On the Conversion between Binary Code and Binary-Reflected Gray Code on Boolean Cubes
|
|
- Griffin Griffith
- 5 years ago
- Views:
Transcription
1 On the Convesion between Binay Code and BinayReflected Gay Code on Boolean Cubes The Havad community has made this aticle openly available. Please shae how this access benefits you. You stoy mattes Citation Johnsson, S. Lennat and ChingTien Ho On the Convesion between Binay Code and BinayReflected Gay Code on Boolean Cubes. Havad Compute Science Goup Technical Repot TR2091. Citable link Tems of Use This aticle was downloaded fom Havad Univesity s DASH epositoy, and is made available unde the tems and conditions applicable to Othe Posted Mateial, as set foth at ns.havad.edu/un:hul.instrepos:dash.cuent.temsofuse#laa
2 On the Convesion between Binay Code and Binay{Reected Gay Code on Boolean Cubes S. Lennat Johnsson ChingTien Ho TR2091 July 1991 Paallel Computing Reseach Goup Cente fo Reseach in Computing Technology Havad Univesity Cambidge, Massachusetts This wok has in pat been suppoted by the Ai Foce Oce of Scientic Reseach unde contact AFOSR89082, and in pat by NSF and DARPA unde contact CCR
3 On the Convesion between Binay Code and Binay{Reected Gay Code on Binay Cubes S. Lennat Johnsson Havad Univesity and Thinking Machines Cop. Cambidge, MA ChingTien Ho IBM Reseach Division Almaden Reseach Cente San Jose, CA Abstact We pesent a new algoithm fo convesion between binay code and binay{eected Gay code that equies appoximately 2K element tansfes in sequence fo K elements pe node, compaed to K element tansfes fo peviously known algoithms. Fo a binay cube of n = 2 dimensions the new algoithm degeneates to yield a complexity of K + 1 element tansfes, 2 which is optimal. The new algoithm is optimal within a facto of 1 with espect to the best known lowe bound fo any outing stategy. We show that the minimum numbe of element tansfes fo minimum path length outing is K with concuent communication on all channels of evey node of a binay cube. 1 Intoduction. Minimizing the equied data motion in memoy hieachies has been cucial in achieving high pefomance almost since the beginning of moden compute technology. In conventional memoy hieachies, minimizing data motion takes the fom of peseving tempoal and spatial locality of efeence in scheduling opeations. Massively paallel pocessos with thousands of pocessing nodes ae cuently the only altenative to exteme pefomance, i.e., a pefomance of a tillion opeations pe second and beyond. Scalability of the design, as well as pefomance, dictates that each pocessing node has its own memoy system, possibly extended with a physically shaed memoy. In such distibuted memoy hieachies, the bandwidth between a pocesso and its local memoy hieachy is often consideably highe than the bandwidth to the memoy units of othe pocessos o shaed memoy if it exists. The latency is anothe impotant pefomance issue. Ou focus is on ecient use of the communications bandwidth. The communication system in distibuted memoy hieachies often epesents a consideable faction of the total system cost. Nevetheless, due to the limited netwok bandwidth, it is often a bottleneck fo pefomance in many computations. Minimizing the demands on the communication system though pope data placement, such that data fequently used togethe eside in the same memoy unit o adjacent memoy units, is impotant. Once the data placement has been made, it is impotant to select paths and schedule the data motion such that the outing time (contention) is minimized. Placing the data among the memoy units fo optimum 1
4 pefomance is a vey had poblem. Often dieent allocations ae pefeable duing dieent phases of the computations. High Pefomance Fotan [2] is an extension of Fotan90 in which diectives have been added fo use contol ove data placement among memoy units. In this pape we addess the outing issues in changing the data placement fom one common placement in binay cubes, binay{eected Gay code [8], to anothe common placement, binay code, o the convese. Computations on egula gids constitute an impotant class of computations in science and engineeing. So called explicit methods fo nite dieence appoximations of patial dieential equations and many signal and image pocessing tasks fall into this class. The computations in explicit methods fo nite dieence appoximations ae often dominated by the evaluation of the dieence appoximation, known as dieence stencil o dieence molecule. The opeation is the same as what is typically known as convolution in signal and image pocessing. Data efeences ae local in Catesian space, which is used to epesent the poblem domain fo most egula discetizations. Howeve, the fast Fouie tansfom, FFT, is a poweful algoithm that is used fequently both fo the solution of patial dieential equations and in signal and image pocessing. The CooleyTukey FFT [1] efeences data that ae adjacent in binay space. Both techniques may be used on the same data set. Thus, if a distibuted memoy hieachy allows fo an ecient placement with espect to efeences both in Catesian and binay space, such a placement may be used. If the memoy hieachy is such that eithe one o the othe efeence patten can be suppoted, but not both simultaneously, then convesion between the two placements become impotant. This is the issue addessed in this pape fo a collection of memoy units inteconnected by a binay cube netwok. Catesian gids with axes' lengths being powes of two ae subgaphs of binay cubes if the total numbe of elements is less than, o equal to, the numbe of nodes in the cube. Binay{ eected Gay codes [8] ae often used fo the embedding of aays in binay cubes, since such codes peseve adjacency in Catesian space. Fo multidimensional aays, the Gay code encoding is typically applied to the dieent axes independently, theeby peseving adjacency along each axis. In this case, distinct subsets of addess bits ae assigned to dieent axes. We efe to each such subset as an addess eld. Fo aays with axes' lengths not being powes of two, peseving adjacency foces an expansion of the numbe of equied cube nodes []. But, even if the adjacency equiement is elaxed fo multidimensional aays with abitay axes' lengths in ode to limit the expansion, binay{eected Gay codes can be used fo the embedding of subsections of aays [5]. A conventional binay encoding of aay axes is suitable fo many divide{and{conque algoithms, such as the FFT. In the binay code, adjacency is peseved in the binay space. The binay cube nodes can be labeled such that adjacent nodes die in pecisely one bit. This pape focuses on the optimal outing of data fo convesion between data placements based on binay{eected Gay codes and binay codes in binay cube netwoks, in paticula in such netwoks allowing concuent communication on all channels of evey node, all{pot communication. The Connection Machine models CM{2 and CM{200 [4, 9] ae examples of achitectues using a binay cube netwok allowing all{pot communication. Ou esults can easily be modied to systems allowing concuent communication on some, but not all, channels of evey pocesso. The esults also povide inteesting insights into the popeties of binay{ eected Gay codes. The algoithms equie vey small tempoay stoage and the contol can be made distibuted, i.e., each node can detemine its own actions based on its addess and the local histoy of events. 2
5 The main new esult is an algoithm that fo K elements pe node of a binay n{cube equies d 2K (n 2) e + (n 2) element tansfes in sequence with all{pot communication. The algoithm is based on pipelining of the element tansfes. It yields an impovement of about K element tansfes in sequence ove pevious algoithms [, 7]. The new algoithm is within a facto of 1 of the best known lowe bound. It is optimal fo n = 2. We also pesent an optimal minimum path length outing algoithm. Though the numbe of element tansfes in sequence is highe than fo the nonminimum path length outing algoithm, the minimum path length algoithm may yield fewe communication statups in a packet{switched system with packet sizes of the same ode as the local data set. We give lowe bounds fo two outing stategies: one fo using any set of outing paths, one fo using only minimum length paths. The outline of the pape is as follows. We st give some popeties of binay cubes and binay{eected Gay codes, state ou assumptions of the communication system, and ou objectives fo the outing algoithms. Then, we pesent an optimal minimum path length outing algoithm, followed by the nonminimum path length outing algoithm that is the main esult of the pape. We conclude with a summay. 2 Peliminaies. A binay n{cube has N = 2 n nodes. Two nodes ae adjacent i thei addesses die in exactly one bit. Thee exist n disjoint paths between any pai of nodes in a binay n{cube. Fo nodes at distance d, d of those paths ae of length d, and n d paths ae of length d + 2. The binay encoding of i is B n (i) = (b n 1 b n 2 b 0 ) and its binay{eected Gay code encoding is G n (i) = (g n 1 g n 2 g 0 ). The pocesso addess bits ae (a n 1 a n 2 : : : a 0 ). Z N = f0; 1; ; N 1g and (1 j ) is a sting of j instances of a bit with value one. \jj" is the concatenation symbol. Fo the complexity estimates we assume bidiectional channels and concuent communication on all channels, all{pot communication. The numbe of elements pe node is K. ^Gn is the sequence of n{bit binay{eected Gay codes fo Z N, i.e., ^G n = (G n (0); G n (1); ; G n (2 n 1)). Denition 1 [8] The binayeected Gay code is dened ecusively as follows. ^G 1 = (G 1 (0); G 1 (1)); whee G 1 (0) = 0; G 1 (1) = 1: 0 1 0jjG n (0) 0jjG n (1). 0jjG n (2 n 2) ^G n+1 = 0jjG n (2 n 1) 1jjG n (2 n 1) : 1jjG n (2 n 2). B 1jjG n (1) A 1jjG n (0) In the following, we efe to the binay{eected Gay code just dened as Gay code. The highest ode bit is the same in the binay code and in the Gay code. The emaining bits in the
6 ?? Routing in a 2cube. 2 5 Routing in a cube. 4 Figue 1: Routing by exchanges in a 2cube and a cube. encoding of i 2 Z N=2 ae dened by G n 1 ((b n 2 b n b 0 )). The emaining bits in the encoding of i 2 Z N Z N=2 ae dened by G n 1 ((b n 2 b n b 0 )). Lemma 1 [8] b m = g n 1 g n 2 g m, m 2 Z n. Convesely, g m = b m b m+1, m 2 Z n with b n = 0. We pesent all of ou esults fo Gay{to{binay convesion, i.e., convesion fom Gay code encoding to binay code encoding. The adaptations of the esults to binay{to{gay convesion is staightfowad. We assume that each addess eld subject to the Gay{to{binay convesion is of length at least two. This assumption avoids the tivial case of a one bit code fo which the Gay code encoding is the same as the binay encoding. Bidiectional communication is assumed fo ou complexity estimates. All communication complexities ae stated in tems of the numbe of element tansfes in sequence. Given B n (i) and G n (i), an exclusive{o opeation on the two elds detemines the cube dimensions though which the data fom each node must be outed. The issues we addess ae: outing with constant stoage (pemutation outing) minimizing the congestion outing without tags A outing that can be viewed as a sequence of pemutations between neighboing nodes is highly desiable since such a outing conseves memoy equiements. Limiting the poblem size due to a demand fo bue space is often met with sevee citicism fom uses. The Gay{to{ binay convesion has the popety that the communication can be pefomed as a sequence of exchanges, theeby ealizing the st objective. This popety of the two codes is illustated in Figue 1 fo the cases n = 2 and n =. Ou minimal path length algoithm is based on Theoem 1 below. Befoe poving the theoem, we pove one citical popety of the binay{eected Gay code. Lemma 2 Let node a in a binay ncube initially contain the element of index G 1 (a). Fo any m 2 f1; 2; ; n 1g, if each node a fo which a n 1 a n 2 a m = 1 exchanges its element with its neighbo acoss dimension m 1, then node a contains the element of index G 1 (a n 1 a m )jjg 1 (a m 1 a 0 ) afte the exchange. 4
7 Poof: It suces to show that G 1 (a n 1 a m )jjg 1 G 1 (a (a m 1 a 0 ) = n 1 a m a 0 ); if a n 1 a m = 0, G 1 (a n 1 a m a 0 ); othewise. Let G 1 (a n 1 a m )jjg 1 (a m 1 a 0 ) = (a 0 n 1 a 0 0), let G 1 (a n 1 a 0 ) = (a 00 n 1 a 00 0) and let G 1 (a n 1 a m a 0 ) = (a 000 n 1 a000 0 ). It follows fom Lemma 1 that a0 j = a n 1 a j = a 00 j = a000 j fo m j n 1. Fo j being in the ange 0 j m 1, conside two cases. (i) Case 1: a n 1 a m = 0. Then, a 00 j = a n 1 a j = a m 1 a j = a 0 j. (ii) Case 2: a n 1 a m = 1. Then, a 000 j = a n 1 a m a j = 1a n 1 a j = a m 1 a j = a 0 j : Lemma 2 is easily genealized to the splitting of one out of seveal addess elds by obseving that the splitting of one addess eld is independent of othe addess elds and thei encoding. Theoem 1 The Gay{to{binay convesion can be pefomed as a sequence of exchanges in dimensions f0; 1; ; n 2g taken in abitay ode. Poof: Pefoming an exchange as in Lemma 2 ceates two independently Gay coded addess elds. The same splitting pocedue can be applied to the ceated subelds in abitay ode until thee ae n 1 addess elds of one bit each. Since the Gay code and binay code fo a one bit eld is identical, the poof is complete. The poof of Lemma 2 povides a way of deiving the exchange contol locally. Let m be the cuent exchange dimension, and let x be the next highe dimension which has aleady appeaed in the exchange sequence. If no such dimension exists, then let x = n 1. Then, the cuent data of node a should be exchanged if and only if a x 1 a m+1 = 1. Figue 2 gives an example of using an algoithm poceeding fom dimension n 2 to dimension 0. Initially, pocesso G 4 (i) contains data of index i. Afte the convesion, i is assigned to pocesso B 4 (i). A pseudocode fo the algoithm is given below. Initially, g m (i)! a m, m 2 Z n 1 and on temination b m (i)! a m, m 2 Z n 1, whee a = (a n 1 a n 2 : : :a 0 ), is the pocesso addess as befoe. /* Conveting Gay code to binay code stating fom the most signicant dimension */ fo j := n 2 downto 0 do if a j+1 = 1 then exchange content with the neighbo in dim. j endif enddo Fo the optimal minimum path length outing algoithm we pesent late, any dimension may be used as a stating dimension. Fo an abitay stating dimension m, m 2 Z n 1 with 5
8 Gay code Exchange Exchange Exchange dim. 2 dim. 1 dim. 0 data padd b data b2 data b1 data Figue 2: Convesion of a Gay code to binay code. exchanges in successive dimensions of deceasing ode, cyclicly, the st exchange equies the computation of the mth bit of G 1 (a), which is a n 1 a m+1. The subsequent steps ae simila to the algoithm above. Figue gives an example. Sequence 2 is the same as in Figue 2. The gue shows the location of i fo each step of the algoithm fo each sequence. /* Conveting Gay code to binay code stating fom dimension m. Dimensions in deceasing ode, cyclically*/ if a n 1 a n 2 a m+1 = 1 then exchange content with the neighbo in dim. m endif fo j := m 1 downto 0 do if a j+1 = 1 then exchange content with the neighbo in dim. j endif enddo fo j := n 2 downto m + 1 do if a j+1 = 1 then exchange content with the neighbo in dim. j endif enddo Lowe bounds. We st conside the case whee Gay{to{binay convesion shall be pefomed on a single pocesso addess eld. We then conside the case fo multiple pocesso addess elds, whee
