SCALE-Sim: Systolic CNN Accelerator

Systolic Arrays are one of the most popular compute substrates within Deep Learning accelerators today, as they provide extremely high efficiency for running dense matrix multiplications. However, the research community lacks tools to insights on both the design trade-offs and efficient mapping strategies for systolic-array based accelerators. We introduce Systolic CNN Accelerator Simulator (SCALE-Sim), which is a configurable systolic array based cycle accurate DNN accelerator simulator. SCALE-Sim exposes various micro-architectural features as well as system integration parameters to the designer to enable comprehensive design space exploration. This is the first systolic-array simulator tuned for running DNNs to the best of our knowledge. Using SCALE-Sim, we conduct a suite of case studies and demonstrate the effect of bandwidth, data flow and aspect ratio on the overall runtime and energy of Deep Learning kernels across vision, speech, text, and games. We believe that these insights will be highly beneficial to architects and ML practitioners.

• Publication year

  2018

• Venue

  arXiv.org

• Publication date

  2018-10-16

• Fields of study

  Computer Science, Engineering

• Identifiers
  arXiv 1811.02883
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• MAERI: Enabling Flexible Dataflow Mapping over DNN Accelerators via Reconfigurable Interconnects

  2018cited by this paper
• Euphrates: Algorithm-SoC Co-Design for Low-Power Mobile Continuous Vision

  2018cited by this paper
• In-datacenter performance analysis of a tensor processing unit

  2017influential reference
• SCNN: An accelerator for compressed-sparse convolutional neural networks

  2017influential reference
• Neural Collaborative Filtering

  2017cited by this paper
• Attention is All you Need

  2017cited by this paper
• TETRIS: Scalable and Efficient Neural Network Acceleration with 3D Memory

  2017cited by this paper
• Mastering the game of Go without human knowledge

  2017cited by this paper
• FlexFlow: A Flexible Dataflow Accelerator Architecture for Convolutional Neural Networks

  2017cited by this paper
• Eyeriss: An Energy-Efficient Reconfigurable Accelerator for Deep Convolutional Neural Networks

  2016influential reference
• Co-designing accelerators and SoC interfaces using gem5-Aladdin

  2016cited by this paper
• EIE: Efficient Inference Engine on Compressed Deep Neural Network

  2016cited by this paper
• Designing Energy-Efficient Convolutional Neural Networks Using Energy-Aware Pruning

  2016cited by this paper
• Stripes: Bit-serial deep neural network computing

  2016cited by this paper
• Deep-learning in Mobile Robotics - from Perception to Control Systems: A Survey on Why and Why not

  2016cited by this paper
• Mobile CPU's rise to power: Quantifying the impact of generational mobile CPU design trends on performance, energy, and user satisfaction

  2016cited by this paper
• A Survey of Deep Network Solutions for Learning Control in Robotics: From Reinforcement to Imitation

  2016cited by this paper
• Minerva: Enabling Low-Power, Highly-Accurate Deep Neural Network Accelerators

  2016cited by this paper
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

  2015cited by this paper
• You Only Look Once: Unified, Real-Time Object Detection

  2015cited by this paper
• ShiDianNao: Shifting vision processing closer to the sensor

  2015cited by this paper
• Deep Speech 2 : End-to-End Speech Recognition in English and Mandarin

  2015cited by this paper
• Deep Learning with Limited Numerical Precision

  2015cited by this paper
• Deep Residual Learning for Image Recognition

  2015cited by this paper
• PuDianNao: A Polyvalent Machine Learning Accelerator

  2015cited by this paper
• Deep Speech: Scaling up end-to-end speech recognition

  2014cited by this paper
• DianNao: a small-footprint high-throughput accelerator for ubiquitous machine-learning

  2014cited by this paper
• Effective Use of Word Order for Text Categorization with Convolutional Neural Networks

