Year: 2022
Type: dissertation
Abstract: Many important algorithms schedule their tasks according to a priority order. While parallel priority schedulerscommonly implement task enqueue and dequeueMin operations, some algorithms need a priority update operation that alters the scheduling metadata for a task, which is not supported by existing hardware. We present Hive, a task-based multicore architecture that extracts fine-grained parallelism from algorithms with priority updates. Like prior hardware systems for ordered parallelism, Hive uses data- and control-dependence speculation and a large speculative window to execute tasks in parallel and out of order. Hive supports updates by identifying the scheduler-carried dependence, and speculating on it with task versioning, distinct from data versioning. Hive enables safe speculative updates to the schedule to better utilize speculation tracking resources and uncover more parallelism. Across nine benchmarks, Hive improves performance at 256 cores by up to 2.8x over the next best hardware solution.
Source: TSpace (University of Toronto)
Author: Posluns, Gilead Wolf
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7133012412
