DOCSLIB.ORG

Fast Search of the Optimal Contraction Sequence in Tensor Networks

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Fast Search of the Optimal Contraction Sequence in Tensor Networks 1 Fast Search of the Optimal Contraction Sequence in Tensor Networks Ling Liang, Jianyu Xu, Lei Deng, Member, IEEE, Mingyu Yan, Xing Hu, Zheng Zhang, Member, IEEE, Guoqi Li, Member, IEEE, Yuan Xie, Fellow, IEEE Abstract—Tensor network and tensor computation are widely [17]. In addition, tensor networks are capable of compressing applied in scientific and engineering domains like quantum the large-size parameters or data in neural networks [18]–[20] physics, electronic design automation, and machine learning. As or signal processing algorithms [21], [22]. one of the most fundamental operations for tensor networks, a tensor contraction eliminates the sharing orders among tensors Tensor contraction, a process of computing a tensor network and produces a compact sub-network. Different contraction by eliminating the sharing orders among pairs of tensors, is sequence usually yields distinct storage and compute costs, and one of the most fundamental operations in tensor network searching the optimal sequence is known as a hard problem. processing [23]. In a tensor network, the contraction operation Prior work have designed heuristic and fast algorithms to solve iteratively merges two nodes into one until the whole network this problem, however, several issues still remain unsolved. For example, the data format and data structure are not efficient, cannot be merged anymore. Different contraction sequence the constraints during modeling are impractical, the search of can result in distinct memory and compute costs. Therefore, the optimal solution might fail, and the search cost is very high. finding the optimal contraction sequence which consumes In this paper, we first introduce a logk order representation and less compute or storage resources is critical for reducing the design an adjacency matrix-based data structure to efficiently consequent contraction cost. accelerate the search of the optimal contraction sequence. Then, we propose an outer product pruning method with acceptable However, this might be a very hard problem. On one hand, overhead to reduce the search space. Finally, we use a multithread to find the contraction sequence with optimal compute cost is optimization in our implementation to further improve the proved to be NP-hard in [24]. On the other hand, for any tensor execution performance. We also present in-depth analysis of network, the optimal storage cost of contraction sequences factors that influence the search time. This work provides a equals the treewidth of its line graph, which has been proved full-stack solution for optimal contraction sequence search from both high-level data structure and search algorithm to low-level by [25]. Since the problem of computing the treewidth of a execution parallelism, and it will benefit a broad range of tensor- graph is NP-hard in general, it is a rational hyphothesis that the related applications. problem of finding the contraction sequence with optimal stor- age cost is also a hard problem. Therefore, designing heuristic Keywords: Tensor Contraction, Adjacency Matrix, BFS al- search algorithms seems the only way to find the contraction gorithm, Search Space Reduction, Multithread Optimization sequence with optimal storage or compute cost. There exist both depth-first constructive search (DFS) algorithms [24] and I. INTRODUCTION breadth-first constructive search (BFS) algorithms [26], [27] to Tensor networks are widely applied in a wide range of appli- do this search. Dynamic programming can also be applied to cations. The most well-known fields are many-body quantum solve this problem [27], [28]. These techniques go through all physics [1]–[5], matrix product states and projected entangled possible contraction sequences before determining the optimal pair states [6]–[11], multiscale entanglement renormalization one. ansatz [12], [13], and quantum circuit design [14], [15]. However, the search space grows exponentially as the Besides the applications in quantum physics, tensor networks network size increases. Some algorithms have also been in- are recently applied in IC modeling [16] and EDA problems vestigated in order to shrink the original search space. For the storage cost, an optimization algorithm is proposed in [29], This work was partially supported by National Science Foundation (Grant but this method does not guarantee an optimal sequence. For No. 1817037). Corresponding author: Lei Deng. Ling Liang, Jianyu Xu, the compute cost, cost capping is used to prune the sequences Xing Hu, Zheng Zhang, and Yuan Xie are with the Department of Elec- trical and Computer Engineering, University of California, Santa Barbara, that cost more than the optimal one, and some outer product CA 93106, USA (email: flingliang, xu jy15, xinghu, [email protected], constraints are added to further reduce the search space [27]. [email protected]). Guoqi Li is with the