Abstract. We study the correctness of shared data structures under reads-write concurrency. A popular approach to ensuring correctness of read-only operations in the presence of concurrent update, is read-set validation, which checks that all read variables have not changed since they were first read. In practice, this approach is often too conserva-tive, which adversely affects performance. In this paper, we introduce a new framework for reasoning about correctness of data structures under reads-write concurrency, which replaces validation of the entire read-set with more general criteria. Namely, instead of verifying that all read conditions over the shared variables, which we call base conditions. We show that reading values that satisfy some base condition at every point in time implies correctness of read-only operations executing in parallel with updates. Somewhat surprisingly, the resulting correctness guarantee is not equivalent to linearizability, and is instead captured through two new conditions: validity and regularity. Roughly speaking, the former re-quires that a read-only operation never reaches a state unreachable in a sequential execution; the latter generalizes Lamport’s notion of regular-ity for arbitrary data structures, and is weaker than linearizability. We further extend our framework to capture also linearizability. We illus-trate how our framework can be applied for reasoning about correctness of a variety of implementations of data structures such as linked lists. 1

Abstract In conjunction with PODC 2012, the TransForm project (Marie Curie Initial Training Network) and EuroTM (COST Action IC1001) supported the 4th edition of the Workshop on the Theory of Transactional Memory (WTTM 2012). The objective of WTTM was to discuss new theoretical challenges and recent achievements in the area of transactional computing. The workshop took place on July 19, 2012, in Madeira, Portugal. This year's WTTM was a milestone event for two reasons. First, because the same year, the two seminal articles on hardware and software transactional memories Transactional memory is a concurrency control mechanism for synchronizing concurrent accesses to shared memory by different threads. It has been proposed as an alternative to lock-based synchronization to simplify concurrent programming while exhibiting good performance. The sequential code is encapsulated in transactions, which are sequences of accesses to shared or local variables that should be executed atomically by a single thread. A transaction ends either by committing, in which case all of its updates take effect, or by aborting, in which case, all its updates are discarded and never become visible to other transactions. Consistency criteria Since the introduction of the transactional memory paradigm, several consistency criteria have being proposed to capture its correct behavior. Some consistency criteria have been inherited from $&0 6,*$&7 1HZV 'HFHPEHU YRO QR the database field (e.g., serializability, strict serializability), others have been proposed to extend these latter to take into account aborted transactions, e.g., opacity, virtual world consistency; some others have been proposed to define the correct behavior when transactions have to be synchronized with non transactional code, e.g., strong atomicity. Among all the criteria, opacity, originally proposed by Guerraoui and Kapalka Victor Luchangco presented his joint work with Mohsen Lesani and Mark Moir provocatively titled "Putting opacity in its place" in which he presented the TMS1 and TMS2 consistency conditions and clarified their relationship with the prefix-closed definition of opacity. Broadly, these conditions ensure that no transaction observes the partial effects of any other transaction in any execution of the STM without having to force a total-order on transactions that participate in the execution. In particular, TMS1 is defined for any object with a well-defined sequential specification while TMS2 is specific for read-write registers. They further show using IO Automata While opacity defines the correctness of transactional memories when shared variables are accessed only inside transactions, understanding the interaction between transactions and locks or between accesses to shared variables in and outside transactions has been a major question. This is motivated by the fact that the code written to work in a transactional system may need to interact with legacy code where locks have been used for synchronization. It has been shown that replacing a lock with a transaction does not always ensure the same behavior. Srivatsan Ravi presented his joint work with Vincent Gramoli and Petr Kuznetsov on the locallyserializable linearizability (ls-linearizability) consistency criterion that applies to both transactionbased and lock-based programs, thus allowing to compare the amount of concurrency of a concurrent program Stephan Diestelhorst presented his preliminary work with Martin Pohlack, "Safely Accessing Timestamps in Transactions". He presented scenarios in which the access to the CPU timestamp counter inside transactions could lead to unexpected behaviors. In particular, this counter is not a transactional variable, so reading it inside a transaction can lead to a violation of the single lock atomicity semantics: multiple accesses to this counter within transactions may lead to a different result than multiple accesses within a critical section. In this talk, a solution to prevent several transactions from accessing the timestamp concurrently was also sketched. Annette Bienusa presented the definition of snapshot trace to simplify the reasoning on the correctness of TMs that ensure snapshot isolation. This is a joint work with Peter Thiemann Finally, Faith Ellen stated that despite the fact that a rich set of consistency criteria have been proposed there is no agreement on the way the semantics of a transactional memory has to be defined: operationally $&0 6,*$&7 1HZV 'HFHPEHU YRO QR Data structures for transactional computing Eliot Moss' talk intended to explore