| J. W. Sias, "Condition awareness support for predicate analysis and optimization," M.S. thesis, Univ. Illinois, Urbana, 1999. |
....predicate P 1 UT (r1 1) hTi and the predicate P 2 UT (r1 = 1) hTi would be recognized as opposites. Previous work treats conditions at this level. In PAS it is possible to represent more complex relations among families of conditions that are based on comparing the same register values [48]. Such a system is able to represent, for example, the exclusivity of (r1 = 1) and (r1 = 2) while indicating that both are subsets of (r1 0) This system has been implemented in BDD form 66 using a technique known as nite domain representation [44] When using this technique, conditions ....
....op15 op18 op21 op22 (b) Schedule for the hyperblock in Figure 6.2(c) Figure 6.3: Comparison of the static schedules for the wc hyperblock before and after logic minimization. from which the set of register values yielding a TRUE evaluation for any member condition can be composed by union [48]. Finally, the relationships between the comparisons are described in BDD form using a nite domain technique [44] Various elements of the optimizer query this BDD to determine the inherent relationships between conditions, which are the decision network s input variables. The overall ....
J. W. Sias, \Condition awareness support for predicate analysis and optimization," M.S. thesis, University of Illinois, Urbana, IL, 1999. 183
....this analysis. D. Construction of the Condition Layer The first step of finding relations among predicates is the definition of relations among condition evaluations. In PAS, these relations are represented together with predicate information in the BDD by providing a set of condition nodes [16]. PAS incorporates arbitrary relations within families of conditions based on comparing the same register values, representing, for example, the exclusivity of and while indicating that both are subsets of . A family is initially represented as a single interval containing all representable ....
....automatically in the BDD as the disjunction representing is formed, one at a time. Fig. 6(d) shows the resulting BDD, in which all relationships among conditions in a family are represented. For example, the expressions for , and , show that implies . This process is described in detail in [16]. This condition layer is used in the optimization phase to determine the relations among condition variables. In this capacity it is separate from the predicate BDD; in later analyses, however, it serves as the foundation for the predicate layer. E. Construction of the Predicate Layer The ....
J. W. Sias, "Condition awareness support for predicate analysis and optimization," M.S. thesis, Univ. Illinois, Urbana, 1999.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC