Search

International audience; Guaranteed classical online algorithms are often designed in order to minimize the competitive ratio, i.e., the performance in the worst case, so usually do not perform well on easy inputs. An orthogonal approach has been developed in the last decade thanks to the progress in machine learning, which allows to predict key parameters of the instance. Its downside is that no guarantee can be given on the prediction quality, for instance if the training set does not represent the current instance. This statement is at the origin of the recently emerging field of learning-augmented algorithms. In this framework, an online algorithm has access to a predictor, which can give relevant information about the future instance, though without any guarantee on the quality. The objective is then to design an algorithm which performs well when the predictor turns out to be accurate, while being robust to imprecise predictions. This tutorial will introduce this new domain, present some recent works and discuss current research questions.

Invasive computing is a paradigm for designing and programming future parallel computing systems. For systems with 1,000 or more cores on a chip, resource-aware programming is of utmost importance to obtain high utilisation as well as computational, energy and power efficiency. Invasive computing provides a programmer explicit handles to specify and argue about resource requirements desired or required in different phases of execution: In an invade phase, an application asks the operating system to allocate a set of processor, memory and communication resources to be claimed. In a subsequent infect phase, the parallel workload is spread and executed on the obtained claim of resources. Finally, if the degree of parallelism should be lower again, a retreat operation frees the claim again, and the application resumes a sequential execution. To support this idea of self-adaptive and resource-aware programming, not only new programming concepts, languages, compilers, and operating systems were needed to be developed, but also revolutionary architectural changes in the design of MPSoCs (multiprocessor systems-on-a-chip) to efficiently support invasion, infection, and retreat operations. This book gives a comprehensive overview of all aspects of invasive computing., Invasives Rechnen ist ein Paradigma für den Entwurf und die Programmierung zukünftiger paralleler Rechensysteme. Für Systeme mit 1.000 oder mehr Kernen auf einem Chip ist eine ressourcengewahre Programmierung von größter Bedeutung, um eine hohe Auslastung sowie Rechen-, Energie- und Leistungseffizienz zu erreichen. Invasives Rechnen bietet dem Programmierer bzw. der Programmiererin explizite Möglichkeiten, die in verschiedenen Phasen der Ausführung gewünschten oder erforderlichen Ressourcenanforderungen zu spezifizieren: In einer Invasionsphase fordert eine Anwendung vom Betriebssystem eine Reihe gewünschter Prozessor-, Speicher- und Kommunikationsressourcen an. In einer anschließenden Infektionsphase wird die parallele Arbeitslast auf die bereitgestellten Ressourcen verteilt und ausgeführt. Nimmt der Grad der Parallelität während der Programmausführung ab, können die allozierten Ressourcen durch eine Rückzugsoperation wieder freigegeben werden, und die Anwendung wird wieder sequenziell ausgeführt. Um diese Idee der selbstadaptiven und ressourcengewahren Programmierung zu realisieren, mussten nicht nur neue Programmierkonzepte, Sprachen, Compiler und Betriebssysteme entwickelt werden, sondern auch revolutionäre architektonische Änderungen im Entwurf von MPSoCs (Multiprozessorsystemen auf einem Chip), um Invasions-, Infektions- und Rückzugsoperationen effizient zu unterstützen. Dieses Buch gibt einen umfassenden Überblick über alle Aspekte des invasiven Rechnens.

Compositrice extrêmement prolifique, la compositrice québécoise Linda Bouchard (née en 1957) est particulièrement active sur la scène musicale aux États-Unis ou elle effectue de longs séjours depuis 1989. Possédant un catalogue d’oeuvres riche et varié, sa musique fait autant appel à l’orchestre symphonique qu’à des instrumentations plus intimistes et aussi au multimédia et à l’improvisation. Cette entrevue aborde principalement les thèmes de l’exil et de l’identité, du point de vue spécifique de la compositrice à la lumière de son parcours, et permet de revivre en condensé sa trajectoire, offrant par ramification plusieurs échos de questionnements brûlant d’actualité sur la place du créateur de musique dans notre société., Quebec composer Linda Bouchard (born 1957) is an extremely prolific composer, particularly active on the new music scene in the u.s. where she has lived for lengthy periods since 1989. She has composed a rich and diverse catalogue of works, ranging from orchestral music to more intimist instrumentations, and also using multimedia or improvisation. In this interview, the composer reflects on the question of exile and identity in light of her own experience, thus reliving her own artistic path while also addressing some still burning questions about the place of music creators in the global culture.

