arxiv: v1 [math.co] 23 Jan 2019

Size: px

Start display at page:

Download "arxiv: v1 [math.co] 23 Jan 2019"

Bathsheba Jackson
5 years ago
Views:

1 Embedding complete multi-partite graphs into Cartesian product of paths and cycles arxiv: v1 [math.co] 23 Jan 2019 R. Sundara Rajan a A. Arul Shantrinal a K. Jagadeesh Kumar a T.M. Rajalaxmi b Jianxi Fan c Weibei Fan c a Department of Mathematics, Hindustan Institute of Technology and Science, Chennai, India, vprsundar@gmail.com shandrinashan@gmail.com jagadeeshgraphs@gmail.com b Department of Mathematics, SSN College of Engineering, Chennai, India, laxmi.raji18@gmail.com c School of Computer Science and Technology, Soochow University, Suzhou, China, jxfan@suda.edu.cn fanweibei@163.com Abstract Graph embedding is a powerful method in parallel computing that maps a guest network G into a host network H. The performance of an embedding can be evaluated by certain parameters, such as the dilation, the edge congestion and the wirelength. In this manuscript, we obtain the wirelength (exact and minimum) of embedding complete multi-partite graphs into Cartesian product of paths and cycles, which include n-cube, n-dimensional mesh (grid), n-dimensional cylinder and n-dimensional torus, etc., as the subfamilies. Keywords: Embedding, edge congestion, wirelength, complete multi-partite graphs, Cartesian product of graphs Mathematics Subject Classification: 05C60, 05C85 1 Introduction and Preliminaries Given two graphs G (guest) and H (host), an embedding from G to H is an injective mapping f : V (G) V (H) and associating a path P f (e) in H for each edge e of G. We, now define the dilation dil(g, H), the edge congestion EC(G, H) and the wirelength W L(G, H) [1] as follows: dil(g, H) = EC(G, H) = W L(G, H) = min f:g H min f:g H min f:g H max dist H(f(x), f(y)) e=xy E(G) max EC f (e) e=xy E(H) dist H (f(x), f(y)) = e=xy E(G) min f:g H e=xy E(H) EC f (e) where dist H (f(x), f(y)) is a distance (need not be a shortest distance) between f(x) and f(y) in H and EC f (e) denote the number of edges e of G such that e = xy is in the path P f (e ) (need not be a shortest path) between f(x) and f(y) in H. Further, EC f (S) = e SEC f (e), where S E(H). 1

2 Figure 1: An embedding f from torus G into a path H with dil f (G, H) = 6, EC f (G, H) = 8 and W L f (G, H) = 48 For example, the dilation, the edge congestion and the wirelength of an embedding f : C 3 C 3 P 9 is given in Figure 1. It is easy to observe that, the above three parameters are different. But, for any embedding g, the sum of the dilations (called dilation sum), the sum of the edge congestion (called edge congestion sum) and the wirelength are all equal. Mathematically, we have the following equality. dist H (g(x), g(y)) = EC g (e) = W L g (G, H). e=xy E(G) e=xy E(H) In this manuscript, we will use the edge congestion sum to estimate the wirelength. For a subgraph M of G of order n, I G (M) = {uv E u, v M}, I G (k) = max M V (G), M =k I G(M) θ G (M) = {uv E u M, v / M}, θ G (k) = min M V (G), M =k θ G(M) The maximum subgraph problem (MSP) for a given k, k [n] is a problem of computing a subset M of V (G) such that M = k and I G (M) = I G (k). Further, the subsets M are called the optimal set [2, 3, 4]. Similarly, we define the minimum cut problem (MCP) for a given k, k [n] is a problem of computing a subset M of V (G) such that M = k and θ G (M) = θ G (k). For a regular graph, say r, we have 2I G (k) + θ G (k) = rk, k [n] [3]. The following lemmas are efficient techniques to find the exact wirelength using MSP and MCP. Lemma 1.1. [5] Let f : G H be an embedding with V (G) = V (H). Let S be set of all edges (or edge cut) of H such that E(H) \ S has exactly two subgraphs H 1 and H 2 and let G i = [f 1 (V (H i ))], i = 1, 2. In other words, G i is the induced subgraph on f 1 (V (H i )) vertices, i = 1, 2. Moreover, S must fulfil the following conditions: 1. For each edge uv E(G i ), i = 1, 2, P f (uv) has no edges in the set S. 2. For each edge uv E(G) with u in V (G 1 ) and v in V (G 2 ), P f (uv) has only one edge in the set S. 3. V (G 1 ) and V (G 2 ) are optimal sets. Then EC f (S) is minimum over all embeddings f : G H and EC f (S) = 2 E(G 1 ) = v V (G 2 ) v V (G 1 ) deg G (v) 2 E(G 2 ), where deg G (v) is the degree of a vertex v in G. deg G (v) 2