9 Gay code Seq 2 Seq 1 Seq 0 assignment Exchange dim. Exchange dim. Exchange dim. Data padd Figue : Concuent convesion of a Gay code to binay code. each eld is subject to a Gay{to{binay convesion..1 A single eld. Theoem 2 With the communication fo Gay{to{binay convesion esticted to minimum length paths only, a lowe bound is max(k; n 1) element tansfes in sequence fo a code eld of n bits on an n{cube with all{pot communication. Poof: With K elements pe node eithe all o no elements must be communicated to some othe node fo code convesion. Fo any n 2 thee always exists a pai of nodes at distance one that must exchange data (nodes 2 and ). Resticting the communication to minimum length paths implies that the code convesion equies at least K element tansfes in sequence. Theoem A lowe bound fo the numbe of element tansfes in sequence equied fo Gay{ to{binay convesion on an n{cube with all{pot communication is max((1 1 n ) K 2 ; n 1). Poof: The total amount of communication equied fo Gay{to{binay convesion using only minimum length paths is 2 P n 1 n 1 i=0 i i = (n 1)2 n 1 K. (It follows fom the denition of the Gay code that the numbe of paths of length i is 2 n 1 i.) The numbe of available edges in an n{cube is n2 n. Hence, the lowe bound follows. Theoem does not give a tight lowe bound, since the only way in which the full bandwidth of the binay cube can be used is by using nonminimum length paths. Fo instance, fo n = 2 Theoem yields the bound K. A bound based on the numbe of disjoint paths is K. An uppe 4 2 bound fo n = 2 is K + 1. Simply oute K + 1 elements along the length one path and K elements ove the length path. Figue 4 illustates the outing of the data. The bound K is also a lowe bound fo K 2, since using two paths implies the use of nonminimum path length outing. 7
10 ?? Routing fom node 2 2 Routing fom node.2 Multiple independent elds. Figue 4: Routing paths in a 2cube. Fo the embedding of multidimensional aays a sepaate encoding is often used fo each addess eld. With d sepaately encoded addess elds and a total of n dimensions used fo the encoding, Theoem 2 is modied as follows. Theoem 4 With the communication fo Gay{to{binay convesion esticted to paths of minimum length only, a lowe bound fo the numbe of element tansfes in sequence is max(k; n d) fo all{pot communication with a code eld of n bits distibuted among d axes. Poof: Thee exists a pai of nodes (say the nodes with indices 2 o fo one axis, and all othe axis indices 0) at distance one that must exchange data. Thus, a time K is equied using only minimum length paths. Theoem 5 A lowe bound fo the numbe of element tansfes in sequence fo Gay{to{binay convesion on an n{cube with all{pot communication is max((1 d ) K ; n d) fo d addess n 2 elds sepaately encoded in n bits. Poof: The outing equiements fo axis i encoded in n i bits is (n i 1)2 n i 1 K. Thee ae 2 n n i such subcubes. Hence, the total outing need is P d 1 i=0 (n i 1)2 n 1 K = (n d)2 n 1 K. 4 Algoithms using only minimum length paths. In this section we use the popety that the code convesion can stat in an abitay dimension to geneate seveal concuent exchange sequences fo all{pot communication. The outing algoithm below uses only minimum length paths. It may be advantageous compaed to the nonminimum path length algoithm in the next section when thee ae othe simultaneous outing needs. The total load on the communications netwok is minimal fo the minimum path length outing, but the contention is not. The minimum path length outing may also be pefeable when thee is a signicant communications ovehead in a packet{switched communications system and the maximum allowable packet size is elatively lage. The minimum path length outing descibed below may yield up to 40% fewe statups than the nonminimum path length K outing fo cubes of high dimension and a maximum packet size of at least elements. n 1 8
11 4.1 Convesion of a single addess eld. We will st conside code convesion of a single addess eld. allocation [] of one{dimensional aays satisfy this constaint. depicted as follows in tems of the addess eld: Consecutive and cyclic data The data allocation can be (g p 1 g p 2 g p n b p n 1 b p n 2 b 0 ) Consecutive {z } {z } padd madd (b p 1 b p 2 b n g n 1 g n 2 g 0 ) Cyclic {z } {z } madd padd In the consecutive allocation, the pocesso addess eld is assigned to the most signicant bits of the aay index domain. In the cyclic allocation, the pocesso addess eld is instead assigned to the least signicant bits of the aay index domain. In the illustation, it is in both cases assumed that the local memoy addesses ae encoded in binay code and that the pocesso addess eld is encoded in Gay code. Theoem The Gay{to{binay convesion fo a eld of n bits can be attained with max(n 1; K) element tansfes in sequence using only minimum length paths and (n 1)pot communication. Poof: The main idea is to geneate seveal independent concuent exchange sequences that ceates a unifom load on all channels used by a minimum path length outing stategy. Fo the poof we conside thee dieent sets of values of K, the numbe of local elements. Case 1 is used fo most elements fo lage values of K and is the only elevant case when K is a multiple of n 1. Case 2 is used to assue optimal tansmission fo the emainde of elements when K is lage but not a multiple of n 1. Case 1. K mod n 1 = 0: Ceate n 1 exchange sequences that ae dieent otations of the dimensions n 2; n ; ; 0. Patition the data set in each pocesso into n 1 sets of the same size and assign one sequence to each data set. If the data set in a node does not equie communication in a dimension, then the communication is simply not pefomed. The numbe of element tansfes in sequence is K. Case 2. n 1 < K < 2n 2: Let x = K (n 1). Ceate K exchange sequences that ae distinct otations of the dimensions n 2; n ; ; 0; 1 ; 2 ; ; x, whee i 's ae dummy dimensions. No exchange is pefomed in a dummy dimension. Duing each step, no dimension in Z n 1 is used by moe than one sequence. The numbe of element exchanges in sequence is K. Case. K < n 1: It is easy to see that n 1 element tansfes in sequence ae necessay and sucient. Fo abitay K n 1, dene the outing by patitioning the data set such that cases 1 and 2 apply. In implementing the algoithm, a table can be set up in each pocesso such that fo each memoy patition thee is a table enty fo each dimension. The enty indicates whethe o not a communication shall be pefomed. 9
12 4.2 Convesion of multiple addess elds. Fo d addess elds encoding a total of n bits thee ae n d dimensions fo which an exchange is equied. Fo K mod (n d) = 0, n d exchange sequences can be geneated by applying n d dieent otations to some basic sequence in a manne analogous to the single addess eld case. The local data set can be divided into n d patitions with each patition assigned one of the exchange sequences. The cases fo K mod (n d) = 0 can also be handled in a way simila to the case with a single addess eld to attain the lowe bound of max(k; n d) element tansfes in sequence. 5 Algoithms using nonminimum length paths. 5.1 A single addess eld. In ode to achieve a time complexity of less than K element tansfes in sequence, it is necessay to use nonminimum length paths fo the convesion. In the following, we efe to the two subcubes with espect to the most signicant dimension as the zeo and one subcube, depending upon whethe o not the leading dimension is zeo o one. Fo n = 2 a lowe bound is K + 1, and we have aleady given an optimal algoithm, Figue 2 4. The citical obsevation fo the algoithm was that of the two edges in dimension 0, only one being used fo the minimum path length outing and none of the edges in dimension 1 being used. Hence, outing pat of the data to the subcube containing the unused edge in dimension 0 (the zeo subcube), pefoming the code convesion fo pat of the data in that subcube, then outing the conveted data back allows the time to be educed, by educing the contention. By consideing the outing paths in Figue 1, it can be obseved that if the {cube is collapsed into a 2{cube along dimension 2 by identifying nodes 0 and 7, 1 and, 2 and 5 and and 4, then all bidiectional links in the esulting 2{cube would be used evenly. Hence, by exchanging data between the zeo and one subcubes, all edges in both subcubes can be used fo the code convesion. Fo the case n = 2 thee is no exchange between the zeo and one subcubes, since thee is no outing equiement in the zeo subcube. Fo n = the edges between nodes 2 and 5, and and 4, espectively, ae used in both diections fo data fom both the zeo and the one subcubes. All othe edges ae used fo \single" exchanges. We will now pove that the outing paths fo Gay{to{binay convesion always can be detemined such that if an n{cube is collapsed into a (n 1){cube along the most signicant dimension, all edges ae used evenly in the collapsed cube. Theoem 7 If the outing paths fo the Gay{to{binay convesion tavese the cube dimensions in ascending ode, then an edge is used fo an exchange in subcube zeo, i the coesponding edge in subcube one is not used. Poof: Let the exchange sequence fo the Gay{to{binay convesion poceed though the cube dimensions in the ode 0; 1; ; n 2. Let the st exchange step be step 0. Fom Lemma 2, an exchange in step 0 is pefomed fo all nodes such that a n 1 a 1 = 1. It is easy to show that the exchange in step i, 0 i n 2, is detemined by a n 1 a i+1. Since 10