Les résidences de création sont pour un compositeur un défi particulier, que celles-ci se déroulent dans un contexte intimiste ou au sein d’une structure plus « lourde » comme un orchestre. Quelles réflexions tirent les compositeurs de ce type d’expérience spécifique, dans lequel la composition en soi n’est souvent qu’une simple partie du travail ? Au travers d’une discussion de ton assez convivial, quelques questions d’ordre plutôt sociologique ont été posées à trois compositeurs ayant participé à des résidences, de façon à en tirer un aperçu au niveau de l’apport concret de cette expérience au métier du compositeur, et aussi générer des pistes de questionnement sur la place du compositeur dans la culture globale d’une société., Composer-in-residency programs are a particular challenge for composers, whether they take place within a more intimistic context or a more “heavy” structure like a symphony orchestra. What kinds of issues are raised in the composer’s mind by such experiences, where composing is just one part of the job? In the form of a convivial discussion, some questions of a sociological nature were put to three composers who have participated in residencies, in order to get an overview of the concrete contributions this challenge made to their craftsmanship, and also to generate some other questions about the place of music creators in the global culture.

We consider the problem of computing a Steiner tree of minimum cost under a k-hop constraint which requires the depth of the tree to be at most k. Our main result is an exact algorithm for metrics induced by graphs of bounded treewidth that runs in time n^O(k). For the special case of a path, we give a simple algorithm that solves the problem in polynomial time, even if k is part of the input. The main result can be used to obtain, in quasi-polynomial time, a near-optimal solution that violates the k-hop constraint by at most one hop for more general metrics induced by graphs of bounded highway dimension and bounded doubling dimension.

Static (offline) techniques for mapping applications given by task graphs to MPSoC systems often deliver overly pessimistic and thus suboptimal results w.r.t. exploiting time slack in order to minimize the energy consumption. This holds true in particular in case computation times of tasks may be workload-dependent and becoming known only at runtime or in case of conditionally executed tasks or scenarios. This paper studies and quantitatively evaluates different classes of algorithms for scheduling periodic applications given by task graphs (i.e., DAGs) with precedence constraints and a global deadline on homogeneous MPSoCs purely at runtime on a per-instance base. We present and analyze algorithms providing provably optimal results as well as approximation algorithms with proven guarantees on the achieved energy savings. For problem instances taken from realistic embedded system benchmarks as well as synthetic scalable problems, we provide results on the computation time and quality of each algorithm to perform a) scheduling and b) voltage/speed assignments for each task at runtime. In our portfolio, we distinguish as well continuous and discrete speed (e.g., DVFS-related) assignment problems. In summary, the presented ties between theory (algorithmic complexity and optimality) and execution time analysis deliver important insights on the practical usability of the presented algorithms for runtime optimization of task scheduling and speed assignment on MPSoCs.

We present a theory of an optically induced valley polarization in an interacting, time-reversal symmetric Weyl semimetal placed under strong magnetic fields. Because the application of a magnetic field reduces the symmetry of the crystal, the optical absorption intensity differs at Weyl nodes that were equivalent by symmetry at zero field. At strong magnetic field, the difference in the absorption intensity reaches 100% for a sizeable frequency interval of the incident light. This complete valley polarization originates from interband transitions involving the chiral Landau level, and can be controlled by changing the directions of the magnetic field and the light propagation. We identify the splitting of $0\to 1$ or $-1\to 0$ inter Landau level transitions as an observable signature of the complete valley polarization, and discuss its manifestation in the TaAs family of materials., Comment: 19 pages, 16 figures

Scientific workflows are frequently modeled as Directed Acyclic Graphs (DAGs) oftasks, which represent computational modules and their dependences in the form of data pro-duced by a task and used by another one. This formulation allows the use of runtime systemswhich dynamically allocate tasks onto the resources of increasingly complex computing plat-forms. However, for some workflows, such a dynamic schedule may run out of memory byprocessing too many tasks simultaneously. This paper focuses on the problem of transform-ing such a DAG to prevent memory shortage, and concentrates on shared memory platforms.We first propose a simple model of DAGs which is expressive enough to emulate complex mem-ory behaviors. We then exhibit a polynomial-time algorithm that computes the maximum peakmemory of a DAG, that is, the maximum memory needed by any parallel schedule. We considerthe problem of reducing this maximum peak memory to make it smaller than a given bound.Our solution consists in adding new fictitious edges, while trying to minimize the critical pathof the graph. After proving that this problem is NP-complete, we provide an ILP solution as wellas several heuristic strategies that are thoroughly compared by simulation on synthetic DAGsmodeling actual computational workflows. We show that on most instances we are able to de-crease the maximum peak memory at the cost of a small increase in the critical path, thus withlittle impact on the quality of the final parallel schedule.

