Discipline scientifique | EPFL Graph Search

Les disciplines scientifiques (aussi appelées « sciences » ou « domaines scientifiques ») sont des subdivisions de la science et des branches du savoir qui utilisent une méthode rigoureuse et systématique pour étudier un domaine particulier de la connaissance. Les disciplines scientifiques se caractérisent par l'utilisation de la méthode scientifique pour tester des hypothèses, recueillir des données et formuler des théories explicatives sur le fonctionnement et l'histoire du monde naturel et social, voire de mondes abstraits. Chaque discipline scientifique possède ses propres objets d'étude, concepts, méthodes et normes de recherche. Mais cette indépendance apparente n'empêche ni de nombreuses collisions entre les différents domaines de la science, ni même l'existence de points communs entre certaines disciplines permettant de dresser des classifications. Les disciplines scientifiques se distinguent des disciplines littéraires. Si les disciplines littéraires peuvent avoir un même o

Performance prediction and race detection in message-passing parallel applications

Basile Schäli

The combination of low cost clusters and multicore processors lowers the barrier for accessing massive amounts of computing power. As computational sciences advance, the use of in silico simulations to complement in vivo experiments promises parallel programming a bright future in multiple scientific fields. It is therefore increasingly important to develop tools helping developers to write efficient and bug-free parallel applications. This thesis focuses on performance prediction and advanced testing tools for distributed memory message-passing parallel applications. The tools have been implemented within the Dynamic Parallel Schedules (DPS) parallelization framework. They have also been partly adapted to applications written using the Message Passing Interface (MPI) standard. The first part presents a parallel application simulator which has been integrated into the DPS framework. We identified a small set of processing and networking parameters that characterize the hardware platform on which the application is running. After parameterizing the hardware platform, the running time of parallel applications can be predicted using direct execution without requiring any change to the application source code. We propose a partial direct execution technique that reduces the execution time and memory consumption of the simulation. Using partial direct execution, the simulation is no longer tied to the platform to be simulated. Simulations may thus run on a desktop computer rather than on the target parallel machine. The proposed parameterization of the application and of the hardware properties enable using the simulator to study the sensitivity of a parallel application to various operating conditions such as the data subdivision granularity, the adopted parallelization strategy and the underlying hardware platform properties. The proposed simulator helps developers identifying the factors having the largest impact on their application's performance, and determining the most suitable cluster hardware configuration. Speed should not come at the expense of correctness. Since improving parallelization efficiency often requires loosening synchronizations or implementing more complex communication patterns, developers need to ensure that their changes do not introduce potential message races or deadlocks. Deadlocks and message races are common sources of problems in parallel applications and stem from the fact that the delivery of messages from different sources is not deterministically ordered. This non-determinism makes such synchronization errors hard to reproduce and debug. The second part of the thesis presents methods for uncovering potential deadlocks and message races by taking advantage of the flow graph structures and checkpointing capabilities of DPS. We developed a debugger for DPS applications that displays an instantaneous graphical view of the global computation state and is able to control the ordering of message delivery in order to explicitly test specific orderings. The number of possible orderings explodes when the number of messages sent by the application increases. Manual testing can only cover a tiny fraction of possible executions. Therefore, we use the simulator's ability to control the execution of a parallel application, in order to automatically detect deadlocks and message races. A first method for reducing the number of orderings to be tested relies on a partial-order reduction of the search space and on the decomposition of the application execution into independently testable subparts. This method relies on a static analysis of an execution trace of the application, and can therefore only be applied to parallel applications that produce a fixed set of messages, i.e. applications producing the same messages for all delivery orderings. In order to overcome this limitation, we propose an approach relying on the dynamic construction of a state graph expressing possible executions. Both methods reduce the testing costs by several orders of magnitude, and can be combined to further improve the results. Nevertheless, testing durations may remain prohibitive for longer running applications. We therefore also define algorithms generating subsets of possible orderings that are likely to reveal erroneous executions. In the recent years, the MPI standard has emerged as the de facto standard for writing message-passing parallel applications. The final part of this thesis therefore focuses on adapting the aforementioned parallel application testing concepts to MPI applications. We first describe the extension of our work on visualizing the execution of parallel applications. We then discuss the limits and the benefits of using partial-order execution graphs to describe MPI application executions, and show that our dynamic message-passing state graph construction approach can be successfully applied.

