• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 22
  • 15
  • 3
  • 1
  • Tagged with
  • 43
  • 43
  • 20
  • 11
  • 10
  • 9
  • 8
  • 6
  • 6
  • 6
  • 6
  • 6
  • 6
  • 6
  • 5
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
11

Routing Algorithms For On Chip Networks

Atagoziyev, Maksat 01 December 2007 (has links) (PDF)
Network-on-Chip (NoC) is communication infrastructure for future multi-core Systems-on-Chip (SoCs). NoCs are expected to overcome scalability and performance limitations of Point-to-Point (P2P) and bus-based communication systems. The routing algorithm of a given NoC affects the performance of the system measured with respect to metrics such as latency, throughput and load distribution. In this thesis, the popular Orthogonal One Turn (O1TURN) and Dimension Order Routing algorithms (DOR) for 2D-meshes are implemented by computer simulation. Investigating the effect of parameters such as packet, buffer and topology sizes on the performance of the network, it is observed that the center of the network is loaded more than the edges. A new routing algorithm is proposed and evaluated to achieve a more balanced load distribution. The results show that this goal is achieved with a trade off in latency and throughput in DOR and O1TURN.
12

Cost-effective Fault Tolerant Routing In Networks On Chip

Adanova, Venera 01 September 2008 (has links) (PDF)
Growing complexity of Systems on Chip (SoC) introduces interconnection problems. As a solution for communication bottleneck the new paradigm, Networks on Chip (NoC), has been proposed. Along with high performance and reliability, NoC brings in area and energy constraints. In this thesis we mainly concentrate on keeping communication-centric design environment fault-tolerant while considering area overhead. The previous researches suggest the adoption solution for fault-tolerance from multiprocessor architectures. However, multiprocessor architectures have excessive reliance on buffering leading to costly solutions. We propose to reconsider general router model by introducing central buffers which reduces buffer size. Besides, we offer a new fault-tolerant routing algorithm which effectively utilizes buffers at hand without additional buffers out of detriment to performance.
13

Resource efficient communication in network-based reconfigurable on-chip systems

Mahr, Philipp January 2012 (has links)
The constantly growing capacity of reconfigurable devices allows simultaneous execution of complex applications on those devices. The mere diversity of applications deems it impossible to design an interconnection network matching the requirements of every possible application perfectly, leading to suboptimal performance in many cases. However, the architecture of the interconnection network is not the only aspect affecting performance of communication. The resource manager places applications on the device and therefore influences latency between communicating partners and overall network load. Communication protocols affect performance by introducing data and processing overhead putting higher load on the network and increasing resource demand. Approaching communication holistically not only considers the architecture of the interconnect, but communication-aware resource management, communication protocols and resource usage just as well. Incorporation of different parts of a reconfigurable system during design- and runtime and optimizing them with respect to communication demand results in more resource efficient communication. Extensive evaluation shows enhanced performance and flexibility, if communication on reconfigurable devices is regarded in a holistic fashion. / Die Leistungsfähigkeit rekonfigurierbarer Rechensysteme steigt kontinuierlich und ermöglicht damit die parallele Ausführung von immer mehr und immer größeren Anwendungen. Die Vielfalt an Anwendungen macht es allerdings unmöglich ein optimales Kommunikationsnetzwerk zu entwickeln, welches die Anforderung jeder denkbaren Anwendung berücksichtigt. Die Performanz des rekonfigurierbaren Rechensystems sinkt. Das Kommunikationsnetzwerk ist jedoch nicht der einzige Teil des Systems, der Einfluss auf die Kommunikation- sperformanz nimmt. Die Ressourcenverwaltung des Systems beeinflusst durch die Platzierung der Anwendungen die Latenz zwischen Kommunikationspartnern und die Kommunikationslast im Netzwerk. Kommunikationsprotokolle beeinträchtigen die Performanz der Kommunikation durch Daten und Rechen- overhead, die ebenso zu erhöhter Netzwerklast sowie Ressourcenanforderungen führen. In einem ganzheitlichen Kommunikationsansatz wird nicht nur das Kommunikationsnetzwerk berücksichtigt, sondern außerdem Ressourcenverwaltung, Kommunikationsprotokolle und die anderweitige Verwendung vorhandener, temporär ungenutzter Kommunikationsressourcen. Durch Einbeziehung dieser Aspekte während Entwurfs- und Laufzeit und durch Optimierung unter Berücksichtigung der Kommunikationsanforderungen, wird eine ressourceneneffizien tere Kommunikation erreicht. Ausführliche Evaluationen zeigen, dass eine ganzheitliche Betrachtung von Kommunikationsfaktoren, Verbesserungen von Performanz und Flexibilität erzielt.
14

Architecture and physical design for advanced networks-on-chip

Jang, Woo Young 01 June 2011 (has links)
The aggressive scaling of the semiconductor technology following the Moore’s Law has delivered true system-on-chip (SoC) integration. Network-on-chip (NoC) has been recently introduced as an effective solution for scalable on-chip communication since dedicated point-to-point (P2P) interconnection and shared bus architecture become performance and power bottlenecks in the SoCs. This dissertation studies three critical NoC challenges such as latency, power, and compatibility with emerging technologies in aspect of an architecture and physical design level. Latency is a key issue in NoC since the performance of applications considerably depends on resource sharing policies employed in an on-chip network. NoCs have been mainly developed to improve network-level performance that captures the inherent performance characteristics of a network itself, but the network-level optimizations are not directly related to application- or system-level performance. In addition, memory latency on NoC critically affects the performance of applications or systems. We propose a synchronous dynamic random access memory (SDRAM) aware NoC design to optimize memory throughput, latency, and design complexity. Furthermore, it is extended to an application-aware NoC design to provide the quality-of-service (QoS) of memory for various applications. NoC provides great on-chip communication. However, it brings no true relief to power budget when the on-chip network scales in terms of complexity/size and signal bandwidth. The combination of NoC and other techniques has the potential to reduce power. We study two power saving research topics for NoC: (a) we propose a voltage-frequency island (VFI) aware NoC optimization framework with a better tradeoff between power efficiency and design complexity to minimize both computation and on-chip communication power. (b) We formulate an application mapping problem to mixed integer quadratic programming (MIQP) with the purpose of reducing power consumption in various hard networks and develop highly efficient algorithms for the MIQP. Regarding NoC compatible with new technologies, we focus on three dimensional (3D) die integration based on through-silicon vias (TSVs). Since an on-chip network design has been subject to not only application constraints but also design/manufacturing constraints, a 3D NoC design is required for innovation in interconnection networks. We propose a chemical-mechanical polishing (CMP) aware application-specific 3D NoC design that minimizes TSV height variation, thus reduces bonding failure, and meanwhile optimizes conventional NoC design objectives such as hop count, wirelength, power, and area. / text
15

Many-core architecture for programmable hardware accelerator

Lee, Junghee 13 January 2014 (has links)
As the further development of single-core architectures faces seemingly insurmountable physical and technological limitations, computer designers have turned their attention to alternative approaches. One such promising alternative is the use of several smaller cores working in unison as a programmable hardware accelerator. It is clear that the vast – and, as yet, largely untapped – potential of hardware accelerators is coming to the forefront of computer architecture. There are many challenges that must be addressed for the programmable hardware accelerator to be realized in practice. In this thesis, load-balancing, on-chip communication, and an execution model are studied. Imbalanced distribution of workloads across the processing elements constitutes wasteful use of resources, which results in degrading the performance of the system. In this thesis, a hardware-based load-balancing technique is proposed, which is demonstrated to be more scalable than state-of-the-art loadbalancing techniques. To facilitate efficient communication among ever increasing number of cores, a scalable communication network is imperative. Packet switching networks-on-chip (NoC) is considered as a viable candidate for scalable communication fabric. The size of flit, which is a unit of flow control in NoC, is one of important design parameters that determine latency, throughput and cost of NoC routers. How to determine an optimal flit size is studied in this thesis and a novel router architecture is proposed, which overcomes a problem related with the flit size. This thesis also includes a new execution model and its supporting architecture. An event-driven model that is an extension of hardware description language is employed as an execution model. The dynamic scheduling and module-level prefetching for supporting the event-driven execution model are evaluated.
16