13 ?? Figue 5: Routing fo code convesion in a 4{cube. a n 1 is always in the expession detemining whethe o not a node shall exchange data, only one node in a pai dieing in thei most signicant bit exchange data in any of the dimensions 0; 1; : : :; n 2. Thus, an edge is used fo an exchange in subcube zeo i the coesponding edge in subcube one is not used. Figue 5 shows the edges used fo Gay{to{binay convesion with outing paths dened by outing in dimensions of ascending ode. Note that by folding the 4{cube onto a {cube, all edges in the {cube ae used evenly in the same way as all edges ae used evenly if a {cube is folded onto a 2{cube. Fo binay to Gay code convesion, the same popety holds fo an exchange sequence using the cube dimensions in descending ode. The basic idea of ou algoithm equiing appoximately 2K as follows. element tansfes in sequence is 1. Divide the local data set into two sets: the st set S 1 of size M 2K and the second set S 2 of size M 0 = K M K. The lage data sets S 1 ae outed ove minimum length outes, shot outes. The smalle data sets S 2 ae outed ove nonminimum length paths, long outes, by st outing the data to the node obtained by complementing the most signicant dimension. 2. Pefom the following two algoithms concuently fo each node i in communicating the data sets S 1 and S 2 to thei destination node G 1 (i). (a) Route each element in S 1 along an ascending ode of cube dimensions 0; 1; ; n 2, as needed. Pipelining the M elements esults in a total of M +n 2 element tansfes in sequence. (b) Route each element in S 2 along cube dimensions n 1; 0; 1; ; n 2; n 1. Note that dimension n 1 needs to be outed twice fo each element since outing in this 11
14 dimension is not pat of the code convesion. Pipelining is applied in the following manne, assuming M 0 n. Pipelining the M 0 elements along dimensions n 1; 0; 1; ; n 2 esults in a total of M 0 + n 1 element tansfes in sequence. Initiating the nal exchange along dimension n 1 fo the M 0 elements afte M 0 time steps of the pevious pocedue esults in M 0 time steps with n 1 steps ovelapped with the pevious pocedue. Note that the set of shot outes is edge{disjoint fom the set of long outes, thus allowing fo the element tansfes to be pipelined without contention. Befoe detemining the optimum size of the small and lage data sets, we state the esult. Theoem 8 The Gay{to{binay convesion equies at most d 2K (n 2) e + (n 2) element tansfes in sequence with all{pot communication fo K > n +. The numbe of element tansfes in sequence fo K n + 2 is n. Fo a pecise optimization of the numbe of element tansfes in sequence, we note that with M elements assigned to each shot oute, the time fo these outes, T s (M), satises the elation T s (M) M + n 2: One simple scheduling fo the long outes, which ae subject to contention in dimension n 1, is to pipeline the tansfes of the elements acoss dimension n 1 with the outing equied fo code convesion in the complemented subcube. The conveted data emains in the complemented subcube until the edges in dimension n 1 become fee, when the conveted data is bought back to the oiginating subcube. Thus, fo M 0 elements assigned to long outes, T l (M 0 ) M 0 + max(m 0 ; n) The optimal patition of the data set K, is detemined by T s (M) = T l (K M). This equality yields M = d 2K (n 2) e fo K > n + 2, and M = 2 fo K n + 2. T min = ( d 2K (n 2) e + (n 2) K > n + 2 n K n Multiple addess elds. With Gay{to{binay convesion equied on multiple addess elds, the numbe of element tansfes in sequence is almost identical to code convesion fo a single addess eld encoded in the same numbe of bits. Fo the outing, the local data sets can be divided into as many patitions as thee ae axes. The code convesion fo the dieent patitions can be pefomed concuently. All patitions must undego code convesion fo each axis. 5. Binay{to{Gay convesion. A binay{to{gay convesion is obtained by unning ou algoithms backwads. Thus, 12
15 Theoem 9 The binay{to{gay convesion equies at most d 2K (n 2) e + (n 2) element tansfes in sequence with all{pot communication fo K > n +. The numbe of element tansfes in sequence fo K n + 2 is n. The theoem follows fom Theoem 8 by using descending ode outing fo the code convesion and combining outing along shot and long outes as in the Gay{to{binay convesion case. Summay. We have given an algoithm fo the convesion between binay{eected Gay code and binay code that equies appoximately 2K element tansfes in sequence. The algoithm oes a eduction in the numbe of element tansfes in sequence by about K compaed to peviously known algoithms. Fo n = 2, the new algoithm is optimal fo any outing stategy. Fo lage values of n, it is nonoptimal by at most a facto of 1 compaed to the best known lowe bound, K We have also given an optimal minimum path length outing algoithm that fo K elements pe node and concuent communication on all channels of evey pocesso in a binay ncube equies max(k; n d) element tansfes in sequence. The algoithm has n 2 element tansfes less than a pipelined algoithm [], when K n 1. The Connection Machine models CM{2 and CM{200 ae distibuted memoy achitectues which allow concuent communication on all channels of evey node, with the nodes inteconnected as a binay cube. The code convesion outine on the Connection Machine was implemented befoe the nonminimum path length outing algoithm was discoveed. The existing outine use the pipelined algoithm in []. The esults hee show that the code convesion time on the Connection Machine can be sped up by as much as 50%. Fo a packet{switched communication system with a maximum packet size B, the minimum K path length outing equies d e(n 1) statups. The nonminimum length path outing (n 1)B equies max(d 2K B e + n 2; 2d K K e + n 1) statups. Fo B = d e, the minimum length path B n 1 outing equies n 1 statups, while the nonminimum path length outing equies appoximately 2d 2(n 1) e + n 1 statups. Fo lage n, the atio of the numbe of statups between minimum path length outing and nonminimum path length outing appoaches 0.. Fo K B 1, the atio instead appoaches 1.5. Acknowledgment This wok has in pat been suppoted by the Ai Foce Oce of Scientic Reseach unde contact AFOSR89082, and in pat by NSF and ARPA unde contact CCR Refeences [1] James C. Cooley and J.W. Tukey. An algoithm fo the machine computation of complex fouie seies. Math. Comp, 19:291{01,
16 [2] High Pefomance Fotan Foum. High pefomance fotan language specication, vesion 0.4. Technical epot, Depatment of Compute Science, Rice Univesity, Novembe [] I. Havel and J. Moavek. Bvaluations of gaphs. Czech. Math. J., 22:8{51, [4] W. Daniel Hillis. The Connection Machine. MIT Pess, Cambidge, MA, [5] ChingTien Ho and S. Lennat Johnsson. Embedding meshes in Boolean cubes by gaph decomposition. J. of Paallel and Distibuted Computing, 8(4):25{9, Apil [] S. Lennat Johnsson. Communication ecient basic linea algeba computations on hypecube achitectues. J. Paallel Distibuted Computing, 4(2):1{172, Apil [7] S. Lennat Johnsson and ChingTien Ho. The complexity of eshaping aays on Boolean cubes. In The Fifth Distibuted Memoy Computing Confeence, pages 70{77. IEEE Compute Society, Apil [8] E.M. Reingold, J. Nievegelt, and N. Deo. Combinatoial Algoithms. PenticeHall, Englewood Clis. NJ, [9] Thinking Machines Cop. CM{200 Technical Summay,
FACE VECTORS OF FLAG COMPLEXES
FACE VECTORS OF FLAG COMPLEXES ANDY FROHMADER Abstact. A conjectue of Kalai and Eckhoff that the face vecto of an abitay flag complex is also the face vecto of some paticula balanced complex is veified.
