MEMMU : Memory Expansion for MMU-Less Embedded Systems

Northwestern University
Electrical Engineering and Computer Science
Lan S. Bai
Lei Yang
Prof. Robert Dick

Summary

MEMMU is an automated runtime memory compression technique for MMU-less embedded systems. It provides application developers with access to more usable RAM and requires no or minor changes to application code and no changes to hardware. MEMMU is composed of a runtime library and LLVM passes for compile-time transformation. The runtime library contains memory compression and management functions. The LLVM passes transform the application to use the MEMMU runtime library functions for memory access. It also applies compile-time memory access optimizations. MEMMU can be used on MMU-less embedded systems that do not have hardware virtual memory support, such as TelosB wireless sensor node. Our experiments using Telosb showed that MEMMU is often capable of expanding usable memory by up to 50% with less than 10% performance penalty on wireless sensor applications. It presently supports applications written in C and nesC languages. It is freely available for academic and non-profit use.

Downloads

The version below contains source code of MEMMU runtime library written in C, MEMMU LLVM passes written in C++, and a Python script to automatically apply the transformation flow of MEMMU on original program.

• Required environment of MEMMU

  1. LLVM version 1.9. It can be downloaded at https://www.llvm.org/
  2. Python. Official website: http://www.python.org/

• MEMMU Source Code

  The tar file for MEMMU version 0.9 is available here. It contains source code, Python scripts, and README file. Please refer to the README file about how to install and use MEMMU.

Publications

• L. S. Bai, L. Yang, and R. P. Dick, “MEMMU: Memory Expansion for MMU-Less Embedded Systems,” ACM Trans. Embedded Computing Systems, vol. 8, no. 3, April 2009.
• L. S. Bai, L. Yang, and R. P. Dick, “Automated Compile-Time and Run-Time Techniques to Increase Usable Memory in MMU-Less Embedded Systems,” Proc. Int. Conf. Compilers, Architecture & Synthesis for Embedded Systems, Oct. 2006.

Related Projects

• CComp - CComp is an offline RAM compression technique developed by Nathan Cooprider and Dr. John Regehr at University of Utah. CComp and MEMMU are complementary.

Page maintained by Lan Bai.