Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682141
L. Rudolph
Several recent architectural trends appear to require some of the special properties of optical interconnection networks. The memory hierarchy will probably get deeper and the hardware will be more dynamic. Reconfigurable hardware and software are the subject of intense research. It is likely that the basic word size and the number of communication streams will be dynamically configurable. Optical interconnection networks, and in particular bit parallel free-space implementations, may provide the necessary flexibility.
{"title":"Do parallel computers really need optical interconnection networks?","authors":"L. Rudolph","doi":"10.1109/MPPOI.1998.682141","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682141","url":null,"abstract":"Several recent architectural trends appear to require some of the special properties of optical interconnection networks. The memory hierarchy will probably get deeper and the hardware will be more dynamic. Reconfigurable hardware and software are the subject of intense research. It is likely that the basic word size and the number of communication streams will be dynamically configurable. Optical interconnection networks, and in particular bit parallel free-space implementations, may provide the necessary flexibility.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"07 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125268897","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682129
C. Zhou, Yuanyuan Yang
Multicast is the ability to transmit information from a single source node to multiple destination nodes. Current trends in networking applications indicate that there will be an increasing demand in future networks for multicast. In this paper, we study multicast communication in a class of optical WDM networks with direct interconnects under some commonly used routing algorithms. We formalize multicast communication patterns by the maximum fanout of a source node, and obtain the necessary and sufficient conditions on the minimum number of wavelengths required for a network to be rearrangeable for arbitrary multicast communication. We give general results covering various multicast communication patterns from permutation to broadcast in rings, linear arrays, meshes, and hypercubes.
{"title":"Multicast communication in a class of rearrangeable optical WDM networks","authors":"C. Zhou, Yuanyuan Yang","doi":"10.1109/MPPOI.1998.682129","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682129","url":null,"abstract":"Multicast is the ability to transmit information from a single source node to multiple destination nodes. Current trends in networking applications indicate that there will be an increasing demand in future networks for multicast. In this paper, we study multicast communication in a class of optical WDM networks with direct interconnects under some commonly used routing algorithms. We formalize multicast communication patterns by the maximum fanout of a source node, and obtain the necessary and sufficient conditions on the minimum number of wavelengths required for a network to be rearrangeable for arbitrary multicast communication. We give general results covering various multicast communication patterns from permutation to broadcast in rings, linear arrays, meshes, and hypercubes.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126742502","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682132
A. Cangellaris
The semiconductor industry appears to be confident that, even without any major breakthroughs in photolithography, it will be able to achieve device feature sizes in the order of 100 nm by the year 2006. This reduction in feature size implies significantly higher device switching speeds and faster circuits. More specifically, microprocessors with 3 GHz on-chip clock frequency are within reach by the year 2006. The only obstacle to such an impressive level of performance, both across the chip and beyond the chip at the system level, is the availability of an interconnection network of unprecedented complexity and capable of supporting multi-GHz-bandwidth, distortion- and interference-free propagation both on-chip and off-chip. This interconnect bottleneck to multi-GHz processor realization is examined in this paper The emphasis is on the evaluation of the most promising and cost-effective electrical ways of overcoming this bottleneck. It is argued that some of these potential solutions involve technologies that are compatible with on-going developments in optical interconnects. Thus, an opportunity is identified for bringing together electrical and optical interconnect technologies at a level in the integration hierarchy where traditionally optics is considered to be at a disadvantage, namely, at the chip and chip-to-package interconnect level.
{"title":"The interconnect bottleneck in multi-GHz processors; new opportunities for hybrid electrical/optical solutions","authors":"A. Cangellaris","doi":"10.1109/MPPOI.1998.682132","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682132","url":null,"abstract":"The semiconductor industry appears to be confident that, even without any major breakthroughs in photolithography, it will be able to achieve device feature sizes in the order of 100 nm by the year 2006. This reduction in feature size implies significantly higher device switching speeds and faster circuits. More specifically, microprocessors with 3 GHz on-chip clock frequency are within reach by the year 2006. The only obstacle to such an impressive level of performance, both across the chip and beyond the chip at the system level, is the availability of an interconnection network of unprecedented complexity and capable of supporting multi-GHz-bandwidth, distortion- and interference-free propagation both on-chip and off-chip. This interconnect bottleneck to multi-GHz processor realization is examined in this paper The emphasis is on the evaluation of the most promising and cost-effective electrical ways of overcoming this bottleneck. It is argued that some of these potential solutions involve technologies that are compatible with on-going developments in optical interconnects. Thus, an opportunity is identified for bringing together electrical and optical interconnect technologies at a level in the integration hierarchy where traditionally optics is considered to be at a disadvantage, namely, at the chip and chip-to-package interconnect level.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131922002","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682138
Chih-Fang Wang, S. Sahni
In this paper we develop algorithms for some basic operations-broadcast, window broadcast, prefix sum, data sum, rank, shift, data accumulation, consecutive sum, adjacent sum, concentrate, distribute, generalize, sorting, random access read and write-on the OTIS-Mesh model. These operations are useful in the development of efficient algorithms for numerous application.
