The continued growth in the processing power of FPGAs coupled with high bandwidth memories (HBM), makes systems like the Xilinx U280 credible platforms for linear solvers which often dominate the run time of scientific and engineering applications. In this paper, we present Callipepla, an accelerator for a preconditioned conjugate gradient linear solver (CG). FPGA acceleration of CG faces three challenges: (1) how to support an arbitrary problem and terminate acceleration processing on the fly, (2) how to coordinate long-vector data flow among processing modules, and (3) how to save off-chip memory bandwidth and maintain double (FP64) precision accuracy. To tackle the three challenges, we present (1) a stream-centric instruction set for efficient streaming processing and control, (2) vector streaming reuse (VSR) and decentralized vector flow scheduling to coordinate vector data flow among modules and further reduce off-chip memory access latency with a double
memory channel design, and (3) a mixed precision scheme to save bandwidth yet still achieve effective double precision quality solutions. To the best of our knowledge, this is the first work to introduce the concept of VSR for data reusing between on-chip modules to reduce unnecessary off-chip accesses and enable modules working in parallel for FPGA accelerators. We prototype the
accelerator on a Xilinx U280 HBM FPGA. Our evaluation shows that compared to the Xilinx HPC product, the XcgSolver, Callipepla achieves a speedup of 3.94×, 3.36× higher throughput, and 2.94× better energy efficiency. Compared to an NVIDIA A100 GPU which has 4× the memory bandwidth of Callipepla, we still achieve 77% of its throughput with 3.34× higher energy efficiency. The code is available at https://github.com/UCLA-VAST/Callipepla.
more »
« less
Observed Memory Bandwidth and Power Usage on FPGA Platforms with OneAPI and Vitis HLS: A Comparison with GPUs
The two largest barriers to adoption of FPGA platforms for HPC applications are the difficulty of programming FPGAs and the performance gap when compared to GPUs. To address the first barrier, new ecosystems like Intel oneAPI, and Xilinx Vitis HLS aim to improve programmability for FPGA platforms. From a performance aspect, FPGAs trade off lower compute frequencies for more customized hardware acceleration and power efficiency when compared to GPUs. The performance for memory-bound applications on recent GPU platforms like NVIDIA’s H100 and AMD’s MI210 has also improved due to the inclusion of high-bandwidth memories (HBM), and newer FPGA platforms are also starting to include HBM in addition to traditional DRAM.
To understand the current state-of-the-art and performance differences between FPGAs and GPUs, we consider realized memory bandwidth for recent FPGA and GPU platforms. We utilize a custom STREAM benchmark to evaluate two Intel FPGA platforms, the Stratix 10 SX PAC and Bittware 520N-MX, two AMD/Xilinx FPGA platforms, the Alveo U250 and Alveo U280, as well as GPU platforms from NVIDIA and AMD. We also extract power measurements and estimate memory bandwidth per Watt ((GB/s)/W) on these platforms to evaluate how FPGAs compare against GPU execution. While the GPUs far exceed the FPGAs in raw performance, the HBM equipped FPGAs demonstrate a competitive performance-power balance for larger data sizes that can be easily implemented with oneAPI and Vitis HLS kernels. These findings suggest a potential sweet spot for this emerging FPGA ecosystem to serve bandwidth limited applications in an energy-efficient fashion.