how the availability of transactions as a programming construct might impact the design of data types. He gave multiple examples on how to design data types having in mind that these types are going to be used in transactions. In a similar direction, Maurice Herlihy considered data types that support high-level methods and their inverse. As an example a set of elements supports the method to add an element, add(x), and the method to remove an element from the set remove(x). For these kind of data types, he discussed the possibility to apply a technique called transactional boosting which provides a modular way to make highly concurrent thread-safe data structures transactional. He suggested to distinguish the transaction-level synchronization with the thread-level synchronization. In particular, to synchronize the access to a linearizable object, non-commutative method calls have to be executed serially (e.g., add(x) and remove(x)). Two method calls are commutative if they can be applied in any order and the final state of the object does not change. For example, add(x) and remove(x) do not commute, while add(x) and add(y) commute. Since methods have inverses, recovery can be done at the granularity of methods. This technique exploits the object semantics to synchronize concurrent accesses to the object. This is expected to be more efficient than STM implementations where consistency is guaranteed by detecting read/write conflicts. Performance Improving the efficiency of TMs has been a key problem for the last few years. In fact, in order for transactional memory to be accepted as a candidate to replace locks, we need to show that it has performance comparable to these latter. In her talk Faith Ellen summarized the theoretical results on the efficiency of TMs. She stated that efficiency has been considered through three axes: properties that state under which circumstances aborts have to be avoided (permissiveness Mykhailo Laremko discussed how to apply known techniques (e.g., combining) to boost the performance of existing STM systems that have a central point of synchronization. In particular, in his joint work with Panagiota Fatourou, Eleftherios Kosmas and Giorgos E. Papadakis, they augment the NOrec transactional memory by combining and replacing the single global lock with a set of locks. They provide preliminary simulation results to compare NOrec and its augmented versions, showing that these latter perform better. Nuno Diegues with João Cachopo [6] study how to extend a transactional memory to support nested transactions efficiently. The difficulty is to take into account the constraints imposed by the baseline algorithm. To investigate different directions in the design space (lazy versus eager conflict detection, multiversion versus single version etc), they consider the following transactional memories: JVSTM[9], NesTM [3] and PNSTM 'HFHPEHU YRO QR Diegues shows that PNSTM's throughput is not affected by parallel nesting, while it is the case for the throughput of JVSTM and NesTM. In particular, NesTM shows the greater degradation of performance w.r.t. the depth of the nesting. Legacy code and hardware transactional memory The keynote by Nir Shavit was about his joint work with Yehuda Afek and Alex Matveev on "Pessimistic transactional lock-elision (PLE)" Distributed transactional memory Distributed transactional memory is the implementation of the transactional memory paradigm in a networked environment where processes communicate by exchanging messages. Differently from transactional memory for multicore machines, the networked environment needs to take into account the non negligible communication delays. To support local accesses to shared objects, distributed transactional memories usually rely on replication. Pawel T. Wojciechowski presented his joint work with Jan Konczak. They consider the problem of recovering the state of the shared data after some node crashes. This requests to write data into stable storage. Their goal was to minimize writes to stable storage, which are slow, or to do them in parallel with the execution of transactions. He presented a crash-recovery model for distributed transactional memory which is based on deferred update replication relying on atomic broadcast. Their model takes into account the tradeoff between performance and fault-tolerance. See Sebastiano Peluso claimed that efficient replication schemes for distributed transactional memory have to follow three design principles: partial replication, genuineness to ensure scalability and support for wait-free read-only transactions. According to genuineness the only nodes involved in the execution of a transaction are ones that maintain a replica of an object accessed by the transaction. He claimed that genuineness is a fundamental property for the scalability of distributed transactional memory. This is a joint work with Paolo Romano and Francesco Quaglia $&0 6,*$&7 1HZV 'HFHPEHU YRO QR Conclusion While transactional memory has become a practical technology integrated in the hardware of the IBM BlueGene/Q supercomputer and the upcoming Intel Haswell processors, the theory of transactional memory still misses good models of computation, good complexity measures, agreement on the right definitions, identification of fundamental and useful algorithmic questions, innovative algorithm designs and lower bounds on problems. Upcoming challenges will likely include the design of new transactional algorithms that exploit the low-overhead of low-level instructions on the one hand, and the concurrency of high-level data types on the other hand. For further information the abstracts and slides of the talks can be found at http://sydney.edu.au/engineering/it/˜gramoli/events/wttm4. Acknowledgements We are grateful to the speakers, to the program committee members of WTTM 2012 for their help in reviewing this year's submissions and to Panagiota Fatourou for her help in the organization of the event. We would like to thank Srivatsan Ravi and Mykhailo Laremko for sharing their notes on the talks of the workshop.

