Safety and Deferred Update in Transactional Memory
- 960 Downloads
Transactional memory allows the user to declare sequences of instructions as speculative transactions that can either commit or abort, providing all-or-nothing semantics. If a transaction commits, it should appear to execute sequentially, so that the committed transactions constitute a correct sequential execution. If a transaction aborts, none of its instructions should affect other transactions. These semantics allow the programmer to incorporate sequential code within transactions and let the transactional memory care about conflicts between concurrent transactions. In this sense, it is important that the memory is safe, i.e., every transaction has a consistent view even if the transaction aborts later. Otherwise, inconsistencies not predicted by the sequential program may cause a fatal irrecoverable error or an infinite loop. Furthermore, in a general setting, where a transaction may be explicitly aborted by the user or an external contention manager, a transaction should not be allowed to read from a not yet committed transaction, which is often called deferred-update semantics. This chapter overviews the scope of consistency criteria proposed so far to capture deferred-update semantics, and shows that—under reasonable conditions—the semantics induces a safety property.
KeywordsSafety Property Read Operation Transactional Memory Opaque History Serial Execution
Unable to display preview. Download preview PDF.
- 2.Attiya, H., Gotsman, A., Hans, S., Rinetzky, N.: Safety of live transactions in transactional memory: TMS is necessary and sufficient. In: Kuhn, F. (ed.) DISC 2014. LNCS, vol. 8784, pp. 376–390. Springer, Heidelberg (2014)Google Scholar
- 3.Attiya, H., Hans, S., Kuznetsov, P., Ravi, S.: Safety of deferred update in transactional memory. In: ICDCS, pp. 601–610 (2013)Google Scholar
- 4.Attiya, H., Welch, J.: Distributed Computing: Fundamentals, Simulations, and Advanced Topics, 2nd edn. Wiley Interscience (2004)Google Scholar
- 7.Guerraoui, R., Kapalka, M.: Principles of Transactional Memory, Synthesis Lectures on Distributed Computing Theory. Morgan and Claypool (2010)Google Scholar
- 8.Guerraoui, R., Ruppert, E.: Linearizability is not always a safety property. In: Noubir, G., Raynal, M. (eds.) NETYS 2014. LNCS, vol. 8539, pp. 57–69. Springer, Heidelberg (2014)Google Scholar
- 12.Imbs, D., Raynal, M.: Virtual world consistency: A condition for STM systems (with a versatile protocol with invisible read operations). Theor. Comput. Sci. 444 (July 2012)Google Scholar
- 13.König, D.: Theorie der Endlichen und Unendlichen Graphen: Kombinatorische Topologie der Streckenkomplexe. Akad. Verlag (1936)Google Scholar
- 14.Kuznetsov, P., Ravi, S.: On the cost of concurrency in transactional memory. CoRR, abs/1103.1302 (2011)Google Scholar
- 15.Lesani, M., Luchangco, V., Moir, M.: Putting opacity in its place. In: WTTM (2012)Google Scholar
- 16.Lynch, N.A.: Distributed Algorithms. Morgan Kaufmann (1996)Google Scholar
- 19.Shavit, N., Touitou, D.: Software transactional memory. In: PODC 1995, pp. 204–213 (1995)Google Scholar