and energy efficiency with OpenCL is still a difficult task. In order to further contribute to the advance of the OpenCL usage for FPGAs, we utilize a realistic application scenario as our case study: the AutoDock molecular docking software. While OpenCL has proven its effectiveness in accelerating molecular docking on GPUs, for FPGA-based AutoDock accelerators it struggles with difficult design patterns. Besides complex multiple-producers to single-consumer datapaths, these include time-intensive loops with variable runtimes. Therefore, this work presents the design and optimization steps for implementing AutoDock in OpenCL targeting an Arria-10 FPGA, as well as a corresponding execution runtime and energy-efficiency evaluation. Applying these techniques improved the performance of the initial OpenCL implementation for FPGAs by three orders of magnitude, with the final version of the code now yielding speed-ups of up to ~2.7x, and energy-efficiency gains of up to ~1.8x over the original serial AutoDock version executing on a current-generation CPU.

The sheer amount of computing resources required to run modern cloud workloads has put a lot of pressure on the design of power efficient cluster nodes. To address this problem, Intel (HARP) and Microsoft (Catapult) have proposed CPU-FPGA integrated architectures that can deliver efficient power-performance executions. Unfortunately, the integration of FPGA acceleration modules to software is a challenging endeavor that does not have a seamless programming model. This paper proposes HardCloud (www.hardcloud.org), an extension of the OpenMP 4.X standard that eases the task of offloading FPGA modules to cluster accelerators.