Enabling efficient, high-performance data access in parallel file systems (PFS) is critical for today's highperformance computing systems. PFS client-side I/O heavily impacts the final I/O performance delivered to individual applications and the entire system. Autotuning the key client-side I/O behaviors has been extensively studied and shows promising results. However, existing work has heavily relied on extensive number of global runtime metrics to monitor and accurate modeling of applications' I/O patterns. Such heavy overheads significantly limit the ability to enable fine-grained, dynamic tuning in practical systems. In this study, we propose DIAL (Decentralized I/O AutoTuning via Learned Client-side Local Metrics) which takes a drastically different approach. Instead of trying to extract the global I/O patterns of applications, DIAL takes a decentralized approach, treating each I/O client as an independent unit and tuning configurations using only its locally observable metrics. With the help of machine learning models, DIAL enables multiple tunable units to make independent but collective decisions, reacting to what is happening in the global storage systems in a timely manner and achieving better I/O performance globally for the application.

We propose a pad-focused, net-by-net, two-stage printed circuit board (PCB) routing approach comprising the global routing using Monte Carlo tree search (MCTS) and the detailed routing using$\mathrm{A}^{*}$:. Compared with conventional PCB routing algorithms, our approach can route PCB components in both BGA and non-BGA packages. To minimize the gap between the global and detailed routing stages, a polygon-based dynamic routable region partitioning mechanism is introduced. Experimental results show that our approach outperforms state-of-the-art routers such as DeepPCB and FreeRouting in terms of success rate or wirelength.

Evaluating machine-generated summaries without a human-written reference summary has been a need for a long time. Inspired by preference labeling in existing work of summarization evaluation, we propose to judge summary quality by learning the preference rank of summaries using the Bradley-Terry power ranking model from inferior summaries generated by corrupting base summaries. Extensive experiments on several datasets show that our weakly supervised scheme can produce scores highly correlated with human ratings.

Forrest Bao, Ge Luo, Hebi Li, Minghui Qiu, Yinfei Yang, Youbiao He, Cen Chen. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.

Graphical PCB design tools like KiCAD lack support for high-level abstraction such as functions and loops. To improve PCB design productivity, we hereby present BHDL, a programming framework for PCB designs. In its compact and declarative syntax, schematics and layouts can be modeled effectively and expressed concisely. Treating all circuits, even a resistor, as functions, BHDL naturally supports modularized development that builds a complex design up from smaller designs hierarchically. As an embedded Domain Specific Language (eDSL), BHDL allows users to leverage the full feature of the host language for customization and extension. Our Jupyter kernel supports web-based, REPL-style development and generates auto-placed PCBs.

Today's high-performance computing (HPC) platforms are still dominated by batch jobs. Accordingly, effective batch job scheduling is crucial to obtain high system efficiency. Existing HPC batch job schedulers typically leverage heuristic priority functions to prioritize and schedule jobs. But, once configured and deployed by the experts, such priority functions can hardly adapt to the changes of job loads, optimization goals, or system settings, potentially leading to degraded system efficiency when changes occur. To address this fundamental issue, we present RLScheduler, an automated HPC batch job scheduler built on reinforcement learning. RLScheduler relies on minimal manual interventions or expert knowledge, but can learn high-quality scheduling policies via its own continuous `trial and error'. We introduce a new kernel-based neural network structure and trajectory filtering mechanism in RLScheduler to improve and stabilize the learning process. Through extensive evaluations, we confirm that RLScheduler can learn high-quality scheduling policies towards various workloads and various optimization goals with relatively low computation cost. Moreover, we show that the learned models perform stably even when applied to unseen workloads, making them practical for production use.