3 Remark 1.2. For a regular graph G, it is easy to check that, V (G 2 ) is optimal if V (G 1 ) is optimal and vice-versa [6]. Lemma 1.3. [5] For an embedding f from G into H, let {P 1, P 2,..., P t } be an edge partition of H such that each P i is an edge cut of H and it satisfies all the conditions of Lemma 1.1. Then W L f (G, H) = t EC f (P i ). Remark 1.4. [7] For an embedding f from G into H and f satisfies Lemma 1.3. Then the wirelength of an embedding from G into H is equal to the wirelength of an embedding from G into H with respect to f. The multipartite graph is one in all the foremost in style convertible and economical topological structures of interconnection networks. The multipartite has several wonderful options and its one in all the most effective topological structure of parallel processing and computing systems. In parallel computing, a large process is often decomposed into a collection of little sub processes which will execute in parallel with communications among these sub processes. Due to these communication relations among these sub processes the multipartite graph can be applied for avoiding conflicts in the network as well as multipartite networks helps to identify the errors occurring areas in easy way. A complete p-partite graph G = K n1,...,n p is a graph that contains p independent sets containing n i, i [p], vertices, and all possible edges between vertices from different parts. The Cartesian product technique is a very powerful technique for create a huger graph from given little graphs and it is very important technique for planning large-scale interconnection networks [8]. Especially, the n-dimensional grid (cylinder and torus) structure of interconnection networks offer a really powerful communication pattern to execute a lot of algorithms in many parallel computing systems [8], which helps to arrange the interconnection network into sequence of sub processors (layers) in uniform distribution manner for transmits the data s in faster way without delay in sending the data packets (messages). Mathematically, we now defined the Cartesian product of graphs as follows: Definition 1.5. [9] Let G and H be two graphs of order n and m respectively. Then the Cartesian product of G and H denoted by G H is the graph with the vertex set V (G) V (H), and two vertices (u, v) and (u, v ) being adjacent if either u = u and vv E(H), or v = v and uu E(G). If G 1, G 2,..., G m are graphs of order n 1, n 2,..., n m respectively, then the Cartesian product G 1 G 2 G m is denoted by m G i. Remark 1.6. The graph n G i is said to be an n-dimensional grid or torus or cylinder if all G i s are paths or cycles or any one of the G i is cycle and the remaining G i s are paths, respectively. The graph embedding problem has been well-studied by many authors with a different networks [1,5,6,7,10 37], and to our knowledge, almost all graphs considered as a host graph is a unique family (for example: path P n, cycle C n, grid P n P m, cylinder P n C m, torus C n C m, hypercube Q r and so on). In this paper, we overcome this by taking Cartesian product of paths and cycles as a host graph. Moreover, we obtain the wirelength of embedding complete 2 p -partite graphs K 2 r p,2 r p,...,2 r p into the Cartesian product of graphs n G i, where G i s are either a path or a cycle, V (G 1 ) V (G 2 ) V (G n ), V (G i ) = 2 r i, V ( n G i ) = 2 r 1+r 2 + +r n = 2 r, r, n 3, p 1 and 1 i n. 3

