A problem on the computational complexity of graph partitions 關(guān)于圖劃分中的一個計算復(fù)雜性問題
For more on improving sparse matrix algorithms, read fast and effective algorithms for graph partitioning and sparse-matrix ordering 如果您想了解更多關(guān)于提高稀疏矩陣算法的信息,請閱讀
With the help of metis, which is a software package in static graph partitioning, we construct a type of block jacobi preconditioner j時十u個1任用問題,分別用該分解作近似川ji’d寸十八’。
Experiments show that the new optimized partition algorithm can improve the parallel simulation performance compared with the ones partitioned by traditional graph partition algorithms 實驗表明,該優(yōu)化劃分方法相對于傳統(tǒng)的圖劃分算法能夠有效地提高模擬性能。
We have studied how to distill a connection matrix between two nodes in a generalized switched network; using tarjan ’ s graph partition algorithm and nonseparable two-terminal graph concept, we proposed an effective algorithm 對從廣義開關(guān)網(wǎng)絡(luò)中提取出兩點間連接矩陣的方法進(jìn)行了研究,利用tarjan的拓?fù)鋱D塊劃分算法、開關(guān)網(wǎng)絡(luò)中的不可分二端圖概念等設(shè)計了較有效的算法。
One of the main problems in pdes is how to partition the network simulation workload to decrease the time needed to complete the simulation and improve performance of simulation . here a new optimized partition algorithm was put forward, which first analyses the performance factors of parallel simulation and then constructs a performance estimation model for partition; based on this model, it mends the graph partition algorithm to consider all factors, including simulation applications and simulation environments . optimized factors are workload balance 、 communication cost and time window of lookhead 鑒于基于傳統(tǒng)圖劃分算法的任務(wù)劃分工具存在諸多不足,本文提出了并行網(wǎng)絡(luò)模擬任務(wù)的優(yōu)化劃分方法,其思想是:首先從pdes機(jī)制出發(fā),分析影響并行網(wǎng)絡(luò)模擬性能的各種因素并建立一個能夠綜合考慮各種因素的并行網(wǎng)絡(luò)模擬性能估計模型;其次,改進(jìn)多級圖劃分算法,使得算法具有綜合考慮模擬應(yīng)用與模擬運行環(huán)境,同時在優(yōu)化過程中使用性能估計模型指導(dǎo)劃分,實現(xiàn)對影響并行網(wǎng)絡(luò)模擬性能的三個因素(包括負(fù)載均衡、通信開銷、安全時間窗口長度)的優(yōu)化能力;最后,結(jié)合并行網(wǎng)絡(luò)模擬性能估計模型與改進(jìn)的多極圖劃分算法,實現(xiàn)了并行網(wǎng)絡(luò)模擬任務(wù)的優(yōu)化劃分。
Some crucial design principles, methods and techniques in petsc are highlighted . propose a general scheme for the parallelization of an unstructured mesh, including graph partitioning, data division and mesh managing, based on message passing programming style . propose a series of comprehensive considerations about how to achieve a well-designed, object-oriented and data-distributed parallel software ?從軟件外在的功能組織、使用模式與內(nèi)在的設(shè)計思想、實現(xiàn)技術(shù)及其因果關(guān)聯(lián)性等多重角度出發(fā),對petsc做了深入的分析探討,尤其從源代碼的層次,重點剖析了petsc的面向?qū)ο蟆⒅行詳?shù)據(jù)結(jié)構(gòu)、上下文環(huán)境、并行設(shè)計與通信等重要核心技術(shù)的實現(xiàn)方法及其對petsc使用模式和性能所產(chǎn)生的意義。
The basis is hierarchical placement based on the model of directed graph partition . introducing timing-driven placement to clustering, it is an excellent method that mixes the characteristics of timing-driven placement, clustering and directed graph partition . integrated with cdp, this algorithm can solve the system-level vfg problems of electronic devices 該算法是基于有向圖分解模型的分級布局算法,將定時驅(qū)動布局思想引入結(jié)群過程,是集定時驅(qū)動布局、基于結(jié)群的布局、基于劃分的布局和有向圖分解思想為一體的快速布局算法。
The basis is hierarchical placement based on the model of directed graph partition . introducing timing-driven placement to clustering, it is an excellent method that mixes the characteristics of timing-driven placement, clustering and directed graph partition . integrated with cdp, this algorithm can solve the system-level vfg problems of electronic devices 該算法是基于有向圖分解模型的分級布局算法,將定時驅(qū)動布局思想引入結(jié)群過程,是集定時驅(qū)動布局、基于結(jié)群的布局、基于劃分的布局和有向圖分解思想為一體的快速布局算法。