Department of Precision However, the search complexity depends on the variance of the Instrument, Center for Brain Inspired Computing Research, Tsinghua Univer- sity, Beijing 100084, China (email: [email protected]). Mingyu sharing orders between tensors. In addition, some algorithms Yan is with the State Key Laboratory of Computer Architecture, Institute of have been proposed to accelerate the search of contraction Computing Technology, Chinese Academy of Sciences, Beijing 100190, China sequences in closed tensor networks (i.e. the tensor network (email: [email protected]), and also with the Department of Electrical and Computer Engineering, University of California, Santa Barbara, CA being contracted into a scalar) [30]. A polynomial search 93106, USA. Lei Deng is with the Department of Electrical and Computer solution is proposed by [31] for considering both storage and Engineering, University of California, Santa Barbara, CA 93106, USA (email: compute costs, however, their solution is only effective for the [email protected]), and also with the Department of Precision Instrument, Center for Brain Inspired Computing Research, Tsinghua University, Beijing tree structure. 100084, China. Based on the observations from prior work, several issues 2 Table I: Variable definition. Variable Definition Variable Definition τi An original single tensor from a tensor network TI A tensor after contraction, where I is a set of i to denote the subscripts of original single tensors FOTI Free order collection of tensor TI SOTI TJ Sharing order collection between tensors TI and TJ , where I \ J = φ SETI Storage expense of tensor TI CETI TJ Compute expense for contracting tensors TI and TJ , where I \ J = φ MS Maximum storage expense of a contraction sequence MC Maximum compute expense of a contraction sequence RTI Row vector of possible tensor TI OTI Outer product vector used in the outer product pruning for possible tensor TI TI1 ; TI2 Two split source tensors to contract possible tensor TI sq A contraction sequence v CV Binomial coefficient Setv A set of possible tensors that are contracted by v original tensors should be considered in the search of the optimal contraction � = 1 � = 2 � = 3 $ $ sequence. First, since the search space is vast, efficient data � �� �% � ( ' format and data structure indeed matter and should be designed �' $ 0 ��# to accelerate the search process. Second, we should propose 1 an algorithm which can find the optimal solution based on the & & 2 ��% ��' data structure, shrink the search space, and fit general tensor 3 networks without specific structure constraints. At last, the 1 search time should be superior. In order to make the optimal 0 1 2 0 �((1,2,0) contraction search more efficient, we propose the following � $ � $ � $ �% �' ( �# ��' techniques. (1) We design a search algorithm based on the �$ �& �( adjacency matrix structure which is friendly to data access & & ��% ��' and network update. (2) Since the outer product between two 3 3 3 2 tensors can be pruned from the search space, we design an � � � efficient algorithm to identify the prunable tensors. (3) Finally, $ & ( 4 4 we adopt multithread optimization for parallel execution of our search algorithm, which can further improve the efficiency. 1 Figure 1: Examples of tensors with different order configu- Our proposed method will benefit a broad range of applications ration. The top subfigures visualize the original tensors; the that rely on tensor computation. middle and bottom subfigures show the graph representation of tensors. II. PRELIMINARIES In this section, we first introduce the background of tensor, where I denotes the subscript set of involved original single tensor contraction, and the problem definition of finding the tensors, e.g. here TI = T01. In this example, T01 is a two- optimal contraction sequence of a tensor network in Section order tensor (i.e. matrix). The element in tensor T01, such as II-A. Then, in Section II-B, we describe the vanilla BFS search T01(2; 1), can be calculated from the contraction between τ0 algorithm that we adopt as the basis of our algorithm design. and τ1 by N 0 N 1 τ0 τ0 A. Tensor Network Contraction X X T01(2; 1) = τ0(α; β; 2) × τ1(α; β; 1): (1) Tensor. The variables commonly used in this paper are α=0 β=0 listed in Table I. We define a tensor in a network as τi. In Figure 2(b), we term N 2 and N 2 as free orders which Tensor can be regarded as generalization of vector and matrix τ0 τ1 to represent high-order data. The number of orders in a tensor have only one end. These orders will be preserved after the contraction. Furthermore, τ0 and τ1 share two orders that have is denoted as M, and the length of the m-th order is denoted 0 1 m m + the same ends, which are remarked as N and N . We as N , where M; N 2 Z . Any element in a tensor τ0τ1 τ0τ1 τi τi m m 0 1 M−1 call them sharing orders and we have N = N . The τ (n ; n ; :::; n ) τiτj τjτi can be represented as i τi τi τi where we have nm 2 f0; 1; :::; N m − 1g contraction between two tensors can be interpreted as eliminat- τi τi .