4 2 Main Result To prove the main result, we need the following result and algorithms. Lemma 2.1. [7] If G is a complete p-partite graph K r,r,...,r of order pr, p, r 2, then k(k 1) 2 ; k p 1 I G (k) = q 2 p(p 1) 2 ; l = qp, 1 q r (q 1) 2 p(p 1) 2 + j(q 1)(p 1) + j(j 1) 2 ; l = (q 1)p + j, 1 j p 1, 2 q r. Guest Graph Algorithm Input: N = 2 r (Total number of elements) p 1, where 2 r p represents the number of elements in the each partite Output: Labeling of complete 2 p -partite graph K 2 r p,2 r p,...,2 r p 1. Begin the algorithm 2. The guest graph is generated by the complete 2 p -partite graph 3. The program contains a function disp 3nr which takes 2 r p, 2 r p,..., 2 r p as partite elements 4. Get the values p and N, where p r p represents the number of elements in a partite 6. 2 p represents the number of partite generated 7. Assign elements in the partite: 8. m = 2 p //Determine number of partite 9. y = for x 0 to n do 11. y Elem val = y 13. for i 0 to p do 14. for j 0 to p do 15. Array[x]+i+j = Elem val 16. Elem val = Elem val + N 17. Print the partite: 18. r = 0 //Initiating array number 19. for x 0 to N do 20. for i 0 to p do 21. for z 0 to p do 22. for j 0 to p do 23. Print (array [r] + i + j)) 4

5 24. Print a tab space 25. r Go to new line 27. z = x%p 28. if z = Print an empty line 30. End the algorithm The Python coding and the corresponding implementation of the above algorithm are given in Annexure I. Host Graph Algorithm Input: Output: The dimension n 3 and the value of r 1, r 2,..., r n Labeling of Cartesian product of graphs n G i, where G i s are either a path or a cycle 1. The host graph is generated by the Python program hostgraph.py. 2. The program contains the main function disp n(a 1, a 2,..., a n ) which takes parameter a 1, a 2,..., a n as a dimension. 3. The a 1 and a 2 parameter represents the two dimensional base in matrix form. 4. The a 3 parameter takes input which produces a 3 copies of the base matrix (a 1 a 2 ). 5. The a 4 parameter takes input which produces a 4 copies of the base matrix (a 1 a 2 a 3 ). 6. Similarly, the a n parameter takes input which produces a n copies of the base matrix (a 1 a 2 a n 1 ). 7. The number of elements N is given by, N = a 1 a 2... a n 8. The input a 1, a 2,..., a n in the order of a 1 a 2... a n. 9. There will be a (i 1) row-wise rotations for a n copies. (i.e) The i th copy of (a 1 a 2... a n 1 a n ) say (a 1 a 2... a n 1 ) i has (i - 1) - row wise rotation in the clockwise sense. 10. The output is generated by considering (a 1 a 2 a n 1 ) base matrix with a n dimension (a 1 a 2 a n ). The Python coding and the corresponding implementation of the above algorithm are given in Annexure II. We, now prove the main result. Theorem 2.2. Let G be the complete 2 p -partite graphs K 2 r p,2 r p,...,2 r p and H be the Cartesian product of graphs n G i, where G i s are either a path or a cycle, V (G 1 ) V (G 2 ) V (G n ), V (G i ) = 2 r i, V ( n G i ) = 2 r 1+r 2 + +r n = 2 r, r, n 3, p 1 and 1 i n. Then the embedding f of G into H given by f(x) = x with minimum (exact) wirelength. 5