More informationA Memory Efficient Array Architecture for Real-Time Motion Estimation
A Memoy Efficient Aay Achitectue fo Real-Time Motion Estimation Vasily G. Moshnyaga and Keikichi Tamau Depatment of Electonics & Communication, Kyoto Univesity Sakyo-ku, Yoshida-Honmachi, Kyoto 66-1, JAPAN
More informationShortest Paths for a Two-Robot Rendez-Vous
Shotest Paths fo a Two-Robot Rendez-Vous Eik L Wyntes Joseph S B Mitchell y Abstact In this pape, we conside an optimal motion planning poblem fo a pai of point obots in a plana envionment with polygonal
More informationJournal of World s Electrical Engineering and Technology J. World. Elect. Eng. Tech. 1(1): 12-16, 2012
2011, Scienceline Publication www.science-line.com Jounal of Wold s Electical Engineeing and Technology J. Wold. Elect. Eng. Tech. 1(1): 12-16, 2012 JWEET An Efficient Algoithm fo Lip Segmentation in Colo
More informationART GALLERIES WITH INTERIOR WALLS. March 1998
ART GALLERIES WITH INTERIOR WALLS Andé Kündgen Mach 1998 Abstact. Conside an at galley fomed by a polygon on n vetices with m pais of vetices joined by inteio diagonals, the inteio walls. Each inteio wall
More informationCommunication vs Distributed Computation: an alternative trade-off curve
Communication vs Distibuted Computation: an altenative tade-off cuve Yahya H. Ezzeldin, Mohammed amoose, Chistina Fagouli Univesity of Califonia, Los Angeles, CA 90095, USA, Email: {yahya.ezzeldin, mkamoose,
More information(a, b) x y r. For this problem, is a point in the - coordinate plane and is a positive number.
Illustative G-C Simila cicles Alignments to Content Standads: G-C.A. Task (a, b) x y Fo this poblem, is a point in the - coodinate plane and is a positive numbe. a. Using a tanslation and a dilation, show
More informationIP Network Design by Modified Branch Exchange Method
Received: June 7, 207 98 IP Netwok Design by Modified Banch Method Kaiat Jaoenat Natchamol Sichumoenattana 2* Faculty of Engineeing at Kamphaeng Saen, Kasetsat Univesity, Thailand 2 Faculty of Management
More informationEmbeddings into Crossed Cubes
Embeddings into Cossed Cubes Emad Abuelub *, Membe, IAENG Abstact- The hypecube paallel achitectue is one of the most popula inteconnection netwoks due to many of its attactive popeties and its suitability
More informationRANDOM IRREGULAR BLOCK-HIERARCHICAL NETWORKS: ALGORITHMS FOR COMPUTATION OF MAIN PROPERTIES
RANDOM IRREGULAR BLOCK-HIERARCHICAL NETWORKS: ALGORITHMS FOR COMPUTATION OF MAIN PROPERTIES Svetlana Avetisyan Mikayel Samvelyan* Matun Kaapetyan Yeevan State Univesity Abstact In this pape, the class
More informationThe International Conference in Knowledge Management (CIKM'94), Gaithersburg, MD, November 1994.
The Intenational Confeence in Knowledge Management (CIKM'94), Gaithesbug, MD, Novembe 994. Hashing by Poximity to Pocess Duplicates in Spatial Databases Walid G. Aef Matsushita Infomation Technology Laboatoy
More informationSegmentation of Casting Defects in X-Ray Images Based on Fractal Dimension
17th Wold Confeence on Nondestuctive Testing, 25-28 Oct 2008, Shanghai, China Segmentation of Casting Defects in X-Ray Images Based on Factal Dimension Jue WANG 1, Xiaoqin HOU 2, Yufang CAI 3 ICT Reseach
More informationConversion Functions for Symmetric Key Ciphers
Jounal of Infomation Assuance and Secuity 2 (2006) 41 50 Convesion Functions fo Symmetic Key Ciphes Deba L. Cook and Angelos D. Keomytis Depatment of Compute Science Columbia Univesity, mail code 0401
More informationANALYTIC PERFORMANCE MODELS FOR SINGLE CLASS AND MULTIPLE CLASS MULTITHREADED SOFTWARE SERVERS
ANALYTIC PERFORMANCE MODELS FOR SINGLE CLASS AND MULTIPLE CLASS MULTITHREADED SOFTWARE SERVERS Daniel A Menascé Mohamed N Bennani Dept of Compute Science Oacle, Inc Geoge Mason Univesity 1211 SW Fifth
More informationControlled Information Maximization for SOM Knowledge Induced Learning
3 Int'l Conf. Atificial Intelligence ICAI'5 Contolled Infomation Maximization fo SOM Knowledge Induced Leaning Ryotao Kamimua IT Education Cente and Gaduate School of Science and Technology, Tokai Univeisity
More informationDEADLOCK AVOIDANCE IN BATCH PROCESSES. M. Tittus K. Åkesson
DEADLOCK AVOIDANCE IN BATCH PROCESSES M. Tittus K. Åkesson Univesity College Boås, Sweden, e-mail: Michael.Tittus@hb.se Chalmes Univesity of Technology, Gothenbug, Sweden, e-mail: ka@s2.chalmes.se Abstact:
More informationReachable State Spaces of Distributed Deadlock Avoidance Protocols
Reachable State Spaces of Distibuted Deadlock Avoidance Potocols CÉSAR SÁNCHEZ and HENNY B. SIPMA Stanfod Univesity We pesent a family of efficient distibuted deadlock avoidance algoithms with applications
More informationMapReduce Optimizations and Algorithms 2015 Professor Sasu Tarkoma
apreduce Optimizations and Algoithms 2015 Pofesso Sasu Takoma www.cs.helsinki.fi Optimizations Reduce tasks cannot stat befoe the whole map phase is complete Thus single slow machine can slow down the
More information4.2. Co-terminal and Related Angles. Investigate
.2 Co-teminal and Related Angles Tigonometic atios can be used to model quantities such as
More informationAn Unsupervised Segmentation Framework For Texture Image Queries
An Unsupevised Segmentation Famewok Fo Textue Image Queies Shu-Ching Chen Distibuted Multimedia Infomation System Laboatoy School of Compute Science Floida Intenational Univesity Miami, FL 33199, USA chens@cs.fiu.edu
More informationSeparability and Topology Control of Quasi Unit Disk Graphs
Sepaability and Topology Contol of Quasi Unit Disk Gaphs Jiane Chen, Anxiao(Andew) Jiang, Iyad A. Kanj, Ge Xia, and Fenghui Zhang Dept. of Compute Science, Texas A&M Univ. College Station, TX 7784. {chen,
More informationAlso available at ISSN (printed edn.), ISSN (electronic edn.) ARS MATHEMATICA CONTEMPORANEA 3 (2010)
Also available at http://amc.imfm.si ISSN 1855-3966 (pinted edn.), ISSN 1855-3974 (electonic edn.) ARS MATHEMATICA CONTEMPORANEA 3 (2010) 109 120 Fulleene patches I Jack E. Gave Syacuse Univesity, Depatment
More informationImage Enhancement in the Spatial Domain. Spatial Domain
8-- Spatial Domain Image Enhancement in the Spatial Domain What is spatial domain The space whee all pixels fom an image In spatial domain we can epesent an image by f( whee x and y ae coodinates along
More informationarxiv: v4 [cs.ds] 7 Feb 2018
Dynamic DFS in Undiected Gaphs: beaking the O(m) baie Suende Baswana Sheejit Ray Chaudhuy Keeti Choudhay Shahbaz Khan axiv:1502.02481v4 [cs.ds] 7 Feb 2018 Depth fist seach (DFS) tee is a fundamental data
More informationLecture 27: Voronoi Diagrams
We say that two points u, v Y ae in the same connected component of Y if thee is a path in R N fom u to v such that all the points along the path ae in the set Y. (Thee ae two connected components in the
