Research On Optimized Layout Transformation Algorithm Based On Hadoop

With the rapid development of integrated circuits(IC),very large scale integration circuit(VLSIC)arises at the historic moment with the development of nanometer technology nodes.In the manufacture of integrated circuits,the yield of integrated circuits has been the spicial concern of the production enterprises,and the optimization of the layout is an important step to improve the yield and reduce the cost of the circuit.Optimizing each one of small image layouts which are converted from the whole CIF layout is an effective way of optimizing layouts,and it is the key and last step in the process of optimizing image layouts.This paper presents the study on the transformation from optimized image layout to CIF layout.The paper proposes a new algorithm which uses the edge feature of image layout to detect the correlation in the nets of CIF layout and merge the related lines based on the existing transformation algorithm from image layout to CIF layout.A complete line network in the original CIF layout is divided into the new two networks which have the same edge features.The edge features have strong correlation,and this characteristic is used to extract the edge feature points to merge split circuit nets and generate CIF layout at last.Compared with the existing CIF layout algorithm,this algorithm not only ensures the integrity of the net of CIF layout from image layout,but also solves the problem of redundant data due to the repeated storage of adjacent net information in the generated CIF layout.For a large number of image layouts converting into a CIF layout,this paper proposes an optimized layout transformation algorithm based on hadoop.In this algorithm the large scale of optimization image layouts are distributed into the different hosts in the Hadoop cluster,and then these image layouts in the different nodes are combined logically into one big image data as a certain size,and at last the job is submitted to start the appropriate number of tasks.After that the layout data are processed in localization and parallel,and then the new generated data are aggregated into the specified CIF layout file with the Map Reduce framework.Compared with the existing conversion algorithm on single machine,this algorithm makes full use of the hardware resources such as computer CPU to process to the image layout in the distributed parallel,and it contributes to improving the conversion efficiency of large scale optimization layout.In addition,the algorithm has the advantages of high reliability,high fault tolerance and high scalability,which can ensure the stability ofthe algorithm and the accuracy of extracting CIF layout.