Jean Pichon-Pharabod's research while affiliated with Aarhus University and other places

... An interesting and short-term one would be to extend FM-JIT with the speculation insertion of CoreJIT [Barrière et al. 2021], now that speculative instruction compilation is provably feasible. Another direction for future work is to extend our JIT to a more realistic input language, such as WebAssembly [Watt et al. 2021] which already has a semantics mechanized in Coq. ...

... Operational semantics furthermore can be divided into those following a microarchitectural style (providing formalizations of the actual hardware architecture) and those trying to abstract from architectures. Most notably, view-based semantics [14,21,31] avoid modelling specific hardware components and instead define the semantics in terms of views of thread on the shared state. Promises [22,24] are employed in operational semantics as a way of capturing out-of-order writes while still executing operations in thread order. ...

... There has been much work on developing axiomatic (declarative) models for memory consistency in parallel systems, at the ISA level [2], [36], [37], the microarchitectural level [12], [16], [38], and the programming language level [19], [39], [40], [41], [42]. There has also been work on constructing equivalent operationalizations for these models, e.g., for Power [2], ARMv8 [10], RA [8], C++ [7], and TSO [18], [9]. ...

... Several works have aimed at improving the security of WebAssembly [4], [9], [13], [14], [15]. CT-wasm [9] proposes a type system to check the constant-time policy. ...

... We note that the difficulties that arise in the attempt to formalise the C memory model stem from the tension between well-established compiler transformations as well as the need to support a multitude of hardware-level memory models seamlessly versus the well-known intricacies of programming correct shared-variable algorithms [48]. This will be an ongoing balancing act that involves many competing factors, especially and including efficiency, and, increasingly, safety and security [50]; if we were to take a position, it would be that sections of code -hopefully, relatively small and localised -can be protected from arbitrary transformations from compiler theory and practice. ...

... Owicki-Grieslike proof calculi have been proposed by a number of researchers [11,12,23,41], and have also recently been given for non-volatile memory [3,33]. Svendsen et al. [37] have developed a separation logic for promises for the C11 memory model. Wright et al. [41] have developed an Owicki-Gries proof system for out-of-order writes (as allowed by promises), but rely on pre-processing via the denotational MRD framework [30]. ...

... Architectures [2,3,39,42] exhibit weak memory concurrency behaviors due to various architectural features such as memory hierarchy, interconnect and so on for performance reason. To gain performance from these architectures, the high level programming languages also introduce primitives and a number of programming models for weak memory concurrency are defined [5,6,11,20,21,25,27,33,34,41,48]. In this paper we follow the C/C++ concurrency model [5,31,34]. ...

... Event structures is a mathematical formalism introduced by Winskel [1] as a semantic domain of concurrent programs. In recent years there has been renewed interest in event structures, with the applications of the theory ranging from relaxed memory models [2][3][4] to model-based mutation testing [5]. The main advantage of event structures compared to traditional interleaving semantics is that they give a more compact and concise representation of programs' behaviors. ...

... 5 The rmw1 and rmw2 constraints are needed for ensuring the right behavior of RMWs including their atomicity. The po-rf constraint is an addition of RC11 on top of C11, which is a conservative solution to the łout-of-thin-air" problem that arises if po ∪ rf-cycles are allowed [Batty et al. 2015;. ...

... -We provide a modular solution to the count down mechanism by supporting a thread-local abstraction on top of the usual global view on its shared counter. While this can be viewed as an instance of fictional separation logic [31], our use of thread-local abstraction goes beyond that by allowing the interference effects of parallel threads to be modularly and precisely aggregated. -We provide interpretations for our abstract predicates. ...