{"title":"Basic operations on the OTIS-Mesh optoelectronic computer","authors":"Chih-Fang Wang, S. Sahni","doi":"10.1109/MPPOI.1998.682138","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682138","url":null,"abstract":"In this paper we develop algorithms for some basic operations-broadcast, window broadcast, prefix sum, data sum, rank, shift, data accumulation, consecutive sum, adjacent sum, concentrate, distribute, generalize, sorting, random access read and write-on the OTIS-Mesh model. These operations are useful in the development of efficient algorithms for numerous application.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"180 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133692963","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682144
K. Teitelbaum
This paper describes an insertion opportunity for optical interconnection technology in embedded massively parallel processors for radar signal processing.
本文描述了在雷达信号处理的嵌入式大规模并行处理器中引入光互连技术的机会。
{"title":"Crossbar tree networks for embedded signal processing applications","authors":"K. Teitelbaum","doi":"10.1109/MPPOI.1998.682144","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682144","url":null,"abstract":"This paper describes an insertion opportunity for optical interconnection technology in embedded massively parallel processors for radar signal processing.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128652783","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682130
X. Yuan, R. Melhem
This paper considers optimal routing and channel assignment (RCA) schemes to realize hypercube communication on optical mesh-like networks. Specifically, we identify lower bounds on the number of channels required to realize hypercube communication on top of array and ring topologies and develop optimal RCA schemes that achieve the lower bounds on these two topologies. We further extend the schemes to mesh and torus topologies and obtain RCA schemes that use at most 2 more channels than the optimal for these topologies.
{"title":"Optimal routing and channel assignments for hypercube communication on optical mesh-like processor arrays","authors":"X. Yuan, R. Melhem","doi":"10.1109/MPPOI.1998.682130","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682130","url":null,"abstract":"This paper considers optimal routing and channel assignment (RCA) schemes to realize hypercube communication on optical mesh-like networks. Specifically, we identify lower bounds on the number of channels required to realize hypercube communication on top of array and ring topologies and develop optimal RCA schemes that achieve the lower bounds on these two topologies. We further extend the schemes to mesh and torus topologies and obtain RCA schemes that use at most 2 more channels than the optimal for these topologies.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125545815","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682148
M. Nakamura, K. Kitayama, Y. Igasaki, K. Kaneda
We experimentally demonstrate 100-m-long image fiber transmission of four-channel multiplexed two-dimensional signals. To upgrade the system throughput, we study several hundred Gb/s-class 2-D optical parallel data link using an image fiber for the transmission and 2-D arrays of vertical-cavity surface-emmitting laser diodes (VCSELs) and p-i-n photo-diodes (PDs) as the transmitter and the receiver respectively. This system employs space code division multiple access (Space-CDMA) to multiplex 2-D optical parallel signals. To establish multi-channel optical link between 2-D VCSEL array and PD array with a high alignment precision and a good repeatability, we develop a novel visual alignment technique using a micro-optic image fiber coupler, which consists of miniature cube beamsplitter and graded index (GRIN) rod lenses. The effectiveness of the visual alignment with the image fiber coupler is experimentally demonstrated. This result will encourage the application of optical space-CDMA using an image fiber and 2-D arrays of VCSELs and PDs to future high-throughput 2-D parallel data links connecting massively parallel processors.
{"title":"Visual alignment using image fiber micro-optics for long-distance, ultra-high throughput Space-CDMA based 2-D parallel optical data link","authors":"M. Nakamura, K. Kitayama, Y. Igasaki, K. Kaneda","doi":"10.1109/MPPOI.1998.682148","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682148","url":null,"abstract":"We experimentally demonstrate 100-m-long image fiber transmission of four-channel multiplexed two-dimensional signals. To upgrade the system throughput, we study several hundred Gb/s-class 2-D optical parallel data link using an image fiber for the transmission and 2-D arrays of vertical-cavity surface-emmitting laser diodes (VCSELs) and p-i-n photo-diodes (PDs) as the transmitter and the receiver respectively. This system employs space code division multiple access (Space-CDMA) to multiplex 2-D optical parallel signals. To establish multi-channel optical link between 2-D VCSEL array and PD array with a high alignment precision and a good repeatability, we develop a novel visual alignment technique using a micro-optic image fiber coupler, which consists of miniature cube beamsplitter and graded index (GRIN) rod lenses. The effectiveness of the visual alignment with the image fiber coupler is experimentally demonstrated. This result will encourage the application of optical space-CDMA using an image fiber and 2-D arrays of VCSELs and PDs to future high-throughput 2-D parallel data links connecting massively parallel processors.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129425024","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682140
A. Benner, L. Rudolph, E. Schenfeld, T. Sterling, T. Szymański
As part of the Department of Energy’s efforts to eliminate the need for nuclear testing, the DOE’s “Accelerated Strategic Computing Initiative” is funding the purchase of a series of extremely-high-performance computers from 1996 through 2004. These machines are expected to help the DOE ensure the safety, reliability, and performance of the nuclear weapons stockpile using computation-based methods, rather than testing-based methods.
