@inbook{16693, author = {{Meyer auf der Heide, Friedhelm and Decker, Thomas}}, booktitle = {{Informatik ’97 Informatik als Innovationsmotor}}, isbn = {{9783540630661}}, issn = {{1431-472X}}, title = {{{Parallel Computing in Paderborn: The SFB 376 “Massive Parallelism — Algorithms, Design Methods, Applications”}}}, doi = {{10.1007/978-3-642-60831-5_22}}, year = {{1997}}, } @article{19958, author = {{Schwarze, Frank and Meyer auf der Heide, Friedhelm and Schröder, Klaus}}, journal = {{Euro-Par 1996}}, pages = {{299--306}}, title = {{{Routing on Networks of Optical Crossbars (Extended Abstract).}}}, volume = {{I}}, year = {{1996}}, } @techreport{17418, author = {{Czumaj, Artur and Meyer auf der Heide, Friedhelm and Stemann, Volker}}, title = {{{Contention Resolution in Hashing Based Shared Memory Simulations}}}, year = {{1996}}, } @inproceedings{17419, abstract = {{We present a parallel algorithm for the rendering of complex three-dimensional scenes. The algorithm runs across heterogeneous architectures of PC-clusters consisting of a visualization-node, equipped with a powerful graphics adapter, and cluster nodes requiring weaker graphics capabilities only. The visualization-node renders a mixture of scene objects and simplified meshes (Reliefboards). The cluster nodes assist the visualization-node by asynchronous computing of Reliefboards, which are used to replace and render distant parts of the scene. Our algorithm is capable of gaining significant speedups if the cluster's nodes provide weak graphics adapters only. We trade the number of cluster nodes off the scene objects' image quality.}}, author = {{Grigoriev, Dima and Karpinski, Marek and Meyer auf der Heide, Friedhelm and Smolensky, Roman}}, booktitle = {{Proc. of 28th ACM-STOC}}, pages = {{612--621}}, publisher = {{Eurographics Symposium on Parallel Graphics and Visualization}}, title = {{{A lower bound for randomized algebraic decision trees}}}, volume = {{65453}}, year = {{1996}}, } @inproceedings{17483, abstract = {{In this paper we develop a model for communication time on parallel computers consisting of processors and a service network, i.e., a network performing services like broadcast, synchronization, and global variables. The implementation of the service network is done on a free configurable Transputer network. Our cost model describes the communication time of accesses to global variables and consists of a multi-linear function. The cost model includes the parameters packet size, send hot spot, and the number of processors accessing global variables. These parameters influence the communication time in a high degree and capture important parameters like contention. We implement a Bitonic Sort and a Connected Components algorithm (among others) and we show that our model is able to predict the communication time within a 10% error if indirect service networks are used. The applications show that it is easy for a programmer to determine the parameter values for our model and that our new cost model precisely predicts the communication time of parallel algorithms. Furthermore, we minimize the communication time of accesses to global variables by finding a balance between the number of messages in the network and their size. Our model predicts the optimal values for these parameters which we validate by experiments. A modified implementation of our routing which determines on-line the optimal parameter values for an access to a global variable achieves good speed ups.}}, author = {{Fischer, Matthias and Rethmann, Jochen and Wachsmann, Alf}}, booktitle = {{3rd Workshop on Abstract Machine Models for Parallel and Distributed Computing (AMW '96)}}, isbn = {{905199267X}}, pages = {{13–27}}, publisher = {{IOS Press}}, title = {{{A Realistic Cost Model for the Communication Time in Parallel Programs}}}, year = {{1996}}, } @inbook{17564, author = {{Bäumker, Armin and Dittrich, Wolfgang and Meyer auf der Heide, Friedhelm and Rieping, Ingo}}, booktitle = {{Lecture Notes in Computer Science}}, isbn = {{9783540616276}}, issn = {{0302-9743}}, pages = {{369--376}}, title = {{{Realistic parallel algorithms: Priority queue operations and selection for the BSP* Model}}}, doi = {{10.1007/bfb0024725}}, year = {{1996}}, } @techreport{18352, abstract = {{In this report, we develop a cost model for the communication time on parallel computers consisting of processors and a service network, i.e., a network performing services like broadcast, synchronization, and global variables. Because we do not have a parallel computer at our disposal that is equipped with a service network, we emulate the service network on a reconfigurable Transputer network. Our cost model describes the communication time of accesses to global variables and consists of a multi­linear function. The cost model includes the parameters packet size, send hot spot (the number of messages sent out by one processor), and number of processors accessing global variables. We show that these parameters influence the communication time in a high degree and capture important parameters like network contention. We implement a Bitonic Sort, Sample Sort, Matrix Multiplication, and Connected Components algorithm, and we show that our model is able to predict the communication time within a 10% error if indirect service networks are used. The applications show that it is easy for a programer to determine the parameter values for our model and that our new cost model precisely predicts the communication time of parallel algorithms. We explore the interaction of hot spots and asynchrony and show that the influence of hot spots to the communication time is not as high as one would expect from theoretical considerations in a synchronous model. Therefore, we do not apprehend the hot spot in our cost model. Furthermore, we minimize the communication time of accesses to global variables by finding a balance between the number of messages in the network and their size. Our model predicts the optimal values for these parameters which we validate by experiments. A modified implementation of our routing which determines on­line the optimal parameter values for an access to a global variable achieves good speed ups. }}, author = {{Fischer, Matthias and Rethmann, Jochen and Wachsmann, Alf}}, title = {{{A Realistic Cost Model for the Communication Time in Parallel Programs on Parallel Computers Using a Service Hardware}}}, year = {{1996}}, } @phdthesis{2181, author = {{Scheideler, Christian}}, publisher = {{University of Paderborn, Germany}}, title = {{{Universal routing strategies}}}, year = {{1996}}, } @article{2182, author = {{Meyer auf der Heide, Friedhelm and Scheideler, Christian and Stemann, Volker}}, journal = {{Theor. Comput. Sci.}}, number = {{2}}, pages = {{245----281}}, title = {{{Exploiting Storage Redundancy to Speed up Randomized Shared Memory Simulations}}}, doi = {{10.1016/0304-3975(96)00032-1}}, year = {{1996}}, } @inproceedings{2183, author = {{Meyer auf der Heide, Friedhelm and Scheideler, Christian}}, booktitle = {{FOCS}}, pages = {{370----379}}, title = {{{Deterministic Routing with Bounded Buffers: Turning Offline into Online Protocols}}}, year = {{1996}}, }