复杂异构计算系统HPL优化研究

当今世界的主流超级计算机越来越多地使用带有加速器的异构系统.随着加速器的浮点性能不断提高，超级计算机内计算节点的CPU、内存、总线、网络以及系统架构都要与之相适应.HPL（High Performance Linpack）是高性能计算机评测的传统基准测试程序，复杂异构系统给HPL评测带来很多机遇与挑战.针对带有GPU的异构超级计算机系统，提出一套新的CPU与加速器计算任务分配方式，提出平衡点理论指导HPL性能优化.为了优化HPL程序，提出了使用CPU与加速器协同工作的look-ahead算法和行交换连续流水算法，实现了加速器、CPU、网络等部件的高度并行.此外，为带有加速器的系统设计了新的panel分解和行交换的实现方法，提高加速器的利用率.在每个节点带有4个GPU的系统上，单节点HPL效率达到79.51%，14884节点效率达到62.22%.

Optimization of HPL on a Complex Heterogeneous Computing System

Nowadays, the mainstream supercomputers in the world adopt heterogeneous systems with accelerators more and more. The increase of float point computation performance of the accelerators requires other components to match its speed, including CPU, memory, bus, and network. HPL(High Performance LINPACK) is the traditional benchmark for high performance computers. Complex heterogeneous systems have brought both opportunities and challenges to the benchmarking with HPL. Therefore, for heterogeneous supercomputers, we propose a new task partitioning scheme between the CPU and the accelerators, using the balance point theory to guide the optimization of HPL. For optimizing HPL, we propose a look-ahead algorithm to coordinate the collaboration of CPU and the accelerators, as well as a contiguous row-swap algorithm, enabling the parallelism among CPU, accelerators, and network. Besides, we have designed new panel factorization and row-swap implementations for the system with accelerators, improving the effectiveness and efficiency of the usage of accelerators. With the configuration of 4 GPUs on each computing node, we have gained HPL efficiency of 79.51% on a single node, and 62,22% on 14884 nodes.