FR2840705A1 - Load balancing method for a parallel-connected PC cluster used in processing a single digital video image, whereby objects in the image are divided into sub-objects in such a way that duplicated processing is minimized - Google Patents

Load balancing method for a parallel-connected PC cluster used in processing a single digital video image, whereby objects in the image are divided into sub-objects in such a way that duplicated processing is minimized Download PDF

Info

Publication number
FR2840705A1
FR2840705A1 FR0206943A FR0206943A FR2840705A1 FR 2840705 A1 FR2840705 A1 FR 2840705A1 FR 0206943 A FR0206943 A FR 0206943A FR 0206943 A FR0206943 A FR 0206943A FR 2840705 A1 FR2840705 A1 FR 2840705A1
Authority
FR
France
Prior art keywords
image
objects
soul
processing
elementary
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
FR0206943A
Other languages
French (fr)
Inventor
Alexis Vartanian
Olivier Bedelet
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
ARTABEL
Original Assignee
ARTABEL
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by ARTABEL filed Critical ARTABEL
Priority to FR0206943A priority Critical patent/FR2840705A1/en
Publication of FR2840705A1 publication Critical patent/FR2840705A1/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T15/003D [Three Dimensional] image rendering
    • G06T15/005General purpose rendering architectures

Abstract

Load balancing method for a parallel-connected PC cluster, with each PC processing a zone or region (R1-Rn) of a digital image, has the following steps: dividing an initial image (I) into a number of regions and meshing of objects in the image using a number of geometric elementary shapes. Additionally for each image: each object is decomposed into a number of sub-objects (SOi, SOj, SOk) each obtained by grouping elementary shapes (Fi), the number of sub-object intersections with the object boundary are determined and the elementary shapes are rearrangement to minimize the number of intersections. An Independent claim is made for a device for balancing the load on a parallel cluster of PCs used in processing a digital image.

Description

op est l'operateur algorithmique.op is the algorithmic operator.

- 1 À< Procede et dispositif d'equilibrage de charge au sein d'un ensemble (cluster) diordinateurs utilises en parallele pour des applications graphiques >' La presente invention concerne un procede d'equilibrage de charge au sein d'un ensemble (cluster) d'ordinateurs utilises en parallele pour des applications graphiques. Wile vise egalement un dispositif pour la mise en  A method and device for load balancing within a cluster used in parallel for graphics applications The present invention relates to a method of load balancing within a cluster (cluster). ) computers used in parallel for graphical applications. Wile also aims at a device for the implementation

ceuvre de ce procede.this process.

- - II existe deja des techniques d'equilibrage de charge au sein d'un cluster d'ordinateurs paralleles utilises pour traiter des portions diune image en temps reel. A chaque ordinateur est associee une portion ou region d'une meme image. Chaque image comporte un ou plusieurs objets, chaque  There are already load balancing techniques within a parallel computer cluster used to process portions of an image in real time. Each computer is associated with a portion or region of the same image. Each image has one or more objects, each

objet pouvant etre decompose en soul-objets.  object that can be broken down into soul-objects.

Une technique existante d'equilibrage est basee sur le deplacement de frontiere entre les differentes regions. Mais on observe un effet de redondance de calculs effectues par plusieurs ordinateurs affectes a des regions adjacentes d'une meme image du fait qu'un meme soul-objet est ainsi calcule plusieurs fois, ce qui occasionne un temps de calcul eleve qui peut s'averer dans certaines configurations prohibitif pour une utilisation  An existing balancing technique is based on border displacement between the different regions. But there is an effect of redundancy calculations made by several computers assigned to adjacent regions of the same image because a same soul-object is thus calculated several times, which causes a high calculation time that can warn in certain prohibitive configurations for use

graphique en temps reel.real time graph.

Le but de la presente invention est de resoudre cet inconvenient en proposant un procede d'equilibrage de charge qui permette de reduire la probabilite qu'un meme objet ou soul-objet soit traite par plusieurs machines. Cet objectif est atteint avec un procede pour equilibrer la charge d'un ensemble (cluster) d'ordinateurs utilises en parallele et affectes chacun au calcul diune zone ou region d'une image numerique, comprenant pour chaque image, un decoupage prealable de ['image en plusieurs regions a chacune desquelles un ordinateur est affecte, et un maillage du ou des objets presents dans ['image selon une pluralite de formes geometriques  The object of the present invention is to solve this disadvantage by proposing a load balancing method which makes it possible to reduce the probability that a same object or soul-object is treated by several machines. This objective is achieved with a method for balancing the load of a set (cluster) of computers used in parallel and each assigned to the calculation of a zone or region of a digital image, comprising for each image, a preliminary decoupage of image in several regions each of which a computer is assigned, and a mesh of the object or objects present in the image according to a plurality of geometric shapes

3 o elementaires.3 o elementary.

