Data underlying the PhD thesis: Accelerating Programmer-Friendly Intermittent Computing (Chapter 6)

This is the archive for Chapter 6 of the PhD thesis named "Accelerating Programmer-Friendly Intermittent Computing" by Vito Kortbeek. <br> <strong>Rationale</strong> Intermittently operating embedded computing platforms powered by energy harvesting must frequently checkpoint their computation state. Using non-volatile main memory significantly reduces the checkpoint size but at the cost of increasing the checkpoint frequency to cover WAR dependencies. Additionally, non-volatile memory is significantly slower to access. Both of these challenges are addressed by the architecture proposed in Chapter 6, greatly increasing performance. <br> <strong>Archive Structure</strong> This archive consists of the software to emulate the system (`icemu/plugins`), and the LLVM toolchain used to compile applications (`llvm`). The code can be built and run using the development docker container in the `docker` directory. To build all the benchmarks, execute the `run.sh` script within the `benchmarks` directory. Run the respective Jupyter Notebook in the `plotting` directory to analyze the results.

本归档文件配套于维托·科特比克（Vito Kortbeek）所著博士论文《加速面向程序员友好型间歇式计算》（Accelerating Programmer-Friendly Intermittent Computing）的第6章。

**设计原理** 由能量采集供电的间歇运行嵌入式计算平台，需频繁对自身计算状态执行检查点（checkpoint）操作。采用非易失性主内存（non-volatile main memory）可显著缩减检查点尺寸，但为此需提高检查点频率以覆盖写后读（WAR，Write-After-Read）依赖关系。此外，非易失性内存的访问速度显著更低。第6章提出的架构可同时解决上述两项挑战，大幅提升系统性能。

**归档文件结构** 本归档包含用于模拟该系统的软件（`icemu/plugins`），以及用于编译应用程序的LLVM工具链（`llvm`）。代码可通过`docker`目录下的开发Docker容器完成构建与运行。如需构建所有基准测试程序，请执行`benchmarks`目录内的`run.sh`脚本。可通过`plotting`目录下对应的Jupyter Notebook分析测试结果。