The Fifth International Workshop on Software Support for Portable Storage

NAND flash memory has been widely used as a nonvolatile storage for data. However, it does not support byte-level random access and hence directly executing program codes from NAND flash memory is not possible. Thus, we have to use RAM as working storage for preloading the program codes and executing them from RAM. This approach, however, requires a large amount of expensive RAM as the program size dramatically increasing. With this regard, our research group has been developing a framework for executing real-time program codes on NAND flash memory with the minimal usage of RAM. This paper overviews the entire framework targeting not only for soft real-time programs but also for hard real-time programs. Specifically, we first explain our baseline design flow called RT-PLRU (Real-Time constrained Pinning and LRU combination) targeting a single tasking soft real-time application. Then, we generalize it to mRT-PLRU (Multi-tasking RT-PLRU) so that the design flow can be applied to multitasking soft real-time application. This mRT-PLRU design flow is further extended to HRT-PLRU (Hard RT-PLRU) targeting multi-tasking hard real-time applications.