In the context of real-time fault-tolerant scheduling in multiprocessor systems, Primary-backup scheme plays an important role. A backup copy is always preferred to be executed as passive backup copy whenever possible...In the context of real-time fault-tolerant scheduling in multiprocessor systems, Primary-backup scheme plays an important role. A backup copy is always preferred to be executed as passive backup copy whenever possible because it can take the advantages of backup copy de-allocation technique and overloading technique to improve schedulability. In this paper, we propose a novel efficient fault-tolerant ratemonotonic best-fit algorithm efficient fault-tolerant rate-monotonic best-fit (ERMBF) based on multiprocessors systems to enhance the schedulability. Unlike existing scheduling algorithms that start scheduling tasks with only one processor. ERMBF pre-allocates a certain amount of processors before starting scheduling tasks, which enlarge the searching spaces for tasks. Besides, when a new processor is allocated, we reassign the task copies that have already been assigned to the existing processors in order to find a superior tasks assignment configuration. These two strategies are all aiming at making as many backup copies as possible to be executed as passive status. As a result, ERMBF can use fewer processors to schedule a set of tasks without losing real-time and fault-tolerant capabilities of the system. Simulation results reveal that ERMBF significantly improves the schedulability over existing, comparable algorithms in literature.展开更多
The high-speed computational performance is gained at the cost of huge hardware resource,which restricts the application of high-accuracy algorithms because of the limited hardware cost in practical use.To solve the p...The high-speed computational performance is gained at the cost of huge hardware resource,which restricts the application of high-accuracy algorithms because of the limited hardware cost in practical use.To solve the problem,a novel method for designing the field programmable gate array(FPGA)-based non-uniform rational B-spline(NURBS) interpolator and motion controller,which adopts the embedded multiprocessor technique,is proposed in this study.The hardware and software design for the multiprocessor,one of which is for NURBS interpolation and the other for position servo control,is presented.Performance analysis and experiments on an X-Y table are carried out,hardware cost as well as consuming time for interpolation and motion control is compared with the existing methods.The experimental and comparing results indicate that,compared with the existing methods,the proposed method can reduce the hardware cost by 97.5% using higher-accuracy interpolation algorithm within the period of 0.5 ms.A method which ensures the real-time performance and interpolation accuracy,and reduces the hardware cost significantly is proposed,and it’s practical in the use of industrial application.展开更多
To decrease the cost of exchanging load information among processors, a dynamic load-balancing (DLB) algorithm which adopts multieast tree technology is proposed. The muhieast tree construction rules are also propos...To decrease the cost of exchanging load information among processors, a dynamic load-balancing (DLB) algorithm which adopts multieast tree technology is proposed. The muhieast tree construction rules are also proposed to avoid wrongly transferred or redundant DLB messages due to the overlapping of multicast trees. The proposed DLB algorithm is distributed controlled, sender initiated and can help heavily loaded processors with complete distribution of redundant loads with minimum number of executions. Experiments were executed to compare the effects of the proposed DLB algorithm and other three ones, the results prove the effectivity and practicability of the proposed algorithm in dealing with great scale compute-intensive tasks.展开更多
P k |fix| C max problem is a new scheduling problem based on the multiprocessor parallel job, and it is proved to be NP hard problem when k ≥3. This paper focuses on the case of k =3. Some new observations and new te...P k |fix| C max problem is a new scheduling problem based on the multiprocessor parallel job, and it is proved to be NP hard problem when k ≥3. This paper focuses on the case of k =3. Some new observations and new techniques for P 3 |fix| C max problem are offered. The concept of semi normal schedulings is introduced, and a very simple linear time algorithm Semi normal Algorithm for constructing semi normal schedulings is developed. With the method of the classical Graham List Scheduling, a thorough analysis of the optimal scheduling on a special instance is provided, which shows that the algorithm is an approximation algorithm of ratio of 9/8 for any instance of P 3|fix| C max problem, and improves the previous best ratio of 7/6 by M.X.Goemans.展开更多
In shared-memory bus-based multiprocessors, when the number of processors grows, the processors spend an increasing amount of time waiting for access to the bus (and shared memory). This contention reduces the perform...In shared-memory bus-based multiprocessors, when the number of processors grows, the processors spend an increasing amount of time waiting for access to the bus (and shared memory). This contention reduces the performance of processors and imposes a limitation of the number of processors that can be used efficiently in bus-based systems. Since the multi-processor’s performance depends upon many parameters which affect the performance in different ways, timed Petri nets are used to model shared-memory bus-based multiprocessors at the instruction execution level, and the developed models are used to study how the performance of processors changes with the number of processors in the system. The results illustrate very well the restriction on the number of processors imposed by the shared bus. All performance characteristics presented in this paper are obtained by discrete-event simulation of Petri net models.展开更多
This paper presents the mechanism of the bus arbitration in PI-MPS multiprocessor sys-tem,describes encode approach,arbiter timing states and uniqueness of master modular ininterconnection bus,and measures and analyse...This paper presents the mechanism of the bus arbitration in PI-MPS multiprocessor sys-tem,describes encode approach,arbiter timing states and uniqueness of master modular ininterconnection bus,and measures and analyses latency of bus arbitration as well.展开更多
The design of parallel algorithms is studied in this paper. These algorithms are applicable to shared memory MIMD machines In this paper, the emphasis is put on the methods for design of the efficient parallel algori...The design of parallel algorithms is studied in this paper. These algorithms are applicable to shared memory MIMD machines In this paper, the emphasis is put on the methods for design of the efficient parallel algorithms. The design of efficient parallel algorithms should be based on the following considerationst algorithm parallelism and the hardware-parallelism; granularity of the parallel algorithm, algorithm optimization according to the underling parallel machine. In this paper , these principles are applied to solve a model problem of the PDE. The speedup of the new method is high. The results were tested and evaluated on a shared memory MIMD machine. The practical results were agree with the predicted performance.展开更多
This paper considers the scheduling problem observed in chip sorting operation of LED manufacturing, where each lot (job) with release time have four operations to be processed on a set of processing stages without pr...This paper considers the scheduling problem observed in chip sorting operation of LED manufacturing, where each lot (job) with release time have four operations to be processed on a set of processing stages without pre-determined necessary route. Each stage has one and more identical sorting machines. The sorting machines scheduling problem can be treated as a four-stage multiprocessor open shop problem with dynamic job release, and the objective is minimizing the makespan in the paper. This problem is formulated into a mixed integer programming (MIP) model and empirically shows its computational intractability. Due to the computational intractability, a particle swarm optimization (PSO) algorithm is proposed. A series of computational experiments are conducted to evaluate the performance of the proposed PSO in comparison with exact solution on various small-size problem instances. The results show that the PSO algorithm could finds most optimal or better solutions in one second.展开更多
FMS is the basic and frontier technology of advanced manufacturing.Its critical compo-nent is FMS control system.Reconstructable fault-tolerant multiprocessor control system,YH-MCS,is the result of the research on the...FMS is the basic and frontier technology of advanced manufacturing.Its critical compo-nent is FMS control system.Reconstructable fault-tolerant multiprocessor control system,YH-MCS,is the result of the research on the high-performance and high-reliable FMS con-trol system.This paper describes its architecture,technology characteristics,academic valueand application potentiality.展开更多
The Fork-Join program consisting of K parallel tasks is a useful model for a large number of computing applications. When the parallel processor has multi-channels, later tasks may finish execution earlier than their ...The Fork-Join program consisting of K parallel tasks is a useful model for a large number of computing applications. When the parallel processor has multi-channels, later tasks may finish execution earlier than their earlier tasks and may join with tasks from other programs. This phenomenon is called exchangeable join (EJ), which introduces correlation to the task’s service time. In this work, we investigate the response time of multiprocessor systems with EJ with a new approach. We analyze two aspects of this kind of systems: exchangeable join (EJ) and the capacity constraint (CC). We prove that the system response time can be effectively reduced by EJ, while the reduced amount is constrained by the capacity of the multiprocessor. An upper bound model is constructed based on this analysis and a quick estimation algorithm is proposed. The approximation formula is verified by extensive simulation results, which show that the relative error of approximation is less than 5%.展开更多