Designing fault tolerant NoCs to improve reliability on SoCs / Projeto de NoCs tolerantes a falhas para o aumento da confiabilidade em SoCs

Frantz, Arthur Pereira January 2007 (has links)
Com a redução das dimensões dos dispositivos nas tecnologias sub-micrônicas foi possível um grande aumento no número de IP cores integrados em um mesmo chip e consequentemente novas arquiteturas de comunicação são usadas bucando atingir os requisitos de desempenho e potência. As redes intra-chip (Networks-on-Chip) foram propostas como uma plataforma alternativa de comunicação capaz de prover interconexões e comunicação entre os cores de um mesmo chip, tratando questões como desempenho, consumo de energia e reusabilidade para grandes sistemas integrados. Por outro lado, a mesma evolução tecnológica dos processos nanométricos reduziu drasticamente a confiabilidade de circuitos integrados, tornando dispositivos e interconexões mais sensíveis a novos tipos de falhas. Erros podem ser gerados por variações no processo de fabricação ou mesmo pela susceptibilidade do projeto, quando este opera em um ambiente hostil. Na comunicação de NoCs as duas principais fontes de erros são falhas de crosstalk e soft errors. No passado, se assumia que interconexões não poderiam ser afetadas por soft errors, por não possuirem circuitos seqüenciais. Porém, quando NoCs são usadas, buffers e circuitos seqüenciais estão presentes nos roteadores e, consequentemente, podem ocorrer soft errors entre a fonte e o destino da comunicação, provocando erros. Técnicas de tolerância a falhas, que tem sido aplicadas em circuitos em geral, podem ser usadas para proteger roteadores contra bit-flips. Neste cenário, este trabalho inicia com a avaliação dos efeitos de soft errors e falhas de crosstalk em uma arquitetura de NoC, através de simulação de injeção de falhas, analisando detalhadamente o impacto de tais falhas no roteador. Os resultados mostram que os efeitos dessas falhas na comunicação do SoC podem ser desastrosos, levando a perda de pacotes e travamento ou indisponibilidade do sistema. Então é proposta e avaliada a aplicação de um conjunto de técnicas de tolerância a falhas em roteadores, possibilitando diminuir os soft errors e falhas de crosstalk no nível de hardware. Estas técnicas propostas foram baseadas em códigos de correção de erros e redundância de hardware. Resultados experimentais mostram que estas técnicas podem obter zero erros com 50% a menos de overhead de área, quando comparadas com a duplicação simples. Entretanto, algumas dessas técnicas têm um grande consumo de potência, pois toda essas técnicas são baseadas na adição de hardware redundante. Considerando que as técnicas de proteção baseadas em software também impõe um considerável overhead na comunicação devido à retransmissão, é proposto o uso de técnicas mistas de hardware e software, que podem oferecer um nível de proteção satisfatório, baseado na análise do ambiente onde o sistema irá operar (soft error rate), fatores relativos ao projeto e fabricação (variações de atraso em interconexões, pontos susceptíveis a crosstalk), a probabilidade de uma falha gerar um erro em um roteador, a carga de comunicação e os limites de potência e energia suportados. / As the technology scales down into deep sub-micron domain, more IP cores are integrated in the same die and new communication architectures are used to meet performance and power constraints. Networks-on-Chip have been proposed as an alternative communication platform capable of providing interconnections and communication among onchip cores, handling performance, energy consumption and reusability issues for large integrated systems. However, the same advances to nanometric technologies have significantly reduced reliability in mass-produced integrated circuits, increasing the sensitivity of devices and interconnects to new types of failures. Variations at the fabrication process or even the susceptibility of a design under a hostile environment might generate errors. In NoC communications the two major sources of errors are crosstalk faults and soft errors. In the past, it was assumed that connections cannot be affected by soft errors because there was no sequential circuit involved. However, when NoCs are used, buffers and sequential circuits are present in the routers, consequently, soft errors can occur between the communication source and destination provoking errors. Fault tolerant techniques that once have been applied in integrated circuits in general can be used to protect routers against bit-flips. In this scenario, this work starts evaluating the effects of soft errors and crosstalk faults in a NoC architecture by performing fault injection simulations, where it has been accurate analyzed the impact of such faults over the switch service. The results show that the effect of those faults in the SoC communication can be disastrous, leading to loss of packets and system crash or unavailability. Then it proposes and evaluates a set of fault tolerant techniques applied at routers able to mitigate soft errors and crosstalk faults at the hardware level. Such proposed techniques were based on error correcting codes and hardware redundancy. Experimental results show that using the proposed techniques one can obtain zero errors with up to 50% of savings in the area overhead when compared to simple duplication. However some of these techniques are very power consuming because all the tolerance is based on adding redundant hardware. Considering that softwarebased mitigation techniques also impose a considerable communication overhead due to retransmission, we then propose the use of mixed hardware-software techniques, that can develop a suitable protection scheme driven by the analysis of the environment that the system will operate in (soft error rate), the design and fabrication factors (delay variations in interconnects, crosstalk enabling points), the probability of a fault generating an error in the router, the communication load and the allowed power or energy budget.
17

