DOCSLIB.ORG

Communication-Optimal Parallel 2.5D Matrix Multiplication and LU Factorization Algorithms

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Communication-Optimal Parallel 2.5D Matrix Multiplication and LU Factorization Algorithms Introduction 2.5D matrix multiplication 2.5D LU factorization Conclusion Communication-optimal parallel 2.5D matrix multiplication and LU factorization algorithms Edgar Solomonik and James Demmel UC Berkeley September 1st, 2011 Edgar Solomonik and James Demmel 2.5D algorithms 1/ 36 Introduction 2.5D matrix multiplication 2.5D LU factorization Conclusion Outline Introduction Strong scaling 2.5D matrix multiplication Strong scaling matrix multiplication Performing faster at scale 2.5D LU factorization Communication-optimal LU without pivoting Communication-optimal LU with pivoting Conclusion Edgar Solomonik and James Demmel 2.5D algorithms 2/ 36 Introduction 2.5D matrix multiplication Strong scaling 2.5D LU factorization Conclusion Solving science problems faster Parallel computers can solve bigger problems I weak scaling Parallel computers can also solve a xed problem faster I strong scaling Obstacles to strong scaling I may increase relative cost of communication I may hurt load balance Edgar Solomonik and James Demmel 2.5D algorithms 3/ 36 Introduction 2.5D matrix multiplication Strong scaling 2.5D LU factorization Conclusion Achieving strong scaling How to reduce communication and maintain load balance? I reduce communication along the critical path Communicate less I avoid unnecessary communication Communicate smarter I know your network topology Edgar Solomonik and James Demmel 2.5D algorithms 4/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion Strong scaling matrix multiplication Matrix multiplication on BG/P (n=65,536) 100 2.5D MM 2D MM 80 60 40 20 Percentage of machine peak 0 256 512 1024 2048 #nodes Edgar Solomonik and James Demmel 2.5D algorithms 5/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion Blocking matrix multiplication B A B A B A B A Edgar Solomonik and James Demmel 2.5D algorithms 6/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion 2D matrix multiplication [Cannon 69], [Van De Geijn and Watts 97] B A 16 CPUs (4x4) B CPU CPU CPU CPU CPU CPU CPU CPU A CPU CPU CPU CPU CPU CPU CPU CPU B A B A Edgar Solomonik and James Demmel 2.5D algorithms 7/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion 3D matrix multiplication [Agarwal et al 95], [Aggarwal, Chandra, and Snir 90], [Bernsten 89] 64 CPUs (4x4x4) CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU B CPU CPU CPU CPU A CPU CPU CPU CPU CPU CPU CPU CPU B CPU CPU CPU CPU CPU CPU CPU CPU A CPU CPU CPU CPU B CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU A B CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU A 4 copies of matrices Edgar Solomonik and James Demmel 2.5D algorithms 8/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion 2.5D matrix multiplication B 32 CPUs (4x4x2) A CPU CPU CPU CPU CPU CPU CPU CPU B CPU CPU CPU CPU CPU CPU CPU CPU A CPU CPU CPU CPU B CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU A 2 copies of matrices B A Edgar Solomonik and James Demmel 2.5D algorithms 9/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion 2.5D strong scaling n = dimension, p = #processors, c = #copies of data I must satisfy 1 c p1=3 I special case: c = 1 yields 2D algorithm I special case: c = p1=3 yields 3D algorithm cost(2.5D MM(p; c)) = O(n3=p) ops p + O(n2= c ¡ p) words moved q + O( p=c3) messages£ *ignoring log(p) factors Edgar Solomonik and James Demmel 2.5D algorithms 10/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion 2.5D strong scaling n = dimension, p = #processors, c = #copies of data I must satisfy 1 c p1=3 I special case: c = 1 yields 2D algorithm I special case: c = p1=3 yields 3D algorithm cost(2D MM(p)) = O(n3=p) ops p + O(n2= p) words moved p + O( p) messages£ = cost(2.5D MM(p; 1)) *ignoring log(p) factors Edgar Solomonik and James Demmel 2.5D algorithms 11/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion 2.5D strong scaling n = dimension, p = #processors, c = #copies of data I must satisfy 1 c p1=3 I special case: c = 1 yields 2D algorithm I special case: c = p1=3 yields 3D algorithm cost(2.5D MM(c ¡ p; c)) = O(n3=(c ¡ p)) ops p + O(n2=(c ¡ p)) words moved p + O( p=c) messages = cost(2D MM(p))=c perfect strong scaling Edgar Solomonik and James Demmel 2.5D algorithms 12/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion 2.5D MM on 65,536 cores Matrix multiplication on 16,384 nodes of BG/P 100 2.5D