Certainty in Lockless Concurrent Algorithms: an Informal Proof of Lace

Lockless concurrent programming brings new challenges to the eld of program verication. These lockless programs require methods such as compare-and-swap and memory fences to ensure correctness. However, their unpredictable behaviour in combination with these methods complicates verifying such algorithms. We use linearisation points[3], i.e. the points in time when the state of the system changes, to abstract these methods. By deducing the possible ordering of these linearisation points we can predict the possible states of the system and draw conclusions about the scrutinised algorithms. This paper uses linearisation points and the control ow of the program to create an informal proof of the Lace[10] algorithm, which implements a work-stealing method for concurrent programs.