Florence Plateau

N -synchronous Kahn networks: a relaxed model of synchrony for real-time systems

The design of high-performance stream-processing systems is a fast growing domain, driven by markets such like high-end TV, gaming, 3D animation and medical imaging. It is also a surprisingly demanding task, with respect to the algorithmic and conceptual simplicity of streaming applications. It needs the close cooperation between numerical analysts, parallel programming experts, real-time control experts and computer architects, and incurs a very high level of quality insurance and optimization.In search for improved productivity, we propose a programming model and language dedicated to high-performance stream processing. This language builds on the synchronous programming model and on domain knowledge -- the periodic evolution of streams -- to allow correct-by-construction properties to be proven by the compiler. These properties include resource requirements and delays between input and output streams. Automating this task avoids tedious and error-prone engineering, due to the combinatorics of the composition of filters with multiple data rates and formats. Correctness of the implementation is also difficult to assess with traditional (asynchronous, simulation-based) approaches. This language is thus provided with a relaxed notion of synchronous composition, called n-synchrony: two processes are n-synchronous if they can communicate in the ordinary (0-)synchronous model with a FIFO buffer of size n.Technically, we extend a core synchronous data-flow language with a notion of periodic clocks, and design a relaxed clock calculus (a type system for clocks) to allow non strictly synchronous processes to be composed or correlated. This relaxation is associated with two sub-typing rules in the clock calculus. Delay, buffer insertion and control code for these buffers are automatically inferred from the clock types through a systematic transformation into a standard synchronous program. We formally define the semantics of the language and prove the soundness and completeness of its clock calculus and synchronization transformation. Finally, the language is compared with existing formalisms.

Lucy-n: a n-synchronous extension of Lustre

Synchronous functional languages such as Lustre or Lucid Synchrone define a restricted class of Kahn Process Networks which can be executed with no buffer. Every expression is associated to a clock indicating the instants when a value is present. A dedicated type system, the clock calculus, checks that the actual clock of a stream equals its expected clock and thus does not need to be buffered. The n-synchrony relaxes synchrony by allowing the communication through bounded buffers whose size is computed at compile-time. It is obtained by extending the clock calculus with a subtyping rule which defines buffering points. This paper presents the first implementation of the n-synchronous model inside a Lustre-like language called Lucy-n. The language extends Lustre with an explicit buffer construct whose size is automatically computed during the clock calculus. This clock calculus is defined as an inference type system and is parametrized by the clock language and the algorithm used to solve subtyping constraints. We detail here one algorithm based on the abstraction of clocks, an idea originally introduced in [5]. The paper presents a simpler, yet more precise, clock abstraction for which the main algebraic properties have been proved in Coq. Finally, we illustrate the language on various examples including a video application.

Abstraction of Clocks in Synchronous Data-Flow Systems

Synchronous data-flow languages such as Lustre manage infinite sequences or streams as basic values. Each stream is associated to a clock which defines the instants where the current value of the stream is present. This clock is a type information and a dedicated type system -- the so-called clock-calculus -- statically rejects programs which cannot be executed synchronously. In existing synchronous languages, it amounts at asking whether two streams have the same clocks and thus relies on clock equality only. Recent works have shown the interest of introducing some relaxed notion of synchrony, where two streams can be composed as soon as they can be synchronized through the introduction of a finite buffer (as done in the SDF model of Edward Lee). This technically consists in replacing typing by subtyping. The present paper introduces a simple way to achieve this relaxed model through the use of clock envelopes . These clock envelopes are sets of concrete clocks which are not necessarily periodic. This allows to model various features in real-time embedded software such as bounded jitter as found in video-systems, execution time of real-time processes and scheduling resources or the communication through buffers. We present the algebra of clock envelopes and its main theoretical properties.

Interactive Programming of Reactive Systems

ReactiveML is a synchronous reactive extension of the general purpose programming language OCaml. It allows to program reactive systems such as video games or simulators. This paper presents rmltop, the ReactiveML counterpart of the OCaml toplevel. This toplevel allows a programmer to interactively write ReactiveML programs which are type-checked, compiled and loaded on the fly. The user can then progressively run concurrent processes and observe the interactions between them. The main strength of rmltop is that all valid ReactiveML expressions are accepted in the toplevel with the same semantics as in the compiler. This allows to use the ReactiveML toplevel as a debugger. Furthermore, the interpreted code is as efficient as if it was compiled. Moreover, a toplevel interpreter being itself a reactive system, another originality of rmltop is its own implementation in ReactiveML which makes it relatively light.

Synchronization of periodic clocks

We propose a programming model dedicated to real-time video-streaming applications for embedded media devices, including high-definition TVs. This model is built on the synchronous programming model extended with domain-specific knowledge --- periodic evolution of streams --- to allow correct-by-construction properties of the application to be proven by the compiler. These properties include buffer requirements and delays between input and output streams.Such properties are tedious to analyze by hand, due to the combinatorics of video filters, multiple data rates and formats. We show how to extend a core synchronous data-flow language with a notion of periodic clocks, and to design a relaxed clock calculus (a type system for clocks) to allow non strictly synchronous processes to be composed. This relaxation is associated with a subtyping rule in the clock calculus. Delay, buffer insertion and control code for these buffers are automatically inferred from the clock types through a systematic program transformation.