# Bit Fusion: Bit-Level Dynamically Composable Architecture for Accelerating Deep Neural Networks

arXiv: https://arxiv.org/abs/1712.01507 · HTML: https://mlsanity.com/paper/1712.01507/

## Key facts

- Cited by 109 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Dec 2017.

## Built on by (most cited first, 30 of 109)

- [BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstruction](https://mlsanity.com/paper/2102.05426/) (2021)
- [OliVe: Accelerating Large Language Models via Hardware-friendly Outlier-Victim Pair Quantization](https://mlsanity.com/paper/2304.07493/) (2023)
- [Towards Mixed-Precision Quantization of Neural Networks via Constrained Optimization](https://mlsanity.com/paper/2110.06554/) (2021)
- [ANT: Exploiting Adaptive Numerical Data Type for Low-bit Deep Neural Network Quantization](https://mlsanity.com/paper/2208.14286/) (2022)
- [MQBench: Towards Reproducible and Deployable Model Quantization Benchmark](https://mlsanity.com/paper/2111.03759/) (2021)
- [BSQ: Exploring Bit-Level Sparsity for Mixed-Precision Neural Network Quantization](https://mlsanity.com/paper/2102.10462/) (2021)
- [Hardware Approximate Techniques for Deep Neural Network Accelerators: A Survey](https://mlsanity.com/paper/2203.08737/) (2022)
- [Enable Deep Learning on Mobile Devices: Methods, Systems, and Applications](https://mlsanity.com/paper/2204.11786/) (2022)
- [SDQ: Stochastic Differentiable Quantization with Mixed Precision](https://mlsanity.com/paper/2206.04459/) (2022)
- [Tender: Accelerating Large Language Models via Tensor Decomposition and Runtime Requantization](https://mlsanity.com/paper/2406.12930/) (2024)
- [MicroScopiQ: Accelerating Foundational Models through Outlier-Aware Microscaling Quantization](https://mlsanity.com/paper/2411.05282/) (2024)
- [A Survey: Collaborative Hardware and Software Design in the Era of Large Language Models](https://mlsanity.com/paper/2410.07265/) (2024)
- [Hardware Acceleration of Fully Quantized BERT for Efficient Natural Language Processing](https://mlsanity.com/paper/2103.02800/) (2021)
- [EcoFormer: Energy-Saving Attention with Linear Complexity](https://mlsanity.com/paper/2209.09004/) (2022)
- [It's All In the Teacher: Zero-Shot Quantization Brought Closer to the Teacher](https://mlsanity.com/paper/2203.17008/) (2022)
- [Transformer Acceleration with Dynamic Sparse Attention](https://mlsanity.com/paper/2110.11299/) (2021)
- [Ditto: Accelerating Diffusion Model via Temporal Value Similarity](https://mlsanity.com/paper/2501.11211/) (2025)
- [Quantization Variation: A New Perspective on Training Transformers with Low-Bit Precision](https://mlsanity.com/paper/2307.00331/) (2023)
- [M-ANT: Efficient Low-bit Group Quantization for LLMs via Mathematically Adaptive Numerical Type](https://mlsanity.com/paper/2502.18755/) (2025)
- [Enabling Design Methodologies and Future Trends for Edge AI: Specialization and Co-design](https://mlsanity.com/paper/2103.15750/) (2021)
- [Sparse Attention Acceleration with Synergistic In-Memory Pruning and On-Chip Recomputation](https://mlsanity.com/paper/2209.00606/) (2022)
- [BAFFLE: A Baseline of Backpropagation-Free Federated Learning](https://mlsanity.com/paper/2301.12195/) (2023)
- [A Precision-Scalable RISC-V DNN Processor with On-Device Learning Capability at the Extreme Edge](https://mlsanity.com/paper/2309.08186/) (2023)
- [FlexBlock: A Flexible DNN Training Accelerator with Multi-Mode Block Floating Point Support](https://mlsanity.com/paper/2203.06673/) (2022)
- [Resource-Efficient Deep Learning: A Survey on Model-, Arithmetic-, and Implementation-Level Techniques](https://mlsanity.com/paper/2112.15131/) (2021)
- [Anda: Unlocking Efficient LLM Inference with a Variable-Length Grouped Activation Data Format](https://mlsanity.com/paper/2411.15982/) (2024)
- [Accelerating Attention through Gradient-Based Learned Runtime Pruning](https://mlsanity.com/paper/2204.03227/) (2022)
- [ULEEN: A Novel Architecture for Ultra Low-Energy Edge Neural Networks](https://mlsanity.com/paper/2304.10618/) (2023)
- [DaCapo: Accelerating Continuous Learning in Autonomous Systems for Video Analytics](https://mlsanity.com/paper/2403.14353/) (2024)
- [Mixed-Precision Neural Networks: A Survey](https://mlsanity.com/paper/2208.06064/) (2022)

## Abstract (arXiv)

Fully realizing the potential of acceleration for Deep Neural Networks (DNNs) requires understanding and leveraging algorithmic properties. This paper builds upon the algorithmic insight that bitwidth of operations in DNNs can be reduced without compromising their classification accuracy. However, to prevent accuracy loss, the bitwidth varies significantly across DNNs and it may even be adjusted for each layer. Thus, a fixed-bitwidth accelerator would either offer limited benefits to accommodate the worst-case bitwidth requirements, or lead to a degradation in final accuracy. To alleviate these deficiencies, this work introduces dynamic bit-level fusion/decomposition as a new dimension in the design of DNN accelerators. We explore this dimension by designing Bit Fusion, a bit-flexible accelerator, that constitutes an array of bit-level processing elements that dynamically fuse to match the bitwidth of individual DNN layers. This flexibility in the architecture enables minimizing the computation and the communication at the finest granularity possible with no loss in accuracy. We evaluate the benefits of BitFusion using eight real-world feed-forward and recurrent DNNs. The proposed microarchitecture is implemented in Verilog and synthesized in 45 nm technology. Using the synthesis results and cycle accurate simulation, we compare the benefits of Bit Fusion to two state-of-the-art DNN accelerators, Eyeriss and Stripes. In the same area, frequency, and process technology, BitFusion offers 3.9x speedup and 5.1x energy savings over Eyeriss. Compared to Stripes, BitFusion provides 2.6x speedup and 3.9x energy reduction at 45 nm node when BitFusion area and frequency are set to those of Stripes. Scaling to GPU technology node of 16 nm, BitFusion almost matches the performance of a 250-Watt Titan Xp, which uses 8-bit vector instructions, while BitFusion merely consumes 895 milliwatts of power.