This paper addresses the Energy-Aware Distributed Hybrid Flow Shop Scheduling Problem with Multiprocessor Tasks(EADHFSPMT)by considering two objectives simultaneously,i.e.,makespan and total energy consumption.It cons...This paper addresses the Energy-Aware Distributed Hybrid Flow Shop Scheduling Problem with Multiprocessor Tasks(EADHFSPMT)by considering two objectives simultaneously,i.e.,makespan and total energy consumption.It consists of three sub-problems,i.e.,job assignment between factories,job sequence in each factory,and machine allocation for each job.We present a mixed inter linear programming model and propose a Novel MultiObjective Evolutionary Algorithm based on Decomposition(NMOEA/D).We specially design a decoding scheme according to the characteristics of the EADHFSPMT.To initialize a population with certain diversity,four different rules are utilized.Moreover,a cooperative search is designed to produce new solutions based on different types of relationship between any solution and its neighbors.To enhance the quality of solutions,two local intensification operators are implemented according to the problem characteristics.In addition,a dynamic adjustment strategy for weight vectors is designed to balance the diversity and convergence,which can adaptively modify weight vectors according to the distribution of the non-dominated front.Extensive computational experiments are carried out by using a number of benchmark instances,which demonstrate the effectiveness of the above special designs.The statistical comparisons to the existing algorithms also verify the superior performances of the NMOEA/D.展开更多
In order to achieve maximization of parallelism, effective distribution of rendering tasks, balance between performance and flexibility in graphics processing pipeline, this article presents design, performance analys...In order to achieve maximization of parallelism, effective distribution of rendering tasks, balance between performance and flexibility in graphics processing pipeline, this article presents design, performance analysis and optimization for multi-core interactive graphics processing unit (MIGPU). This processor integrates twelve processing cores with specific instruction set architecture and many sophisticated application-specific accelerators into a 3D graphics engine. It is implemented on XC6VLX550T field programmable gate array (FPGA). MIGPU supports OpenGL2.0 with programmable front-end processor, vertex shader, plane clipper, geometry transformer, three-D clippers and pixel shaders. For boosting the performance of MIGPU, the relationship model is established between primitive types, vertices, pixels, and the effect of culling, clipping, and memory access, and shows a way to improve the speed up of the graphics pipeline. It is capable of assigning graphics rendering tasks to different processors for efficiency and flexibility. The pixel filling rate can reach to 40 Mpixel/s at its peak performance.展开更多
Efficiency of batch processing is becoming increasingly important for many modern commercial service centers, e.g., clusters and cloud computing datacenters. However, periodical resource contentions have become the ma...Efficiency of batch processing is becoming increasingly important for many modern commercial service centers, e.g., clusters and cloud computing datacenters. However, periodical resource contentions have become the major performance obstacles for concurrently running applications on mainstream CMP servers. I/O contention is such a kind of obstacle, which may impede both the co-running performance of batch jobs and the system throughput seriously. In this paper, a dynamic I/O-aware scheduling algorithm is proposed to lower the impacts of I/O contention and to enhance the co-running performance in batch processing. We set up our environment on an 8-socket, 64-core server in Dawning Linux Cluster. Fifteen workloads ranging from 8 jobs to 256 jobs are evaluated. Our experimental results show significant improvements on the throughputs of the workloads, which range from 7% to 431%. Meanwhile, noticeable improvements on the slowdown of workloads and the average runtime for each job can be achieved. These results show that a well-tuned dynamic I/O-aware scheduler is beneficial for batch-mode services. It can also enhance the resource utilization via throughput improvement on modern service platforms.展开更多
The Multiprocessor Priority Ceiling Protocol (MPCP) is a classic suspension-based real-time locking protocol for partitioned fixed-priority (P-FP) scheduling. However, existing blocking time analysis is pessimisti...The Multiprocessor Priority Ceiling Protocol (MPCP) is a classic suspension-based real-time locking protocol for partitioned fixed-priority (P-FP) scheduling. However, existing blocking time analysis is pessimistic under the P-FP + MPCP scheduling, which negatively impacts the schedulability for real-time tasks. In this paper, we model each task as an alternating sequence of normal and critical sections, and use both the best-case execution time (BCET) and the worst-case execution time (WCET) to describe the execution requirement for each section. Based on this model, a novel analysis is proposed to bound shared resource requests. This analysis uses BCET to derive the lower bound on the inter-arrival time for shared resource requests, and uses WCET to obtain the upper bound on the execution time of a task on critical sections during an arbitrary time interval of △t. Based on this analysis, improved blocking analysis and its associated worst-case response time (WCRT) analysis are proposed for P-FP + MPCP scheduling. Schedulability experiments indicate that the proposed method outperforms the existing methods and improves the schedulability significantly.展开更多
