Software transactional memory (STM) has proven to be a useful abstraction for developing concurrent applications where programmers denote transactions with an atomic construct that delimits a collection of reads and writes to shared mutable references. The runtime system then guarantees that all transactions are observed to execute atomically with respect to each other. Traditionally, when the runtime system detects that one transaction conflicts with another, it aborts one of the transactions and restarts its execution from the beginning. This can lead to problems with both execution time and throughput.
This thesis presents a novel approach that uses first-class continuations to restart a conflicting transaction at the point of a conflict, avoiding the re-execution of any work from the beginning of the transaction that has not been compromised. In practice, this allows transactions to complete more quickly, decreasing execution time and increasing throughput. The ideas presented in this thesis have been implemented in the context of the Manticore project, an ML-family language with support for parallelism and concurrency. Crucially, this work relies on constant-time continuation capturing via a continuation-passing-style (CPS) transformation and heap-allocated continuations. The partial abort scheme has been implemented as a part of three modern STM implementations: TL2, TinySTM, and NOrec. Experimental results show that, while no base STM implementation is universally best, each partial-abort implementation compares favorably to its full-abort counterpart.
In addition to an implementation, this thesis presents a formal semantics for partial aborts. A proof of correctness is given by relating the partial-abort semantics to an analogous full-abort semantics via a simulation. All proofs have been formally verified using the Coq Theorem Prover.