MM 2D MM 80 2.7X faster 60 Using c=16 matrix copies 40 12X faster 20 Percentage of machine peak 0 8192 131072 n Edgar Solomonik and James Demmel 2.5D algorithms 13/ 36 Introduction 2.5D matrix multiplication Strong scaling matrix multiplication 2.5D LU factorization Performing faster at scale Conclusion Cost breakdown of MM on 65,536 cores Matrix multiplication on 16,384 nodes of BG/P 1.4 communication 1.2 idle 95% reduction in comm computation 1 0.8 0.6 0.4 0.2 Execution time normalized by 2D 0 n=8192, 2Dn=8192, 2.5D n=131072, 2Dn=131072, 2.5D Edgar Solomonik and James Demmel 2.5D algorithms 14/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2.5D LU strong scaling (without pivoting) LU without pivoting on BG/P (n=65,536) 100 ideal scaling 2.5D LU 2D LU 80 60 40 20 Percentage of machine peak 0 256 512 1024 2048 #nodes Edgar Solomonik and James Demmel 2.5D algorithms 15/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2D blocked LU factorization A Edgar Solomonik and James Demmel 2.5D algorithms 16/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2D blocked LU factorization U₀₀ L₀₀ Edgar Solomonik and James Demmel 2.5D algorithms 17/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2D blocked LU factorization U L Edgar Solomonik and James Demmel 2.5D algorithms 18/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2D blocked LU factorization U L S=A-LU Edgar Solomonik and James Demmel 2.5D algorithms 19/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2D block-cyclic decomposition 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 Edgar Solomonik and James Demmel 2.5D algorithms 20/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2D block-cyclic LU factorization Edgar Solomonik and James Demmel 2.5D algorithms 21/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2D block-cyclic LU factorization U L Edgar Solomonik and James Demmel 2.5D algorithms 22/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2D block-cyclic LU factorization U L S=A-LU Edgar Solomonik and James Demmel 2.5D algorithms 23/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2.5D LU factorization U₀₀ L₀₀ U₀₀ U₀₃ L₀₀ U₀₀ U₀₃ L₀₀ U₀₀ U₀₁ L₀₀ (A) L₃₀ L₂₀ L₁₀ Edgar Solomonik and James Demmel 2.5D algorithms 24/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2.5D LU factorization U₀₀ L₀₀ (B) U₀₀ U₀₃ L₀₀ U₀₀ U₀₃ L₀₀ U₀₀ U₀₁ L₀₀ (A) L₃₀ L₂₀ L₁₀ Edgar Solomonik and James Demmel 2.5D algorithms 25/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2.5D LU factorization U₀₀ L₀₀ (B) U₀₀ U₀₃ L₀₀ U₀₀ U₀₃ L₀₀ U₀₀ U₀₁ L₀₀ (A) L₃₀ L₂₀ L₁₀ U (D) (C) L Edgar Solomonik and James Demmel 2.5D algorithms 26/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2.5D LU factorization Look at how this update is distributed. What does it remind you of? Edgar Solomonik and James Demmel 2.5D algorithms 27/ 36 Introduction 2.5D matrix multiplication Communication-optimal LU without pivoting 2.5D LU factorization Communication-optimal LU with pivoting Conclusion 2.5D LU factorization Look at how this update is distributed.
Load more

Recommended publications
  • Scaling Algorithms and Tropical Methods in Numerical Matrix Analysis
    Scaling Algorithms and Tropical Methods in Numerical Matrix Analysis: Application to the Optimal Assignment Problem and to the Accurate Computation of Eigenvalues Meisam Sharify To cite this version: Meisam Sharify. Scaling Algorithms and Tropical Methods in Numerical Matrix Analysis: Application to the Optimal Assignment Problem and to the Accurate Computation of Eigenvalues. Numerical Analysis [math.NA]. Ecole Polytechnique X, 2011. English. pastel-00643836 HAL Id: pastel-00643836 https://pastel.archives-ouvertes.fr/pastel-00643836 Submitted on 24 Nov 2011 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Th`esepr´esent´eepour obtenir le titre de DOCTEUR DE L'ECOLE´ POLYTECHNIQUE Sp´ecialit´e: Math´ematiquesAppliqu´ees par Meisam Sharify Scaling Algorithms and Tropical Methods in Numerical Matrix Analysis: Application to the Optimal Assignment Problem and to the Accurate Computation of Eigenvalues Jury Marianne Akian Pr´esident du jury St´ephaneGaubert Directeur Laurence Grammont Examinateur Laura Grigori Examinateur Andrei Sobolevski Rapporteur Fran¸coiseTisseur Rapporteur Paul Van Dooren Examinateur September 2011 Abstract Tropical algebra, which can be considered as a relatively new field in Mathemat- ics, emerged in several branches of science such as optimization, synchronization of production and transportation, discrete event systems, optimal control, oper- ations research, etc.
