Explore a groundbreaking family of non-blocking memory reclamation schemes called Crystalline in this 21-minute conference talk from PLDI 2024. Delve into how Crystalline addresses the challenges of high performance, memory efficiency, and wait-freedom simultaneously. Learn about its ability to guarantee complete wait-freedom even with dynamically recycled threads, asynchronously reclaim memory, and ensure balanced reclamation workload across threads. Discover how Crystalline overcomes unique challenges to support x86-64 and ARM64 architectures while achieving superior throughput compared to previous fast schemes like epoch-based reclamation (EBR). Gain insights into Crystalline's advantages over other approaches such as hazard pointers (HP), WFE, and Hyaline, particularly in providing strict non-blocking guarantees when used with multiple data structures. Examine the supplementary materials, including the available and reusable artifacts, to further understand this innovative approach to memory management in concurrent systems.
A Family of Fast and Memory Efficient Lock- and Wait-Free Reclamation - PLDI 2024
ACM SIGPLAN via YouTube
Overview
Syllabus
[PLDI24] A Family of Fast and Memory Efficient Lock- and Wait-Free Reclamation
Taught by
ACM SIGPLAN