Significant advances in field-programmable gate arrays (FPGAs) have made it viable to explore innovative multiprocessor solutions on a single FPGA chip. For multiprocessors, an efficient communication network that m...Significant advances in field-programmable gate arrays (FPGAs) have made it viable to explore innovative multiprocessor solutions on a single FPGA chip. For multiprocessors, an efficient communication network that matches the needs of the target application is always critical to the overall performance. Wormhole packet-switching network-on-chip (NoC) solutions are replacing conventional shared buses to deal with scalability and complexity challenges coming along with the increasing number of processing elements (PEs). However, the quest for high performance networks has led to very complex and resource-expensive NoC designs, leaving little room for the real computing force, i.e., PEs. Moreover, many techniques offer very small performance gains or none at all when network traffic is light while increasing the resource usage of routers. We argue that computation is still the primary task of multiprocessors and sufficient resources should be reserved for PEs. This paper presents our novel design and implementation of a resource-efficient communication network for multiprocessors on FPGAs. We reduce not only the required number of routers for a given number of PEs by introducing a new PE-router topology, but also the resource requirement of each router. Our communication network relies on the NEWS channels to transfer packets in a pipelined fashion following the path determined by the routing network, The implementation results on various Xilinx FPGAs show good performance in the typical range of network load for multiprocessor applications.展开更多
Synchronization in parallel programs is a major performance bottleneck in multiprocessor systems. Shared data is protected by locks and a lot of time is spent on the competition arising at the lock hand-off. In order ...Synchronization in parallel programs is a major performance bottleneck in multiprocessor systems. Shared data is protected by locks and a lot of time is spent on the competition arising at the lock hand-off. In order to be serialized, requests to the same cache line can either be bounced (NACKed) or buffered in the coherence controller. In this paper, we focus mainly on systems whose coherence controllers buffer requests. In a lock hand-off, a burst of requests to the same line arrive at the coherence controller. During lock hand-off only the requests from the winning processor contribute to progress of the computation, since the winning processor is the only one that will advance the work. This key observation leads us to propose a hardware mechanism we call request bypassing, which allows requests from the winning processor to bypass the requests buffered in the coherence controller keeping the lock line. We present an inexpensive implementation of request bypassing that reduces the time spent on all the execution phases of a critical section (acquiring the lock, accessing shared data, and releasing the lock) and which, as a consequence, speeds up the whole parallel computation. This mechanism requires neither compiler or programmer support nor ISA or coherence protocol changes. By simulating a 32-processor system, we show that using request bypassing does not degrade but rather improves performance in three applications with low synchronization rates, while in those having a large amount of synchronization activity (the remaining four), we see reductions in execution time and in lock stall time ranging from 14% to 39% and from 52% to 7170, respectively. We compare request bypassing with a previously proposed technique called read combining and with a system that bounces requests, observing a significantly lower execution time with the bypassing scheme. Finally, we analyze the sensitivity of our results to some key hardware and software parameters.展开更多
As the number of cores in chip multiprocessors (CMPs) increases, cache coherence protocol has become a key issue in integration of chip multiprocessors. Supporting cache coherence protocol in large chip multiprocess...As the number of cores in chip multiprocessors (CMPs) increases, cache coherence protocol has become a key issue in integration of chip multiprocessors. Supporting cache coherence protocol in large chip multiprocessors still faces three hurdles: design complexity, performance and scalability. This paper proposes Cache Coherent Network on Chip (CCNoC), a scheme that decouples cache coherency maintenance from processors and shared L2 caches and implements it completely in network on chip to free up processors and shared L2 caches from the chore of maintaining coherency, thereby reduces design complexity of CMPs. In this way, CCNoC also improves the performance of cache coherence