    [Show full text]
  • THESIS MULTIDIMENSIONAL SCALING: INFINITE METRIC MEASURE SPACES Submitted by Lara Kassab Department of Mathematics in Partial Fu
    THESIS MULTIDIMENSIONAL SCALING: INFINITE METRIC MEASURE SPACES Submitted by Lara Kassab Department of Mathematics In partial fulfillment of the requirements For the Degree of Master of Science Colorado State University Fort Collins, Colorado Spring 2019 Master’s Committee: Advisor: Henry Adams Michael Kirby Bailey Fosdick Copyright by Lara Kassab 2019 All Rights Reserved ABSTRACT MULTIDIMENSIONAL SCALING: INFINITE METRIC MEASURE SPACES Multidimensional scaling (MDS) is a popular technique for mapping a finite metric space into a low-dimensional Euclidean space in a way that best preserves pairwise distances. We study a notion of MDS on infinite metric measure spaces, along with its optimality properties and goodness of fit. This allows us to study the MDS embeddings of the geodesic circle S1 into Rm for all m, and to ask questions about the MDS embeddings of the geodesic n-spheres Sn into Rm. Furthermore, we address questions on convergence of MDS. For instance, if a sequence of metric measure spaces converges to a fixed metric measure space X, then in what sense do the MDS embeddings of these spaces converge to the MDS embedding of X? Convergence is understood when each metric space in the sequence has the same finite number of points, or when each metric space has a finite number of points tending to infinity. We are also interested in notions of convergence when each metric space in the sequence has an arbitrary (possibly infinite) number of points. ii ACKNOWLEDGEMENTS We would like to thank Henry Adams, Mark Blumstein, Bailey Fosdick, Michael Kirby, Henry Kvinge, Facundo Mémoli, Louis Scharf, the students in Michael Kirby’s Spring 2018 class, and the Pattern Analysis Laboratory at Colorado State University for their helpful conversations and support throughout this project.
    [Show full text]
  • Affine Transformations  Foley, Et Al, Chapter 5.1-5.5
    Reading Optional reading: Angel 4.1, 4.6-4.10 Angel, the rest of Chapter 4 Affine transformations Foley, et al, Chapter 5.1-5.5. David F. Rogers and J. Alan Adams, Mathematical Elements for Computer Graphics, 2nd Ed., McGraw-Hill, New York, Daniel Leventhal 1990, Chapter 2. Adapted from Brian Curless CSE 457 Autumn 2011 1 2 Linear Interpolation More Interpolation 3 4 1 Geometric transformations Vector representation Geometric transformations will map points in one We can represent a point, p = (x,y), in the plane or space to points in another: (x', y‘, z‘ ) = f (x, y, z). p=(x,y,z) in 3D space These transformations can be very simple, such as column vectors as scaling each coordinate, or complex, such as non-linear twists and bends. We'll focus on transformations that can be represented easily with matrix operations. as row vectors 5 6 Canonical axes Vector length and dot products 7 8 2 Vector cross products Inverse & Transpose 9 10 Two-dimensional Representation, cont. transformations We can represent a 2-D transformation M by a Here's all you get with a 2 x 2 transformation matrix matrix M: If p is a column vector, M goes on the left: So: We will develop some intimacy with the If p is a row vector, MT goes on the right: elements a, b, c, d… We will use column vectors. 11 12 3 Identity Scaling Suppose we choose a=d=1, b=c=0: Suppose we set b=c=0, but let a and d take on any positive value: Gives the identity matrix: Gives a scaling matrix: Provides differential (non-uniform) scaling in x and y: Doesn't move the points at all 1 2 12 x y 13 14 ______________ ____________ Suppose we keep b=c=0, but let either a or d go Now let's leave a=d=1 and experiment with b.