International audience; This study proposes a novel design of a laboratory built static collector using the 3D printing technology. This new collector produces aligned‐to‐random nanofibers in nanofibrous scaffold through electrospinning process. A design of experiment (DOE), based on response surface, analyzes the effect of the main process parameters; concentration, voltage, and distance; on the responses diameter and orientation. A quantifying approach has been used to investigate the orientation of nanofibers in the produced patterned scaffold through Fourier transforms method. The obtained results have proven a good potential to be used in tissue engineering application, especially for cells requiring specific guidance.

This thesis deals with three main themes linked to task graph scheduling on modern computing platforms. A graph of tasks is a classical model of a program to be executed, for instance a scientific application. The decomposition of an application into several tasks allows to exploit the potential parallelism of this application without adaptating the program to the computing platform. The graph describes the tasks as well as their dependences, some tasks cannot be initiated before others are completed. The execution of an application is then determined by a schedule of the graph, computed by a dedicated software, which in particular describes which resources should be allocated to each task at which time. The three studied themes are the following: exploit inner task parallelism, use accelerators such as GPUs, and cope with a limited memory.For some applications, two types of parallelism can be exploited: several tasks can be executed concurrently, and each task may be executed on several processors, which reduces its processing time. We propose and study two models allowing to describe this processing time acceleration, in order to efficiently exploit both types of parallelism.We then study how to efficiently use accelerators such as GPUs, in a dynamic context in which the future tasks to schedule are unknown. The main difficulty consists in deciding whether a task should be executed on one of the rare available accelerators or on one of the many classical processors. The last theme covered in this thesis deals with a available main memory of limited size, and the resort to expensive data transfers. We focused on two scenarios. If it is possible to avoid such transfers, we propose to modify the graph in order to guarantee that any execution fits in memory, which allows to dynamically schedule the graph at runtime. If every schedule needs transfers, we studied how to minimize their quantity.The work on these three themes has led to a better understanding of the underlying complexities. The proposed theoretical solutions will influence future software implementations.; Cette thèse porte sur trois thématiques principales liées à l'ordonnancement de graphes de tâches sur des plates-formes de calcul modernes. Un graphe de tâches est une modélisation classique d'un programme à exécuter, par exemple une application de calcul scientifique. La décomposition d'une application en différentes tâches permet d'exploiter le parallélisme potentiel de cette application sans adapter le programme à la plate-forme de calcul visée. Le graphe décrit ces tâches ainsi que leurs dépendances, certaines tâches ne pouvant être exécutées avant que d'autres ne soient terminées. L'exécution d'une application est alors déterminée par un ordonnancement du graphe, calculé par un logiciel dédié, qui décrit entre autres quelles ressources sont allouées à chaque tâche et à quel moment. Les trois thèmes étudiés sont les suivants: exploiter le parallélisme intrinsèque des tâches, utiliser des accélérateurs tels que des GPU, et prendre en compte une mémoire limitée.Certaines applications présentent deux types de parallélisme que l'on peut exploiter: plusieurs tâches peuvent être exécutées simultanément, et chaque tâche peut être exécutée sur plusieurs processeurs afin de réduire son temps de calcul. Nous proposons et étudions deux modèles permettant de régir ce temps de calcul, afin d'exploiter ces deux types de parallélisme.Nous étudions ensuite comment utiliser efficacement des accélérateurs de calcul tels que des GPU, dans un contexte dynamique où les futures tâches à ordonnancer ne sont pas connues. La difficulté principale consiste à décider si une tâche doit être exécutée sur l'un des rares accélérateurs disponibles ou sur l'un des nombreux processeurs classiques. La dernière thématique abordée concerne le problème d'une mémoire principale limitée, et le recours à des transferts de données coûteux. Nous avons traité ce problème via deux scénarios. S'il est possible d'éviter de tels transferts, nous avons proposé de modifier le graphe afin de garantir que toute exécution ne dépasse pas la mémoire disponible, ce qui permet d'ordonnancemer les tâches dynamiquement au moment de l'exécution. Si tout ordonnancement nécessite des transferts, nous avons étudié le problème consistant à minimiser leur quantité.L'étude de ces trois thèmes a permis de mieux comprendre la complexité de ces problèmes. Les solutions proposées dans le cadre d'étude théorique pourront influencer de futures implémentations logicielles.