protocol through reducing directory access latency and enhances scalability by avoiding massive directories overhead in shared L2 caches. In CCNoC, coherence state caches and active directory caches are implemented in the network interface components of network on chip to maintain cache coherence states for blocks in L1 caches and manage directory information for recently accessed blocks in L2 caches respectively. CCNoC provides a scalable CMP framework to tackle cache coherency which is the foundation of CMP. This paper evaluates the performance of CCNoC. Experimental results show that for a 16-core system, CCNoC improves performance by 3% on average over the conventional chip multiprocessor and by 10% at best, while reduces storage overhead by 1.8% and saves directory storage by 88%, showing good scalability.展开更多
基金Supported by the National Basic Reseach Program of China (973 Program 2004 CB318200)
文摘In the context of real-time fault-tolerant scheduling in multiprocessor systems, Primary-backup scheme plays an important role. A backup copy is always preferred to be executed as passive backup copy whenever possible because it can take the advantages of backup copy de-allocation technique and overloading technique to improve schedulability. In this paper, we propose a novel efficient fault-tolerant ratemonotonic best-fit algorithm efficient fault-tolerant rate-monotonic best-fit (ERMBF) based on multiprocessors systems to enhance the schedulability. Unlike existing scheduling algorithms that start scheduling tasks with only one processor. ERMBF pre-allocates a certain amount of processors before starting scheduling tasks, which enlarge the searching spaces for tasks. Besides, when a new processor is allocated, we reassign the task copies that have already been assigned to the existing processors in order to find a superior tasks assignment configuration. These two strategies are all aiming at making as many backup copies as possible to be executed as passive status. As a result, ERMBF can use fewer processors to schedule a set of tasks without losing real-time and fault-tolerant capabilities of the system. Simulation results reveal that ERMBF significantly improves the schedulability over existing, comparable algorithms in literature.
基金supported by National Key Basic Research Program of China(973 ProgramGrant No.2011CB706804)+1 种基金Shanghai Municipal Science and Technology Commission of China(Grant No.11QH1401400)Research Project of State Key Laboratory of Mechanical System & Vibration of China(Grant No.MSVMS201102)
文摘The high-speed computational performance is gained at the cost of huge hardware resource,which restricts the application of high-accuracy algorithms because of the limited hardware cost in practical use.To solve the problem,a novel method for designing the field programmable gate array(FPGA)-based non-uniform rational B-spline(NURBS) interpolator and motion controller,which adopts the embedded multiprocessor technique,is proposed in this study.The hardware and software design for the multiprocessor,one of which is for NURBS interpolation and the other for position servo control,is presented.Performance analysis and experiments on an X-Y table are carried out,hardware cost as well as consuming time for interpolation and motion control is compared with the existing methods.The experimental and comparing results indicate that,compared with the existing methods,the proposed method can reduce the hardware cost by 97.5% using higher-accuracy interpolation algorithm within the period of 0.5 ms.A method which ensures the real-time performance and interpolation accuracy,and reduces the hardware cost significantly is proposed,and it’s practical in the use of industrial application.
基金the National Natural Science Foundation of China(69973007)
文摘To decrease the cost of exchanging load information among processors, a dynamic load-balancing (DLB) algorithm which adopts multieast tree technology is proposed. The muhieast tree construction rules are also proposed to avoid wrongly transferred or redundant DLB messages due to the overlapping of multicast trees. The proposed DLB algorithm is distributed controlled, sender initiated and can help heavily loaded processors with complete distribution of redundant loads with minimum number of executions. Experiments were executed to compare the effects of the proposed DLB algorithm and other three ones, the results prove the effectivity and practicability of the proposed algorithm in dealing with great scale compute-intensive tasks.
文摘P k |fix| C max problem is a new scheduling problem based on the multiprocessor parallel job, and it is proved to be NP hard problem when k ≥3. This paper focuses on the case of k =3. Some new observations and new techniques for P 3 |fix| C max problem are offered. The concept of semi normal schedulings is introduced, and a very simple linear time algorithm Semi normal Algorithm for constructing semi normal schedulings is developed. With the method of the classical Graham List Scheduling, a thorough analysis of the optimal scheduling on a special instance is provided, which shows that the algorithm is an approximation algorithm of ratio of 9/8 for any instance of P 3|fix| C max problem, and improves the previous best ratio of 7/6 by M.X.Goemans.