    [Show full text]
  • 1 Review of Inner Products 2 the Approximation Problem and Its Solution Via Orthogonality
    Approximation in inner product spaces, and Fourier approximation Math 272, Spring 2018 Any typographical or other corrections about these notes are welcome. 1 Review of inner products An inner product space is a vector space V together with a choice of inner product. Recall that an inner product must be bilinear, symmetric, and positive definite. Since it is positive definite, the quantity h~u;~ui is never negative, and is never 0 unless ~v = ~0. Therefore its square root is well-defined; we define the norm of a vector ~u 2 V to be k~uk = ph~u;~ui: Observe that the norm of a vector is a nonnegative number, and the only vector with norm 0 is the zero vector ~0 itself. In an inner product space, we call two vectors ~u;~v orthogonal if h~u;~vi = 0. We will also write ~u ? ~v as a shorthand to mean that ~u;~v are orthogonal. Because an inner product must be bilinear and symmetry, we also obtain the following expression for the squared norm of a sum of two vectors, which is analogous the to law of cosines in plane geometry. k~u + ~vk2 = h~u + ~v; ~u + ~vi = h~u + ~v; ~ui + h~u + ~v;~vi = h~u;~ui + h~v; ~ui + h~u;~vi + h~v;~vi = k~uk2 + k~vk2 + 2 h~u;~vi : In particular, this gives the following version of the Pythagorean theorem for inner product spaces. Pythagorean theorem for inner products If ~u;~v are orthogonal vectors in an inner product space, then k~u + ~vk2 = k~uk2 + k~vk2: Proof.
    [Show full text]
  • A Geometric Model of Allometric Scaling
    1 Site model of allometric scaling and fractal distribution networks of organs Walton R. Gutierrez* Touro College, 27West 23rd Street, New York, NY 10010 *Electronic address: [email protected] At the basic geometric level, the distribution networks carrying vital materials in living organisms, and the units such as nephrons and alveoli, form a scaling structure named here the site model. This unified view of the allometry of the kidney and lung of mammals is in good agreement with existing data, and in some instances, improves the predictions of the previous fractal model of the lung. Allometric scaling of surfaces and volumes of relevant organ parts are derived from a few simple propositions about the collective characteristics of the sites. New hypotheses and predictions are formulated, some verified by the data, while others remain to be tested, such as the scaling of the number of capillaries and the site model description of other organs. I. INTRODUCTION Important Note (March, 2004): This article has been circulated in various forms since 2001. The present version was submitted to Physical Review E in October of 2003. After two rounds of reviews, a referee pointed out that aspects of the site model had already been discovered and proposed in a previous article by John Prothero (“Scaling of Organ Subunits in Adult Mammals and Birds: a Model” Comp. Biochem. Physiol. 113A.97-106.1996). Although modifications to this paper would be required, it still contains enough new elements of interest, such as (P3), (P4) and V, to merit circulation. The q-bio section of www.arxiv.org may provide such outlet.
    [Show full text]
  • Eigenvalues, Eigenvectors and the Similarity Transformation
    Eigenvalues, Eigenvectors and the Similarity Transformation Eigenvalues and the associated eigenvectors are ‘special’ properties of square matrices. While the eigenvalues parameterize the dynamical properties of the system (timescales, resonance properties, amplification factors, etc) the eigenvectors define the vector coordinates of the normal modes of the system. Each eigenvector is associated with a particular eigenvalue. The general state of the system can be expressed as a linear combination of eigenvectors. The beauty of eigenvectors is that (for square symmetric matrices) they can be made orthogonal (decoupled from one another). The normal modes can be handled independently and an orthogonal expansion of the system is possible. The decoupling is also apparent in the ability of the eigenvectors to diagonalize the original matrix, A, with the eigenvalues lying on the diagonal of the new matrix, . In analogy to the inertia tensor in mechanics, the eigenvectors form the principle axes of the solid object and a similarity transformation rotates the coordinate system into alignment with the principle axes. Motion along the principle axes is decoupled. The matrix mechanics is closely related to the more general singular value decomposition. We will use the basis sets of orthogonal eigenvectors generated by SVD for orbit control problems. Here we develop eigenvector theory since it is more familiar to most readers. Square matrices have an eigenvalue/eigenvector equation with solutions that are the eigenvectors xand the associated eigenvalues : Ax = x The special property of an eigenvector is that it transforms into a scaled version of itself under the operation of A. Note that the eigenvector equation is non-linear in both the eigenvalue () and the eigenvector (x).