EPFL2009

Fluid Dynamics and Outliers

Bernard André Vincent Calisti

The current craze for unicorns and other hyper growth companies brings a novel interest in what makes the exceptional performance of a firm. Literature about technological innovation explains the bond between the performance level reached by a company and its ability to promote radical innovations. But if the theory of disruption enables to validate various levels of innovation, it faces difficulties to define more accurately the drivers of hyper growth. This thesis comprises five main moments to better understand what drives fast growing young technological companies. To survey this question we have defined one single scientific field, namely microfluidics, focusing on active clusters and spinoffs from the same parent in this sector. The first phase aims at highlighting evolutionary prospects about scientific enhancements in microfluidics leading to successful innovations. It shows that there is no innovation âdue to succeedâ and that technical innovations do not make progress as linearly as usual technological trees found in academic literature could suggest. That leads to a kind of rewriting in the evolution of inkjet technology. The second phase intends to explore performance among groups of spinoffs from the same parent. We consider the 2 European clusters in microfluidics. It appears that all these firms are developing, but they do not reach the same degree of performance. From far, yet, the differences in performance do not clearly appear. To explain variations, we study each spinoffâs growth during the 5 first years after launching before comparing with the growth during the 5 first years of the mother company. We have to look in details to the European clusters, to see also technology variants. A fractal approach is thus enabling to understand that performance of firms is like hidden in technology layers. The third moment elaborates on the relationship between a superior technology and an outstanding performance level. In the case of spin off performance we are not at a stage where we could put the phenomenon under equation! That is why it was appropriate to survey the scientific and technological clusters in Europe from a case study perspective. It appears then that seemingly similarly disrupting technologies do not end up capturing similar value. In fact, slight differences in the technologies used are leading to massive differences in performance. We discover in the end that one single variable is rooting Outstanding Performance. In a fourth phase we restate the initial question to better understand what makes Outliers and Outstanding Performance. Where we show that exceptional value capture is coming from exceptionally fitting of the technology considered to a very specific application. Outstanding Performance is thus attached to an extraordinary EBITDA. The last moment of the dissertation is devoted to ultimately check findings. The whole checking process eventually led to the conclusion that outliers and outstanding performance are permanently referring to a kind of âuniquenessâ embedded in the various cases of exceptional value capture. Such a unique feature, qualifying outliers among technological startups, makes by itself a research topic. As more and more attention is given to specific statistical populations, not only could this research contribute to improve the knowledge about hyper growth companies but it could also impact methods in statistics; and beyond methods: statistics themselves as a discipline.

EPFL2017

Activities of daily living with reverse prostheses: importance of scapular compensation for functional mobility of the shoulder

Alain Farron, Dominique Pioletti, Patricia Scheuber, Alexandre Terrier

Hypothesis: The nonanatomical design of reverse shoulder prostheses induce medial displacement of the center of rotation, impingements and may reduce the mobility of the shoulder. The aim of this study is to test the hypothesis that during activities of daily living functional mobility of the shoulder can be restored by scapular compensation. Material and methods: A numerical 3-dimensional model was developed to reproduce the movement of the scapula and humerus, during 4 activities of daily living measured experimentally. This hypothesis was tested in 4 configurations of the aequalis reverse prosthesis (standard 36-mm glenosphere, 42-mm glenosphere, lateralized 36-mm glenosphere, lateralized Bony Increased-Offset Reverse Shoulder Arthroplasty [BIO-RSA]), which were implanted in the virtual model. All impingement positions were evaluated, as the required scapular compensation to avoid impingements. Results: With the 36-mm glenosphere, impingements occurred only for rest of hand to back-pocket positions. The 42-mm partly improved the mobility. The 2 lateralized glenospheres were free of impingement. When impingements occurred, the scapular compensation was less than 10 degrees. Conclusion: Most reverse prostheses impingements reported in clinical and biomechanical studies can be avoided, either by scapular compensation or by a glenosphere lateralization. After reverse shoulder arthroplasty, a fraction of the mobility of the gleno-humeral is transferred to the scapulo-thoracic joint. Level of evidence: Basic Science Study, Computer Modeling. (C) 2013 Journal of Shoulder and Elbow Surgery Board of Trustees.