文摘In shared-memory bus-based multiprocessors, when the number of processors grows, the processors spend an increasing amount of time waiting for access to the bus (and shared memory). This contention reduces the performance of processors and imposes a limitation of the number of processors that can be used efficiently in bus-based systems. Since the multi-processor’s performance depends upon many parameters which affect the performance in different ways, timed Petri nets are used to model shared-memory bus-based multiprocessors at the instruction execution level, and the developed models are used to study how the performance of processors changes with the number of processors in the system. The results illustrate very well the restriction on the number of processors imposed by the shared bus. All performance characteristics presented in this paper are obtained by discrete-event simulation of Petri net models.
文摘This paper presents the mechanism of the bus arbitration in PI-MPS multiprocessor sys-tem,describes encode approach,arbiter timing states and uniqueness of master modular ininterconnection bus,and measures and analyses latency of bus arbitration as well.
文摘The design of parallel algorithms is studied in this paper. These algorithms are applicable to shared memory MIMD machines In this paper, the emphasis is put on the methods for design of the efficient parallel algorithms. The design of efficient parallel algorithms should be based on the following considerationst algorithm parallelism and the hardware-parallelism; granularity of the parallel algorithm, algorithm optimization according to the underling parallel machine. In this paper , these principles are applied to solve a model problem of the PDE. The speedup of the new method is high. The results were tested and evaluated on a shared memory MIMD machine. The practical results were agree with the predicted performance.
文摘This paper considers the scheduling problem observed in chip sorting operation of LED manufacturing, where each lot (job) with release time have four operations to be processed on a set of processing stages without pre-determined necessary route. Each stage has one and more identical sorting machines. The sorting machines scheduling problem can be treated as a four-stage multiprocessor open shop problem with dynamic job release, and the objective is minimizing the makespan in the paper. This problem is formulated into a mixed integer programming (MIP) model and empirically shows its computational intractability. Due to the computational intractability, a particle swarm optimization (PSO) algorithm is proposed. A series of computational experiments are conducted to evaluate the performance of the proposed PSO in comparison with exact solution on various small-size problem instances. The results show that the PSO algorithm could finds most optimal or better solutions in one second.
基金the Commission of science,Technology and Industry for National Defence
文摘FMS is the basic and frontier technology of advanced manufacturing.Its critical compo-nent is FMS control system.Reconstructable fault-tolerant multiprocessor control system,YH-MCS,is the result of the research on the high-performance and high-reliable FMS con-trol system.This paper describes its architecture,technology characteristics,academic valueand application potentiality.
基金Project supported by the National Natural Science Foundation of0 China (Nos. 60274011 and 60574067), and the Program for NewCentury Excellent Talents in University (No. NCET-04-0094), China
文摘The Fork-Join program consisting of K parallel tasks is a useful model for a large number of computing applications. When the parallel processor has multi-channels, later tasks may finish execution earlier than their earlier tasks and may join with tasks from other programs. This phenomenon is called exchangeable join (EJ), which introduces correlation to the task’s service time. In this work, we investigate the response time of multiprocessor systems with EJ with a new approach. We analyze two aspects of this kind of systems: exchangeable join (EJ) and the capacity constraint (CC). We prove that the system response time can be effectively reduced by EJ, while the reduced amount is constrained by the capacity of the multiprocessor. An upper bound model is constructed based on this analysis and a quick estimation algorithm is proposed. The approximation formula is verified by extensive simulation results, which show that the relative error of approximation is less than 5%.
基金supported by the National Natural Science Fund for Distinguished Young Scholars of China(No.61525304)the National Natural Science Foundation of China(No.61873328)。
文摘This paper addresses the Energy-Aware Distributed Hybrid Flow Shop Scheduling Problem with Multiprocessor Tasks(EADHFSPMT)by considering two objectives simultaneously,i.e.,makespan and total energy consumption.It consists of three sub-problems,i.e.,job assignment between factories,job sequence in each factory,and machine allocation for each job.We present a mixed inter linear programming model and propose a Novel MultiObjective Evolutionary Algorithm based on Decomposition(NMOEA/D).We specially design a decoding scheme according to the characteristics of the EADHFSPMT.To initialize a population with certain diversity,four different rules are utilized.Moreover,a cooperative search is designed to produce new solutions based on different types of relationship between any solution and its neighbors.To enhance the quality of solutions,two local intensification operators are implemented according to the problem characteristics.In addition,a dynamic adjustment strategy for weight vectors is designed to balance the diversity and convergence,which can adaptively modify weight vectors according to the distribution of the non-dominated front.Extensive computational experiments are carried out by using a number of benchmark instances,which demonstrate the effectiveness of the above special designs.The statistical comparisons to the existing algorithms also verify the superior performances of the NMOEA/D.
