Pub Date : 2008-10-01DOI: 10.1109/ICCD.2008.4751872
Marc-André Daigneault, J. Langlois, J. David
This paper presents a novel application specific instruction set processor specialized for block motion estimation. The proposed architecture includes an efficient register file system in terms of data reuse and parallel processing. Performances and area costs are presented for different levels of parallelism and register file dimensions. Various FPGA implementations of the architecture are further studied in order to present the most important factors affecting performance and hardware resource utilization. The proposed instruction extension block architecture enables acceleration by 3 orders of magnitude for full-search block matching algorithms.
{"title":"Application Specific Instruction set processor specialized for block motion estimation","authors":"Marc-André Daigneault, J. Langlois, J. David","doi":"10.1109/ICCD.2008.4751872","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751872","url":null,"abstract":"This paper presents a novel application specific instruction set processor specialized for block motion estimation. The proposed architecture includes an efficient register file system in terms of data reuse and parallel processing. Performances and area costs are presented for different levels of parallelism and register file dimensions. Various FPGA implementations of the architecture are further studied in order to present the most important factors affecting performance and hardware resource utilization. The proposed instruction extension block architecture enables acceleration by 3 orders of magnitude for full-search block matching algorithms.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"105 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134630833","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 : 2008-10-01DOI: 10.1109/ICCD.2008.4751850
Keisuke Inoue, M. Kaneko, T. Iwagaki
For recent and future nanometer-technology VLSIs, static and dynamic delay variations become a serious problem. In many cases, the hold constraint, as well as the setup constraint, becomes critical for latching a correct signal under delay variations. While the timing violation due to the fail of the setup constraint can be fixed by tuning a clock frequency or using a delayed latch, the timing violation due to the fail of the hold constraint cannot be fixed by those methods in general. Our approach to delay variations (in particular, the hold constraint) proposed in this paper is a novel register assignment strategy in high-level synthesis, which guarantees safe clocking by contra-data-direction (CDD) clocking. After the formulation of this new register assignment problem, we prove NP-hardness of the problem, and then derive an integer linear programming formulation for the problem. The proposed method receives a scheduled data flow graph, and generates a datapath having (1) robustness against delay variations, which is ensured by CDD-based register assignment, and (2) the minimum possible number of registers. Experimental results show the effectiveness of the proposed method for some benchmark circuits.
{"title":"Safe clocking register assignment in datapath synthesis","authors":"Keisuke Inoue, M. Kaneko, T. Iwagaki","doi":"10.1109/ICCD.2008.4751850","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751850","url":null,"abstract":"For recent and future nanometer-technology VLSIs, static and dynamic delay variations become a serious problem. In many cases, the hold constraint, as well as the setup constraint, becomes critical for latching a correct signal under delay variations. While the timing violation due to the fail of the setup constraint can be fixed by tuning a clock frequency or using a delayed latch, the timing violation due to the fail of the hold constraint cannot be fixed by those methods in general. Our approach to delay variations (in particular, the hold constraint) proposed in this paper is a novel register assignment strategy in high-level synthesis, which guarantees safe clocking by contra-data-direction (CDD) clocking. After the formulation of this new register assignment problem, we prove NP-hardness of the problem, and then derive an integer linear programming formulation for the problem. The proposed method receives a scheduled data flow graph, and generates a datapath having (1) robustness against delay variations, which is ensured by CDD-based register assignment, and (2) the minimum possible number of registers. Experimental results show the effectiveness of the proposed method for some benchmark circuits.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125373118","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 : 2008-10-01DOI: 10.1109/ICCD.2008.4751873
Dong Ye, Aravind Pavuluri, Carl A. Waldspurger, Brian Tsang, Bohuslav Rychlik, Steven Woo
We use a novel virtualization-based approach for computer architecture performance analysis. We present a case study analyzing a hypothetical hybrid main memory, which consists of a first-level DRAM augmented by a 10-100x slower second-level memory. This architecture is motivated by the recent emergence of lower-cost, higher-density, and lower-power alternative memory technologies. To model such a system, we customize a virtual machine monitor (VMM) with delay-simulation and instrumentation code. Benchmarks representing server, technical computing, and desktop productivity workloads are evaluated in virtual machines (VMs). Relative to baseline all-DRAM systems, these workloads experience widely varying performance degradation when run on hybrid main memory systems which have significant amounts of second-level memory.