Modern computing platforms commonly include accelerators. We target the problem of scheduling applications modeled as task graphs on hybrid platforms made of two types of resources, such as CPUs and GPUs. We consider that task graphs are uncovered dynamically, and that the scheduler has information only on the available tasks, i.e., tasks whose predecessors have all been completed. Each task can be processed by either a CPU or a GPU, and the corresponding processing times are known. Our study extends a previous $4\sqrt{m/k}-competitive$ online algorithm [3], where m is the number of CPUs and k the number of GPUs (m≥k). We prove that no online algorithm can have a competitive ratio smaller than $\sqrt{m/k}$. We also study how adding flexibility on task processing, such as task migration or spoliation, or increasing the knowledge of the scheduler by providing it with information on the task graph, influences the lower bound. We provide a $(2\sqrt{m/k}+1)$-competitive algorithm as well as a tunable combination of a system-oriented heuristic and a competitive algorithm; this combination performs well in practice and has a competitive ratio in $Θ(\sqrt{m/k})$. We extend our results to more types of processors. Finally, simulations on different sets of task graphs illustrate how the instance properties impact the performance of the studied algorithms and show that our proposed tunable algorithm performs the best among the online algorithms in almost all cases and has even performance close to an offline algorithm.; Les plateformes de calcul modernes comportent souvent des accélérateurs. Nous nous intéressons au problème d’ordonnancement d’applications modélisées par des graphes de tâches, sur de telles plateformes composées de deux types de processeurs, par exemple des CPU et des GPU. On considère que les tâches sont dévoilées dynamiquement, et que l’ordonnanceur ne connaît que les tâches disponibles, i.e., les tâches dont les prédecesseurs ont tous été exécutés. Chaque tâche peut être traitée soit par un CPU soit par un GPU, et les temps de calculs correspondants sont connus. Notre étude étend un précédent algorithme online $4\sqrt{m/k}$-competitif, où m est le nombre de CPU et k le nombre de GPU (m≥k). Nous prouvons qu’aucun algorithme online ne peut avoir un facteur de compétitivité plus petit que $\sqrt{m/k}$. Nous étudions également comment cette borne inférieure est influencée par l’ajout de flexibilité sur le traitement des tâches (migration ou spoliation) ou par une meilleure connaissance du graphe par l’ordonnanceur. Nous fournissons un algorithme $(2\sqrt{m/k}+1)$-compétitif ainsi qu’une combinaison paramètrable avec un algorithme efficace en pratique qui permet d’obtenir un facteur de compétitivité en $Θ(\sqrt{m/k})$. Nous étendons nos résultats pour plus de deux types de processeurs. Enfin, des simulations sur plusieurs ensembles de graphes de tâches illustrent les performances des algorithmes étudiés

It has been recently predicted that the interplay between Coulomb interactions and Berry curvature can produce interesting optical phenomena in topologically nontrivial two-dimensional insulators. Here, we present a theory of the interband optical absorption for three-dimensional, doped Weyl semimetals. We find that the Berry curvature, Coulomb interactions, and the nonlinearity in the single-particle energy spectrum can together enable a light-induced valley polarization. We support and supplement our numerical results with an analytical toy model calculation, which unveils topologically nontrivial Mahan excitons with nonzero vorticity.

Semaine d'Etude Maths-Entreprises, Lyon, 30 janvier - 3 février 2017; Ce rapport conclut une collaboration avec le groupe Sanofi-Pasteur dans le cadre de la SEME (Semaine d’Étude Maths Entreprise). Pour le développement de nouveaux vaccins, l'entreprise compare des données expérimentales obtenues après injection de candidats vaccins afin de déterminer celui qui générerait la meilleure réponse immunitaire. Cependant, les valeurs des réponses immunitaires déterminées sont censurées à gauche en raison de la limite de quantification de l'appareil de mesure utilisé. Notre objectif est de prendre en compte les non-répondeurs (sujets dont la réponse immunitaire est sous le seuil de l'appareil) dans notre comparaison de vaccins.