基金supported by the Key National Natural Science Foundation of China under Grant No.61136002the National Natural Science Foundation of China under Grant No.61272120the Natural Science Basic Research Plan in Shaanxi Province of China under Grant No.2013JC2-32
文摘In order to achieve maximization of parallelism, effective distribution of rendering tasks, balance between performance and flexibility in graphics processing pipeline, this article presents design, performance analysis and optimization for multi-core interactive graphics processing unit (MIGPU). This processor integrates twelve processing cores with specific instruction set architecture and many sophisticated application-specific accelerators into a 3D graphics engine. It is implemented on XC6VLX550T field programmable gate array (FPGA). MIGPU supports OpenGL2.0 with programmable front-end processor, vertex shader, plane clipper, geometry transformer, three-D clippers and pixel shaders. For boosting the performance of MIGPU, the relationship model is established between primitive types, vertices, pixels, and the effect of culling, clipping, and memory access, and shows a way to improve the speed up of the graphics pipeline. It is capable of assigning graphics rendering tasks to different processors for efficiency and flexibility. The pixel filling rate can reach to 40 Mpixel/s at its peak performance.
基金Supported by the National High Technology Research and Development 863 Program of China under Grant No.2012AA010902the National Basic Research 973 Program of China under Grant No.2011CB302504the National Natural Science Foundation of China under Grant Nos.61202055,60925009,60921002,61100011
文摘Efficiency of batch processing is becoming increasingly important for many modern commercial service centers, e.g., clusters and cloud computing datacenters. However, periodical resource contentions have become the major performance obstacles for concurrently running applications on mainstream CMP servers. I/O contention is such a kind of obstacle, which may impede both the co-running performance of batch jobs and the system throughput seriously. In this paper, a dynamic I/O-aware scheduling algorithm is proposed to lower the impacts of I/O contention and to enhance the co-running performance in batch processing. We set up our environment on an 8-socket, 64-core server in Dawning Linux Cluster. Fifteen workloads ranging from 8 jobs to 256 jobs are evaluated. Our experimental results show significant improvements on the throughputs of the workloads, which range from 7% to 431%. Meanwhile, noticeable improvements on the slowdown of workloads and the average runtime for each job can be achieved. These results show that a well-tuned dynamic I/O-aware scheduler is beneficial for batch-mode services. It can also enhance the resource utilization via throughput improvement on modern service platforms.
基金supported by the National Natural Science Foundation of China under Grant No.61103041the National High Technology Research and Development 863 Program of China under Grant No.2012AA010904+2 种基金the Fundamental Research Funds for the Central Universities of China under Grant No.ZYGX2012J070the Huawei Technology Foundation under Grant No.IRP-2012-02-07the Excellent Ph.D.Student Academic Support Program of UESTC under Grant No.YBXSZC20131028
文摘The Multiprocessor Priority Ceiling Protocol (MPCP) is a classic suspension-based real-time locking protocol for partitioned fixed-priority (P-FP) scheduling. However, existing blocking time analysis is pessimistic under the P-FP + MPCP scheduling, which negatively impacts the schedulability for real-time tasks. In this paper, we model each task as an alternating sequence of normal and critical sections, and use both the best-case execution time (BCET) and the worst-case execution time (WCET) to describe the execution requirement for each section. Based on this model, a novel analysis is proposed to bound shared resource requests. This analysis uses BCET to derive the lower bound on the inter-arrival time for shared resource requests, and uses WCET to obtain the upper bound on the execution time of a task on critical sections during an arbitrary time interval of △t. Based on this analysis, improved blocking analysis and its associated worst-case response time (WCRT) analysis are proposed for P-FP + MPCP scheduling. Schedulability experiments indicate that the proposed method outperforms the existing methods and improves the schedulability significantly.
