In this paper, we propose MRIMA, as a novel MRAM-based In-Memory Accelerator for non-volatile, flexible, and efficient in-memory computing. MRIMA transforms current Spin Transfer Torque Magnetic Random Access Memory (STT-MRAM) arrays to massively parallel computational units capable of working as both non-volatile memory and in-memory logic. Instead of integrating complex logic units in cost-sensitive memory, MRIMA exploits hardware-friendly bit-line computing methods to implement complete Boolean logic functions between operands within a memory array in a single clock cycle, overcoming the multi-cycle logic issue in contemporary Processing-In-Memory (PIM) platforms. We present practical case studies to demonstrate MRIMA’s acceleration for binary-weight and low bit-width Convolutional Neural Networks (CNN) as well as data encryption. Our device-to-architecture co-simulation results on CNN acceleration demonstrate that MRIMA can obtain 1.7× better energy-efficiency and 11.2× speed-up compared to ASICs, and, 1.8× better energy-efficiency and 2.4× speed-up over the best DRAM-based PIM solutions. As an AES in-memory encryption engine, MRIMA shows 77% and 21% lower energy consumption compared to CMOS-ASIC and recent domain wall-based design, respectively.
SMART: A Heterogeneous Scratchpad Memory Architecture for Superconductor SFQ-based Systolic CNN Accelerators
Ultra-fast & low-power superconductor single-flux-quantum (SFQ)-based CNN systolic accelerators are built to enhance the CNN inference throughput. However, shift-register (SHIFT)-based scratchpad memory (SPM) arrays prevent a SFQ CNN accelerator from exceeding 40% of its peak throughput, due to the lack of random access capability. This paper first documents our study of a variety of cryogenic memory technologies, including Vortex Transition Memory (VTM), Josephson-CMOS SRAM, MRAM, and Superconducting Nanowire Memory, during which we found that none of the aforementioned technologies made a SFQ CNN accelerator achieve high throughput, small area, and low power simultaneously. Second, we present a heterogeneous SPM architecture, SMART, composed of SHIFT arrays and a random access array to improve the inference throughput of a SFQ CNN systolic accelerator. Third, we propose a fast, low-power and dense pipelined random access CMOS-SFQ array by building SFQ passive-transmission-line-based H-Trees that connect CMOS sub-banks. Finally, we create an ILP-based compiler to deploy CNN models on SMART. Experimental results show that, with the same chip area overhead, compared to the latest SHIFT-based SFQ CNN accelerator, SMART improves the inference throughput by 3.9 × (2.2 ×), and reduces the inference energy by 86% (71%) when inferring a single image (a batch of images).
- Award ID(s):
- Publication Date:
- NSF-PAR ID:
- Journal Name:
- 54th Annual IEEE/ACM International Symposium on Microarchitecture
- Page Range or eLocation-ID:
- 912 to 924
- Sponsoring Org:
- National Science Foundation
More Like this
Large Convolutional Neural Networks (CNNs) are often pruned and compressed to reduce the amount of parameters and memory requirement. However, the resulting irregularity in the sparse data makes it difficult for FPGA accelerators that contains systolic arrays of Multiply-and-Accumulate (MAC) units, such as Intel’s FPGA-based Deep Learning Accelerator (DLA), to achieve their maximum potential. Moreover, FPGAs with low-bandwidth off-chip memory could not satisfy the memory bandwidth requirement for sparse matrix computation. In this paper, we present 1) a sparse matrix packing technique that condenses sparse inputs and filters before feeding them into the systolic array of MAC units in the Intel DLA, and 2) a customization of the Intel DLA which allows the FPGA to efficiently utilize a high bandwidth memory (HBM2) integrated in the same package. For end-to-end inference with randomly pruned ResNet-50/MobileNet CNN models, our experiments demonstrate 2.7x/3x performance improvement compared to an FPGA with DDR4, 2.2x/2.1x speedup against a server-class Intel SkyLake CPU, and comparable performance with 1.7x/2x power efficiency gain as compared to an NVidia V100 GPU.
With the success of deep neural networks (DNN), many recent works have been focusing on developing hardware accelerator for power and resource-limited embedded system via model compression techniques, such as quantization, pruning, low-rank approximation, etc. However, almost all existing DNN structure is fixed after deployment, which lacks runtime adaptive DNN structure to adapt to its dynamic hardware resource, power budget, throughput requirement, as well as dynamic workload. Correspondingly, there is no runtime adaptive hardware platform to support dynamic DNN structure. To address this problem, we first propose a dynamic channel-adaptive deep neural network (CA-DNN) which can adjust the involved convolution channel (i.e. model size, computing load) at run-time (i.e. at inference stage without retraining) to dynamically trade off between power, speed, computing load and accuracy. Further, we utilize knowledge distillation method to optimize the model and quantize the model to 8-bits and 16-bits, respectively, for hardware friendly mapping. We test the proposed model on CIFAR-10 and ImageNet dataset by using ResNet. Comparing with the same model size of individual model, our CA-DNN achieves better accuracy. Moreover, as far as we know, we are the first to propose a Processing-in-Memory accelerator for such adaptive neural networks structure based on Spin Orbitmore »
In this paper, we pave a novel way towards the concept of bit-wise In-Memory Convolution Engine (IMCE) that could implement the dominant convolution computation of Deep Convolutional Neural Networks (CNN) within memory. IMCE employs parallel computational memory sub-array as a fundamental unit based on our proposed Spin Orbit Torque Magnetic Random Access Memory (SOT-MRAM) design. Then, we propose an accelerator system architecture based on IMCE to efficiently process low bit-width CNNs. This architecture can be leveraged to greatly reduce energy consumption dealing with convolutional layers and also accelerate CNN inference. The device to architecture co-simulation results show that the proposed system architecture can process low bit-width AlexNet on ImageNet data-set favorably with 785.25μJ/img, which consumes ~3× less energy than that of recent RRAM based counterpart. Besides, the chip area is ~4× smaller.
The control of cryogenic qubits in today’s super-conducting quantum computer prototypes presents significant scalability challenges due to the massive costs of generating/routing the analog control signals that need to be sent from a classical controller at room temperature to the quantum chip inside the dilution refrigerator. Thus, researchers in industry and academia have focused on designing in-fridge classical controllers in order to mitigate these challenges. Due to the maturity of CMOS logic, many industrial efforts (Microsoft, Intel) have focused on Cryo-CMOS as a near-term solution to design in-fridge classical controllers. Meanwhile, Supercon-ducting Single Flux Quantum (SFQ) is an alternative, less mature classical logic family proposed for large-scale in-fridge controllers. SFQ logic has the potential to maximize scalability thanks to its ultra-high speed and very low power consumption. However, architecture design for SFQ logic poses challenges due to its unconventional pulse-driven nature and lack of dense memory and logic. Thus, research at the architecture level is essential to guide architects to design SFQ-based classical controllers for large-scale quantum machines.In this paper, we present DigiQ, the first system-level design of a Noisy Intermediate Scale Quantum (NISQ)-friendly SFQ-based classical controller. We perform a design space exploration of SFQ-based controllers and co-design the quantummore »