- 2 Suivant ['invention, le procede d'equilibrage de charge comprend en outre, pour chaque image: - une decomposition audit objet en une pluralite de soul-objets obtenus chacun par regroupement de formes geometriques elementaires, - un calcul de l'empreinte de cheque soul- objet comprenant une determination du nombre d'intersections audit soul- objet avec les frontieres de regions, et - un ajustement des regroupements de formes geometriques elementaires de  According to the invention, the method of load balancing further comprises, for each image: a decomposition to said object in a plurality of soul-objects each obtained by grouping elementary geometric shapes, a calculation of the imprint a cross-reference check comprising a determination of the number of intersections to said subject-object with region boundaries, and - an adjustment of the groupings of elementary geometrical forms of

fa,con a minimiser le nombre d'intersections.  to minimize the number of intersections.

Ainsi, avec le procede d'equilibrage de charge selon ['invention, on reduit significativement la probabilite de redondance puisque les soulobjets vent construits de fa,con a n'etre inclus que dans une seule region et ne vent done  Thus, with the load balancing method according to the invention, the likelihood of redundancy is significantly reduced since the wind loads are constructed so that they are included in only one region and therefore do not wind up.

traites que par la machine affectee a cette region determinee.  processed only by the machine assigned to that particular area.

Les formes geometriques elementaires peuvent comprendre des polyedres. D'autres avantages et caracteristiques de ['invention apparaftront a  The elementary geometric shapes may include polyhedra. Other advantages and features of the invention will be apparent from

l'examen de la description detaillee d'un mode de mise en aeuvre nullement  the examination of the detailed description of a mode of implementation by no means

limitatif, et des dessins annexes sur lesquels: - la figure 1 represente schematiquement une image traitee avec le procede d'equilibrage de charge selon ['invention; et - la figure 2 illustre schematiquement un cluster d'ordinateurs pour lequel le  and FIG. 1 schematically represents an image processed with the load balancing method according to the invention; and - Figure 2 schematically illustrates a computer cluster for which the

procede d'equilibrage de charge est mis en aeuvre.  load balancing method is implemented.

On va maintenant decrire, en reference aux figures precitees, un exemple de mise en cauvre du procede d'equilibrage selon ['invention, en  We will now describe, with reference to the above figures, an example of setting the balance of the balancing process according to the invention in

meme temps que le dispositif permettant cette mise en oeuvre.  same time as the device allowing this implementation.

On considere un cluster C d'ordinateurs M1,,Mi, Mn pilote par une unite de pilotage U a laquelle vent connectes des peripheriques et equipements informatiques au nombre desquels un ecran E. L'unite de pilotage U est adaptee pour exploiter un traitement graphique d'images numeriques en affectant a chacune des regions d'une meme image un 3 0 ordinateur du cluster C. L'unite U inclut notamment un dispositif D d'equilibrage de la charge de calcul assuree par chacun des ordinateurs du cluster. Ce dispositif D est concu pour mettre en couvre le procede  A cluster C of computers M1,, Mi, Mn driven by a control unit U is connected to which connected devices and computer equipment, including a screen E. The pilot unit U is adapted to operate a graphics processing. digital image by assigning to each of the regions of the same image a computer of the cluster C. The unit U includes in particular a D device for balancing the computing load provided by each of the computers in the cluster. This device D is designed to cover the process

d'equilibrage de charge selon ['invention.  load balancing system according to the invention.

Pour chaque image numerique 1, un decoupage prealable est effectue de facon 3 fournir un ensemble de regions R1,..,Ri,..Rn, par exemple sous la forme de rectangles de taille reguliere. Chaque objet O present dans ['image I est decompose en un ensemble d'objets elementaires ou soulobjets SO1,...SO\,...,SON obtenus par agregation ou regroupement de formes geometriques elementaires Fi,  For each digital image 1, a preliminary cut is performed so as to provide a set of regions R1, .., Ri, .. Rn, for example in the form of rectangles of regular size. Each object O present in ['image I is decomposed into a set of elementary objects or object objects SO1,... SO \,..., SON obtained by aggregation or grouping of elementary geometric forms Fi,

par exemple des triangles elementaires.  for example elementary triangles.