More informationA New and Efficient 2D Collision Detection Method Based on Contact Theory Xiaolong CHENG, Jun XIAO a, Ying WANG, Qinghai MIAO, Jian XUE
5th Intenational Confeence on Advanced Mateials and Compute Science (ICAMCS 2016) A New and Efficient 2D Collision Detection Method Based on Contact Theoy Xiaolong CHENG, Jun XIAO a, Ying WANG, Qinghai
More informationa Not yet implemented in current version SPARK: Research Kit Pointer Analysis Parameters Soot Pointer analysis. Objectives
SPARK: Soot Reseach Kit Ondřej Lhoták Objectives Spak is a modula toolkit fo flow-insensitive may points-to analyses fo Java, which enables expeimentation with: vaious paametes of pointe analyses which
More informationOptical Flow for Large Motion Using Gradient Technique
SERBIAN JOURNAL OF ELECTRICAL ENGINEERING Vol. 3, No. 1, June 2006, 103-113 Optical Flow fo Lage Motion Using Gadient Technique Md. Moshaof Hossain Sake 1, Kamal Bechkoum 2, K.K. Islam 1 Abstact: In this
More informationParametric Query Optimization for Linear and Piecewise Linear Cost Functions
Paametic Quey Oimization fo Linea and Piecewise Linea Cost Functions Avind Hulgei S. Sudashan Indian Institute of Technology, Bombay {au, sudasha}@cse.iitb.ac.in Abstact The of a quey an depends on many
More informationLecture # 04. Image Enhancement in Spatial Domain
Digital Image Pocessing CP-7008 Lectue # 04 Image Enhancement in Spatial Domain Fall 2011 2 domains Spatial Domain : (image plane) Techniques ae based on diect manipulation of pixels in an image Fequency
More informationGARBAGE COLLECTION METHODS. Hanan Samet
gc0 GARBAGE COLLECTION METHODS Hanan Samet Compute Science Depatment and Cente fo Automation Reseach and Institute fo Advanced Compute Studies Univesity of Mayland College Pak, Mayland 07 e-mail: hjs@umiacs.umd.edu
More informationInput Layer f = 2 f = 0 f = f = 3 1,16 1,1 1,2 1,3 2, ,2 3,3 3,16. f = 1. f = Output Layer
Using the Gow-And-Pune Netwok to Solve Poblems of Lage Dimensionality B.J. Biedis and T.D. Gedeon School of Compute Science & Engineeing The Univesity of New South Wales Sydney NSW 2052 AUSTRALIA bbiedis@cse.unsw.edu.au
More informationAny modern computer system will incorporate (at least) two levels of storage:
1 Any moden compute system will incopoate (at least) two levels of stoage: pimay stoage: andom access memoy (RAM) typical capacity 32MB to 1GB cost pe MB $3. typical access time 5ns to 6ns bust tansfe
More informationAutomatically Testing Interacting Software Components
Automatically Testing Inteacting Softwae Components Leonad Gallaghe Infomation Technology Laboatoy National Institute of Standads and Technology Gaithesbug, MD 20899, USA lgallaghe@nist.gov Jeff Offutt
More informationA Two-stage and Parameter-free Binarization Method for Degraded Document Images
A Two-stage and Paamete-fee Binaization Method fo Degaded Document Images Yung-Hsiang Chiu 1, Kuo-Liang Chung 1, Yong-Huai Huang 2, Wei-Ning Yang 3, Chi-Huang Liao 4 1 Depatment of Compute Science and
More informationn If S is in convex position, then thee ae exactly k convex k-gons detemined by subsets of S. In geneal, howeve, S may detemine fa fewe convex k-gons.
Counting Convex Polygons in Plana Point Sets Joseph S. B. Mitchell a;1, Günte Rote b, Gopalakishnan Sundaam c, and Gehad Woeginge b a Applied Mathematics and Statistics, SUNY Stony Book, NY 11794-3600.
More informationCardiac C-Arm CT. SNR Enhancement by Combining Multiple Retrospectively Motion Corrected FDK-Like Reconstructions
Cadiac C-Am CT SNR Enhancement by Combining Multiple Retospectively Motion Coected FDK-Like Reconstuctions M. Pümme 1, L. Wigstöm 2,3, R. Fahig 2, G. Lauitsch 4, J. Honegge 1 1 Institute of Patten Recognition,
More informationdc - Linux Command Dc may be invoked with the following command-line options: -V --version Print out the version of dc
- CentOS 5.2 - Linux Uses Guide - Linux Command SYNOPSIS [-V] [--vesion] [-h] [--help] [-e sciptexpession] [--expession=sciptexpession] [-f sciptfile] [--file=sciptfile] [file...] DESCRIPTION is a evese-polish
More informationA Shape-preserving Affine Takagi-Sugeno Model Based on a Piecewise Constant Nonuniform Fuzzification Transform
A Shape-peseving Affine Takagi-Sugeno Model Based on a Piecewise Constant Nonunifom Fuzzification Tansfom Felipe Fenández, Julio Gutiéez, Juan Calos Cespo and Gacián Tiviño Dep. Tecnología Fotónica, Facultad
More informationAssessment of Track Sequence Optimization based on Recorded Field Operations
Assessment of Tack Sequence Optimization based on Recoded Field Opeations Matin A. F. Jensen 1,2,*, Claus G. Søensen 1, Dionysis Bochtis 1 1 Aahus Univesity, Faculty of Science and Technology, Depatment
More informationConfiguring RSVP-ATM QoS Interworking
Configuing RSVP-ATM QoS Intewoking Last Updated: Januay 15, 2013 This chapte descibes the tasks fo configuing the RSVP-ATM QoS Intewoking featue, which povides suppot fo Contolled Load Sevice using RSVP
More informationModeling a shared medium access node with QoS distinction
Modeling a shaed medium access node with QoS distinction Matthias Gies, Jonas Geutet Compute Engineeing and Netwoks Laboatoy (TIK) Swiss Fedeal Institute of Technology Züich CH-8092 Züich, Switzeland email:
More informationThe Internet Ecosystem and Evolution
The Intenet Ecosystem and Evolution Contents Netwok outing: basics distibuted/centalized, static/dynamic, linkstate/path-vecto inta-domain/inte-domain outing Mapping the sevice model to AS-AS paths valley-fee
More informationModelling, simulation, and performance analysis of a CAN FD system with SAE benchmark based message set
Modelling, simulation, and pefomance analysis of a CAN FD system with SAE benchmak based message set Mahmut Tenuh, Panagiotis Oikonomidis, Peiklis Chachalakis, Elias Stipidis Mugla S. K. Univesity, TR;
More informationIP Multicast Simulation in OPNET
IP Multicast Simulation in OPNET Xin Wang, Chien-Ming Yu, Henning Schulzinne Paul A. Stipe Columbia Univesity Reutes Depatment of Compute Science 88 Pakway Dive South New Yok, New Yok Hauppuage, New Yok
More informationWIRELESS sensor networks (WSNs), which are capable
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. XX, NO. XX, XXX 214 1 Lifetime and Enegy Hole Evolution Analysis in Data-Gatheing Wieless Senso Netwoks Ju Ren, Student Membe, IEEE, Yaoxue Zhang, Kuan
More informationAnnales UMCS Informatica AI 2 (2004) UMCS
Pobane z czasopisma Annales AI- Infomatica http://ai.annales.umcs.pl Annales Infomatica AI 2 (2004) 33-340 Annales Infomatica Lublin-Polonia Sectio AI http://www.annales.umcs.lublin.pl/ Embedding as a
More informationModule 6 STILL IMAGE COMPRESSION STANDARDS
Module 6 STILL IMAE COMPRESSION STANDARDS Lesson 17 JPE-2000 Achitectue and Featues Instuctional Objectives At the end of this lesson, the students should be able to: 1. State the shotcomings of JPE standad.