Redes-em-chip para sistemas embarcados visando a otimização de medidas de qualidade de serviço para aplicações de tempo real / Networks on chip in embedded systems for optimization of quality of service measurement for real time applications

Corrêa, Edgard de Faria January 2007 (has links)
O avanço da tecnologia, com a possibilidade de inclusão de um número cada vez maior de transistores em uma única pastilha de silício, tem permitido integração de diversos blocos, formando sistemas completos em um único chip. Esses sistemas em chip possuem uma maior capacidade, mas também uma maior complexidade de projeto. Um dos aspectos a ser resolvido no projeto é que infra-estrutura de comunicação será utilizada na interconexão dos diversos blocos do sistema. Nos últimos anos, as propostas têm apontado para a utilização de redes em chip (NoC – do inglês, Network on Chip) para solucionar este problema de comunicação. Essas redes possuem capacidade de reuso de componentes, escalabilidade, paralelismo, embora apresentem maiores custos e latência que outras soluções. Entretanto, a latência pode ser atenuada, em alguns casos, através de ajustes na configuração da rede, tais como: topologia, arbitragem, mecanismos de controle de fluxo, política de roteamento, tamanho dos buffers. Por outro lado, os sistemas embarcados apresentam, geralmente, requisitos cada vez mais rígidos em relação à qualidade de serviço (QoS – do inglês, Quality of Service) e a restrições temporais. Dessa forma, esses requisitos temporais e de QoS aumentam ainda mais a complexidade do projeto de sistemas embarcados. Em virtude desse aumento da complexidade, o ideal é que a exploração do espaço de projeto seja feita no nível de abstração mais alto possível. Com isso, espera-se manter o tempo de projeto dentro dos níveis adequados, além de permitir uma exploração de espaço de projeto mais ampla e rápida. Nessa exploração, a configuração da rede têm impacto direto sobre os requisitos temporais e de QoS. Esta tese situa-se no contexto de investigar a influência da estrutura de comunicação no atendimento aos requisitos de QoS das aplicações de tempo real. Frente aos requisitos dessas aplicações, especificamente em relação ao atendimento dos deadlines das tarefas e a latência das comunicações, este trabalho apresenta mecanismos de ajustes no planejamento e configuração da NoC em sistemas embarcados, objetivando a garantia desses requisitos. As estratégias utilizadas nos ajustes das características da NoC objetivam permitir o uso mínimo de recursos para atender os requisitos das aplicações de tempo real, dentro das exigências de QoS. Os resultados apresentados comprovam que o ajuste correto nos parâmetros da estrutura de comunicação tem impacto direto no desempenho do sistema, especificamente em relação ao atendimento dos deadlines das mensagens e na redução da latência das comunicações. / With the technology advancing, a huge number of transistors can be included in a single chip. As a consequence, it is possible to integrate many blocks to build a complete system on a chip (SoC). These SoCs have more capacity, but their designs are more complex. One of the problems to solve is the design of the communication infrastructure to interconnect the systems blocks. In the last years, the utilization of networks as a solution for the communication problem has been proposed. These Networks-on-Chip (NoCs) have some interesting characteristics, such as reuse of components, scalability, and parallelism. On the other side, NoCs have higher costs and latency if compared to others solutions. The latency can be reduced, in some cases, by the adaptation of the network configuration, for instance adjusting topology, arbitration, flow control mechanisms, routing policy, size of buffers, etc. However, in general, embedded systems have increasingly rigid requirements regarding quality of service (QoS) and timing constraints. These timing and QoS requirements increase the complexity of embedded systems design. Due to this increased complexity, it is better that the design space exploration is performed at the highest possible abstraction level. With this, it is expected that the design time can be kept within adequate values, besides allowing a faster and broader design space exploration. In this exploration, the network configuration has direct impact upon timing and QoS requirements. The context of this thesis is the investigation of the influence of the communication structure on meeting QoS requirements in real time applications, in particular with respect to the fulfillment of task deadlines and latencies. This work shows mechanisms for adaptation of the NoC configuration for embedded systems, in order to meet the application requirements. The strategies used in the adjustment of the NoC characteristics allow the minimum use of resources to meet the real time application constraints, among the QoS requirements. The presented results demonstrate that the correct adjustment in the communication structure parameters has direct impact on the system performance, specifically with respect to the fulfillment of message deadlines and to the reduction of the communication latencies.
18

