Smart Templates for Assisting Portability Layers (STAPL)

Achieving exascale is critical for improving America’s economic competitiveness and making scientific breakthroughs that will have profound effects on America and the world. The road to exascale is evolving toward advanced computer architectures with diverse processors. Advanced software solutions on programming models and runtimes are needed to obtain the best performance from these new hardware. The advanced programming paradigms are however complex, impose verbose expressions of algorithms, and demand expensive and rare expertise. The goal of this proposal is to develop an exascale software tool as a productive solution towards achieving portable performance across modern and emerging exascale hardware architectures. Our focus will be on developing automatic mapping to a performance portable programming model, namely, Kokkos. Our approach will be to extend an existing high performance computing compiler tool that can automatically parallelize and optimize programs and generate code for the new programming models. In Phase I, we will provide capabilities for automatic transformations and code generation for supporting advanced Kokkos concepts for exascale mapping, namely, heterogeneous mapping, hierarchical parallelism, custom data layout, and scratchpad memory management and data movement. We will also optionally validate and retarget an existing Kokkos code with potential advanced transformations. We will demonstrate the performance and productivity benefits of the tool through important DOE application kernels, namely kernels from E3SM application. STAPL will be a key tool in the exascale ecosystem to attain performance, programmability, and portability. This will enable DOE and the users of exascale ecosystem to push the frontiers of science and technology to enable scientific discovery, enhance national security, and improve competitiveness and global leadership of US economy. During the Phase II time frame, the Government labs will have large installations of new machines (e.g., Aurora21 and CORAL2) and will be striving to maximize the utilization of the Leadership Computing Facilities. A mature product built using the STAPL technology will be useful to the users mapping their applications to such machines. Commercial sectors deal with frequent code modernizations and face the need to port their applications (or write new applications) to modern powerful computing systems. The financial sector, biosciences industry, and the oil and gas industry are relevant examples that will potentially benefit from STAPL. STAPL will be also be a useful tool for the embedded computing market (involving Government and commercial sectors) to adopt embedded applications to the rapidly evolving embedded supercomputers.