Scientific workflows are frequently modeled as Directed Acyclic Graphs (DAG) oftasks, which represent computational modules and their dependencies, in the form of dataproduced by a task and used by another one. This formulation allows the use of runtime sys-tems which dynamically allocate tasks onto the resources of increasingly complex and hetero-geneous computing platforms. However, for some workflows, such a dynamic schedule mayrun out of memory by exposing too much parallelism. This paper focuses on the problem oftransforming such a DAG to prevent memory shortage, and concentrates on shared memoryplatforms. We first propose a simple model of DAG which is expressive enough to emulate com-plex memory behaviors. We then exhibit a polynomial-time algorithm that computes the max-imum peak memory of a DAG, that is, the maximum memory needed by any parallel schedule.We consider the problem of reducing this maximum peak memory to make it smaller than agiven bound by adding new fictitious edges, while trying to minimize the critical path of thegraph. After proving this problem NP-complete, we provide an ILP solution as well as severalheuristic strategies that are thoroughly compared by simulation on synthetic DAGs modelingactual computational workflows. We show that on most instances, we are able to decrease themaximum peak memory at the cost of a small increase in the critical path, thus with little im-pact on quality of the final parallel schedule.; Les applications de calcul scientifique sont souvent modélisées pardes graphes de tâches orientés acycliques (DAG), qui représentent les tâchesde calcul et leurs dépendances, sous la forme de données produites par unetâche et utilisées par une autre. Cette formulation permet l’utilisation d’APIqui allouent dynamiquement les tâches sur les ressources de plateformes decalcul hétérogènes de plus en plus complexes. Cependant, pour certaines ap-plications, un tel ordonnancement dynamique peut manquer de mémoire enexploitant trop de parallélisme. Cet article porte sur le problème consistant àtransformer un tel DAG pour empêcher toute pénurie de mémoire, en se con-centrant sur les plateformes à mémoire partagée. On propose tout d’abord unmodèle simple de graphe qui est assez expressif pour émuler des comporte-ments mémoires complexes. On expose ensuite un algorithme polynomial quicalcule le pic mémoire maximum d’un DAG, qui représente la mémoire maxi-male requise par tout ordonnancement parallèle. On considère ensuite le prob-lème consistant à réduire ce pic mémoire maximal pour qu’il devienne plus pe-tit qu’une borne donnée en rajoutant des arêtes fictives, tout en essayant deminimiser le chemin critique du graphe. Après avoir prouvé ce problème NP-complet, on fournit un programme linéaire en nombres entiers le résolvant,ainsi que plusieurs stratégies heuristiques qui sont minitieusement comparées-sur des graphes synthétiques modélisant des applications de calcul réelles. Onmontre que sur la plupart des instances, on arrive à diminuer le pic mémoiremaximal, au prix d’une légère augmentation du chemin critique, et donc avecpeu d’impact sur la qualité de l’ordonnancement parallèle final.

International audience; This paper formalizes a resource-allocation problem that is all too familiar to the seasoned program-committee member. For each submission j that the PC member has the honor of reviewing, there is a choice. The PC member can spend the time to review submission j in detail on his/her own at a cost of Cj. Alternatively, the PC member can spend the time to identify and contact peers, hoping to recruit them as subreviewers, at a cost of 1 per subreviewer. These potential subreviewers have a certain probability of rejecting each review request, and this probability increases as time goes on. Once the PC member runs out of time or unasked experts, he/she is forced to review the paper without outside assistance. This paper gives optimal solutions to several variations of the scheduling-reviewers problem. Most of the solutions from this paper are based on an iterated log function of Cj. In particular, with k rounds, the optimal solution sends the k-iterated log of Cj requests in the first round, the (k − 1)-iterated log in the second round, and so forth. One of the contributions of this paper is solving this problem exactly, even when rejection probabilities may increase. Naturally, PC members must make an integral number of subreview requests. This paper gives, as an intermediate result, a linear-time algorithm to transform the artificial problem in which one can send fractional requests into the less-artificial problem in which one sends an integral number of requests. Finally, this paper considers the case where the PC member knows nothing about the probability that a potential subreviewer agrees to review the paper. This paper gives an approximation algorithm for this case, whose bounds improve as the number of rounds increases.