Métodos de teste de redes-em-chip (NoCs)

Hervé, Marcos Barcellos January 2009 (has links)
Este trabalho tem como objetivo estudar e propor métodos de teste funcional visando a detecção e localização de falhas na infra-estrutura das redes-em-chip. Para isso, o trabalho apresenta, inicialmente, uma descrição das principais características das redes-em-chip, explicando o que elas são e para que elas servem. Em seguida são apresentados conceitos de teste de circuitos integrados, bem como trabalhos relacionados ao teste das redes-em-chip. Um método de teste visando a detecção de falhas nas interconexões de dados de uma NoC é apresentado no trabalho, sendo este método posteriormente estendido para incluir as interconexões de controle. Os circuitos de teste necessários para implementar a estratégia de teste proposta também são descritos. A partir do método de teste apresentado, é feito um estudo sobre sua capacidade de localização de falhas, onde alterações visando o aumento dessa capacidade de localização de falhas são propostas. Por fim o método de teste é estendido para detecção de falhas nos roteadores da rede. / The purpose of this work is to study and propose functional test methods that aim the detection and location of faults in the NoC’s infrastructure. In order to do so, this work presents, initially, a description of the main characteristics of networks-on-chip, explaining what are NoCs and what is their purpose. Fallowing this description, some concepts related to the test of integrated circuits are presented as well as related works on NoC testing. A method aiming the detection of data interconnect faults in a NoC is presented in this work. This method is later extended to include faults in the control interconnections as well. The circuits used to implement the proposed strategy are also described here. Based on the proposed test strategy, the method’s capability to locate faults is studied. Changes are proposed to the test method in order to increase this fault location capability. Finally, the test method is extended to include faults inside the router’s logic.
19

Designing fault tolerant NoCs to improve reliability on SoCs / Projeto de NoCs tolerantes a falhas para o aumento da confiabilidade em SoCs