    [Show full text]
  • Scaling Symmetry Meets Topology ⇑ Pengfei Zhang, Hui Zhai
    Science Bulletin 64 (2019) 289–290 Contents lists available at ScienceDirect Science Bulletin journal homepage: www.elsevier.com/locate/scib Research Highlight Scaling symmetry meets topology ⇑ Pengfei Zhang, Hui Zhai Institute for Advanced Study, Tsinghua University, Beijing 100084, China In 1970, Vitaly Efimov found an interesting phenomenon in a also been proposed and observed in ultracold atomic gases [7]. quantum three-body problem, which is now known as the Efimov Very recently, it has also been pointed out that the Efimov effect effect [1]. Efimov found that when the two-body interaction poten- can be linked to the fractal behavior in the time domain [8,9]. Nev- tial is short-ranged and is tuned to the vicinity of an s-wave reso- ertheless, these studies have so far been limited to atomic and nance, an infinite number of three-body bound states emerge and nuclear systems. The recently published experiment by Wang their eigenenergies En form a geometric sequence as et al. [12] brings the Efimov physics into condensed matter system En ¼jE0j expð2pn=s0Þ, where s0 is a universal constant. This by observing such a universal discrete scaling symmetry in the effect is directly related to a symmetry called the scaling symme- topological semi-metals. The topological semi-metal has received try. In short, the three-body problem with a resonant two-body considerable attentions in the past decades. The interplay between interacting potential can be reduced to the following one-dimen- this discrete scaling symmetry and topology introduces a new sional Schrödinger equation as twist to the physics of topological material and will make its ! physics even richer.
    [Show full text]
  • Computer Graphics Programming: Matrices and Transformations Outline
    Computer Graphics Programming: Matrices and Transformations Outline • Computer graphics overview • Object /Geometry modlideling • 2D modeling transformations and matrices • 3D modeling transformations and matrices • Relevant Unity scripting features Computer Graphics • Algorithmically generating a 2D image from 3D data (models, textures, lighting) • Also called rendering • Raster graphics – Array of pixels – About 25x25 in the example ‐> • Algorithm tradeoffs: – Computation time – Memory cost – Image quality Computer Graphics • The graphics pipeline is a series of conversions of points into different coordinate systems or spaces Computer Graphics • Virtual cameras in Unity will handle everything from the viewing transformation on OpenGL specifying geometry Legacy syntax example: glBegin(GL_POLYGON); glVertex2f(-0.5, -0.5); glVertex2f(-0.5, 0.5); glVertex2f(0. 5, 05);0.5); glVertex2f(0.5, -0.5); glEnd(); Unity specifying geometry – Mesh class • Requires two types of values – Vertices (specified as an array of 3D points) – Triangles (specified as an array of Vector3s whose values are indices in the vertex array) • Documentation and Example – http://docs.unityy/3d.com/Documentation ///Manual/Ge neratingMeshGeometryProcedurally.html – http://docs.unity3d.com/Documentation/ScriptRefere nce/Mesh.html • The code on the following slides is attached to a cube game object (rather than an EmptyObject) Mesh pt. 1 – assign vertices Mesh mesh = new Mesh(); gameObject.GetComponent<MeshFilter>().mesh = mesh; Vector3[] vertices = new Vector3[4]; vertices[0]
    [Show full text]
  • CMSC427 Transformations I
    CMSC427 Transformations I Credit: slides 9+ from Prof. Zwicker Transformations: outline • Types of transformations • Specific: translation, rotation, scaling, shearing • Classes: rigid, affine, projective • Representing transformations • Unifying representation with homogeneous coordinates • Transformations represented as matrices • Composing transformations • Sequencing matrices • Sequencing using OpenGL stack model • Transformation examples • Rotating or scaling about a point • Rotating to a new coordinate frame • Applications • Modeling transformations (NOW) • Viewing transformations (LATER) Modeling with transformations • Create instance of • Object coordinate object in object space coordinate space • Create circle at origin • Transform object to • World coordinate space world coordinate space • Scale by 1.5 • Move down by 2 unit • Do so for other objects • Two rects make hat • Three circles make body • Two lines make arms Classes of transformations • Rigid • Affine • Translate, rotate, • Translate, rotate, scale uniform scale (non-uniform), shear, • No distortion to object reflect • Limited distortions • Preserve parallel lines Classes of transformations • Affine • Projective • Preserves parallel lines • Foreshortens • Lines converge • For viewing/rendering Classes of transformations: summary • Affine • Reshape, size object • Rigid • Place, move object • Projective • View object • Later … • Non-linear, arbitrary • Twists, pinches, pulls • Not in this unit First try: scale and rotate vertices in vector notation • Scale a point p by
    [Show full text]
  • CSCI 445: Computer Graphics