We present a TM system that executes transactions without ever causing any aborts. The system uses a set of t-var lists, one for each transactional variable. A scheduler undertakes the task of placing the instructions of each transaction in the appropriate t-var lists based on which t-variable each of them accesses. A set of worker threads are responsible to execute these instructions. Because of the way instructions are inserted in and removed from the lists, by the way the worker threads work, and by the fact that the scheduler places all the instructions of a transaction in the appropriate t-var lists before doing so for the instructions of any subsequent transaction, it follows that no conflict will ever occur. Parallelism is fine-grained since it is achieved at the level of transactional instructions instead of transactions themselves (i.e., the instructions of a transaction may be executed concurrently). 1.

Multithreaded programming, which well fits the structure of modern shared memory systems, is becoming one of the most popular parallel programming models. Two important supporting components of multithreaded programming, concurrent data structures and transactional memory, can benefit from speculation. However, traditional speculation mechanisms, which do not exploit high level knowledge about the program, are inefficient in concurrent data structures and transactional memory systems. In this proposal, we bridge the gap between the high-level program knowledge and the speculation in multithreaded programming with the compiler’s assistance. We propose a language extension to incorporate fast speculation into concurrent data structure design, and show how the language extension helps the speculation in newly emerged HTM processors. Besides, in order to improve the speculation in transaction memory systems under the circumstance with data high contention, two compiler optimization techniques are also proposed for STM and HTM, respectively. iii

As usual, I conclude the year with an annual review of distributed computing awards and conferences. I begin by reporting on two prestigious awards- the Dijkstra Prize and the Principles of Distributed Computing Doctoral Dissertation Award. I then proceed with reviews of the main two distributed computing conferences, PODC – the ACM Symposium on Principles of Distributed

Abstract-The long-lived renaming problem appears in shared-memory systems where a set of threads need to register and deregister frequently from the computation, while concurrent operations scan the set of currently registered threads. Instances of this problem show up in concurrent implementations of transactional memory, flat combining, thread barriers, and memory reclamation schemes for lock-free data structures. In this paper, we analyze a randomized solution for long-lived renaming. The algorithmic technique we consider, called the LevelArray, has previously been used for hashing and one-shot (single-use) renaming. Our main contribution is to prove that, in long-lived executions, where processes may register and deregister polynomially many times, the technique guarantees constant steps on average and O(log log n) steps with high probability for registering, unit cost for deregistering, and O(n) steps for collect queries, where n is an upper bound on the number of processes that may be active at any point in time. We also show that the algorithm has the surprising property that it is self-healing: under reasonable assumptions on the schedule, operations running while the data structure is in a degraded state implicitly help the data structure re-balance itself. This subtle mechanism obviates the need for expensive periodic rebuilding procedures. Our benchmarks validate this approach, showing that, for typical use parameters, the average number of steps a process takes to register is less than two and the worst-case number of steps is bounded by six, even in executions with billions of operations. We contrast this with other randomized implementations, whose worst-case behavior we show to be unreliable, and with deterministic implementations, whose cost is linear in n.