Frantz, Arthur Pereira January 2007 (has links)
Com a redução das dimensões dos dispositivos nas tecnologias sub-micrônicas foi possível um grande aumento no número de IP cores integrados em um mesmo chip e consequentemente novas arquiteturas de comunicação são usadas bucando atingir os requisitos de desempenho e potência. As redes intra-chip (Networks-on-Chip) foram propostas como uma plataforma alternativa de comunicação capaz de prover interconexões e comunicação entre os cores de um mesmo chip, tratando questões como desempenho, consumo de energia e reusabilidade para grandes sistemas integrados. Por outro lado, a mesma evolução tecnológica dos processos nanométricos reduziu drasticamente a confiabilidade de circuitos integrados, tornando dispositivos e interconexões mais sensíveis a novos tipos de falhas. Erros podem ser gerados por variações no processo de fabricação ou mesmo pela susceptibilidade do projeto, quando este opera em um ambiente hostil. Na comunicação de NoCs as duas principais fontes de erros são falhas de crosstalk e soft errors. No passado, se assumia que interconexões não poderiam ser afetadas por soft errors, por não possuirem circuitos seqüenciais. Porém, quando NoCs são usadas, buffers e circuitos seqüenciais estão presentes nos roteadores e, consequentemente, podem ocorrer soft errors entre a fonte e o destino da comunicação, provocando erros. Técnicas de tolerância a falhas, que tem sido aplicadas em circuitos em geral, podem ser usadas para proteger roteadores contra bit-flips. Neste cenário, este trabalho inicia com a avaliação dos efeitos de soft errors e falhas de crosstalk em uma arquitetura de NoC, através de simulação de injeção de falhas, analisando detalhadamente o impacto de tais falhas no roteador. Os resultados mostram que os efeitos dessas falhas na comunicação do SoC podem ser desastrosos, levando a perda de pacotes e travamento ou indisponibilidade do sistema. Então é proposta e avaliada a aplicação de um conjunto de técnicas de tolerância a falhas em roteadores, possibilitando diminuir os soft errors e falhas de crosstalk no nível de hardware. Estas técnicas propostas foram baseadas em códigos de correção de erros e redundância de hardware. Resultados experimentais mostram que estas técnicas podem obter zero erros com 50% a menos de overhead de área, quando comparadas com a duplicação simples. Entretanto, algumas dessas técnicas têm um grande consumo de potência, pois toda essas técnicas são baseadas na adição de hardware redundante. Considerando que as técnicas de proteção baseadas em software também impõe um considerável overhead na comunicação devido à retransmissão, é proposto o uso de técnicas mistas de hardware e software, que podem oferecer um nível de proteção satisfatório, baseado na análise do ambiente onde o sistema irá operar (soft error rate), fatores relativos ao projeto e fabricação (variações de atraso em interconexões, pontos susceptíveis a crosstalk), a probabilidade de uma falha gerar um erro em um roteador, a carga de comunicação e os limites de potência e energia suportados. / As the technology scales down into deep sub-micron domain, more IP cores are integrated in the same die and new communication architectures are used to meet performance and power constraints. Networks-on-Chip have been proposed as an alternative communication platform capable of providing interconnections and communication among onchip cores, handling performance, energy consumption and reusability issues for large integrated systems. However, the same advances to nanometric technologies have significantly reduced reliability in mass-produced integrated circuits, increasing the sensitivity of devices and interconnects to new types of failures. Variations at the fabrication process or even the susceptibility of a design under a hostile environment might generate errors. In NoC communications the two major sources of errors are crosstalk faults and soft errors. In the past, it was assumed that connections cannot be affected by soft errors because there was no sequential circuit involved. However, when NoCs are used, buffers and sequential circuits are present in the routers, consequently, soft errors can occur between the communication source and destination provoking errors. Fault tolerant techniques that once have been applied in integrated circuits in general can be used to protect routers against bit-flips. In this scenario, this work starts evaluating the effects of soft errors and crosstalk faults in a NoC architecture by performing fault injection simulations, where it has been accurate analyzed the impact of such faults over the switch service. The results show that the effect of those faults in the SoC communication can be disastrous, leading to loss of packets and system crash or unavailability. Then it proposes and evaluates a set of fault tolerant techniques applied at routers able to mitigate soft errors and crosstalk faults at the hardware level. Such proposed techniques were based on error correcting codes and hardware redundancy. Experimental results show that using the proposed techniques one can obtain zero errors with up to 50% of savings in the area overhead when compared to simple duplication. However some of these techniques are very power consuming because all the tolerance is based on adding redundant hardware. Considering that softwarebased mitigation techniques also impose a considerable communication overhead due to retransmission, we then propose the use of mixed hardware-software techniques, that can develop a suitable protection scheme driven by the analysis of the environment that the system will operate in (soft error rate), the design and fabrication factors (delay variations in interconnects, crosstalk enabling points), the probability of a fault generating an error in the router, the communication load and the allowed power or energy budget.
20

Redes-em-chip para sistemas embarcados visando a otimização de medidas de qualidade de serviço para aplicações de tempo real / Networks on chip in embedded systems for optimization of quality of service measurement for real time applications