  2014cited by this paper
• GemDroid: a framework to evaluate mobile platforms

  2014cited by this paper
• DaDianNao: A Machine-Learning Supercomputer

  2014cited by this paper
• ImageNet Large Scale Visual Recognition Challenge

  2014cited by this paper
• Small-footprint keyword spotting using deep neural networks

  2014cited by this paper
• DRAMSim2: A Cycle Accurate Memory System Simulator

  2011cited by this paper

• SpikeX: Exploring Accelerator Architecture and Network-Hardware Co-Optimization for Sparse Spiking Neural Networks

  2025cites this paper
• UniZK: Accelerating Zero-Knowledge Proof with Unified Hardware and Flexible Kernel Mapping

  2025cites this paper
• Ensuring the Accuracy of CNN Accelerators Supplied at Ultra-Low Voltage

  2024cites this paper
• Expanding hardware accelerator system design space exploration with gem5-SALAMv2

  2024cites this paper
• Embedded Distributed Inference of Deep Neural Networks: A Systematic Review

  2024cites this paper
• TransAxx: Efficient Transformers With Approximate Computing

  2024cites this paper
• SGCN: Exploiting Compressed-Sparse Features in Deep Graph Convolutional Network Accelerators

  2023cites this paper
• On Fault-Tolerant Microarchitectural Techniques for Voltage Underscaling in On-Chip Memories of CNN Accelerators

  2023cites this paper
• An Area Efficient Superconducting Unary CNN Accelerator

  2023cites this paper
• Heterogeneous Integration of In-Memory Analog Computing Architectures with Tensor Processing Units

  2023influential citation
• An Analytical Model of Configurable Systolic Arrays to find the Best-Fitting Accelerator for a given DNN Workload

  2023cites this paper
• VPU-EM: An Event-based Modeling Framework to Evaluate NPU Performance and Power Efficiency at Scale

  2023cites this paper
• TiC-SAT: Tightly-coupled Systolic Accelerator for Transformers

  2023cites this paper
• Dynamic Thermal-Predicted Workload Movement with Three-Dimensional DRAM-RRAM Hybrid Memories for Convolutional Neural Network Applications

  2022cites this paper
• Hardware Flexible Systolic Architecture for Convolution Accelerator in Convolutional Neural Networks

  2022cites this paper
• Design of High-Throughput Mixed-Precision CNN Accelerators on FPGA

  2022cites this paper
• Canopy: A CNFET-based Process Variation Aware Systolic DNN Accelerator

  2022influential citation
• QUIDAM: A Framework for Quantization-aware DNN Accelerator and Model Co-Exploration

  2022influential citation
• Analysis of the Effect of Off-chip Memory Access on the Performance of an NPU System

  2022cites this paper
• MOCCA: A Process Variation Tolerant Systolic DNN Accelerator using CNFETs in Monolithic 3D

  2022cites this paper
• Cascading structured pruning: enabling high data reuse for sparse DNN accelerators

  2022cites this paper
• DRAGON (Differentiable Graph Execution) : A suite of Hardware Simulation and Optimization tools for Modern AI/Non-AI Workloads

  2022influential citation
• QADAM: Quantization-Aware DNN Accelerator Modeling for Pareto-Optimality

  2022cites this paper
• ENCORE Compression: Exploiting Narrow-width Values for Quantized Deep Neural Networks

  2022cites this paper
• QAPPA: Quantization-Aware Power, Performance, and Area Modeling of DNN Accelerators

  2022cites this paper
• Integration of Single-Port Memory (ISPM) for Multiprecision Computation in Systolic-Array-Based Accelerators

  2022cites this paper
• An FPGA Overlay for CNN Inference with Fine-grained Flexible Parallelism

  2022cites this paper
• ARK: Fully Homomorphic Encryption Accelerator with Runtime Data Generation and Inter-Operation Key Reuse

  2022cites this paper
• Bifrost: End-to-End Evaluation and optimization of Reconfigurable DNN Accelerators

  2022cites this paper
• Exploring the Use of Novel Spatial Accelerators in Scientific Applications