Le dispositif d'equilibrage de charge D calcule alors ltempreinte de chaque soul-objet sur liecran, definie ici comme un rectangle englobant la projection du soul-objet sur l'ecran, et determine le nombre diintersections  The load balancing device D then calculates the fingerprint of each soul-object on the screen, defined here as a rectangle encompassing the projection of the soul-object on the screen, and determines the number of intersections.

de ce soul-objet avec les frontieres des regions de ['image.  of this soul-object with the borders of the regions of the image.

Les regroupements des formes geometriques elementaires vent ajustees de facon a minimiser le nombre d'intersections pour chacun des soul-objets. On obtient alors un ensemble optimise de soul-objets completement inclus dans une region unique et done traites chacun par un  The groupings of the basic geometric shapes are adjusted to minimize the number of intersections for each of the soul-objects. We then obtain an optimized set of soul-objects completely included in a single region and thus each treated by a

seul ordinateur.only computer.

Bien sGr, I'invention n'est pas limitee aux exemples qui viennent d'etre decrits et de nombreux amenagements peuvent etre apportes a ces  Of course, the invention is not limited to the examples just described and many improvements can be made to these examples.

exemples sans sortir du cadre de ['invention.  examples without departing from the scope of the invention.

- 4 -- 4 -

Claims (5)

REVENDICATIONS 1. Procede pour equilibrer la charge d'un ensemble (cluster) (C) d'ordinateurs (M 1,..,Mi,..Mn) utilises en parallele et affectes chacun au calcul d'une zone ou region (R1,..,Ri,..Rn) diune image numerique (I), comprenant pour chaque image: - un decoupage prealable de ['image (I) en plusieurs regions (R1,..,Ri,..,Rn) a chacune desquelles un ordinateur est affecte, - un maillage du ou des objets (O) presents dans ['image (I) selon une pluralite de formes geometriques elementaires (F), caracterise en ce qu'il comprend en outre, pour chaque image: - une decomposition audit objet (O) en une pluralite de soul-objets (SOi, SOj, SO k) obtenus chacun par reg roupement de formes geometriques elementaires (F), - un calcul de ltempreinte de chaque soul-objet comprenant une determination du nombre d'intersections audit soul-objet avec les frontieres de regions, et - un ajustement des regroupements de formes geometriques  1. Method for balancing the load of a set (cluster) (C) of computers (M 1, .., Mi, .. Mn) used in parallel and each assigned to the calculation of a zone or region (R1, .., Ri, .. Rn) diune digital image (I), comprising for each image: - a preliminary decoupage of ['image (I) in several regions (R1, .., Ri, .., Rn) each of which a computer is assigned, - a mesh of the object or objects (O) present in ['image (I) according to a plurality of elementary geometrical shapes (F), characterized in that it further comprises, for each image: a decomposition to said object (O) into a plurality of soul-objects (SOi, SOj, SO k) each obtained by a group of elementary geometric shapes (F), - a calculation of the fingerprint of each soul-object comprising a determination of the number of intersections auditing the soul-object with the boundaries of regions, and - an adjustment of the groupings of geometric forms elementaires de facon a minimiser le nombre d'intersections.  in order to minimize the number of intersections. 2. Procede selon la revendication 1, caracterise en ce que les formes  2. Method according to claim 1, characterized in that the forms geometriques elementaires comprennent des polyedres.  Basic geometrics include polyhedra. 3. Procede selon l'une des revendications 1 ou 2, caracterise en ce que les  3. Method according to one of claims 1 or 2, characterized in that regions vent de forme rectangulaire.  regions of rectangular shape. 4. Dispositif (D) pour equilibrer la charge d'un ensemble (C) (cluster) d'ordinateurs (M 1,..,Mi,..,Mn) utilises en parallele et affectes chacun au calcul d'une zone ou region (Ri) d'une image numerique (I), comprenant: - 5 - - des moyens pour realiser pour chaque image un decoupage prealable de ladite image en plusieurs regions 3 chacune desquelles un ordinateur est affecte, - des moyens pour realiser un maillage du ou des objets presents dans ['image selon une pluralite de formes geometriques elementaires, caracterise en ce qu'il comprend en outre: - des moyens pour realiser une decomposition audit objet en une pluralite de soul-objets obtenus chacun par regroupement de formes geometriques elementaires, - des moyens pour calculer l'empreinte de chaque soul-objet comprenant une determination du nombre d'intersections audit soul-objet avec les frontieres de regions, et - des moyens pour ajuster les regroupements de formes geometriques elementaires de facon a minimiser le nombre d'intersections.  4. Device (D) for balancing the load of a set (C) (cluster) of computers (M 1, .., Mi, .., Mn) used in parallel and each assigned to the calculation of a zone or region (Ri) of a digital image (I), comprising: - means for realizing for each image a preliminary cutting of said image into several regions 3 each of which a computer is affected, - means for making a mesh of the object or objects present in the image according to a plurality of elementary geometric shapes, characterized in that it further comprises: means for performing a decomposition to said object in a plurality of soul-objects each obtained by grouping geometric shapes elementary means for calculating the footprint of each soul-object comprising a determination of the number of intersections to said soul-object with the region borders, and means for adjusting the groupings of elementary geometric shapes so as to minimize the number of inter sections. 5. Application du procede d'equilibrage de charge selon l'une des5. Application of the load balancing method according to one of the revendications 1 a 3, pour un cluster d'ordinateurs adapte au traitement  1 to 3, for a computer cluster adapted to the treatment
FR0206943A 2002-06-06 2002-06-06 Load balancing method for a parallel-connected PC cluster used in processing a single digital video image, whereby objects in the image are divided into sub-objects in such a way that duplicated processing is minimized Pending FR2840705A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
FR0206943A FR2840705A1 (en) 2002-06-06 2002-06-06 Load balancing method for a parallel-connected PC cluster used in processing a single digital video image, whereby objects in the image are divided into sub-objects in such a way that duplicated processing is minimized

