An official website of the United States government
FPGAs are well-suited for accelerating deep learning (DL) applications owing to the rapidly changing algorithms, network architectures and computation requirements in this field. However, the generic building blocks available on traditional FPGAs limit the acceleration that can be achieved. Many modifications to FPGA architecture have been proposed and deployed including adding specialized artificial intelligence (AI) processing engines, adding support for smaller precision math like 8-bit fixed point and IEEE half-precision (fp16) in DSP slices, adding shadow multipliers in logic blocks, etc. In this paper, we describe replacing a portion of the FPGA’s programmable logic area with Tensor Slices. These slices have a systolic array of processing elements at their heart that support multiple tensor operations, multiple dynamically-selectable precisions and can be dynamically fractured into individual multipliers and MACs (multiply-and-accumulate). These slices have a local crossbar at the inputs that helps with easing the routing pressure caused by a large block on the FPGA. Adding these DL-specific coarse-grained hard blocks to FPGAs increases their compute density and makes them even better hardware accelerators for DL applications, while still keeping the vast majority of the real estate on the FPGA programmable at fine-grain.
more »
« less
CoMeFa: Deploying Compute-in-Memory on FPGAs for Deep Learning Acceleration
Block random access memories (BRAMs) are the storage houses of FPGAs, providing extensive on-chip memory bandwidth to the compute units implemented using logic blocks and digital signal processing slices. We propose modifying BRAMs to convert them to CoMeFa (Compute-in-Memory Blocks forFPGAs) random access memories (RAMs). These RAMs provide highly parallel compute-in-memory by combining computation and storage capabilities in one block. CoMeFa RAMs utilize the true dual-port nature of FPGA BRAMs and contain multiple configurable single-bit bit-serial processing elements. CoMeFa RAMs can be used to compute with any precision, which is extremely important for applications like deep learning (DL). Adding CoMeFa RAMs to FPGAs significantly increases their compute density while also reducing data movement. We explore and propose two architectures of these RAMs: CoMeFa-D (optimized for delay) and CoMeFa-A (optimized for area). Compared to existing proposals, CoMeFa RAMs do not require changing the underlying static RAM technology like simultaneously activating multiple wordlines on the same port, and are practical to implement. CoMeFa RAMs are especially suitable for parallel and compute-intensive applications like DL, but these versatile blocks find applications in diverse applications like signal processing and databases, among others. By augmenting an Intel Arria 10–like FPGA with CoMeFa-D (CoMeFa-A) RAMs at the cost of 3.8% (1.2%) area, and with algorithmic improvements and efficient mapping, we observe a geomean speedup of 2.55× (1.85×) across microbenchmarks from various applications and a geomean speedup of up to 2.5× across multiple deep neural networks. Replacing all or some BRAMs with CoMeFa RAMs in FPGAs can make them better accelerators of DL workloads.
more »
« less
- Award ID(s):
- 1763848
- PAR ID:
- 10488312
- Publisher / Repository:
- ACM
- Date Published:
- Journal Name:
- ACM Transactions on Reconfigurable Technology and Systems
- Volume:
- 16
- Issue:
- 3
- ISSN:
- 1936-7406
- Page Range / eLocation ID:
- 50:1-50:34
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Deep-Learning has become a dominant computing paradigm across a broad range of application domains. Different architectures of Deep-Networks like CNN, MLP, and RNN have emerged as the prominent machine-learning approaches for today’s application domains. These architectures are heavily data-dependent, requiring frequent access to memory. As a result, these applications suffer the most from the memory bottleneck of the von Neumann architectures. There is an imminent need for memory-centric architectures for deep-learning and big-data analytic applications that are memory intensive. Modern Field Programmable Gate Arrays (FPGAs) are ideal programmable substrates for creating customized Processor in/near Memory (PIM) accelerators. Modern FPGAs contain 100s of Mbits of dual-ported SRAM in the form of disaggregated, configurable Block RAMs (BRAMs). These BRAMs contain TB/s of available internal bandwidth. Unfortunately, developing FPGA-based accelerators for deep learning is not a simple task and demands the utilization of specialized tools provided by the FPGA vendors. It requires expertise in low-level hardware microarchitecture design. These are often not available to most researchers in the field of deep learning. Even with the ongoing improvements in High-Level Synthesis (HLS) tools, the requirement for hardware-specific design knowledge cannot be completely eliminated. This research developed a new reconfigurable memory-centric architecture and design approach that opens the advantages of FPGAs and Processor-in-Memory architecture to memory-intensive applications. Due to its high-performance and scalable memory-centric design, this architecture can deliver the highest speed and the lowest latency achievable from an FPGA overcoming the memory bottleneck.more » « less
-
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
-
Masking has become one of the most effective approaches for securing hardware designs against side-channel attacks. Regardless of the effort put into correctly implementing masking schemes on a field-programmable gate array (FPGA), leakage can be unexpectedly observed. This is due to the fact that the assumption underlying all masked designs, i.e., the leakages of different shares are independent of each other, may no longer hold in practice. In this regard, extreme temperatures have been shown to be an important factor in inducing leakage, even in correctlymasked designs. This has previously been verified using an external heat generator (i.e., a climate chamber). In this paper, we examine whether the leakage can be induced using the circuit components themselves without making any changes to the design. Specifically, we target masked neural networks (NNs) in FPGAs, one of the main building blocks of which is block random access memory (BRAM). In this respect, thanks to the inherent characteristics of NNs, our novel internal heat generators leverage solely the memories devoted to storing the user’s input, especially when frequently writing alternating patterns into BRAMs. The possibility of observing first-order leakage is evaluated by considering one of the most recent and successful first-order secure masked NNs, namely ModuloNET. ModuloNET is specifically designed for FPGAs, where BRAMs are used to store inputs and intermediate computations. Our experimental results demonstrate that undesirable first-order leakage can be observed and exploited by increasing the temperature when an alternating input is applied to the masked NN. To give a better understanding of the impact of extreme heat, we further perform a similar test on the design using an external heat generator, where a similar conclusion can be drawn.more » « less
-
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
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1145/3603504
