Taskflow: A Lightweight Parallel and Heterogeneous Task Graph Computing System

Taskflow aims to streamline the building of parallel and heterogeneous applications using a lightweight task graph-based approach. introduces an expressive graph programming model assist developers in implementation decomposition strategies on computing platform. Our distinguishes itself as very general class parallelism with in-graph control flow enable end-to-end optimization. To support our high performance, we design efficient system runtime that solves many new scheduling challenges arising out models optimizes performance across latency, energy efficiency, throughput. We have demonstrated promising real-world applications. As example, large-scale machine learning workload up 29% faster, 1.5× less memory, 1.9× higher throughput than industrial system, oneTBB, 40 CPUs 4 GPUs. opened source deployed it large numbers users open-source community.