Load more

Recommended publications
  • Tensor Manipulation in GPL Maxima
    Tensor Manipulation in GPL Maxima Viktor Toth http://www.vttoth.com/ February 1, 2008 Abstract GPL Maxima is an open-source computer algebra system based on DOE-MACSYMA. GPL Maxima included two tensor manipulation packages from DOE-MACSYMA, but these were in various states of disrepair. One of the two packages, CTENSOR, implemented component-based tensor manipulation; the other, ITENSOR, treated tensor symbols as opaque, manipulating them based on their index properties. The present paper describes the state in which these packages were found, the steps that were needed to make the packages fully functional again, and the new functionality that was implemented to make them more versatile. A third package, ATENSOR, was also implemented; fully compatible with the identically named package in the commercial version of MACSYMA, ATENSOR implements abstract tensor algebras. 1 Introduction GPL Maxima (GPL stands for the GNU Public License, the most widely used open source license construct) is the descendant of one of the world’s first comprehensive computer algebra systems (CAS), DOE-MACSYMA, developed by the United States Department of Energy in the 1960s and the 1970s. It is currently maintained by 18 volunteer developers, and can be obtained in source or object code form from http://maxima.sourceforge.net/. Like other computer algebra systems, Maxima has tensor manipulation capability. This capability was developed in the late 1970s. Documentation is scarce regarding these packages’ origins, but a select collection of e-mail messages by various authors survives, dating back to 1979-1982, when these packages were actively maintained at M.I.T. When this author first came across GPL Maxima, the tensor packages were effectively non-functional.
    [Show full text]
  • Electromagnetic Fields from Contact- and Symplectic Geometry
    ELECTROMAGNETIC FIELDS FROM CONTACT- AND SYMPLECTIC GEOMETRY MATIAS F. DAHL Abstract. We give two constructions that give a solution to the sourceless Maxwell's equations from a contact form on a 3-manifold. In both construc- tions the solutions are standing waves. Here, contact geometry can be seen as a differential geometry where the fundamental quantity (that is, the con- tact form) shows a constantly rotational behaviour due to a non-integrability condition. Using these constructions we obtain two main results. With the 3 first construction we obtain a solution to Maxwell's equations on R with an arbitrary prescribed and time independent energy profile. With the second construction we obtain solutions in a medium with a skewon part for which the energy density is time independent. The latter result is unexpected since usually the skewon part of a medium is associated with dissipative effects, that is, energy losses. Lastly, we describe two ways to construct a solution from symplectic structures on a 4-manifold. One difference between acoustics and electromagnetics is that in acoustics the wave is described by a scalar quantity whereas an electromagnetics, the wave is described by vectorial quantities. In electromagnetics, this vectorial nature gives rise to polar- isation, that is, in anisotropic medium differently polarised waves can have different propagation and scattering properties. To understand propagation in electromag- netism one therefore needs to take into account the role of polarisation. Classically, polarisation is defined as a property of plane waves, but generalising this concept to an arbitrary electromagnetic wave seems to be difficult. To understand propaga- tion in inhomogeneous medium, there are various mathematical approaches: using Gaussian beams (see [Dah06, Kac04, Kac05, Pop02, Ral82]), using Hadamard's method of a propagating discontinuity (see [HO03]) and using microlocal analysis (see [Den92]).