La spiritualité en musique se limite-t-elle à la représentation d’un dogme et l’utilisation des diverses composantes de sa liturgie, ou s’agit-il aussi d’une certaine approche du son que l’on peut retrouver dans les stratégies compositionnelles du compositeur ? Telle est la question que veut poser cet article, où il est souhaité de générer, par l’interprétation de certains de leurs écrits, une sorte de dialogue imaginaire entre le « compositeur théologique » Olivier Messiaen et le « mystique agnostique » Morton Feldman. Il est aussi souhaité de réfléchir sur les rapports entre spiritualité et musique en général dans la musique du xxe siècle, ainsi que sur les diverses implications qu’une quête spirituelle peut avoir sur la production et la pensée esthétique d’un compositeur, que celui-ci soit croyant ou non, et que cette quête se concrétise par des oeuvres dites « sacrées » ou « profanes »., Is spirituality in music confined to the representation of a dogma and the use of liturgical components, or is it also a certain approach to sound inherent to the composer’s compositional strategies? This paper raises this question in hopes of generating a kind of imaginary dialogue between the ‘theological’ composer Olivier Messiaen, and the ‘agnostic mystic’ Morton Feldman by interpreting some of their writings. A further aim is to consider the relationship between spirituality and music in general in 20th century music, and the implications of a spiritual quest on a composer’s production and aesthetic thinking, whether or not the composer is a believer, and how this quest is actualized in works called “sacred” or “profane”.

Scientific workloads are often described by Directed Acyclic task Graphs.Indeed, DAGs represent both a model frequently studied in theoretical literature and the structure employed by dynamic runtime schedulers to handle HPC applications. A natural problem is then to compute a makespan-minimizing schedule of a given graph. In this paper, we are motivated by task graphs arising from multifrontal factorizations of sparsematrices and therefore work under the following practical model. We focus on malleable tasks (i.e., a single task can be allotted a time-varying number of processors) and specifically on a simple yet realistic speedup model: each task can be perfectly parallelized, but only up to a limited number of processors. We first prove that the associated decision problem of minimizing the makespan is NP-Complete. Then, we study a widely used algorithm, PropScheduling, under this practical model and propose a new strategy GreedyFilling. Even though both strategies are 2-approximations, experiments on real and synthetic data sets show that GreedyFilling achieves significantly lower makespans.

La communication « Highly Oriented Nanofibers Produced by Electrospinning » a été présentée sous forme de POSTER à la Fiber Society Spring 2016 Conference (Mulhouse 2016); International audience; This project consists of production of Polyamide-6 (PA-6) nanofiber with patterned structure by using novel collectors in electrospinning. These collectors (under patent process) will allow producing both oriented and non-oriented nanofibers in one scaffold for biomedical applications. These static collectors make it possible to have different nanofibers' structure and produce tunable oriented electrospun nanofibers scaffolds.

International audience; La microscopie tomographique diffractive est une technique d'imagerie combinant imagerie en amplitude-phase avec une variation des conditions d'observations, suivies d'une reconstruction numérique du spécimen observé. L'intérêt de cette approche, par rapport à la microscopie optique classique est double. L'information enregistrée est liée à la distribution des indices optiques complexes dans le spécimen observé, une quantité difficilement, voire non accessible en imagerie incohérente. De plus, cette approche permet de doubler la résolution par rapport à l'holographie. Nous avons développé cette technique en transmission, principalement pour l'observation de spécimens biologiques non-marqués, et en réflexion, pour l'observation de surfaces et interfaces (sciences des matériaux, métrologie). Récemment, nous avons aussi développé un système de rotation des échantillons, permettant une résolution 3D isotrope. Cette présentation a pour but de présenter les principes fondamentaux de cette technique, et certains résultats récents obtenus (en transmission) sur des spécimens biologiques et en réflexion sur des échantillons réfléchissants, ainsi que les améliorations que l'on peut en attendre.

Solving sparse linear systems can lead to processing tree workflows on a platform of processors. In this study, we use the model of malleable tasks motivated in [Prasanna96,Beaumont07] in order to study tree workflow schedules under two contradictory objectives: makespan minimization and memory minization. First, we give a simpler proof of the result of [Prasanna96] which allows to compute a makespan-optimal schedule for tree workflows. Then, we study a more realistic speed-up function and show that the previous schedules are not optimal in this context. Finally, we give complexity results concerning the objective of minimizing both makespan and memory.