A.W. Aho, R. Sethi and J.D. Ullmann,COMPILERS Principles, Techniques and Tools, Bell Telephone Laboratories, 1987.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....that one that is supposed to be implemented on the ASIC core) Step 2 tests whether the preceding cluster might probably be already part of the ASIC core such that the estimation can take that into account accordingly. The estimation of commu 8 We use gen #####and use #####as it is defined in [16]. 3 nication effort for the ASIC core (steps 3 and 4) follows the same principle as described steps 1 and 2. Finally, the total amount of energy due to core core partitioning is calculated in 5) using an energy amount E bus read=write for a bus access 9 . 3.4 Determining the Utilization Rate ....

A.W. Aho, R. Sethi and J.D. Ullmann,COMPILERS Principles, Techniques and Tools, Bell Telephone Laboratories, 1987.

....results for the SYMTA approach. 2 Program Path analysis 2.1 Problems and previous work For path analysis techniques [10] 7] a program is divided in basic blocks. Definition 1 A basic block is a program segment which is only entered at the first statement and only left at the last statement [1]. Function calls are considered as single statements. Any program can be partitioned into disjoint basic blocks. The program structure is represented on a directed program flow graph with basic blocks as nodes. Fig. 1 shows an example. For each basic block the worst or best case execution time ....

....and b6, are in MFP path segments. Fig. 3(c) shows how the remaining program path segments are clustered to SFP nodes. What we need for this clustering approach is an algorithm to determine condition (see above) i.e. input data dependency of conditions. This requires a global data flow analysis [1] with a transitive closure over all data dependencies of variables in control statements. A global data flow analysis, however, does typically not cover dependencies over array elements and operation on data. Therefore, the global data flow analysis is complemented with symbolic simulation of ....

A.V. Aho, R. Sethi, J.D. Ullman, Compiler principles, Techniques and Tools, Bell Telephone Laboratories, Inc. 1987.

....but semantically equivalent queries, which produce the same result for all database instances that satisfy the integrity constraints and functional dependencies. The objective of a semantic query optimizer is to find a semantically equivalent query which yields a more efficient execution plan[1 9]. The success of a semantic query optimizer depends on the availability of an effective set of query transformation rules which can reduce the execution cost of a large set of queries. The set of query transformation rules usually includes user defined integrity constraints which are true for all ....

A. V. Aho, J. E. Hopcroft, and J. D. Ullman, The Design and Analysis of Computer Algorithms, Bell Telephone Laboratories (1974).

....of optimization methods makes it possible to add, delete and modify individual methods, without affecting other methods. Since this system is rule based it provides an extensible system for maintaining and managing optimization. Semantic query optimization is well motivated in the literature[1, 2, 8, 9] as adding a new dimension to conventional query optimization. The success of a semantic query optimizer depends on the availability of an effective set of query transformation rules which can reduce the execution cost of a large set of queries. The set of query transformation rules usually ....

A. V. Aho, J. E. Hopcroft, and J. D. Ullman, The Design and Analysis of Computer Algorithms, Bell Telephone Laboratories (1974).

....within each b i . The number of paths grows exponentially with the number of operations within a b i in the worst case. Since the complexity of the combinatorial optimization problem for partitioning grows exponentially with n (O(2 n ) 10 The term basic block is used in the same manner as in [1]. 11 n is the number of all b i s in B. the requirements concerning n seem to be contradictory. We have overcome this problem by using a path based estimation technique [11] So the user can determine the number n of pieces b i as one constraint for clustering. The quality of pieces b i ....

....estimation can take place afterwards, T pre Gammaexec HW;b i is determined by a simple list schedule. The transfer time for data from software to hardware and vice versa is obtained by a data flow analysis that determines the IN Set(b i ) and OUT Set(b i ) similar to the method described in [1]. A possible overlap of b i can get expressed by T misc b i (then there is T misc b i 0) So the quality metric gives a hint whether b i could be a good candidate for the HW SW partitioning with the difference that in HW SW partitioning itself a more sophisticated cost function is used ( 9, ....

A.W. Aho, R. Sethi and J.D. Ullmann,COMPILERS Principles, Techniques and Tools, Bell Telephone Laboratories, 1987.

....approaches. Loops and conditional branches will be presented by special nodes i.e. LoopBegin and LoopEnd Nodes and edges to connect them (Fig. 4) Data flow is structured by basic blocks. Basic blocks which are code sequences which are performed in connection are defined in compiler theory [Aho87]. This structuring supports the easy recognition of maximal parallel degree. In contrast to other systems the data flow graph connects the basic blocks into one global data flow graph of a system. This allows global data analysis by presenting an overall data base to variables, which gives ....

A.V. Aho, R.Sethi, and J.D. Ullmann. COMPILERS Principles, Techniques and Tools. Bell Telephone Laboratories, Inc., 1987.

....which underline the unprecedented analysis accuracy. 3 Timing analysis problem 3.1 Path analysis For path analysis techniques [Pa92] MaWo95] a program is divided in basic blocks. A basic block is a program segment which is only entered at the first statement and only left at the last statement [AhSeUl87]. Function calls are considered as single statements. Any program can be divided into disjoint basic blocks. The program structure is represented as a directed program flow graph with basic blocks as nodes. Fig. 1 shows an example for a part of an FFT algorithm. n = nn 1; j=1; for ....

....and b6, are on MFP path segments. Fig. 3c shows how the remaining program path segments are clustered to SFP nodes. What we need for this clustering approach is an algorithm to determine condition (see above) i.e. input data dependency of conditions. This requires a global data flow analysis [AhSeUl87] with a transitive closure over all data dependencies of variables in control statements. A global data flow analysis, however, does typically not cover dependencies over array elements and operation on data. Therefore, the global data flow analysis is complemented with symbolic simulation of ....

A.V. Aho, R. Sethi, J.D. Ullman, Compiler principles, Techniques and Tools, Bell Telephone Laboratories, Inc. 1987.

....indicating the end of a basic block. An overview of these instructions can be found in Groningen (1990) of determining the minimal number of registers is rather simple. An algorithm (which makes some assumptions about the registers and instructions of the target machine) has been given in Aho et al. 1986). The problem with common subexpressions is that the results of these expressions have to be stored somewhere until they are used the last time. This implies that after evaluating a certain sub expression, the number of registers in use does not always increase by exactly one (due to the ....

.... necessary to hold the value of that sub expression) It is possible that this increase is greater than one (if the dag contains CSE s that were not evaluated yet) or even smaller than one (if registers containing values of CSE s were used for the last time) Furthermore, the algorithm presented in Aho et al. 1986) cannot deal with values kept in registers at the beginning and at the end of a basic block. As with CSE s, such registers can be released as soon as their contents are not needed anymore. To deal with CSE s when determining the evaluation order we first have to introduce some notions. A (rooted) ....

Aho, A.V., R. Sethi and J.D. Ullman (1986). Compilers, Principles, Techniques and Tools, Bell Telephone Laboratories, Incorporated, Addison-Wesley.

No context found.

A. V. Aho, J. E. Hopcroft, and J. D. Ullman, Data Structures and Algorithms. Bell Telephone Laboratories, 1983.

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