With the rise of large language models, distributed execution across multiple accelerators has become commonplace. Current ML systems must adopt complex distributed execution strategies for efficiency, but do so at the cost of extensibility. We believe that it is time to introduce a general-purpose distributed runtime for programming clusters of accelerators that enables: (1) placement flexibility, and (2) interoperability, without sacrificing (3) codesign. We propose using the DAFT API: distributed actors, futures, and tasks. To enable a smooth tradeoff between flexibility vs. performance, we introduce two execution modes: interpreted vs. compiled. We show how current applications in LLM inference and training can be executed as interpreted and compiled DAFT programs and discuss open questions and challenges.

Operating system (OS) extensions are more popular than ever. For example, Linux BPF is marketed as a "superpower" that allows user programs to be downloaded into the kernel, verified to be safe and executed at kernel hook points. So, BPF extensions have high performance and are often placed at performance-critical paths for tracing and filtering.

For processor cores, software-based self-test (SBST) is a promising complement to scan-based testing, especially for applications that demand high reliability. However, most prior SBST techniques only target stuck-at faults and thus fall short in detecting aging induced timing violations. In this paper, we propose an automatic test program generator for detection of transition delay faults (TDFs) in pipelined processors. The key technique is the conversion of scan-based launch-on-capture (LoC) TDF test patterns to instruction sequences, which are combined to form the self-test program. In the field, the processor under test can execute the test program on demand, in its functional mode, to detect TDFs. To facilitate the pattern-to-instruction conversion, a test program template is developed. Derived from the pipelined processor operation, the template helps systematically and efficiently set the flip-flop values specified in LoC test patterns. The proposed technique is validated on a MIPS32 processor and achieves 97.82% transition delay fault coverage.

Hadoop MapReduce is a software framework for processing vast amounts of data in parallel on large clusters. As data size increases, a need arises to resolve the significant increase in the disk I/O of reducer nodes, which may cause runtime bottlenecks. In response, we propose Bucket MapReduce, a system that utilizes bucket sort and pipeline parallelism to improve the performance of Hadoop MapReduce. Several parameters may influence the performance in Bucket MapReduce, and thus we will further discuss parameter tuning. In this paper, we perform experiments on TeraSort benchmark. Bucket MapReduce successfully reduces local disk I/O by 61% and improves the runtime by 1. 39× in 800GB TeraSort benchmark.

Reconvergence has been recognized as the main reason for ATPG backtrack. It induces not only more, but also prolonged backtracks and causes more severe performance degradation than expected. In this paper, we propose a reconvergence-aware testability measure to better guide the ATPG justification process. Experiment results show that the proposed method significantly decreases the ATPG runtime, especially for circuits with deep logic level, by up to 76%.