CIMFlow Documentation
An Integrated Framework for SRAM-based digital Compute-in-Memory Architectures
Welcome to the temporary home for CIMFlow’s documentation! We’re hard at work developing detailed guides and references for this powerful framework.
About CIMFlow
CIMFlow is a comprehensive framework for designing and analyzing SRAM-based Compute-in-Memory (CIM) architectures. It integrates:
- A custom Instruction Set Architecture (ISA)
- An MLIR-based compiler for efficient code generation
- A SystemC-based simulator for architectural exploration and validation
Documentation Status
Our documentation is actively evolving. We plan to provide in-depth guidance on setup, configuration, optimization, and more. Although it’s still under construction, here’s a glimpse of what’s on the horizon:
- Getting Started Guide – Step-by-step instructions to set up CIMFlow
- API Reference – Descriptions of functions and interfaces
- Architecture Guide – Insights into CIMFlow’s architecture abstraction and modular design
- Examples & Tutorials – Practical, hands-on resources
Coming Soon
- Additional demos showcasing how CIMFlow integrates with diverse DNN workloads
- Updated tutorials for best practices and design considerations
- Tools for performance benchmarking and visualization
Stay Tuned
We appreciate your patience as we work to finalize the codebase and documentation. Check back soon for new updates, or follow along to learn about upcoming releases and features. We look forward to sharing more details as development progresses.