    [Show full text]
  • Automated Operation Minimization of Tensor Contraction Expressions in Electronic Structure Calculations
    Automated Operation Minimization of Tensor Contraction Expressions in Electronic Structure Calculations Albert Hartono,1 Alexander Sibiryakov,1 Marcel Nooijen,3 Gerald Baumgartner,4 David E. Bernholdt,6 So Hirata,7 Chi-Chung Lam,1 Russell M. Pitzer,2 J. Ramanujam,5 and P. Sadayappan1 1 Dept. of Computer Science and Engineering 2 Dept. of Chemistry The Ohio State University, Columbus, OH, 43210 USA 3 Dept. of Chemistry, University of Waterloo, Waterloo, Ontario N2L BG1, Canada 4 Dept. of Computer Science 5 Dept. of Electrical and Computer Engineering Louisiana State University, Baton Rouge, LA 70803 USA 6 Computer Sci. & Math. Div., Oak Ridge National Laboratory, Oak Ridge, TN 37831 USA 7 Quantum Theory Project, University of Florida, Gainesville, FL 32611 USA Abstract. Complex tensor contraction expressions arise in accurate electronic struc- ture models in quantum chemistry, such as the Coupled Cluster method. Transfor- mations using algebraic properties of commutativity and associativity can be used to significantly decrease the number of arithmetic operations required for evalua- tion of these expressions, but the optimization problem is NP-hard. Operation mini- mization is an important optimization step for the Tensor Contraction Engine, a tool being developed for the automatic transformation of high-level tensor contraction expressions into efficient programs. In this paper, we develop an effective heuristic approach to the operation minimization problem, and demonstrate its effectiveness on tensor contraction expressions for coupled cluster equations. 1 Introduction Currently, manual development of accurate quantum chemistry models is very tedious and takes an expert several months to years to develop and debug. The Tensor Contrac- tion Engine (TCE) [2, 1] is a tool that is being developed to reduce the development time to hours/days, by having the chemist specify the computation in a high-level form, from which an efficient parallel program is automatically synthesized.
    [Show full text]
  • The Riemann Curvature Tensor
    The Riemann Curvature Tensor Jennifer Cox May 6, 2019 Project Advisor: Dr. Jonathan Walters Abstract A tensor is a mathematical object that has applications in areas including physics, psychology, and artificial intelligence. The Riemann curvature tensor is a tool used to describe the curvature of n-dimensional spaces such as Riemannian manifolds in the field of differential geometry. The Riemann tensor plays an important role in the theories of general relativity and gravity as well as the curvature of spacetime. This paper will provide an overview of tensors and tensor operations. In particular, properties of the Riemann tensor will be examined. Calculations of the Riemann tensor for several two and three dimensional surfaces such as that of the sphere and torus will be demonstrated. The relationship between the Riemann tensor for the 2-sphere and 3-sphere will be studied, and it will be shown that these tensors satisfy the general equation of the Riemann tensor for an n-dimensional sphere. The connection between the Gaussian curvature and the Riemann curvature tensor will also be shown using Gauss's Theorem Egregium. Keywords: tensor, tensors, Riemann tensor, Riemann curvature tensor, curvature 1 Introduction Coordinate systems are the basis of analytic geometry and are necessary to solve geomet- ric problems using algebraic methods. The introduction of coordinate systems allowed for the blending of algebraic and geometric methods that eventually led to the development of calculus. Reliance on coordinate systems, however, can result in a loss of geometric insight and an unnecessary increase in the complexity of relevant expressions. Tensor calculus is an effective framework that will avoid the cons of relying on coordinate systems.
    [Show full text]
  • Appendix a Relations Between Covariant and Contravariant Bases
    Appendix A Relations Between Covariant and Contravariant Bases The contravariant basis vector gk of the curvilinear coordinate of uk at the point P is perpendicular to the covariant bases gi and gj, as shown in Fig. A.1.This contravariant basis gk can be defined as or or a gk g  g ¼  ðA:1Þ i j oui ou j where a is the scalar factor; gk is the contravariant basis of the curvilinear coordinate of uk. Multiplying Eq. (A.1) by the covariant basis gk, the scalar factor a results in k k ðgi  gjÞ: gk ¼ aðg : gkÞ¼ad ¼ a ÂÃk ðA:2Þ ) a ¼ðgi  gjÞ : gk gi; gj; gk The scalar triple product of the covariant bases can be written as pffiffiffi a ¼ ½¼ðg1; g2; g3 g1  g2Þ : g3 ¼ g ¼ J ðA:3Þ where Jacobian J is the determinant of the covariant basis tensor G. The direction of the cross product vector in Eq. (A.1) is opposite if the dummy indices are interchanged with each other in Einstein summation convention. Therefore, the Levi-Civita permutation symbols (pseudo-tensor components) can be used in expression of the contravariant basis. ffiffiffi p k k g g ¼ J g ¼ðgi  gjÞ¼Àðgj  giÞ eijkðgi  gjÞ eijkðgi  gjÞ ðA:4Þ ) gk ¼ pffiffiffi ¼ g J where the Levi-Civita permutation symbols are defined by 8 <> þ1ifði; j; kÞ is an even permutation; eijk ¼ > À1ifði; j; kÞ is an odd permutation; : A:5 0ifi ¼ j; or i ¼ k; or j ¼ k ð Þ 1 , e ¼ ði À jÞÁðj À kÞÁðk À iÞ for i; j; k ¼ 1; 2; 3 ijk 2 H.