{"title":"Prototyping a hybrid main memory using a virtual machine monitor","authors":"Dong Ye, Aravind Pavuluri, Carl A. Waldspurger, Brian Tsang, Bohuslav Rychlik, Steven Woo","doi":"10.1109/ICCD.2008.4751873","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751873","url":null,"abstract":"We use a novel virtualization-based approach for computer architecture performance analysis. We present a case study analyzing a hypothetical hybrid main memory, which consists of a first-level DRAM augmented by a 10-100x slower second-level memory. This architecture is motivated by the recent emergence of lower-cost, higher-density, and lower-power alternative memory technologies. To model such a system, we customize a virtual machine monitor (VMM) with delay-simulation and instrumentation code. Benchmarks representing server, technical computing, and desktop productivity workloads are evaluated in virtual machines (VMs). Relative to baseline all-DRAM systems, these workloads experience widely varying performance degradation when run on hybrid main memory systems which have significant amounts of second-level memory.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134400656","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 : 2008-10-01DOI: 10.1109/ICCD.2008.4751906
Amit Hadke, Tony Benavides, R. Amirtharajah, M. Farrens, V. Akella
We present OCDIMM (Optically Connected DIMM), a CPU-DRAM interface that uses multiwavelength optical interconnects. We show that OCDIMM is more scalable and offers higher bandwidth and lower latency than FBDIMM (Fully-Buffered DIMM), a state-of-the-art electrical alternative. Though OCDIMM is more power efficient than FBDIMM, we show that ultimately the total power consumption in the memory subsystem is a key impediment to scalability and thus to achieving truly balanced computing systems in the terascale era.
{"title":"Design and evaluation of an optical CPU-DRAM interconnect","authors":"Amit Hadke, Tony Benavides, R. Amirtharajah, M. Farrens, V. Akella","doi":"10.1109/ICCD.2008.4751906","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751906","url":null,"abstract":"We present OCDIMM (Optically Connected DIMM), a CPU-DRAM interface that uses multiwavelength optical interconnects. We show that OCDIMM is more scalable and offers higher bandwidth and lower latency than FBDIMM (Fully-Buffered DIMM), a state-of-the-art electrical alternative. Though OCDIMM is more power efficient than FBDIMM, we show that ultimately the total power consumption in the memory subsystem is a key impediment to scalability and thus to achieving truly balanced computing systems in the terascale era.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131987309","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 : 2008-10-01DOI: 10.1109/ICCD.2008.4751920
Shrirang M. Yardi, M. Hsiao
Adaptive micro-architectures aim to achieve greater energy efficiency by dynamically allocating computing resources to match the workload performance. The decisions of when to adapt (temporal dimension) and what to adapt (spatial dimension) are taken by a control algorithm based on an analysis of the power/performance tradeoffs in both dimensions. We perform a rigorous analysis to quantify the energy efficiency limits of fine-grained temporal and coordinated spatial adaptation of multiple architectural resources by casting the control algorithm as a constrained optimization problem. Our study indicates that coordinated adaptation can potentially improve energy efficiency by up to 60% as compared to static architectures and by up to 33% over algorithms that adapt resources in isolation. We also analyze synergistic application of coarse and fine grained adaptation and find modest improvements of up to 18% over optimized dynamic voltage/frequency scaling. Finally, we analyze several previous control algorithms to understand the underlying reasons for their inefficiency.
