<div>Data-processing programs are becoming increasingly important in the Big-data era. However, two notable problems of these programs may cause sub-optimal data- processing results. On one hand, these programs contain large number of floating-point computations. Due to the limited precision of floating-point representations, errors are introduced, propagated and accumulated in series of computations, making the computation results unreliable. We call this problem as floating-point instability. On the other hand, these programs are heavily parameterized. As no universal optimal parameter configuration exists for all possible inputs, the setting of program parameters should be carefully chosen and tuned for each input. Otherwise, the result would be sub-optimal. Manual tuning is infeasible because the number of parameters and the range of each parameter value may be big.</div><div><br></div><div>We try to address these two challenges in this dissertation. For floating-point instability problem, we develop a novel runtime technique to capture different output variations in the presence of instability. It features the idea of transforming every floating point value to a vector of multiple values $-$ the values added to create the vector are obtained by introducing artificial errors that are upper bounds of actual errors. The propagation of artificial errors models the propagation of actual errors. When values in vectors result in discrete execution differences (e.g., following different paths), the execution is forked to capture the resulting output variations.</div><div><br></div><div>For parameterized data-processing programs, we develop a white-box program tuning framework to tune the program parameter configuration for optimal data-processing result of each program input. </div><div>To further reduce the parameter configuration overhead, we propose the first general framework to inject artificial intelligence (AI) in the program, so the intelligent program is able to predict the parameter configuration for each incoming input directly. However, similar to many other ML/AI applications, the crucial challenge lies in feature selection, i.e., selection of the feature variables for predicting the target parameter specified by the users.</div><div>Thus, we propose a novel approach by combining program analysis and statistical analysis for better program feature variables selection which further helps better target parameter prediction and improves the result.</div>

No / As unconventional computation matures and non-standard programming frameworks are demonstrated, the need for formal verification will become more prevalent. This is so because “programming” in unconventional substrates is difficult. In this paper we show how conventional verification tools can be used to verify unconventional programs implementing a logical XOR gate. / SK and MG acknowledge the EPSRC support (grant number: EP/I031812/1) support; NK’s work is supported by EPSRC (grant numbers: EP/I031642/1, EP/J004111/1, EP/L001489/1). MG and FI are partially supported by CNCS UEFISCDI (grant number: PN-II-ID-PCE-2011-3-0688). CD acknowledges an EPSRC studentship.