More informationTransmission Lines Modeling Based on Vector Fitting Algorithm and RLC Active/Passive Filter Design
Tansmission Lines Modeling Based on Vecto Fitting Algoithm and RLC Active/Passive Filte Design Ahmed Qasim Tuki a,*, Nashien Fazilah Mailah b, Mohammad Lutfi Othman c, Ahmad H. Saby d Cente fo Advanced
More informationA Novel Automatic White Balance Method For Digital Still Cameras
A Novel Automatic White Balance Method Fo Digital Still Cameas Ching-Chih Weng 1, Home Chen 1,2, and Chiou-Shann Fuh 3 Depatment of Electical Engineeing, 2 3 Gaduate Institute of Communication Engineeing
More informationTopological Characteristic of Wireless Network
Topological Chaacteistic of Wieless Netwok Its Application to Node Placement Algoithm Husnu Sane Naman 1 Outline Backgound Motivation Papes and Contibutions Fist Pape Second Pape Thid Pape Futue Woks Refeences
More informationHISTOGRAMS are an important statistic reflecting the
JOURNAL OF L A T E X CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 1 D 2 HistoSketch: Disciminative and Dynamic Similaity-Peseving Sketching of Steaming Histogams Dingqi Yang, Bin Li, Laua Rettig, and Philippe
More information5 4 THE BERNOULLI EQUATION
185 CHATER 5 the suounding ai). The fictional wok tem w fiction is often expessed as e loss to epesent the loss (convesion) of mechanical into themal. Fo the idealied case of fictionless motion, the last
More informationA Family of Distributed Deadlock Avoidance Protocols and their Reachable State Spaces
A Family of Distibuted Deadlock Avoidance Potocols and thei Reachable State Spaces Césa Sánchez, Henny B. Sipma, and Zoha Manna Compute Science Depatment Stanfod Univesity, Stanfod, CA 94305-9025 {cesa,sipma,manna}@cs.stanfod.edu
More informationBinary Dissection: Variants & Applications
Binay Dissection: Vaiants & Applications Shahid H. Bokhai Depatment of Electical Engineeing Univesity of Engineeing & Technology Lahoe, Pakistan Thomas W. Cockett ICASE NASA Langley Reseach Cente Hampton,
More informationOPTIMAL KINEMATIC SYNTHESIS OF CRANK & SLOTTED LEVER QUICK RETURN MECHANISM FOR SPECIFIC STROKE & TIME RATIO
OPTIMAL KINEMATIC SYNTHESIS OF CRANK & SLOTTED LEVER QUICK RETURN MECHANISM FOR SPECIFIC STROKE & TIME RATIO Zeeshan A. Shaikh 1 and T.Y. Badguja 2 1,2 Depatment of Mechanical Engineeing, Late G. N. Sapkal
More informationDynamic Topology Control to Reduce Interference in MANETs
Dynamic Topology Contol to Reduce Intefeence in MANETs Hwee Xian TAN 1,2 and Winston K. G. SEAH 2,1 {stuhxt, winston}@i2.a-sta.edu.sg 1 Depatment of Compute Science, School of Computing, National Univesity
More informationIllumination methods for optical wear detection
Illumination methods fo optical wea detection 1 J. Zhang, 2 P.P.L.Regtien 1 VIMEC Applied Vision Technology, Coy 43, 5653 LC Eindhoven, The Nethelands Email: jianbo.zhang@gmail.com 2 Faculty Electical
More informationIntroduction to Medical Imaging. Cone-Beam CT. Introduction. Available cone-beam reconstruction methods: Our discussion:
Intoduction Intoduction to Medical Imaging Cone-Beam CT Klaus Muelle Available cone-beam econstuction methods: exact appoximate Ou discussion: exact (now) appoximate (next) The Radon tansfom and its invese
More informationADDING REALISM TO SOURCE CHARACTERIZATION USING A GENETIC ALGORITHM
ADDING REALISM TO SOURCE CHARACTERIZATION USING A GENETIC ALGORITHM Luna M. Rodiguez*, Sue Ellen Haupt, and Geoge S. Young Depatment of Meteoology and Applied Reseach Laboatoy The Pennsylvania State Univesity,
More informationCOSC 6385 Computer Architecture. - Pipelining
COSC 6385 Compute Achitectue - Pipelining Sping 2012 Some of the slides ae based on a lectue by David Culle, Pipelining Pipelining is an implementation technique wheeby multiple instuctions ae ovelapped
More informationDYNAMIC STORAGE ALLOCATION. Hanan Samet
ds0 DYNAMIC STORAGE ALLOCATION Hanan Samet Compute Science Depatment and Cente fo Automation Reseach and Institute fo Advanced Compute Studies Univesity of Mayland College Pak, Mayland 07 e-mail: hjs@umiacs.umd.edu
More informationTHE THETA BLOCKCHAIN
THE THETA BLOCKCHAIN Theta is a decentalized video steaming netwok, poweed by a new blockchain and token. By Theta Labs, Inc. Last Updated: Nov 21, 2017 esion 1.0 1 OUTLINE Motivation Reputation Dependent
More informationSeveral algorithms exist to extract edges from point. system. the line is computed using a least squares method.
Fast Mapping using the Log-Hough Tansfomation B. Giesle, R. Gaf, R. Dillmann Institute fo Pocess Contol and Robotics (IPR) Univesity of Kalsuhe D-7618 Kalsuhe, Gemany fgieslejgafjdillmanng@ia.uka.de C.F.R.
More informationIntroduction To Pipelining. Chapter Pipelining1 1
Intoduction To Pipelining Chapte 6.1 - Pipelining1 1 Mooe s Law Mooe s Law says that the numbe of pocessos on a chip doubles about evey 18 months. Given the data on the following two slides, is this tue?
More informationCombinatorial Mobile IP: A New Efficient Mobility Management Using Minimized Paging and Local Registration in Mobile IP Environments
Wieless Netwoks 0, 3 32, 200 200 Kluwe Academic Publishes. Manufactued in The Nethelands. Combinatoial Mobile IP: A New Efficient Mobility Management Using Minimized Paging and Local Registation in Mobile
More informationThe University of North Carolina. its production and consumption. Embedded signal processing applications are naturally
Feasibility concens in PGM gaphs with bounded bues Sanjoy Bauah Depatment of Compute Science The Univesity of Vemont Email: sanjoy@cs.uvm.edu Steve Goddad Kevin Jeay Depatment of Compute Science The Univesity
More informationEfficient Maximal Poisson-Disk Sampling
Efficient Maximal Poisson-Disk Sampling Mohamed S. Ebeida Sandia National Laboatoies Andew A. Davidson Univesity of Califonia, Davis Anjul Patney Univesity of Califonia, Davis Patick M. Knupp Sandia National
More informationarxiv: v2 [physics.soc-ph] 30 Nov 2016
Tanspotation dynamics on coupled netwoks with limited bandwidth Ming Li 1,*, Mao-Bin Hu 1, and Bing-Hong Wang 2, axiv:1607.05382v2 [physics.soc-ph] 30 Nov 2016 1 School of Engineeing Science, Univesity
More informationClass 21. N -body Techniques, Part 4
Class. N -body Techniques, Pat Tee Codes Efficiency can be inceased by gouping paticles togethe: Neaest paticles exet geatest foces diect summation. Distant paticles exet smallest foces teat in goups.