6 Proof. Label the vertices of G using Guest Graph Algorithm from 1 to 2 r. Since the graph H contains an n-dimensional grid and label the vertices of n-dimensional grid using Host Graph Algorithm from 1 to 2 r. For illustration, see Figures 2, 3, 4 and 5. Let us assume that, the label represent each of the vertex, which is allocated by the above algorithms. Let f : G H be an embedding and let f(v) = v for all v V (G) and for uv E(G), let P f (uv) be a path (shortest) between f(u) and f(v) in H. Now, we have the following 3 cases. Case 1 (All G i s are paths): It is clear that, the graph H becomes an n-dimensional grid P 2 r 1 P 2 r 2 P 2 rn. For all i, j, 1 i n and 1 j 2 r i 1, let S j i be the edge cut of P 2 r 1 P 2 r 2 P 2 rn consisting of the edges between the j th and j + 1 th copies of P 2 r 1 P 2 r 2 P r 2 (i 1) P r 2 (i+1) P 2 rn. Then {S j i : 1 i n and 1 j 2 r i 1} is an edge-partition of P 2 r 1 P 2 r 2 P 2 rn. For all i, j, 1 i n and 1 j 2 r i 1, E(H)\S j i has two components H j i and H j i, where V (H j i ) = (2r r i )j and V (H j i ) = 2 r r i (2 r i j). Let G j i and Gj i be the induced subgraph of the inverse images of V (H j i ) and V (Hj i ) under f respectively. By the guest graph algorithm, deg G (v) = 2 r ri p (2 p 1)j, for all v V (G j i ) and hence I G((2 r r i )j) = 2 2r 2ri p (2 p 1)j 2 /2. By Case 2 of Lemma 2.1, E(G j i ) is the maximum subgraph on V (Gj i ) = (2r r i )j vertices in G. Thus the edge cut S j i fulfil all the conditions of Lemma 1.1. Therefore ( EC f (S j i ) = 2r p (2 p 1)(2 r r i 2 2r 2r i p (2 p 1)j 2 ) )j 2 2 = 2 2r 2r i p (2 p 1)(2 r i j)j is minimum for 1 i n and 1 j 2 r i 1. Then by Lemma 1.3, W L(G, H) = = n n EC f (S j i ) 2 r i 1 j=1 2 2r 2ri p (2 p 1)(2 r i j)j 2 r i 1 j=1 = 22r p (2 p 1) 6 [ ( (2 r r rn ) 2 r r )]. 2 2 rn Case 2 (All G i s are cycles): It is clear that, the graph H becomes an n-dimensional torus C 2 r 1 C 2 r 2 C 2 rn. For all i, j, 1 i n and 1 j 2 ri 1, let T j i be the edge cut of C 2 r 1 C 2 r 2 C 2 rn consisting of the edges between the (2 ri 1 i+j) th & (2 ri 1 i+j +1) th and (2 r i i + j) th & (2 r i i + j + 1) th copies of C 2 r 1 C 2 r 2 C r 2 (i 1) P 2 r i 1 C r 2 (i+1) C 2 rn. Then {T j i : 1 i n and 1 j 2 ri 1 } is an edge-partition of C 2 r 1 C 2 r 2 C 2 rn. For all i, j, 1 i n and 1 j 2 ri 1, E(H)\T j i has two components H j i and H j i, where V (H j i ) = V (Hj i ) = 2 r 1. Let G j i and Gj i be the induced subgraph of the inverse images of V (H j i ) and V (H j i ) under f respectively. By the guest graph algorithm, deg G (v) = 2 r p 1 (2 p 1), for all v V (G j i ) and hence I G(2 r 1 ) = 2 2r 2p 2 2 p (2 p 1)/2. By Case 2 of Lemma 2.1, E(G j i ) is the maximum subgraph on V (G j i ) = 2r 1 vertices in G. Thus the edge cut T j i fulfil all the conditions 6