{"title":"Quantifying the energy efficiency of coordinated micro-architectural adaptation for multimedia workloads","authors":"Shrirang M. Yardi, M. Hsiao","doi":"10.1109/ICCD.2008.4751920","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751920","url":null,"abstract":"Adaptive micro-architectures aim to achieve greater energy efficiency by dynamically allocating computing resources to match the workload performance. The decisions of when to adapt (temporal dimension) and what to adapt (spatial dimension) are taken by a control algorithm based on an analysis of the power/performance tradeoffs in both dimensions. We perform a rigorous analysis to quantify the energy efficiency limits of fine-grained temporal and coordinated spatial adaptation of multiple architectural resources by casting the control algorithm as a constrained optimization problem. Our study indicates that coordinated adaptation can potentially improve energy efficiency by up to 60% as compared to static architectures and by up to 33% over algorithms that adapt resources in isolation. We also analyze synergistic application of coarse and fine grained adaptation and find modest improvements of up to 18% over optimized dynamic voltage/frequency scaling. Finally, we analyze several previous control algorithms to understand the underlying reasons for their inefficiency.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115680189","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 : 2008-10-01DOI: 10.1109/ICCD.2008.4751832
T. Panhofer, W. Friesenbichler, M. Delvai
The trend towards higher integration and faster operating speed leads to decreasing feature sizes and lower supply voltages in modern integrated circuits. These properties make the circuits more error-prone, requiring a fault tolerant implementation for applications demanding high reliability, e.g. space missions. In previous work we presented a concept how to obtain fault tolerant digital circuits by using asynchronous four-state logic (FSL). This type of logic already exhibits a high degree of fault tolerance where most faults simply halt the circuit (deadlock). The remaining types of faults are handled by temporal redundancy. Adding a deadlock detection unit and introducing the concept of self-healing cells (SHCs) leads to a highly reliable circuit that is able to tolerate even multiple faults. However our experiments revealed that some specific fault constellations neither cause a deadlock nor are they detected by a redundant calculation. We present two improved ways of error detection, which allow to capture even these types of faults. Further, a comparison between the size of an SHC and the achieved fault tolerance wrt. multiple faults is performed.
{"title":"Fault tolerant Four-State Logic by using Self-Healing Cells","authors":"T. Panhofer, W. Friesenbichler, M. Delvai","doi":"10.1109/ICCD.2008.4751832","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751832","url":null,"abstract":"The trend towards higher integration and faster operating speed leads to decreasing feature sizes and lower supply voltages in modern integrated circuits. These properties make the circuits more error-prone, requiring a fault tolerant implementation for applications demanding high reliability, e.g. space missions. In previous work we presented a concept how to obtain fault tolerant digital circuits by using asynchronous four-state logic (FSL). This type of logic already exhibits a high degree of fault tolerance where most faults simply halt the circuit (deadlock). The remaining types of faults are handled by temporal redundancy. Adding a deadlock detection unit and introducing the concept of self-healing cells (SHCs) leads to a highly reliable circuit that is able to tolerate even multiple faults. However our experiments revealed that some specific fault constellations neither cause a deadlock nor are they detected by a redundant calculation. We present two improved ways of error detection, which allow to capture even these types of faults. Further, a comparison between the size of an SHC and the achieved fault tolerance wrt. multiple faults is performed.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123934769","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 : 2008-10-01DOI: 10.1109/ICCD.2008.4751849
V. Honkote, B. Taskin
Timing closure and power envelopes for contemporary multi-core chips with high speed clock networks make the clock distribution design a challenging task. Resonant rotary clocking is a novel clocking technology for multi-gigahertz rate clock generation that provides minimal power dissipation. Rotary clocking implementations can easily provide independent synchronization of multiple cores as well. The traditional rotary clock design involves a regular array topology of oscillatory rings. In this paper, the rotary clock networks are designed and implemented using a custom ring topology. Custom ring topologies are advantageous as they reduce the total tapping wirelength for the registers tapping onto the oscillatory rings. A maze router based algorithm is developed for the implementation of custom topology rotary rings. In experiments performed on UCLA IBM R1-R5 benchmark circuits with the Elmore delay model, an improvement of 11.04% for register tapping wirelength is achieved on average.
