strand-persistency · research · also persist strand

Strand persistency

Strand persistency divides a thread into logical segments that can make independent persistence progress while retaining specified within-strand and conflicting-access ordering constraints.

A new strand relaxes implicit dependencies; it does not erase all obligations on shared addresses. Pelley and colleagues retain strong persist atomicity. [1] Han and Kimura describe a later TSO-based semantics with newStrand, joinStrand and pbarrier primitives. These research interfaces are not portable CPU instructions or equivalent to enabling eADR. [2]

Relations

related

Memory persistency, Strict persistency, Epoch persistency, Persistence ordering, Extended Asynchronous DRAM Refresh

Citations

1. Pelley, Chen and Wenisch. Memory Persistency. ISCA 2014, pp. 265-276; sections 4-5 and 7-8, Table 1. (accessed 2026-10-19)
2. Han and Kimura. Durable Queue Implementations Built on a Formally Defined Strand Persistency Model. JIP 29 (2021), pp. 823-838, sections 2-3. (accessed 2026-10-19)

Further reading

3. Steven Pelley, Peter M. Chen and Thomas F. Wenisch. RETROSPECTIVE: Memory Persistency. ISCA retrospective, 2023; sections I-II. (accessed 2026-10-19)
4. Hesela. Persistent memory: DAX is not a durable commit. (accessed 2026-10-19)

JSON: /data/terms/strand-persistency.json