7 of Lemma 1.1. Therefore ( EC f (T j 2 i ) = 2r p (2 p 1)2 r 1 2r 2p 2 2 p (2 p ) 1) 2 2 = 2 2r p 2 (2 p 1) is minimum for 1 i n and 1 j 2 r i 1. Then by Lemma 1.3, W L(G, H) = = n n 2 r i 1 j=1 2 r i 1 EC f (T j i ) 2 2r p 2 (2 p 1) j=1 = 2 2r p 3 (2 p 1)(2 r r rn ). Case 3 (Some G i s are paths and the remaining G i s are cycles): For given a and b of positive integers with n = a + b, the host graph H is defined as G 1 G 1 G 2 G 2 G k G q, where G i s are path on 2 r i vertices or a null graph (it is a graph with V = 0) and G j s are cycle on 2 r j vertices or a null graph, 1 i k n, 1 j q n, k, q n, 2 r 1 2 r 1 2 r 2 2 r 2 2 r k 2 r q, r 1 + r r k + r 1 + r r q = r, r 1 and n 3. Proceeding along the same lines of Case 1 and Case 2, we describe the edge cuts Si l and T j k, if G i and G j will exists, (we mean G i and G j are not a null graph) respectively, where 1 i k n, 1 j q n, 1 l 2 r i 1 and 1 m 2 r j 1. Then, it is clear that {Si l : 1 i k n, 1 l 2 r i 1} {Tj k : 1 j q n, 1 m 2r j 1 } is an edge-partition of G 1 G 1 G 2 G 2 G k G q. Again by similar arguments in Case 1 and Case 2, we get EC f (S l i) = 2 2r 2r i p (2 p 1)(2 r i l)l is minimum for 1 i k and 1 l 2 r i 1 and is minimum for 1 j q and 1 m 2 r j 1. Then by Lemma 1.3, EC f (T k j ) = 2 2r p 2 (2 p 1) W L(G, H) = k i = 1 & G i exist 2 r i 1 l=1 EC f (S l i) + q j = 1 & G j exist 2 r j 1 m=1 EC f (T m j ) = = 1 6 k i = 1 & G i exist k i = 1 & G i exist 2 r i 1 l=1 2 2r 2r i p (2 p 1)(2 r i l)l + 2 2r r i p (2 p 1)(2 r i 1) + q j = 1 & G j exist q j = 1 & G j exist 2 r j 1 m=1 2 2r p 2 (2 p 1) 2 2r+r j p 3 (2 p 1). 7

8 Corollary 2.3. If G 1 is a cycle on 2 r 1 vertices and G i is a path on 2 r i vertices, 2 i n, then the host graph n G i becomes an n-dimensional cylinder C 2 r 1 P 2 r 2 P 2 rn and the wirelength of an embedding f from G into H is given by W L(G, H) = 22r p (2 p 1) 6 +(2 p 1)2 2r+r 1 p 3. 3 Conclusion and Future Work [ (2 r r rn ) ( 1 2 r r )] 3 2 rn In this manuscript, we found the wirelength (exact and minimum) of an embedding complete multipartite graphs into Cartesian product of paths and cycles. Computing the dilation and the edge congestion of embedding complete multi-partite graphs into Cartesian product and other product of graphs are under investigation. Acknowledgments The work of R. Sundara Rajan was partially supported by Project no. ECR/2016/1993, Science and Engineering Research Board (SERB), Department of Science and Technology (DST), Government of India. Further, we thank Prof. Gregory Gutin and Prof. Stefanie Gerke, Royal Holloway, University of London, TW20 0EX, UK; Prof. Indra Rajasingh, School of Advanced Sciences, Vellore Institute of Technology, Chennai, India and Dr. N. Parthiban, Department of Computer Science and Engineering, SRM Institute of Science and Technology, Chennai, India, for their fruitful suggestions. References [1] S.L. Bezrukov, J.D. Chavez, L.H. Harper, M. Röttger and U.P. Schroeder, Embedding of hypercubes into grids, MFCS, , [2] M.R. Garey and D.S. Johnson, Computers and Intractability, A Guide to the Theory of NP-Completeness, Freeman, San Francisco, [3] S.L. Bezrukov, S.K. Das and R. Elsässer, An edge-isoperimetric problem for powers of the Petersen graph, Annals of Combinatorics, Vol. 4, , [4] L.H. Harper, Global Methods for Combinatorial Isoperimetric Problems, Cambridge University Press, [5] M. Miller, R.S. Rajan, N. Parthiban and I. Rajasingh, Minimum linear arrangement of incomplete hypercubes, The Computer Journal, Vol. 58, no.2, , [6] P. Manuel, I. Rajasingh, B. Rajan and H. Mercy, Exact wirelength of hypercube on a grid, Discrete Applied Mathematics, Vol. 157, no. 7, , [7] R.S. Rajan, T.M. Rajalaxmi, J.B. Liu and G. Sethuraman, Wirelength of embedding complete multipartite graphs into certain graphs, Discrete Applied Mathematics, [8] J.M. Xu, Topological Structure and Analysis of Interconnection Networks, Kluwer Academic Publishers,