文摘Significant advances in field-programmable gate arrays (FPGAs) have made it viable to explore innovative multiprocessor solutions on a single FPGA chip. For multiprocessors, an efficient communication network that matches the needs of the target application is always critical to the overall performance. Wormhole packet-switching network-on-chip (NoC) solutions are replacing conventional shared buses to deal with scalability and complexity challenges coming along with the increasing number of processing elements (PEs). However, the quest for high performance networks has led to very complex and resource-expensive NoC designs, leaving little room for the real computing force, i.e., PEs. Moreover, many techniques offer very small performance gains or none at all when network traffic is light while increasing the resource usage of routers. We argue that computation is still the primary task of multiprocessors and sufficient resources should be reserved for PEs. This paper presents our novel design and implementation of a resource-efficient communication network for multiprocessors on FPGAs. We reduce not only the required number of routers for a given number of PEs by introducing a new PE-router topology, but also the resource requirement of each router. Our communication network relies on the NEWS channels to transfer packets in a pipelined fashion following the path determined by the routing network, The implementation results on various Xilinx FPGAs show good performance in the typical range of network load for multiprocessor applications.
基金supported in part by Spanish Government and European ERDF under Grant Nos. TIN2007-66423, TIN2010-21291-C02-01 and TIN2007-60625gaZ:T48 research group (Arag'on Government and European ESF)+1 种基金Consolider CSD2007-00050 (Spanish Government)HiPEAC-2 NoE (European FP7/ICT 217068)
文摘Synchronization in parallel programs is a major performance bottleneck in multiprocessor systems. Shared data is protected by locks and a lot of time is spent on the competition arising at the lock hand-off. In order to be serialized, requests to the same cache line can either be bounced (NACKed) or buffered in the coherence controller. In this paper, we focus mainly on systems whose coherence controllers buffer requests. In a lock hand-off, a burst of requests to the same line arrive at the coherence controller. During lock hand-off only the requests from the winning processor contribute to progress of the computation, since the winning processor is the only one that will advance the work. This key observation leads us to propose a hardware mechanism we call request bypassing, which allows requests from the winning processor to bypass the requests buffered in the coherence controller keeping the lock line. We present an inexpensive implementation of request bypassing that reduces the time spent on all the execution phases of a critical section (acquiring the lock, accessing shared data, and releasing the lock) and which, as a consequence, speeds up the whole parallel computation. This mechanism requires neither compiler or programmer support nor ISA or coherence protocol changes. By simulating a 32-processor system, we show that using request bypassing does not degrade but rather improves performance in three applications with low synchronization rates, while in those having a large amount of synchronization activity (the remaining four), we see reductions in execution time and in lock stall time ranging from 14% to 39% and from 52% to 7170, respectively. We compare request bypassing with a previously proposed technique called read combining and with a system that bounces requests, observing a significantly lower execution time with the bypassing scheme. Finally, we analyze the sensitivity of our results to some key hardware and software parameters.
基金supported by the National Natural Science Foundation of China under Grant Nos.60970002,60833004,60773146, and 60673145.
文摘As the number of cores in chip multiprocessors (CMPs) increases, cache coherence protocol has become a key issue in integration of chip multiprocessors. Supporting cache coherence protocol in large chip multiprocessors still faces three hurdles: design complexity, performance and scalability. This paper proposes Cache Coherent Network on Chip (CCNoC), a scheme that decouples cache coherency maintenance from processors and shared L2 caches and implements it completely in network on chip to free up processors and shared L2 caches from the chore of maintaining coherency, thereby reduces design complexity of CMPs. In this way, CCNoC also improves the performance of cache coherence protocol through reducing directory access latency and enhances scalability by avoiding massive directories overhead in shared L2 caches. In CCNoC, coherence state caches and active directory caches are implemented in the network interface components of network on chip to maintain cache coherence states for blocks in L1 caches and manage directory information for recently accessed blocks in L2 caches respectively. CCNoC provides a scalable CMP framework to tackle cache coherency which is the foundation of CMP. This paper evaluates the performance of CCNoC. Experimental results show that for a 16-core system, CCNoC improves performance by 3% on average over the conventional chip multiprocessor and by 10% at best, while reduces storage overhead by 1.8% and saves directory storage by 88%, showing good scalability.