Corrêa, Edgard de Faria January 2007 (has links)
O avanço da tecnologia, com a possibilidade de inclusão de um número cada vez maior de transistores em uma única pastilha de silício, tem permitido integração de diversos blocos, formando sistemas completos em um único chip. Esses sistemas em chip possuem uma maior capacidade, mas também uma maior complexidade de projeto. Um dos aspectos a ser resolvido no projeto é que infra-estrutura de comunicação será utilizada na interconexão dos diversos blocos do sistema. Nos últimos anos, as propostas têm apontado para a utilização de redes em chip (NoC – do inglês, Network on Chip) para solucionar este problema de comunicação. Essas redes possuem capacidade de reuso de componentes, escalabilidade, paralelismo, embora apresentem maiores custos e latência que outras soluções. Entretanto, a latência pode ser atenuada, em alguns casos, através de ajustes na configuração da rede, tais como: topologia, arbitragem, mecanismos de controle de fluxo, política de roteamento, tamanho dos buffers. Por outro lado, os sistemas embarcados apresentam, geralmente, requisitos cada vez mais rígidos em relação à qualidade de serviço (QoS – do inglês, Quality of Service) e a restrições temporais. Dessa forma, esses requisitos temporais e de QoS aumentam ainda mais a complexidade do projeto de sistemas embarcados. Em virtude desse aumento da complexidade, o ideal é que a exploração do espaço de projeto seja feita no nível de abstração mais alto possível. Com isso, espera-se manter o tempo de projeto dentro dos níveis adequados, além de permitir uma exploração de espaço de projeto mais ampla e rápida. Nessa exploração, a configuração da rede têm impacto direto sobre os requisitos temporais e de QoS. Esta tese situa-se no contexto de investigar a influência da estrutura de comunicação no atendimento aos requisitos de QoS das aplicações de tempo real. Frente aos requisitos dessas aplicações, especificamente em relação ao atendimento dos deadlines das tarefas e a latência das comunicações, este trabalho apresenta mecanismos de ajustes no planejamento e configuração da NoC em sistemas embarcados, objetivando a garantia desses requisitos. As estratégias utilizadas nos ajustes das características da NoC objetivam permitir o uso mínimo de recursos para atender os requisitos das aplicações de tempo real, dentro das exigências de QoS. Os resultados apresentados comprovam que o ajuste correto nos parâmetros da estrutura de comunicação tem impacto direto no desempenho do sistema, especificamente em relação ao atendimento dos deadlines das mensagens e na redução da latência das comunicações. / With the technology advancing, a huge number of transistors can be included in a single chip. As a consequence, it is possible to integrate many blocks to build a complete system on a chip (SoC). These SoCs have more capacity, but their designs are more complex. One of the problems to solve is the design of the communication infrastructure to interconnect the systems blocks. In the last years, the utilization of networks as a solution for the communication problem has been proposed. These Networks-on-Chip (NoCs) have some interesting characteristics, such as reuse of components, scalability, and parallelism. On the other side, NoCs have higher costs and latency if compared to others solutions. The latency can be reduced, in some cases, by the adaptation of the network configuration, for instance adjusting topology, arbitration, flow control mechanisms, routing policy, size of buffers, etc. However, in general, embedded systems have increasingly rigid requirements regarding quality of service (QoS) and timing constraints. These timing and QoS requirements increase the complexity of embedded systems design. Due to this increased complexity, it is better that the design space exploration is performed at the highest possible abstraction level. With this, it is expected that the design time can be kept within adequate values, besides allowing a faster and broader design space exploration. In this exploration, the network configuration has direct impact upon timing and QoS requirements. The context of this thesis is the investigation of the influence of the communication structure on meeting QoS requirements in real time applications, in particular with respect to the fulfillment of task deadlines and latencies. This work shows mechanisms for adaptation of the NoC configuration for embedded systems, in order to meet the application requirements. The strategies used in the adjustment of the NoC characteristics allow the minimum use of resources to meet the real time application constraints, among the QoS requirements. The presented results demonstrate that the correct adjustment in the communication structure parameters has direct impact on the system performance, specifically with respect to the fulfillment of message deadlines and to the reduction of the communication latencies.

Page generated in 0.381 seconds