Abstract:Technology mapping involves mapping logical circuits to a library of cells. Traditionally, the full technology library is used, leading to a large search space and potential overhead. Motivated by randomly sampled technology mapping case studies, we propose MapTune framework that addresses this challenge by utilizing reinforcement learning to make design-specific choices during cell selection. By learning from the environment, MapTune refines the cell selection process, resulting in a reduced search space and potentially improved mapping quality.
The effectiveness of MapTune is evaluated on a wide range of benchmarks, different technology libraries and technology mappers. The experimental results demonstrate that MapTune achieves higher mapping accuracy and reducing delay/area across diverse circuit designs, technology libraries and mappers. The paper also discusses the Pareto-Optimal exploration and confirms the perpetual delay-area trade-off. Conducted on benchmark suites ISCAS 85/89, ITC/ISCAS 99, VTR8.0 and EPFL benchmarks, the post-technology mapping and post-sizing quality-of-results (QoR) have been significantly improved, with average Area-Delay Product (ADP) improvement of 22.54\% among all different exploration settings in MapTune. The improvements are consistently remained for four different technologies (7nm, 45nm, 130nm, and 180 nm) and two different mappers.
What problem does this paper attempt to address?
### Problems Addressed by the Paper
The paper aims to address the challenges in the process of Technology Mapping in hardware design. Specifically, traditional technology mapping methods typically use a complete Standard Cell Library, which results in a vast search space and potential high overhead. To tackle this issue, the authors propose a framework named **MapTune**, which leverages Reinforcement Learning (RL) to guide the selection of the standard cell library, thereby reducing the search space and improving mapping quality.
### Main Problems and Solutions
1. **Vast Search Space**:
- Traditional technology mapping methods use a complete standard cell library, leading to a very large search space, increasing computational complexity and runtime.
- **Solution**: By using a partially selected standard cell library, the search space is reduced, thereby lowering computational complexity and runtime.
2. **Mapping Quality and Performance Optimization**:
- Using a complete standard cell library for technology mapping may result in suboptimal outcomes, affecting the final circuit's performance (e.g., delay and area).
- **Solution**: Utilizing reinforcement learning algorithms to dynamically select the optimal subset of standard cells based on specific design requirements, thereby improving mapping quality and performance.
3. **Design-Specific Optimization**:
- Different design goals and technical requirements necessitate different standard cell library selection strategies.
- **Solution**: The MapTune framework can adaptively select the optimal subset of standard cells based on specific design goals and characteristics, achieving design-specific optimization.
### Experimental Validation
The authors validated the effectiveness of the MapTune framework through experiments on multiple benchmark suites (including ISCAS 85/89, ITC/ISCAS 99, VTR8.0, and EPFL benchmarks). The experimental results show that MapTune significantly improves mapping accuracy, reduces delay, and area across different technology libraries and mappers. Specifically, the average Area-Delay Product (ADP) improved by 22.54%, and consistent improvements were maintained across four different technology nodes (7nm, 45nm, 130nm, and 180nm) and two different mappers.
### Conclusion
The MapTune framework effectively addresses the issues of excessive search space and insufficient performance optimization in the technology mapping process by leveraging reinforcement learning techniques, providing a more efficient and flexible solution for hardware design.