more »
« less
- Award ID(s):
- 2016701
- NSF-PAR ID:
- 10518455
- Publisher / Repository:
- Springer
- Date Published:
- Journal Name:
- Springer Lecture Notes in Computer Science
- Volume:
- 13999
- ISBN:
- 978-3-031-40843-4
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Emerging FPGA systems are providing higher external memory bandwidth to compete with GPU performance. However, because FPGAs often achieve parallelism through deep pipelines, traditional FPGA design strategies do not necessarily scale well to large amounts of replicated pipelines that can take advantage of higher bandwidth. We show that sliding-window applications, an important subset of digital signal processing, demonstrate this scalability problem. We introduce a window generator architecture that enables replication to over 330 GB/s, which is an 8.7x improvement over previous work. We evaluate the window generator on the Intel Broadwell+Arria10 system for 2D convolution and show that for traditional convolution (one filter per image), our approach outperforms a 12-core Xeon Broadwell E5 by 81x and a high-end Nvidia P6000 GPU by an order of magnitude for most input sizes, while improving energy by 15.7x. For convolutional neural nets (CNNs), we show that although the GPU and Xeon typically outperform existing FPGA systems, projected performances of the window generator running on FPGAs with sufficient bandwidth can outperform high-end GPUs for many common CNN parameters.more » « less
-
Deep Neural Networks (DNNs) have been successfully applied in many fields. Considering performance, flexibility, and energy efficiency, Field Programmable Gate Array (FPGA) based accelerator for DNNs is a promising solution. The existing frameworks however lack the possibility of reusability and friendliness to design a new network with minimum efforts. Modern high-level synthesis (HLS) tools greatly reduce the turnaround time of designing and implementing complex FPGA-based accelerators. This paper presents a framework for hardware accelerator for DNNs using high level specification. A novel architecture is introduced that maximizes data reuse and external memory bandwidth. This framework allows to generate a scalable HLS code for a given pre-trained model that can be mapped to different FPGA platforms. Various HLS compiler optimizations have been applied to the code to produce efficient implementation and high resource utilization. The framework achieves a peak performance of 23 frames per second for SqueezeNet on Xilinx Alveo u250 board.more » « less
-
null (Ed.)While FPGAs have been traditionally considered hard to program, recently there have been efforts aimed to allow the use of high-level programming models and libraries intended for multi-core CPUs and GPUs to program FPGAs. For example, both Intel and Xilinx are now providing toolchains to deploy OpenCL code onto FPGA. However, because the nature of the parallelism offered by GPU and FPGA devices is fundamentally different, OpenCL code optimized for GPU can prove very inefficient on FPGA, in terms of both performance and hardware resource utilization. This paper explores this problem on finite automata traversal. In particular, we consider an OpenCL NFA traversal kernel optimized for GPU but exhibiting FPGA-friendly characteristics, namely: limited memory requirements, lack of synchronization, and SIMD execution. We explore a set of structural code changes, custom and best-practice optimizations to retarget this code to FPGA. We showcase the effect of these optimizations on an Intel Stratix V FPGA board using various NFA topologies from different application domains. Our evaluation shows that, while the resource requirements of the original code exceed the capacity of the FPGA in use, our optimizations lead to significant resource savings and allow the transformed code to fit the FPGA for all considered NFA topologies. In addition, our optimizations lead to speedups up to 4x over an already optimized code-variant aimed to fit the NFA traversal kernel on FPGA. Some of the proposed optimizations can be generalized for other applications and introduced in OpenCL-to-FPGA compiler.more » « less
-
Random Forests (RFs) are a commonly used machine learning method for classification and regression tasks spanning a variety of application domains, including bioinformatics, business analytics, and software optimization. While prior work has focused primarily on improving performance of the training of RFs, many applications, such as malware identification, cancer prediction, and banking fraud detection, require fast RF classification. In this work, we accelerate RF classification on GPU and FPGA. In order to provide efficient support for large datasets, we propose a hierarchical memory layout suitable to the GPU/FPGA memory hierarchy. We design three RF classification code variants based on that layout, and we investigate GPU- and FPGA-specific considerations for these kernels. Our experimental evaluation, performed on an Nvidia Xp GPU and on a Xilinx Alveo U250 FPGA accelerator card using publicly available datasets on the scale of millions of samples and tens of features, covers various aspects. First, we evaluate the performance benefits of our hierarchical data structure over the standard compressed sparse row (CSR) format. Second, we compare our GPU implementation with cuML, a machine learning library targeting Nvidia GPUs. Third, we explore the performance/accuracy tradeoff resulting from the use of different tree depths in the RF. Finally, we perform a comparative performance analysis of our GPU and FPGA implementations. Our evaluation shows that, while reporting the best performance on GPU, our code variants outperform the CSR baseline both on GPU and FPGA. For high accuracy targets, our GPU implementation yields a 5-9 × speedup over CSR, and up to a 2 × speedup over Nvidia’s cuML library.more » « less
