Tree traversals with task-memory affinities

نویسندگان

  • Julien Herrmann
  • Loris Marchal
  • Yves Robert
چکیده

We study the complexity of traversing tree-shaped workflows whose tasks require large I/O files. We target a heterogeneous architecture with two resource types, each with a different memory, such as a multicore node equipped with a dedicated accelerator (FPGA or GPU). The tasks in the workflow are colored according to their type and can be processed if all their input and output files can be stored in the corresponding memory. The amount of used memory of each type at a given execution step strongly depends upon the ordering in which the tasks are executed, and upon when communications between both memories are scheduled. The objective is to determine an efficient traversal that minimizes the maximum amount of memory of each type needed to traverse the whole tree. In this paper, we establish the complexity of this two-memory scheduling problem, and provide inapproximability results. In addition, we design several heuristics, based on both post-order and general traversals, and we evaluate them on a comprehensive set of tree graphs, including random trees as well as assembly trees arising in the context of sparse matrix factorizations. Key-words: Scheduling, Memory-aware, Sparse matrix factorization, Multifrontal method, Tree traversal, Bi-objective optimization ∗ Ecole Normale Supérieure de Lyon, CNRS & INRIA, France † University of Tennessee Knoxville, USA ha l-0 07 87 75 3, v er si on 1 22 M ar 2 01 3 Ordonnancement d’arbres de tâches avec affinité mémoire Résumé : Dans ce rapport, nous nous intéressons à la complexité du traitement d’arbres de tâches utilisant de gros fichiers d’entrée et de sortie. Nous nous focalisons sur une architecture hétérogène avec deux types de ressources, utilisant chacune une mémoire spécifique, comme par exemple un noeud multicore équipé d’un accélérateur (FPGA ou GPU). Les tâches de l’arbre sont colorées suivant leur type et peuvent être exécutées si tous leurs fichiers d’entrée et de sortie peuvent être stockés dans la mémoire correspondante. La quantité de mémoire de chaque type utilisée à une étape donnée de l’exécution dépend fortement de l’ordre dans lequel les tâches sont exécutées et du moment où sont effectuées les communications entre les deux mémoires. L’objectif est de déterminer un ordonnancement efficace qui minimise la quantité de mémoire de chaque type nécessaire pour traiter l’arbre entier. Dans ce rapport, nous établissons la complexité de ce problème d’ordonnancement à deux mémoires et nous fournissons des résultats d’inapproximabilité. De plus, nous proposons plusieurs heuristiques, fondées à la fois sur des traversées d’arbres en profondeur et des traversées générales, que nous évaluons sur un ensemble complet d’arbres, comprenant des arbres aléatoires ainsi que des arbres rencontrés dans le domaine de la factorisation de matrices creuses. Mots-clés : Ordonnancement, Contrainte Mémoire, Factorisation de matrices creuses, Méthode multifrontale, Traversée d’arbres, Optimisation bi-critère ha l-0 07 87 75 3, v er si on 1 22 M ar 2 01 3 Tree traversals with task-memory affinities 3

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

An improved algorithm to reconstruct a binary tree from its inorder and postorder traversals

It is well-known that, given inorder traversal along with one of the preorder or postorder traversals of a binary tree, the tree can be determined uniquely. Several algorithms have been proposed to reconstruct a binary tree from its inorder and preorder traversals. There is one study to reconstruct a binary tree from its inorder and postorder traversals, and this algorithm takes running time of...

متن کامل

An improved algorithm to reconstruct a binary tree from its inorder and postorder traversals

It is well-known that, given inorder traversal along with one of the preorder or postorder traversals of a binary tree, the tree can be determined uniquely. Several algorithms have been proposed to reconstruct a binary tree from its inorder and preorder traversals. There is one study to reconstruct a binary tree from its inorder and postorder traversals, and this algorithm takes running time of...

متن کامل

A novel algorithm to determine the leaf (leaves) of a binary tree from its preorder and postorder traversals

Binary trees are essential structures in Computer Science. The leaf (leaves) of a binary tree is one of the most significant aspects of it. In this study, we prove that the order of a leaf (leaves) of a binary tree is the same in the main tree traversals; preorder, inorder, and postorder. Then, we prove that given the preorder and postorder traversals of a binary tree, the leaf (leaves) of a bi...

متن کامل

Memory-aware tree traversals with pre-assigned tasks

We study the complexity of traversing tree-shaped workflows whose tasks require large I/O files. We target a heterogeneous architecture with two resource types, each with a different memory, such as a multicore node equipped with a dedicated accelerator (FPGA or GPU). The tasks in the workflow are colored according to their type and can be processed if all there input and output files can be st...

متن کامل

Model and Complexity Results for Tree Traversals on Hybrid Platforms

We study the complexity of traversing tree-shaped workflows whose tasks require large I/O files. We target a heterogeneous architecture with two resources of different types, each equipped with its own memory, such as a multicore node equipped with a dedicated accelerator (FPGA or GPU). Tasks in the workflow are tagged with the type of resource that is needed for their processing. Besides, a ta...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2013