  2022influential citation
• Scale-out Systolic Arrays

  2022cites this paper
• Spatial Data Dependence Graph Based Pre-RTL Simulator for Convolutional Neural Network Dataflows

  2022influential citation
• Compiler-Driven Simulation of Reconfigurable Hardware Accelerators

  2022influential citation
• Thermal-constrained memory management for three-dimensional DRAM-PCM memory with deep neural network applications

  2022cites this paper
• A dynamic computational memory address architecture for systolic array CNN accelerators

  2022cites this paper
• Slice-and-Forge: Making Better Use of Caches for Graph Convolutional Network Accelerators

  2022cites this paper
• Hardware-Software Codesign of a CNN Accelerator

  2022cites this paper
• FSA: An Efficient Fault-tolerant Systolic Array-based DNN Accelerator Architecture

  2022influential citation
• Integration of a systolic array based hardware accelerator into a DNN operator auto-tuning framework

  2022cites this paper
• PhotoFourier: A Photonic Joint Transform Correlator-Based Neural Network Accelerator

  2022cites this paper
• RDO-Q: Extremely Fine-Grained Channel-Wise Quantization via Rate-Distortion Optimization

  2022cites this paper
• An Indoor-Monitoring LiDAR Sensor for Patients with Alzheimer Disease Residing in Long-Term Care Facilities

  2022cites this paper
• Scalable Coherent Optical Crossbar Architecture using PCM for AI Acceleration

  2022cites this paper
• A Fast and Flexible FPGA-based Accelerator for Natural Language Processing Neural Networks

  2022cites this paper
• A Fast, Accurate, and Comprehensive PPA Estimation of Convolutional Hardware Accelerators

  2022cites this paper
• Thermal-Aware Design for Approximate DNN Accelerators

  2022cites this paper
• Enable Simultaneous DNN Services Based on Deterministic Operator Overlap and Precise Latency Prediction

  2021cites this paper
• TENT: Efficient Quantization of Neural Networks on the tiny Edge with Tapered FixEd PoiNT

  2021cites this paper
• Self-Adaptive Reconfigurable Arrays (SARA): Using ML to Assist Scaling GEMM Acceleration

  2021cites this paper
• Control Variate Approximation for DNN Accelerators

  2021cites this paper
• Mind mappings: enabling efficient algorithm-accelerator mapping space search

  2021cites this paper
• GNNerator: A Hardware/Software Framework for Accelerating Graph Neural Networks

  2021cites this paper
• Barrier-Free Large-Scale Sparse Tensor Accelerator (BARISTA) For Convolutional Neural Networks

  2021cites this paper
• Rate-Distortion Optimized Coding for Efficient CNN Compression

  2021cites this paper
• HyCA: A Hybrid Computing Architecture for Fault-Tolerant Deep Learning

  2021influential citation
• NASGuard: A Novel Accelerator Architecture for Robust Neural Architecture Search (NAS) Networks

  2021cites this paper
• S2 Engine: A Novel Systolic Architecture for Sparse Convolutional Neural Networks

  2021cites this paper
• Energy Efficient Computing with Heterogeneous DNN Accelerators

  2021cites this paper
• A Construction Kit for Efficient Low Power Neural Network Accelerator Designs

  2021cites this paper
• S2TA: Exploiting Structured Sparsity for Energy-Efficient Mobile CNN Acceleration

  2021cites this paper
• CREW: Computation Reuse and Efficient Weight Storage for Hardware-accelerated MLPs and RNNs

  2021cites this paper
• Mapping and virtual neuron assignment algorithms for MAERI accelerator

  2021cites this paper
• Auto-Split: A General Framework of Collaborative Edge-Cloud AI

  2021cites this paper
• Design and Scaffolded Training of an Efficient DNN Operator for Computer Vision on the Edge

  2021influential citation
• An Electro-Photonic System for Accelerating Deep Neural Networks