Dans le domaine des plates-formes ouvertes, l’utilisation des cadriciels (frameworks) enrichis par des déclarations pour exprimer les permissions de l’application est de plus en plus répandue. Ceci est une réaction logique au fait qu’il y a une explosion d’adoption des appareils embarqués et mobiles. Leur omniprésence dans notre vie quotidienne engendre des craintes liées à la sécurité et à la vie privée, car l’usager partage de plus en plus ses données et ressources privées avec des tiers qui développent des applications auxquelles on n’a pas de raison de faire confiance. Malheureusement, la manière dont ces langages de spécification ainsi que ces cadres d’applications sont développés est généralement assez ad hoc et repose sur un domaine d’application et un langage de programmation fixes. De plus, ces cadriciels ne sont pas assez restrictifs pour régler le problème de la fuite de données privées et ne donnent souvent pas non plus assez d’informations à l’usager sur le comportement attendu de l’application. Cette thèse présente une méthodologie généraliste pour développer des cadriciels dirigés par des déclarations, qui cible un spectre large de langages de programmation. Nous montrons comment des langages de déclaration expressifs permettent de spécifier avec modularité les droits d’accès aux ressources ainsi que le flux de contrôle d’une telle application. Ces langages peuvent ensuite être compilés en un cadriciel garantissant à l’usager final le respect de ces permissions. Par rapport aux cadriciels existants, notre méthodologie permet de guider la personne qui développe des applications à partir des spécifications ainsi que d’informer l’usager final sur l’usage des ressources sensibles. Contrairement aux travaux existants, la méthodologie présentée dans cette thèse ne repose par sur un langage de programmation particulier. Nous montrons comment mettre en oeuvre de tels cadriciels dans un spectre de langages : des langages avec typage statique ou dynamique, et suivant le paradigme objet ou fonctionnel. L’efficacité de l’approche est montrée à travers des prototypes dans le domaine des applications mobiles dans deux langages très différents, à savoir Java et Racket, ce qui montre la généralité de notre approche. / In the domain of open platforms, it has become common to use application programming frameworks extended with declarations that express permissions of applications. This is a natural reaction to ever more widespread adoption of mobile and pervasive computing devices. Their wide adoption raises privacy and safety concerns for users, as a result of the increasing number of sensitive resources a user is sharing with non-certified third-party application developers. However, the approach to designing these declaration languages and the frameworks that enforce their requirements is often ad hoc, and limited to a specific combination of application domain and programming language. Moreover, most widely used frameworks fail to address serious privacy leaks, and, crucially, do not provide the user with insight into application behaviour. This dissertation presents a generalised methodology for developing declaration-driven frameworks in a wide spectrum of host programming languages. We show that rich declaration languages, which express modularity, resource permissions and application control flow, can be compiled into frameworks that provide strong guarantees to end users. Compared to other declaration-driven frameworks, our methodology provides guidance to the application developer based on the specifications, and clear insight to the end user regarding the use of their private resources. Contrary to previous work, the methodology we propose does not depend on a specific host language, or even on a specific programming paradigm. We demonstrate how to implement declaration-driven frameworks in languages with static type systems, completely dynamic languages, object-oriented languages, or functional languages. The efficacy of our approach is shown through prototypes in the domain of mobile computing, implemented in two widely differing host programming languages, demonstrating the generality of our approach.

The effective use of GPUs for accelerating applications depends on a number of factors including effective asynchronous use of heterogeneous resources, reducing data transfer between CPU and GPU, increasing occupancy of GPU kernels, overlapping data transfers with computations, reducing GPU idling and kernel optimizations. Overcoming these challenges require considerable effort on the part of the application developers. Most optimization strategies are often proposed and tuned specifically for individual applications. Message-driven executions with over-decomposition of tasks constitute an important model for parallel programming and provide multiple benefits including communication-computation overlap and reduced idling on resources. Charm++ is one such message-driven language which employs over decomposition of tasks, computation-communication overlap and a measurement-based load balancer to achieve high CPU utilization. This research has developed an adaptive runtime framework for efficient executions of Charm++ message-driven parallel applications on GPU systems. In the first part of our research, we have developed a runtime framework, G-Charm with the focus primarily on optimizing regular applications. At runtime, G-Charm automatically combines multiple small GPU tasks into a single larger kernel which reduces the number of kernel invocations while improving CUDA occupancy. G-Charm also enables reuse of existing data in GPU global memory, performs GPU memory management and dynamic scheduling of tasks across CPU and GPU in order to reduce idle time. In order to combine the partial results obtained from the computations performed on CPU and GPU, G-Charm allows the user to specify an operator using which the partial results are combined at runtime. We also perform compile time code generation to reduce programming overhead. For Cholesky factorization, a regular parallel application, G-Charm provides 14% improvement over a highly tuned implementation. In the second part of our research, we extended our runtime to overcome the challenges presented by irregular applications such as a periodic generation of tasks, irregular memory access patterns and varying workloads during application execution. We developed models for deciding the number of tasks that can be combined into a kernel based on the rate of task generation, and the GPU occupancy of the tasks. For irregular applications, data reuse results in uncoalesced GPU memory access. We evaluated the effect of altering the global memory access pattern in improving coalesced access. We’ve also developed adaptive methods for hybrid execution on CPU and GPU wherein we consider the varying workloads while scheduling tasks across the CPU and GPU. We demonstrate that our dynamic strategies result in 8-38% reduction in execution times for an N-body simulation application and a molecular dynamics application over the corresponding static strategies that are amenable for regular applications.