    [Show full text]
  • Gicheru3012018jamcs43211.Pdf
    Journal of Advances in Mathematics and Computer Science 30(1): 1-15, 2019; Article no.JAMCS.43211 ISSN: 2456-9968 (Past name: British Journal of Mathematics & Computer Science, Past ISSN: 2231-0851) Decomposition of Riemannian Curvature Tensor Field and Its Properties James Gathungu Gicheru1* and Cyrus Gitonga Ngari2 1Department of Mathematics, Ngiriambu Girls High School, P.O.Box 46-10100, Kianyaga, Kenya. 2Department of Mathematics, Computing and Information Technology, School of Pure and Applied Sciences, University of Embu, P.O.Box 6-60100, Embu, Kenya. Authors’ contributions This work was carried out in collaboration between both authors. Both authors read and approved the final manuscript. Article Information DOI: 10.9734/JAMCS/2019/43211 Editor(s): (1) Dr. Dragos-Patru Covei, Professor, Department of Applied Mathematics, The Bucharest University of Economic Studies, Piata Romana, Romania. (2) Dr. Sheng Zhang, Professor, Department of Mathematics, Bohai University, Jinzhou, China. Reviewers: (1) Bilal Eftal Acet, Adıyaman University, Turkey. (2) Çiğdem Dinçkal, Çankaya University, Turkey. (3) Alexandre Casassola Gonçalves, Universidade de São Paulo, Brazil. Complete Peer review History: http://www.sciencedomain.org/review-history/27899 Received: 07 July 2018 Accepted: 22 November 2018 Original Research Article Published: 21 December 2018 _______________________________________________________________________________ Abstract Decomposition of recurrent curvature tensor fields of R-th order in Finsler manifolds has been studied by B. B. Sinha and G. Singh [1] in the publications del’ institute mathematique, nouvelleserie, tome 33 (47), 1983 pg 217-220. Also Surendra Pratap Singh [2] in Kyungpook Math. J. volume 15, number 2 December, 1975 studied decomposition of recurrent curvature tensor fields in generalised Finsler spaces.
    [Show full text]
  • Tensor Calculus
    A Primer on Tensor Calculus David A. Clarke Saint Mary’s University, Halifax NS, Canada [email protected] June, 2011 Copyright c David A. Clarke, 2011 Contents Preface ii 1 Introduction 1 2 Definition of a tensor 3 3 The metric 9 3.1 Physical components and basis vectors ..................... 11 3.2 The scalar and inner products .......................... 14 3.3 Invariance of tensor expressions ......................... 17 3.4 The permutation tensors ............................. 18 4 Tensor derivatives 21 4.1 “Christ-awful symbols” .............................. 21 4.2 Covariant derivative ............................... 25 5 Connexion to vector calculus 30 5.1 Gradient of a scalar ................................ 30 5.2 Divergence of a vector .............................. 30 5.3 Divergence of a tensor .............................. 32 5.4 The Laplacian of a scalar ............................. 33 5.5 Curl of a vector .................................. 34 5.6 The Laplacian of a vector ............................ 35 5.7 Gradient of a vector ............................... 35 5.8 Summary ..................................... 36 5.9 A tensor-vector identity ............................. 37 6 Cartesian, cylindrical, spherical polar coordinates 39 6.1 Cartesian coordinates ............................... 40 6.2 Cylindrical coordinates .............................. 40 6.3 Spherical polar coordinates ............................ 41 7 An application to viscosity 42 i Preface These notes stem from my own need to refresh my memory on the fundamentals
    [Show full text]
  • GEMM-Like Tensor-Tensor Contraction