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
FR0206943A FR2840705A1 (en) 2002-06-06 2002-06-06 Load balancing method for a parallel-connected PC cluster used in processing a single digital video image, whereby objects in the image are divided into sub-objects in such a way that duplicated processing is minimized

Publications (1)

Publication Number Publication Date
FR2840705A1 true FR2840705A1 (en) 2003-12-12

Family

ID=29558999

Family Applications (1)

Application Number Title Priority Date Filing Date
FR0206943A Pending FR2840705A1 (en) 2002-06-06 2002-06-06 Load balancing method for a parallel-connected PC cluster used in processing a single digital video image, whereby objects in the image are divided into sub-objects in such a way that duplicated processing is minimized

Country Status (1)

Country Link
FR (1) FR2840705A1 (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2864318A1 (en) * 2003-12-23 2005-06-24 Alexis Vartanian Data organization device for use in graphic display system, has device to alternatively allocate variables related to light and color or depth and transparency of designed object in frame buffer based on selection function
WO2005116930A1 (en) * 2004-05-24 2005-12-08 Koninklijke Philips Electronics N.V. Tile based graphics rendering
FR2901049A1 (en) * 2006-05-12 2007-11-16 Techviz Soc Par Actions Simpli ENCODING METHOD AND DISPLAY SYSTEM ON A SCREEN OF A DIGITAL MODEL OF AN OBJECT IN THE FORM OF A SYNTHESIS IMAGE

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5764243A (en) * 1995-03-24 1998-06-09 3Dlabs Inc. Ltd. Rendering architecture with selectable processing of multi-pixel spans

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5764243A (en) * 1995-03-24 1998-06-09 3Dlabs Inc. Ltd. Rendering architecture with selectable processing of multi-pixel spans