当代多核芯片高速时钟网络的时序封闭和电源封装使得时钟分配设计成为一项具有挑战性的任务。谐振旋转时钟是一种新型的多千兆赫频率时钟产生技术，提供了最小的功耗。旋转时钟实现也可以很容易地提供多核的独立同步。传统的旋转时钟设计涉及振荡环的规则阵列拓扑结构。本文采用自定义环拓扑结构设计并实现了旋转时钟网络。自定义环拓扑是有利的，因为它们减少了敲入振荡环的寄存器的总敲击声长。提出了一种基于迷宫路由器的自定义拓扑旋转环实现算法。采用Elmore延迟模型在UCLA IBM R1-R5基准电路上进行实验，平均提高了11.04%的寄存器分接长度。
{"title":"Custom rotary clock router","authors":"V. Honkote, B. Taskin","doi":"10.1109/ICCD.2008.4751849","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751849","url":null,"abstract":"Timing closure and power envelopes for contemporary multi-core chips with high speed clock networks make the clock distribution design a challenging task. Resonant rotary clocking is a novel clocking technology for multi-gigahertz rate clock generation that provides minimal power dissipation. Rotary clocking implementations can easily provide independent synchronization of multiple cores as well. The traditional rotary clock design involves a regular array topology of oscillatory rings. In this paper, the rotary clock networks are designed and implemented using a custom ring topology. Custom ring topologies are advantageous as they reduce the total tapping wirelength for the registers tapping onto the oscillatory rings. A maze router based algorithm is developed for the implementation of custom topology rotary rings. In experiments performed on UCLA IBM R1-R5 benchmark circuits with the Elmore delay model, an improvement of 11.04% for register tapping wirelength is achieved on average.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124721054","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 : 2008-10-01DOI: 10.1109/ICCD.2008.4751934
Chuanjun Zhang, Bing Xue
High associativity is important for level-two cache designs [9]. Implementing CAM-based highly associative caches (CAM-HAC), however, is both costly in hardware and exhibits poor scalability. We propose to implement the CAM-HAC in macro-blocks to improve scalability. Each macro-block contains 128-row and 8-column of cache blocks. We name it Two dimensional Cache, or T-Cache. Each macro-block has an associativity equivalent to 128times8=1024-way. Twelve bits of the T-Cachepsilas tag are implemented by using CAM, while the remaining tag uses SRAM; Furthermore, random replacement is used in rows to balance cache sets usage while LRU is used in columns to select the victim from a row. The hardware complexity for replacement is reduced greatly compared to a traditional CAM-HAC using LRU solely. Experimental results show that the T-Cache achieves a 16% miss rate reduction over a traditional 8-way unified L2 cache. This translates into an average IPC improvement of 5% and as high as 18%. The T-Cache exhibits a 4% total memory access-related energy savings due to the reduction to applicationspsila execution time.
{"title":"Two dimensional highly associative level-two cache design","authors":"Chuanjun Zhang, Bing Xue","doi":"10.1109/ICCD.2008.4751934","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751934","url":null,"abstract":"High associativity is important for level-two cache designs [9]. Implementing CAM-based highly associative caches (CAM-HAC), however, is both costly in hardware and exhibits poor scalability. We propose to implement the CAM-HAC in macro-blocks to improve scalability. Each macro-block contains 128-row and 8-column of cache blocks. We name it Two dimensional Cache, or T-Cache. Each macro-block has an associativity equivalent to 128times8=1024-way. Twelve bits of the T-Cachepsilas tag are implemented by using CAM, while the remaining tag uses SRAM; Furthermore, random replacement is used in rows to balance cache sets usage while LRU is used in columns to select the victim from a row. The hardware complexity for replacement is reduced greatly compared to a traditional CAM-HAC using LRU solely. Experimental results show that the T-Cache achieves a 16% miss rate reduction over a traditional 8-way unified L2 cache. This translates into an average IPC improvement of 5% and as high as 18%. The T-Cache exhibits a 4% total memory access-related energy savings due to the reduction to applicationspsila execution time.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121530197","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 : 2008-10-01DOI: 10.1109/ICCD.2008.4751921
C. Strydis
This paper evaluates various instruction- and data-cache organizations in terms of performance, power, energy and area on a suitably selected biomedical benchmark suite. The benchmark suite consists of compression, encryption and data-integrity algorithms as well as real implant applications, all executed on biomedical input datasets. Results are used to drive the (micro)architectural design of a novel microprocessor targeting microelectronic implants. Our profiling study has revealed a L1 instruction-cache of 8 KB size (when relaxed area constraints are imposed) and a L1 data-cache of 4 KB size, both structured as 2-way associative caches, as optimal organizations for the envisioned implant processor.