    GEMM-like Tensor-Tensor Contraction Paul Springer Paolo Bientinesi AICES, RWTH Aachen University AICES, RWTH Aachen University Aachen, Germany Aachen, Germany [email protected] [email protected] 1 INTRODUCTION 2 GEMM-LIKE TENSOR CONTRACTION Tensor contractions (TCs)—a generalization of matrix-matrix mul- Let A, B, and C be tensors (i.e., multi-dimensional arrays), a tensor tiplications (GEMM) to multiple dimensions—arise in a wide range contraction is: of scientific computations such as machine learning1 [ ], spectral CI α × AI × BI + β × CI ; (1) element methods [10], quantum chemistry calculations [2], mul- C A B C tidimensional Fourier transforms [4] and climate simulations [3]. where IA, IB, and IC denote symbolic index sets. We further define Despite the close connection between TCs and matrix-matrix prod- three important index sets: Im := IA \ IC, In := IB \ IC, and ucts, the performance of the former is in general vastly inferior to Ik := IA \ IB, respectively denoting the free indices of A, the that of an optimized GEMM. To close such a gap, we propose a novel free indices B, and the contracted indices. Moreover, we adopt the approach: GEMM-like Tensor-Tensor (GETT) contraction [8]. Einstein notation, indicating that all indices of Ik are implicitly GETT captures the essential features of a high-performance summed (i.e., contracted). GEMM. In line with a highly optimized GEMM, and in contrast to Adhering to this definition, we can reformulate Equation (1) previous approaches to TCs, our high-performance implementa- such that its similarity to a matrix-matrix multiplication becomes tion of the GETT approach is fully vectorized, exploits the CPU’s apparent: cache hierarchy, avoids explicit preprocessing steps, and operates C C α ×A A ×B B +β ×C C ; (2) Π ¹Im [In º Π ¹Im [Ik º Π ¹In [Ik º Π ¹Im [In º on arbitrary dimensional subtensors while preserving stride-one A B C memory accesses.
    [Show full text]
  • Conformal Differential Geometry of a Subspace
    CONFORMAL DIFFERENTIAL GEOMETRY OF A SUBSPACE BY AARON FIALKOW Table of contents Introduction. 310 I. Preliminary matters 1. Notation. 315 2. Riemann spaces conformai to Vm. 316 3. Conformai tensors. 318 4. Algebraic and metric properties. 319 5. Generalized covariant differentiation and the second fundamental form of V„. 323 II. Conform al tensor algebra 6. The relative conformai curvature of Vn. 325 7. The conformai measure tensors. 329 III. Conformal tensor analysis 8. Conformal differentiation. 331 9. The replacement principle. 335 10. Conformal geodesic and conformal normal coordinates and parameters. 341 11. The Weyl tensors. 342 12. The Bianchi identities for the Weyl tensors. 346 13. The deviation tensor. 350 14. The conformal Riemann tensor of V„. 352 15. Variation of the mean curvature normal. 354 16. The conformal Riemann tensor of Vm. 355 17. The Ricci identities. 359 IV. The conformal vector spaces of Vn 18. The conformal osculating and normal vector spaces of V„. 359 19. The conformal normal measure tensors. 362 20. The conformal Frenet equations. 365 V. The fundamental equations 21. The boundary, algebraic, reality and derived conditions. 370 22. The conformal Gauss-Codazzi equations. 373 23. Higher integrability conditions. 376 24. The third derivatives of the relative conformal curvature. 379 25. The second derivatives of the mean curvature. Discussion for a hypersurface. 383 26. The second derivatives of the mean curvature normal. Discussion for the general case 385 VI. SUBSPACES IN A CONFORMALLY EUCLIDEAN SPACE Rm 27. Symmetry relations. 392 28. Determination of &£$,, U$Ü¿>,£«„>I<W. 396 Presented to the Society, October 26, 1940 and February 22, 1941; received by the editors February 1, 1943, and, in revised form, December 15, 1943.