  2021influential citation
• Configurable Multi-directional Systolic Array Architecture for Convolutional Neural Networks

  2021influential citation
• Data scheduling and placement in deep learning accelerator

  2021influential citation
• POMMEL: Exploring Off-Chip Memory Energy & Power Consumption in Convolutional Neural Network Accelerators

  2021cites this paper
• Energon: Toward Efficient Acceleration of Transformers Using Dynamic Sparse Attention

  2021cites this paper
• Special Session: When Dataflows Converge: Reconfigurable and Approximate Computing for Emerging Neural Networks

  2021cites this paper
• FLAT: An Optimized Dataflow for Mitigating Attention Bottlenecks

  2021cites this paper
• Analyzing and improving MAESTRO's analytical modeling APPROVED BY SUPERVISING COMMITTEE:

  2021cites this paper
• An Edge 3D CNN Accelerator for Low-Power Activity Recognition

  2020cites this paper
• SIES: A Novel Implementation of Spiking Convolutional Neural Network Inference Engine on Field-Programmable Gate Array

  2020cites this paper
• Runtime Design Space Exploration and Mapping of DCNNs for the Ultra-Low-Power Orlando SoC

  2020cites this paper
• XBAROPT - Enabling Ultra-Pipelined, Novel STT MRAM Based Processing-in-Memory DNN Accelerator

  2020influential citation
• Energy Predictive Models for Convolutional Neural Networks on Mobile Platforms

  2020cites this paper
• Hardware-Aware AutoML for Efficient Deep Learning Applications

  2020cites this paper
• ASTRA-SIM: Enabling SW/HW Co-Design Exploration for Distributed DL Training Platforms

  2020cites this paper
• DNNGuard: An Elastic Heterogeneous DNN Accelerator Architecture against Adversarial Attacks

  2020cites this paper
• DNN-Chip Predictor: An Analytical Performance Predictor for DNN Accelerators with Various Dataflows and Hardware Architectures

  2020cites this paper
• A Gradient-Interleaved Scheduler for Energy-Efficient Backpropagation for Training Neural Networks

  2020influential citation
• Reliable and Energy Efficient MLC STT-RAM Buffer for CNN Accelerators

  2020cites this paper
• Low-Latency Proactive Continuous Vision

  2020cites this paper
• GrateTile: Efficient Sparse Tensor Tiling for CNN Processing

  2020cites this paper
• E-BATCH: Energy-Efficient and High-Throughput RNN Batching

  2020cites this paper
• An Energy-Efficient Deep Convolutional Neural Network Training Accelerator for In Situ Personalization on Smart Devices

  2020cites this paper
• Impact of the Array Shape and Memory Bandwidth on the Execution Time of CNN Systolic Arrays

  2020influential citation
• O-2A: Low Overhead DNN Compression with Outlier-Aware Approximation

  2020cites this paper
• A Systolic Accelerator for Neuromorphic Visual Recognition

  2020cites this paper
• Phonocardiography Signals Compression with Deep Convolutional Autoencoder for Telecare Applications

  2020cites this paper
• ASIE

  2020cites this paper
• A Fast Design Space Exploration Framework for the Deep Learning Accelerators: Work-in-Progress

  2020cites this paper
• Ax-BxP: Approximate Blocked Computation for Precision-reconfigurable Deep Neural Network Acceleration

  2020cites this paper
• CMSA: Configurable Multi-directional Systolic Array for Convolutional Neural Networks

  2020influential citation
• Dynamic Heterogeneous Voltage Regulation for Systolic Array-Based DNN Accelerators

  2020influential citation
• Low-power accelerators for cognitive computing

  2020cites this paper
• Model Compression on Faulty Array-based Neural Network Accelerator

  2020cites this paper
• Squeezing Correlated Neurons for Resource-Efficient Deep Neural Networks

  2020cites this paper
• Enabling Timing Error Resilience for Low-Power Systolic-Array Based Deep Learning Accelerators

  2020cites this paper