| id |
ndltd-OhioLink-oai-etd.ohiolink.edu-ucin1134584775
|
| record_format |
oai_dc
|
| spelling |
ndltd-OhioLink-oai-etd.ohiolink.edu-ucin11345847752021-08-03T06:10:42Z EXPLOITING INCREMENTAL SPARSE MATRIX UPDATES TO IMPROVE EVALUATION SPEED OF CONTINUOUS TIME SYSTEMS SRIDHARAN, MAHESH Simulation provides the means to validate electrical circuits and also to perform behavioral analysis. As the circuit size gets larger the time consumed by simulation becomes a major bottleneck in performance. There are several possible approaches to solve the equations which describe a given circuit. Typically equations are solved using the direct methods. The purpose of this research is to examine the feasibility of increasing the speed of solving sparse matrices, keeping continuous time systems in mind. Evaluation of such systems is performed by breaking the time period into small time steps and evaluating the matrices at each time step. A matrix describing a circuit is generally sparse and consists of several latent points. At every time step of the evaluation, this matrix is updated. It is observed that not much of the matrix changes per iteration. Thus only certain non-zero elements of the matrix are updated at each time step. This thesis aims to explore the reduction in computation time achieved by applying iterative methods to the sparse matrices, and exploiting incremental matrix updates to achieve speedup. Iterative algorithms often converge faster if the "initial guess" is intelligent. But if an intelligent guess is to be made at every time step of analog simulation, it becomes too expensive. Two algorithms are proposed in this write up, which use the solution of the previous time step is used as the "intelligent guess" of the next time step and the speed up thus achieved is compared to that of the direct method. The algorithms have been prototyped and their behavior has been analyzed with regard to specific test cases. To further ensure robustness, the second algorithm is modified and has a direct engine wrapped by iterative algorithms. Thus non convergence issues are addressed. A polynomial gain of order 2 is observed for certain algorithms within the domain of the test cases considered, the comparison being made against the direct method. 2005 English text University of Cincinnati / OhioLINK http://rave.ohiolink.edu/etdc/view?acc_num=ucin1134584775 http://rave.ohiolink.edu/etdc/view?acc_num=ucin1134584775 unrestricted This thesis or dissertation is protected by copyright: all rights reserved. It may not be copied or redistributed beyond the terms of applicable copyright laws.
|
| collection |
NDLTD
|
| language |
English
|
| sources |
NDLTD
|
| author |
SRIDHARAN, MAHESH
|
| spellingShingle |
SRIDHARAN, MAHESH
EXPLOITING INCREMENTAL SPARSE MATRIX UPDATES TO IMPROVE EVALUATION SPEED OF CONTINUOUS TIME SYSTEMS
|
| author_facet |
SRIDHARAN, MAHESH
|
| author_sort |
SRIDHARAN, MAHESH
|
| title |
EXPLOITING INCREMENTAL SPARSE MATRIX UPDATES TO IMPROVE EVALUATION SPEED OF CONTINUOUS TIME SYSTEMS
|
| title_short |
EXPLOITING INCREMENTAL SPARSE MATRIX UPDATES TO IMPROVE EVALUATION SPEED OF CONTINUOUS TIME SYSTEMS
|
| title_full |
EXPLOITING INCREMENTAL SPARSE MATRIX UPDATES TO IMPROVE EVALUATION SPEED OF CONTINUOUS TIME SYSTEMS
|
| title_fullStr |
EXPLOITING INCREMENTAL SPARSE MATRIX UPDATES TO IMPROVE EVALUATION SPEED OF CONTINUOUS TIME SYSTEMS
|
| title_full_unstemmed |
EXPLOITING INCREMENTAL SPARSE MATRIX UPDATES TO IMPROVE EVALUATION SPEED OF CONTINUOUS TIME SYSTEMS
|
| title_sort |
exploiting incremental sparse matrix updates to improve evaluation speed of continuous time systems
|
| publisher |
University of Cincinnati / OhioLINK
|
| publishDate |
2005
|
| url |
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1134584775
|
| work_keys_str_mv |
AT sridharanmahesh exploitingincrementalsparsematrixupdatestoimproveevaluationspeedofcontinuoustimesystems
|
| _version_ |
1719432245219950592
|