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
POUNTAIN D: "PIXEL FLOW: SCALABLE IMAGE PROCESSING", BYTE, MCGRAW-HILL INC. ST PETERBOROUGH, US, vol. 20, no. 11, 1 November 1995 (1995-11-01), pages 249 - 250, XP000535302, ISSN: 0360-5280 *
R. SAMANTA AND T. FUNKHOUSER: "dynamic algorithms for sorting primitives among scree-space tiles in parallel rendering systems", TECHNICAL REPORT, DEPARTMENT OF COMPUTER SCIENCE, PRINCETON, October 1998 (1998-10-01), pages 1 - 11, XP002239580, Retrieved from the Internet <URL:http://www.cs.princeton.edu/gfx/publications.html> [retrieved on 20030428] *
SAMANTA R ET AL: "HYBRID SORT-FIRST AND SORT-LAST PARALLEL RENDERING WITH A CLUSTER OF PCS", PROCEEDINGS 2000 SIGGRAPH/EUROGRAPHICS WORKSHOP ON GRAPHICS HARDWARE. INTERLAKEN, SWITZERLAND, AUG. 21 - 22, 2000, SIGGRAPH / EUROGRAPHICS WORKSHOP ON GRAPHICS HARDWARE, NEW YORK, NY: ACM, US, 21 August 2000 (2000-08-21), pages 97 - 108, XP000964478, ISBN: 1-58113-257-3 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2864318A1 (en) * 2003-12-23 2005-06-24 Alexis Vartanian Data organization device for use in graphic display system, has device to alternatively allocate variables related to light and color or depth and transparency of designed object in frame buffer based on selection function
WO2005116930A1 (en) * 2004-05-24 2005-12-08 Koninklijke Philips Electronics N.V. Tile based graphics rendering
FR2901049A1 (en) * 2006-05-12 2007-11-16 Techviz Soc Par Actions Simpli ENCODING METHOD AND DISPLAY SYSTEM ON A SCREEN OF A DIGITAL MODEL OF AN OBJECT IN THE FORM OF A SYNTHESIS IMAGE
WO2007135309A2 (en) * 2006-05-12 2007-11-29 Techviz Method of coding and system for displaying on a screen a numerical mock-up of an object in the form of a synthesis image
JP2010500634A (en) * 2006-05-12 2010-01-07 テクヴィズ Method for coding a digital mockup of an object in the form of a composite image and a system for display on a screen
WO2007135309A3 (en) * 2006-05-12 2010-04-29 Techviz Method of coding and system for displaying on a screen a numerical mock-up of an object in the form of a synthesis image
JP4933611B2 (en) * 2006-05-12 2012-05-16 テクヴィズ Method for coding a digital mockup of an object in the form of a composite image and a system for display on a screen
US8212811B2 (en) 2006-05-12 2012-07-03 Techviz Method of coding and system for displaying on a screen a digital mock-up of an object in the form of a synthesis image

Similar Documents

Publication Publication Date Title
US8401327B2 (en) Apparatus, computer program product and associated methodology for video analytics
EP2390835A3 (en) Dynamic load balancing in multiple video processing (VPU) systems
DE112019002981T5 (en) PARALLEL COMPUTATION ARCHITECTURE WITH RECONFIGURABLE CORE LEVEL AND VECTOR LEVEL PARALLELITY
EP1345120A3 (en) Method and apparatus for distributed processing control
Pogorelov et al. GPU-accelerated real-time gastrointestinal diseases detection
FR2840705A1 (en) Load balancing method for a parallel-connected PC cluster used in processing a single digital video image, whereby objects in the image are divided into sub-objects in such a way that duplicated processing is minimized
EP0177160A2 (en) Apparatus and method for implementing dilation and erosion transformation in grayscale image processing
Webb Implementation and performance of fast parallel multi-baseline stereo vision
CN114127807A (en) System and method for performing object analysis
MacCormac On-line image processing for tobacco grading in Zimbabwe
CN111581255A (en) Distribution scheduling system of high-density image data stream based on big data mining
EP0378816A3 (en) Production line method and apparatus for instruction execution
CN108782312B (en) Sewage treatment system based on image processing
TW200405176A (en) Method and apparatus for performing a pixel averaging instruction
Qasaimeh et al. A runtime configurable hardware architecture for computing histogram-based feature descriptors
JP2004102904A (en) Filtering device and filtering method
JP7252158B2 (en) LEARNING METHOD, LEARNING DEVICE, IMAGE ANALYSIS DEVICE, AND PROGRAM
KR20090019174A (en) Method for vessel enhancement using decimation-free directional filter bank
FR2840753A1 (en) Digital image computer assembly processing having video signals decoded producing pixel/clock synchronisation signal and pixel signals generated applying OR logic two pixel signals
CN110400249A (en) A kind of image processing method based on CPU
Shang et al. Single Image Dehazing on Mobile Device Based on GPU Rendering Technology.
CN113892081A (en) Hardware accelerator for integral image computation
CN1098509C (en) Picture extraction apparatus
JP2000509170A (en) Image processing device
JP2004248065A (en) Image processor