    CSCI 445: Computer Graphics Qing Wang Assistant Professor at Computer Science Program The University of Tennessee at Martin Angel and Shreiner: Interactive Computer Graphics 7E © 1 Addison-Wesley 2015 Classical Viewing Angel and Shreiner: Interactive Computer Graphics 7E © 2 Addison-Wesley 2015 Objectives • Introduce the classical views • Compare and contrast image formation by computer with how images have been formed by architects, artists, and engineers • Learn the benefits and drawbacks of each type of view Angel and Shreiner: Interactive Computer Graphics 7E © 3 Addison-Wesley 2015 Classical Viewing • Viewing requires three basic elements • One or more objects • A viewer with a projection surface • Projectors that go from the object(s) to the projection surface • Classical views are based on the relationship among these elements • The viewer picks up the object and orients it how she would like to see it • Each object is assumed to constructed from flat principal faces • Buildings, polyhedra, manufactured objects Angel and Shreiner: Interactive Computer Graphics 7E © 4 Addison-Wesley 2015 Planar Geometric Projections • Standard projections project onto a plane • Projectors are lines that either • converge at a center of projection • are parallel • Such projections preserve lines • but not necessarily angles • Nonplanar projections are needed for applications such as map construction Angel and Shreiner: Interactive Computer Graphics 7E © 5 Addison-Wesley 2015 Classical Projections Angel and Shreiner: Interactive Computer Graphics 7E
    [Show full text]
  • Ellipsoid Method for Linear Programming Made Simple
    Ellipsoid Method for Linear Programming made simple Sanjeev Saxena∗ Dept. of Computer Science and Engineering, Indian Institute of Technology, Kanpur, INDIA-208 016 June 17, 2018 Abstract In this paper, ellipsoid method for linear programming is derived using only minimal knowledge of algebra and matrices. Unfortunately, most authors first describe the algorithm, then later prove its correctness, which requires a good knowledge of linear algebra. 1 Introduction Ellipsoid method was perhaps the first polynomial time method for linear programming[4]. However, it is hardly ever covered in Computer Science courses. In fact, many of the existing descriptions[1, 6, 3] first describe the algorithm, then later prove its correctness. Moreover, to understand one require a good knowledge of linear algebra (like properties of semi- definite matrices and Jacobean)[1, 6, 3, 2]. In this paper, ellipsoid method arXiv:1712.04637v1 [cs.DS] 13 Dec 2017 for linear programming is derived using only minimal knowledge of algebra and matrices. We are given a set of linear equations Ax ≥ B and have to find a feasible point. Ellipsoid method can check whether the system Ax ≥ B has a solution or not, and find a solution if one is present. ∗E-mail: [email protected] 1 The algorithm generates a sequence of ellipsoids[1], E0;E1; : : :; with cen- tres x0; x1;::: such that the solution space (if there is a feasible solution) is is inside each of these ellipsoids. If the centre xi of the (current) ellipsoid is not feasible, then some constraint (say) aT x ≥ b is violated (for some row T a of A), i.e., a xi < b.
    [Show full text]
  • Affine Transformations and Rotations
    CMSC 425 Dave Mount & Roger Eastman CMSC 425: Lecture 6 Affine Transformations and Rotations Affine Transformations: So far we have been stepping through the basic elements of geometric programming. We have discussed points, vectors, and their operations, and coordinate frames and how to change the representation of points and vectors from one frame to another. Our next topic involves how to map points from one place to another. Suppose you want to draw an animation of a spinning ball. How would you define the function that maps each point on the ball to its position rotated through some given angle? We will consider a limited, but interesting class of transformations, called affine transfor- mations. These include (among others) the following transformations of space: translations, rotations, uniform and nonuniform scalings (stretching the axes by some constant scale fac- tor), reflections (flipping objects about a line) and shearings (which deform squares into parallelograms). They are illustrated in Fig. 1. rotation translation uniform nonuniform reflection shearing scaling scaling Fig. 1: Examples of affine transformations. These transformations all have a number of things in common. For example, they all map lines to lines. Note that some (translation, rotation, reflection) preserve the lengths of line segments and the angles between segments. These are called rigid transformations. Others (like uniform scaling) preserve angles but not lengths. Still others (like nonuniform scaling and shearing) do not preserve angles or lengths. Formal Definition: Formally, an affine transformation is a mapping from one affine space to another (which may be, and in fact usually is, the same space) that preserves affine combi- nations.
    [Show full text]