    [Show full text]
  • Lecture Notes on General Relativity Sean M
    Lecture Notes on General Relativity Sean M. Carroll Institute for Theoretical Physics University of California Santa Barbara, CA 93106 [email protected] December 1997 Abstract These notes represent approximately one semester’s worth of lectures on intro- ductory general relativity for beginning graduate students in physics. Topics include manifolds, Riemannian geometry, Einstein’s equations, and three applications: grav- itational radiation, black holes, and cosmology. Individual chapters, and potentially updated versions, can be found at http://itp.ucsb.edu/~carroll/notes/. arXiv:gr-qc/9712019v1 3 Dec 1997 NSF-ITP/97-147 gr-qc/9712019 i Table of Contents 0. Introduction table of contents — preface — bibliography 1. Special Relativity and Flat Spacetime the spacetime interval — the metric — Lorentz transformations — spacetime diagrams — vectors — the tangent space — dual vectors — tensors — tensor products — the Levi-Civita tensor — index manipulation — electromagnetism — differential forms — Hodge duality — worldlines — proper time — energy-momentum vector — energy- momentum tensor — perfect fluids — energy-momentum conservation 2. Manifolds examples — non-examples — maps — continuity — the chain rule — open sets — charts and atlases — manifolds — examples of charts — differentiation — vectors as derivatives — coordinate bases — the tensor transformation law — partial derivatives are not tensors — the metric again — canonical form of the metric — Riemann normal coordinates — tensor densities — volume forms and integration 3. Curvature
    [Show full text]
  • M.Sc. (MATHEMATICS) MAL-633
    M.Sc. (MATHEMATICS) MECHANICS OF SOLIDS-I MAL-633 DIRECTORATE OF DISTANCE EDUCATION GURU JAMBHESHWAR UNIVERSITY OF SCIENCE & TECHNOLOGY HISAR, HARYANA-125001. MAL-633 MECHANICS OF SOLIDS-I Contents: Chapter -1 ………………………………………………………………… Page: 1-21 1.1. Introduction 01 1.1.1.1. Rank/Order of tensor 02 1.1.1.2. Characteristics of the tensors 02 1.2 Notation and Summation Convention 03 1.3 Law of Transformation 04 1.4 Some properties of tensor 11 1.5 Contraction of a Tensor 15 1.6 Quotient Law of Tensors 16 1.7 Symmetric and Skew Symmetric tensors 18 Chapter - II ……………………………………………………………… Page: 18-32 2.1 The Symbol ij 19 2.1.1. Tensor Equation 20 2.2 The Symbol ijk 24 2.3 Isotropic Tensors 26 2.4 Contravariant tensors(vectors) 29 2.5 Covariant Vectors 30 Chapter –III …………………………………………………………….. Page: 33-38 3.1 Eigen Values and Eigen Vectors 33 Chapter –IV ............................................................................................ Page: 39-65 4.1 Introduction 39 4.2 Transformation of an Elastic Body 40 4.3 Linear Transformation or Affine Transformation 42 4.4 Small/Infinitesimal Linear Deformation 44 4.5 Homogeneous Deformation 45 4.6 Pure Deformation and Components of Strain Tensor 52 4.7 Geometrical Interpretation of the Components of Strain 56 4.8 Normal and Tangential Displacement 64 Chapter –V ..……………………………………………………………… Page: 66-99 5.1 Strain Quadric of Cauchy 66 5.2 Strain Components at a point in a rotation of coordinate axes 69 5.3 Principal Strains and Invariants 71 5.4 General Infinitesimal Deformation 82 5.5 Saint-Venant’s
    [Show full text]
  • A Code Generator for High-Performance Tensor Contractions on Gpus
    A Code Generator for High-Performance Tensor Contractions on GPUs Jinsung Kim∗, Aravind Sukumaran-Rajam∗, Vineeth Thumma∗, Sriram Krishnamoorthy†, Ajay Panyala†, Louis-Noel¨ Pouchet‡, Atanas Rountev∗, P. Sadayappan∗ ∗The Ohio State University, Ohio, USA {kim.4232, sukumaranrajam.1, thumma.6, rountev.1, sadayappan.1}@osu.edu †Pacific Northwest National Laboratory, Washington, USA {sriram, ajay.panyala}@pnnl.gov ‡Colorado State University, Colorado, USA [email protected] Abstract—Tensor contractions are higher dimensional general- The common solution in use today for performing tensor izations of matrix-matrix multiplication. They form the compute- contractions is to first perform suitable index permutation, i.e., intensive core of many applications in computational science layout transformation of one or both input tensors, to move and data science. In this paper, we describe a high-performance GPU code generator for arbitrary tensor contractions. It exploits all contraction indices to the left end or the right end, so domain-specific properties about data reuse in tensor contrac- that the needed sums of products for the contraction can be tions to devise an effective code generation schema, coupled with performed using matrix-matrix multiplication. A further final an effective model-driven search, to determine parameters for index permutation of the result from the matrix multiplication mapping of computation to threads and staging of data through may also be required. However, as elaborated in the next the GPU memory hierarchy. Experimental evaluation using a set of tensor contraction benchmarks demonstrates performance section, this approach, referred to as the TTGT (Transpose- improvement and/or significantly reduced code generation time Transpose-GEMM-Transpose), incurs several disadvantages.
    [Show full text]