{"responseHeader":{"status":0,"QTime":5,"params":{"q":"{!q.op=AND}id:\"95914\"","hl":"true","hl.simple.post":"","hl.fragsize":"5000","fq":"!embargo_tdt:[NOW TO *]","hl.fl":"ocr_t","hl.method":"unified","wt":"json","hl.simple.pre":""}},"response":{"numFound":1,"start":0,"docs":[{"modified_tdt":"2015-11-02T00:00:00Z","thumb_s":"/8d/f2/8df29028f9809212e372d6535da6b748606b8496.jpg","oldid_t":"compsci 3732","setname_s":"ir_computersa","file_s":"/3a/e9/3ae92df5c5a1dc18c71ccf838719af56befc089f.pdf","title_t":"Page 82","ocr_t":"6.1.1 Data Partitioning for Unstructured Data Sets Data partition strategies for unstructured data sets falJ into two at gori s. Th first one is to divide cells into disjoint groups. Since the data sets ar unstructur d , th geometric shapes of the resulting groups are generally irregular. The advantage of this strategy is that no data redundancy is introduced by the data partition. Th weakness of this method is that the spatial information of the groups is difficult to obtain. It is not easy to tell whether two groups are adjacent, and it is also hard to identify the group where a specified point is contained. The other technique is to partition a data set by super-imposing a regular framework on it. A subset is formed by grouping the cells which are intersected with or contained in a region of the framework. The framework could be a regular 3-D mesh, a k-way tree or an octree [35]. Since the data sets are unstructured, a cell may intersect with several regions of a regular framework. This cell will be assigned to more than one subset; thus data redundancy is inevitable by using this method. A major advantage of the second strategy is that the spatial information of subsets can be easily retrieved. For example, if an octree is employed as the framework of data partition, the subset (or the octant) containing a point can be identified by searching the octree from the root to the leaves within O(logN) steps, where N is the number of the octants. The neighboring subsets of a subset can be found by applying this technique, too. If the framework is a regular 3-D mesh, the above searchings may be completed in constant time. In our out-of-core visualization algorithm, octrees are used as the frameworks for the data partition. The reason of adopting octrees as frameworks is that most meshes used in computational fluid dynamics simulations are highly adaptive. By using octrees as frameworks, we can refine the data partition in the regions where the meshes are dense; thus the sizes of resulting subsets are relatively equal. In Figure 6.1, a simple example of octree is shown. 6.1.2 Data Partitioning In our data partition algorithm, the data set is subdivided in a top-down manner. At first, the whole data set is considered as an octant. Then this octant is decomposed into eight child octants by using three cutting planes perpendicular to the x, y, and z axes. If the number of cells in a child octant exceeds a predefined limit, this child octree is decomposed further. The predefined limit will be called the maximum octant size in the rest of the chapter. The above procedure is performed recursively until all oct ants contain","restricted_i":0,"id":95914,"created_tdt":"2015-11-02T00:00:00Z","format_t":"application/pdf","parent_i":95941,"_version_":1642982571756224515}]},"highlighting":{"95914":{"ocr_t":[]}}}