# DISC: A Dynamic Shape Compiler for Machine Learning Workloads

arXiv: https://arxiv.org/abs/2103.05288 · HTML: https://mlsanity.com/paper/2103.05288/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 9 Mar 2021.

## Built on by (most cited first, 10 of 10)

- [FunASR: A Fundamental End-to-End Speech Recognition Toolkit](https://mlsanity.com/paper/2305.11013/) (2023)
- [Hidet: Task-Mapping Programming Paradigm for Deep Learning Tensor Programs](https://mlsanity.com/paper/2210.09603/) (2022)
- [Transformer-Lite: High-efficiency Deployment of Large Language Models on Mobile Phone GPUs](https://mlsanity.com/paper/2403.20041/) (2024)
- [PAI-Diffusion: Constructing and Serving a Family of Open Chinese Diffusion Models for Text-to-image Synthesis on the Cloud](https://mlsanity.com/paper/2309.05534/) (2023)
- [SoD2: Statically Optimizing Dynamic Deep Neural Network](https://mlsanity.com/paper/2403.00176/) (2024)
- [Tilus: A Tile-Level GPGPU Programming Language for Low-Precision Computation](https://mlsanity.com/paper/2504.12984/) (2025)
- [Data Overfitting for On-Device Super-Resolution with Dynamic Algorithm and Compiler Co-Design](https://mlsanity.com/paper/2407.02813/) (2024)
- [FTuner: A Fast Dynamic Shape Tensors Program Auto-Tuner for Deep Learning Compilers](https://mlsanity.com/paper/2407.21418/) (2024)
- [DVM: A Bytecode Virtual Machine Approach for Dynamic Tensor Computation](https://mlsanity.com/paper/2603.24239/) (2026)
- [DuplexCadence: Exact State and Execution from a Speech Model's Declared Timelines](https://mlsanity.com/paper/2609.35115/) (2026)

## Abstract (arXiv)

Many recent machine learning models show dynamic shape characteristics. However, existing AI compiler optimization systems suffer a lot from problems brought by dynamic shape models, including compilation overhead, memory usage, optimization pipeline and deployment complexity. This paper provides a compiler system to natively support optimization for dynamic shape workloads, named DISC. DISC enriches a set of IR to form a fully dynamic shape representation. It generates the runtime flow at compile time to support processing dynamic shape based logic, which avoids the interpretation overhead at runtime and enlarges the opportunity of host-device co-optimization. It addresses the kernel fusion problem of dynamic shapes with shape propagation and constraints collecting methods. This is the first work to demonstrate how to build an end-to-end dynamic shape compiler based on MLIR infrastructure. Experiments show that DISC achieves up to 3.3x speedup than TensorFlow/PyTorch, and 1.8x than Nimble.