9 [9] M.H. Khalifeh, H.Y.-Azari and A.R. Ashrafi, The hyper-wiener index of graph operations, Computers and Mathematics with Applications, Vol. 56, , [10] P.-L. Lai and C.-H. Tsai, Embedding of tori and grids into twisted cubes, Theoretical Computer Science, Vol. 411, no , , [11] Y. Han, J. Fan, S. Zhang, J. Yang and P. Qian, Embedding meshes into locally twisted cubes, Information Sciences, Vol. 180, no. 19, , [12] L. Auletta, A.A. Rescigno and V. Scarano, Embedding graphs onto the supercube, IEEE Transactions on Computers, Vol. 44, no. 4, , [13] M.Y. Chan and S.-J. Lee, Fault-tolerant embedding of complete binary trees in hypercubes, IEEE Transactions on Parallel and Distributed Systems, Vol. 4, no. 3, , [14] Y.-C. Tseng, S.-H. Chang and J.-P. Sheu, Fault-tolerant ring embedding in a star graph with both link and node failures, IEEE Transactions on Parallel and Distributed Systems, Vol. 8, no. 12, , [15] J. Fan and X. Jia, Embedding meshes into crossed cubes, Information Sciences, Vol. 177, no. 15, , [16] J. Ellis, S. Chow and D. Manke, Many to one embeddings from grids into cylinders, tori and hypercubes, SIAM Journal on Computing, Vol. 32, no. 2, , [17] P.-J. Yang, S.-B. Tien and C.S. Raghavendra, Embedding of rings and meshes onto faulty hypercubes using free dimensions, IEEE Transactions on Computers, Vol. 43, no. 5, , [18] J. Fan, X. Jia and X. Lin, Complete path embeddings in crossed cubes, Information Sciences, Vol. 176, no. 22, , [19] J.-F. Fang and K.-C. Lai, Embedding the incomplete hypercube in books, Information Processing Letters, Vol. 96, 1 6, [20] R.-S. Lo and G.-H. Chen, Embedding Hamiltonian paths in faulty arrangement graphs with the backtracking method, IEEE Transactions on Parallel and Distributed Systems, Vol. 12, no. 2, , [21] A. Wagner, D.G. Corneil, Embedding trees in a hypercube is np-complete, SIAM Journal on Computing, Vol. 19, no. 4, , [22] I. Rajasingh, J. Quadras, P. Manuel and A. William, Embedding of cycles and wheels into arbitrary trees, Networks, Vol. 44, , [23] X. Yang, Q. Dong and Y.Y. Tan, Embedding meshes/tori in faulty crossed cubes, Information Processing Letters, Vol. 110, no , , [24] P. Kulasinghe and S. Bettayeb, Embedding binary trees into crossed cubes, IEEE Transactions on Computers, Vol. 44, no. 7, , [25] M.Y. Chan, Embedding of grids into optimal hypercubes, SIAM Journal on Computing, Vol. 20, no. 5, ,

10 [26] F.R.K. Chung, F.T. Leighton and A.L. Rosenberg, Embedding graphs in books: A layout problem with applications to VLSI design, SIAM Journal on Algebraic Discrete Methods, Vol. 8, no. 1, 33-58, [27] M.-C. Yang, Path embedding in star graphs, Applied Mathematics and Computation, Vol. 207, no. 2, , [28] H. Rostami and J. Habibi, Minimum linear arrangement of chord graphs, Applied Mathematics and Computation, Vol. 203, , [29] S.A. Choudum and U. Nahdini, Complete binary trees in folded and enhanced cubes, Networks, Vol. 43, no. 4, , [30] I. Rajasingh, M. Arockiaraj, B. Rajan and P. Manuel, Minimum wirelength of hypercubes into n-dimensional grid networks, Information Processing Letters, Vol. 112, , [31] I. Rajasingh, B. Rajan and R.S. Rajan, Embedding of special classes of circulant networks, hypercubes and generalized Petersen graphs, International Journal of Computer Mathematics, Vol. 89, no. 15, , [32] D. Adolphson and T.C. Hu, Optimal linear ordering, SIAM Journal on Applied Mathematics, Vol. 25, no. 3, , [33] S.L. Bezrukov and U.-P. Schroeder, The cyclic wirelength of trees, Discrete Applied Mathematics, Vol. 87, no. 1-3, , [34] J. Diaz, J. Petit and M. Serna, A Survey of Graph Layout Problems, ACM Computing Surveys, Vol. 34, no. 3, , [35] I. Rajasingh and R.S. Rajan, Exact wirelength of embedding circulant networks into necklace and windmill Graphs, Ars Combinatoria, Vol. 130, , [36] I. Rajasingh, R.S. Rajan and P. Manuel, A linear time algorithm for embedding Christmas trees into certain trees, Parallel Processing Letters, Vol. 25, no. 4, 1 17, [37] C.-J. Guu, The Circular Wirelength Problem for Hypercubes, Ph.D. dissertation, University of California, Riverside,