{"title":"Design Options for Interconnecting a 100+ TFlop/sec Parallel Supercomputer in 2004","authors":"A. Benner, L. Rudolph, E. Schenfeld, T. Sterling, T. Szymański","doi":"10.1109/MPPOI.1998.682140","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682140","url":null,"abstract":"As part of the Department of Energy’s efforts to eliminate the need for nuclear testing, the DOE’s “Accelerated Strategic Computing Initiative” is funding the purchase of a series of extremely-high-performance computers from 1996 through 2004. These machines are expected to help the DOE ensure the safety, reliability, and performance of the nuclear weapons stockpile using computation-based methods, rather than testing-based methods.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124015428","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682131
C. Salisbury, Rami Melhem
Circuit-switching can be particularly useful in optical multiprocessor interconnection networks, where it may be desirable to provide all-optical connections. Circuit-switched Banyan networks can be built from simple switching elements that do not have logical processing or buffering capabilities. Distributed, dynamic control can be provided to establish circuits in the network in response to the changing needs of a parallel application. These control techniques interleave control and data in the same network, thus avoiding the need for a separate control network. When the switching elements provide broadcast connections, the control techniques can provide multicasting communication capabilities. Time division multiplexing (TDM) or wavelength division multiplexing (WDM) can be used to increase the number of data circuits provided and thus reduce the frequency of control operations.
{"title":"Multicast control in optical circuit-switched Banyan networks","authors":"C. Salisbury, Rami Melhem","doi":"10.1109/MPPOI.1998.682131","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682131","url":null,"abstract":"Circuit-switching can be particularly useful in optical multiprocessor interconnection networks, where it may be desirable to provide all-optical connections. Circuit-switched Banyan networks can be built from simple switching elements that do not have logical processing or buffering capabilities. Distributed, dynamic control can be provided to establish circuits in the network in response to the changing needs of a parallel application. These control techniques interleave control and data in the same network, thus avoiding the need for a separate control network. When the switching elements provide broadcast connections, the control techniques can provide multicasting communication capabilities. Time division multiplexing (TDM) or wavelength division multiplexing (WDM) can be used to increase the number of data circuits provided and thus reduce the frequency of control operations.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"94 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132222914","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1998-06-15DOI: 10.1109/MPPOI.1998.682147
Y. Li, J. Popelek, J. Rhee, L. Wang, T. Wang, K. Shum
Embedding of end-tapered, thin-cladding fiber bundles for board-level large bandwidth optical clock distributions is proposed and demonstrated. Fan-outs of up to 128-nodes on a printed circuit board of area as large as 13/spl times/19 cm/sup 2/ are experimentally demonstrated. Both silica and polymer fibers were tested. Performance parameters, such as fiber bending loss, attenuations, dispersions, coupling loss and uniformity were measured. Embedding of fibers on a double-layered chip-populated real PCB was performed.
{"title":"Demonstration of fiber-based board-level optical clock distributions","authors":"Y. Li, J. Popelek, J. Rhee, L. Wang, T. Wang, K. Shum","doi":"10.1109/MPPOI.1998.682147","DOIUrl":"https://doi.org/10.1109/MPPOI.1998.682147","url":null,"abstract":"Embedding of end-tapered, thin-cladding fiber bundles for board-level large bandwidth optical clock distributions is proposed and demonstrated. Fan-outs of up to 128-nodes on a printed circuit board of area as large as 13/spl times/19 cm/sup 2/ are experimentally demonstrated. Both silica and polymer fibers were tested. Performance parameters, such as fiber bending loss, attenuations, dispersions, coupling loss and uniformity were measured. Embedding of fibers on a double-layered chip-populated real PCB was performed.","PeriodicalId":248808,"journal":{"name":"Proceedings. Fifth International Conference on Massively Parallel Processing (Cat. No.98EX182)","volume":"14 2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1998-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129509017","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：