{"title":"Suitable cache organizations for a novel biomedical implant processor","authors":"C. Strydis","doi":"10.1109/ICCD.2008.4751921","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751921","url":null,"abstract":"This paper evaluates various instruction- and data-cache organizations in terms of performance, power, energy and area on a suitably selected biomedical benchmark suite. The benchmark suite consists of compression, encryption and data-integrity algorithms as well as real implant applications, all executed on biomedical input datasets. Results are used to drive the (micro)architectural design of a novel microprocessor targeting microelectronic implants. Our profiling study has revealed a L1 instruction-cache of 8 KB size (when relaxed area constraints are imposed) and a L1 data-cache of 4 KB size, both structured as 2-way associative caches, as optimal organizations for the envisioned implant processor.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"62 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121918192","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 : 2008-10-01DOI: 10.1109/ICCD.2008.4751853
Shan Yan, Bill Lin
The increasing viability of three dimensional (3D) silicon integration technology has opened new opportunities for chip design innovations, including the prospect of extending emerging systems-on-chip (SoC) design paradigms based on networks-on-chip (NoC) interconnection architectures to 3D chip designs. In this paper, we consider the problem of designing application-specific 3D-NoC architectures that are optimized for a given application. We present novel 3D-NoC synthesis algorithms that make use of accurate power and delay models for 3D wiring with through-silicon vias. In particular, we present a very efficient 3D-NoC synthesis algorithm called ripup-reroute-and-router-merging (RRRM), that is based on a rip-up and reroute formulation for routing flows and a router merging procedure for network optimization. Experimental results on 3D-NoC design cases show that our synthesis results can on average achieve a 74% reduction in power consumption and a 17% reduction in hop count over regular 3D mesh implementations and a 52% reduction in power consumption and a 17% reduction in hop count over optimized 3D mesh implementations.
{"title":"Design of application-specific 3D Networks-on-Chip architectures","authors":"Shan Yan, Bill Lin","doi":"10.1109/ICCD.2008.4751853","DOIUrl":"https://doi.org/10.1109/ICCD.2008.4751853","url":null,"abstract":"The increasing viability of three dimensional (3D) silicon integration technology has opened new opportunities for chip design innovations, including the prospect of extending emerging systems-on-chip (SoC) design paradigms based on networks-on-chip (NoC) interconnection architectures to 3D chip designs. In this paper, we consider the problem of designing application-specific 3D-NoC architectures that are optimized for a given application. We present novel 3D-NoC synthesis algorithms that make use of accurate power and delay models for 3D wiring with through-silicon vias. In particular, we present a very efficient 3D-NoC synthesis algorithm called ripup-reroute-and-router-merging (RRRM), that is based on a rip-up and reroute formulation for routing flows and a router merging procedure for network optimization. Experimental results on 3D-NoC design cases show that our synthesis results can on average achieve a 74% reduction in power consumption and a 17% reduction in hop count over regular 3D mesh implementations and a 52% reduction in power consumption and a 17% reduction in hop count over optimized 3D mesh implementations.","PeriodicalId":345501,"journal":{"name":"2008 IEEE International Conference on Computer Design","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124170662","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：