11 Annexure I Python program for labeling of the guest graph def printline(num, boxes, boxesinrow): i = 0; string = for i in range(0, boxesinrow): for j in range(0, 4): string = string + {:< 3d}.format(num + boxes * j) + string = string + num = num + 1 print(string) def printbox(num, boxes, elements per box, boxes in row): value=num temp = elements per box // 4 while temp > 0: printline(num, boxes, boxes in row) temp = temp - 1 num = num + boxes * 4 print( \n ) return (value+boxes in row) def printpattern(numl): boxes = len(numl) elements P er box = numl[0] num = 1 temp = boxes while temp > 0: if temp >= 4: num = printbox(num, boxes, elements P er box, 4) else: num = printbox(num, boxes, elements P er box, temp) temp = temp - 4 printpattern([32,32,...,32]) 11

12 Implementation of the above Python program Output 1: Figure 2: Complete 8-partite graph K 32,32,...,32 12

13 Output 2: Figure 3: Complete 32-partite graph K 8,8,...,8 13

14 Annexure II Python program for labeling of the host graph n=0 def disp 3nr(numl, n, irange): for i in irange: string = str(i+n) for j in range(1, numl[1]): string=string+ +str(i+n+numl[0]*j) for k in range(1, numl[2]): string=string+ for j in range(0, numl[1]): string = string + + str(i+ n + numl[0] * j+ k*numl[0]*numl[1]) print(string) def disp n(numl): base=numl[0]*numl[1]*numl[2] para num=len(numl) order=numl[3:] global n n = -base tnum1=numl[:3] loopri(order, tnum1, base) def rotate(irange): temp=irange[0] for i in range(0, len(irange)-1): irange[i] = irange[i+1] irange[len(irange)-1]=temp return(irange) def loopri(order, tnuml, base): irange = list(range(1, tnuml[0]+1)) if len(order) == 1: for i in range(0, order[0]): global n print(i+1) n = n + base disp 3nr(tnuml, n, irange) irange=rotate(irange) print( ) else: for i in range(0, order[-1]): print(i) loopr(order[:-1], tnuml, base, irange) irange=rotate(irange) def loopr(order, tnuml, base, irange): if len(order) == 1: for i in range(0, order[0]): global n n = n + base disp 3nr(tnuml, n, irange) print( ) else: for i in range(0, order[-1]): print(i) loopr(order[:-1], tnuml, base, irange) disp n([4,8,16]) 14

15 Implementation of the above Python program Output 1: Figure 4: 3-dimensional grid P 4 P 8 P 16 15

16 Output 2: Figure 5: 4-dimensional cylinder C 4 P 4 P 4 P 8 16

arxiv: v1 [cs.gr] 25 Nov 2015

arxiv: v1 [cs.gr] 25 Nov 2015 EMBEDDING OF HYPERCUBE INTO CYLINDER WEIXING JI, QINGHUI LIU, GUIZHEN WANG, AND ZHUOJIA SHEN arxiv:1511.07932v1 [cs.gr] 25 Nov 2015 Abstract. Task mapping in modern high performance parallel computers