More informationProf. Feng Liu. Fall /17/2016
Pof. Feng Liu Fall 26 http://www.cs.pdx.edu/~fliu/couses/cs447/ /7/26 Last time Compositing NPR 3D Gaphics Toolkits Tansfomations 2 Today 3D Tansfomations The Viewing Pipeline Mid-tem: in class, Nov. 2
More informationScaling Location-based Services with Dynamically Composed Location Index
Scaling Location-based Sevices with Dynamically Composed Location Index Bhuvan Bamba, Sangeetha Seshadi and Ling Liu Distibuted Data Intensive Systems Laboatoy (DiSL) College of Computing, Geogia Institute
More informationGCC-AVR Inline Assembler Cookbook Version 1.2
GCC-AVR Inline Assemble Cookbook Vesion 1.2 About this Document The GNU C compile fo Atmel AVR isk pocessos offes, to embed assembly language code into C pogams. This cool featue may be used fo manually
More informationLifetime and Energy Hole Evolution Analysis in Data-Gathering Wireless Sensor Networks
788 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 12, NO. 2, APRIL 2016 Lifetime and Enegy Hole Evolution Analysis in Data-Gatheing Wieless Senso Netwoks Ju Ren, Student Membe, IEEE, Yaoxue Zhang,
More informationA VECTOR PERTURBATION APPROACH TO THE GENERALIZED AIRCRAFT SPARE PARTS GROUPING PROBLEM
Accepted fo publication Intenational Jounal of Flexible Automation and Integated Manufactuing. A VECTOR PERTURBATION APPROACH TO THE GENERALIZED AIRCRAFT SPARE PARTS GROUPING PROBLEM Nagiza F. Samatova,
More informationSwitching and Fault Transients of Unit Transformers of a Combined-Cycle Gas Turbine Generator
Switching and Fault Tansients of Unit Tansfomes of a Combined-Cycle Gas Tubine Geneato M. Kizilcay, S. Demmig Abstact A new combined-cycle powe plant is to be connected to a 110-kV gid though a powe cable
More informationAccurate Diffraction Efficiency Control for Multiplexed Volume Holographic Gratings. Xuliang Han, Gicherl Kim, and Ray T. Chen
Accuate Diffaction Efficiency Contol fo Multiplexed Volume Hologaphic Gatings Xuliang Han, Gichel Kim, and Ray T. Chen Micoelectonic Reseach Cente Depatment of Electical and Compute Engineeing Univesity
More informationUser Specified non-bonded potentials in gromacs
Use Specified non-bonded potentials in gomacs Apil 8, 2010 1 Intoduction On fist appeaances gomacs, unlike MD codes like LAMMPS o DL POLY, appeas to have vey little flexibility with egads to the fom of
More informationNODAL AND LOOP ANALYSIS TECHNIQUES
NODAL AND LOOP ANALYSIS TECHNIQUES LEANING GOALS NODAL ANALYSIS LOOP ANALYSIS Deelop systematic techniques to determine all the oltages and currents in a circuit NODE ANALYSIS One of the systematic ways
More informationA Minutiae-based Fingerprint Matching Algorithm Using Phase Correlation
A Minutiae-based Fingepint Matching Algoithm Using Phase Coelation Autho Chen, Weiping, Gao, Yongsheng Published 2007 Confeence Title Digital Image Computing: Techniques and Applications DOI https://doi.og/10.1109/dicta.2007.4426801
More informationAn Extension to the Local Binary Patterns for Image Retrieval
, pp.81-85 http://x.oi.og/10.14257/astl.2014.45.16 An Extension to the Local Binay Pattens fo Image Retieval Zhize Wu, Yu Xia, Shouhong Wan School of Compute Science an Technology, Univesity of Science
More informationA Mathematical Implementation of a Global Human Walking Model with Real-Time Kinematic Personification by Boulic, Thalmann and Thalmann.
A Mathematical Implementation of a Global Human Walking Model with Real-Time Kinematic Pesonification by Boulic, Thalmann and Thalmann. Mashall Badley National Cente fo Physical Acoustics Univesity of
More informationTowards Adaptive Information Merging Using Selected XML Fragments
Towads Adaptive Infomation Meging Using Selected XML Fagments Ho-Lam Lau and Wilfed Ng Depatment of Compute Science and Engineeing, The Hong Kong Univesity of Science and Technology, Hong Kong {lauhl,
More informationTESSELLATIONS. This is a sample (draft) chapter from: MATHEMATICAL OUTPOURINGS. Newsletters and Musings from the St. Mark s Institute of Mathematics
TESSELLATIONS This is a sample (daft) chapte fom: MATHEMATICAL OUTPOURINGS Newslettes and Musings fom the St. Mak s Institute of Mathematics James Tanton www.jamestanton.com This mateial was and can still
More informationOn using circuit-switched networks for file transfers
On using cicuit-switched netwoks fo file tansfes Xiuduan Fang, Malathi Veeaaghavan Univesity of Viginia Email: {xf4c, mv5g}@viginia.edu Abstact High-speed optical cicuit-switched netwoks ae being deployed
More informationA Consistent, User Friendly Interface for Running a Variety of Underwater Acoustic Propagation Codes
Poceedings of ACOUSTICS 6 - Novembe 6, Chistchuch, New Zealand A Consistent, Use Fiendly Inteface fo Running a Vaiety of Undewate Acoustic Popagation Codes Alec J Duncan, Amos L Maggi Cente fo Maine Science
More informationConservation Law of Centrifugal Force and Mechanism of Energy Transfer Caused in Turbomachinery
Poceedings of the 4th WSEAS Intenational Confeence on luid Mechanics and Aeodynamics, Elounda, Geece, August 1-3, 006 (pp337-34) Consevation Law of Centifugal oce and Mechanism of Enegy Tansfe Caused in
More informationUndecidability of Static Analysis. William Landi. Siemens Corporate Research Inc. 755 College Rd East.
Undecidability of Static Analysis William Landi Siemens Copoate Reseach Inc 755 College Rd East Pinceton, NJ 08540 wlandi@sc.siemens.com Abstact Static Analysis of pogams is indispensable to any softwae
More informationAn Improved Resource Reservation Protocol
Jounal of Compute Science 3 (8: 658-665, 2007 SSN 549-3636 2007 Science Publications An mpoved Resouce Resevation Potocol Desie Oulai, Steven Chambeland and Samuel Piee Depatment of Compute Engineeing
More informationHigh performance CUDA based CNN image processor
High pefomance UDA based NN image pocesso GEORGE VALENTIN STOIA, RADU DOGARU, ELENA RISTINA STOIA Depatment of Applied Electonics and Infomation Engineeing Univesity Politehnica of Buchaest -3, Iuliu Maniu
More informationDetection and Recognition of Alert Traffic Signs
Detection and Recognition of Alet Taffic Signs Chia-Hsiung Chen, Macus Chen, and Tianshi Gao 1 Stanfod Univesity Stanfod, CA 9305 {echchen, macuscc, tianshig}@stanfod.edu Abstact Taffic signs povide dives
More informationImprovement of First-order Takagi-Sugeno Models Using Local Uniform B-splines 1
Impovement of Fist-ode Takagi-Sugeno Models Using Local Unifom B-splines Felipe Fenández, Julio Gutiéez, Gacián Tiviño and Juan Calos Cespo Dep. Tecnología Fotónica, Facultad de Infomática Univesidad Politécnica
More informationWorst-Case Delay Bounds for Uniform Load-Balanced Switch Fabrics
Wost-Case Delay Bounds fo Unifom Load-Balanced Switch Fabics Spyidon Antonakopoulos, Steven Fotune, Rae McLellan, Lisa Zhang Bell Laboatoies, 600 Mountain Ave, Muay Hill, NJ 07974 fistname.lastname@alcatel-lucent.com
More informationzeo, i.e., each watchman emains stationay, and we have the at galley poblem. At the othe limit, if m = 1, we have the watchman oute poblem. Anothe con
Optimum Guad Coves and m-watchmen Routes fo Resticted Polygons Svante Calsson y Bengt J. Nilsson yz Simeon Ntafos x Abstact A watchman, in the teminology of at galleies, is a mobile guad. We conside seveal
More informationThe Reconfigurable Mesh: Programming Model, Self-Simulation, Adaptability, Optimality, and Applications Mohammad Manzur Murshed
The Reconfiguable Mesh: Pogamming Model, Self-Simulation, Adaptability, Optimality, and Applications Mohammad Manzu Mushed A thesis submitted fo the degee of Docto of Philosophy at The Austalian National
More informationPipes, connections, channels and multiplexors
Pipes, connections, channels and multiplexos Fancisco J. Ballesteos ABSTRACT Channels in the style of CSP ae a poeful abstaction. The ae close to pipes and connections used to inteconnect system and netok
More informationParallel processing model for XML parsing
Recent Reseaches in Communications, Signals and nfomation Technology Paallel pocessing model fo XML pasing ADRANA GEORGEVA Fac. Applied Mathematics and nfomatics Technical Univesity of Sofia, TU-Sofia
More informationWe are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors
We ae IntechOpen, the wold s leading publishe of Open Access books Built by scientists, fo scientists,800 6,000 0M Open access books available Intenational authos and editos Downloads Ou authos ae among
More informationFrequency Domain Approach for Face Recognition Using Optical Vanderlugt Filters
Optics and Photonics Jounal, 016, 6, 94-100 Published Online August 016 in SciRes. http://www.scip.og/jounal/opj http://dx.doi.og/10.436/opj.016.68b016 Fequency Domain Appoach fo Face Recognition Using
More information= dv 3V (r + a 1) 3 r 3 f(r) = 1. = ( (r + r 2
Random Waypoint Model in n-dimensional Space Esa Hyytiä and Joma Vitamo Netwoking Laboatoy, Helsinki Univesity of Technology, Finland Abstact The andom waypoint model (RWP) is one of the most widely used
More information