Loveman, D. B., "Program Improvement by Source-to-Source Transformation," JACM 20(1) pp. 121-145 (January 1977).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Inputs to graphics pipelines typically lists of vertices are usually many thousands of elements long. In our system, these inputs are divided into batches so that only a subset of them are processed at one time. This common vector technique is called stripmining [Lamport, 1974; Loveman, 1977] it allows the hardware to best take advantage of the producer consumer locality in the application by keeping the working set of data small enough to fit in the SRF. The second optimization we perform is used at both the kernel and stream levels. This optimization is applied to loops and is ....

David B. Loveman. Program Improvement by Source-toSource Transformation. Journal of the ACM, 24(1):121-- 145, January 1977.

....transformations; for other systems the composition of the theta and once transformations is the best option. The approach adopted in this paper is in the long tradition of source to source program transformations: changes at the source level that can improve eciency without altering correctness [21]. The suggestions here by no means exhaust the transformations of this type. Within ILP, a related approach is described in [2] where a set of queries is restructured so that they can be executed more eciently, without changing the individual queries however. The two approaches are obviously ....

D. B. Loveman. Program improvement by source-to-source transformation. Journal of the Association for Computing Machinery, 24(1):121-145, 1977.

.... visible to the user [Schneck72] These systems would like to perform the standard optimizations done by a compiler [Allen72] and exploit standard rules of exchange for the operators of general purpose languages [Standish76a] Recent interest in optimizing transformations was renewed by Loveman [Loveman77] in his attempt to define source to source Software Construction Using Components 49 transformations which group FORTRAN program features for execution on a parallel machine. Rutter [Rutter77] describes a sourceto source transformation system for LISP programs and examines the problems of ....

Loveman, D.B., Program Improvement by Source-to-Source Transformation, Journal of the ACM, 24(1):121-145, January, 1977.

....area optimisation has not yet attracted much attention. Local resource utilisation optimisation has been done using transformations steered by stochastic techniques [7] Sometimes area optimisation is a secondary goal in throughput optimisation [8] Because most transformations are well known [9, 10, 11], selecting a set of transformations is relatively simple. It is however extremely important to reduce the set to a minimal size to limit the search space. This requires an appropriate model. A transformation can be executed when all its preconditions are satisfied. However, doing so may enable or ....

....executed when all its preconditions are satisfied. However, doing so may enable or disable other transformations. This dependency makes finding an efficient and effective ordering both desirable and very difficult in almost all cases. Therefore, most research relies on manually selected orderings [11, 12, 13, 14]. One way to tackle the ordering problem is by executing random moves using simulated annealing [7] However, a stochastic method such as simulated annealing is based on the assumption that many small and independent moves can be performed very fast. But transformations are dependent, and their ....

D.B. Loveman, "Program Improvement by Sourceto -Source Transformation," Journal of the ACM, Vol. 24, No. 1, pp. 121-145, Jan. 1977.

....a novel method for the verification of transformations on CDFGs. 5.5. Related Work The use of transformations for the design of both software and hardware is not new. While formal program refinement techniques never really caught on, the use of transformations in software compilers is common [Lov77, ASU86] There are, however, several aspects that make hardware design different from software design and the use of formal techniques more attractive. In hardware design, rebuilding hardware normally has a very high cost and is timeconsuming. Furthermore the efficiency and real time requirements ....

D.B. Loveman, Program Improvement by Source-to-Source transformations, Journal of the ACM, VOl 24, No. 1, pp. 121145, January 1977.

....with similar characteristics, and with other ILP systems that use possibly di erent enumeration strategies. The approach adopted in this paper is in the long tradition source to source program transformations: changes at the source level that can improve eciency without altering correctness [13]. While the two suggestions here by no means exhaust the transformations of this type, it is only one side of the story. E ciency gains are as much to be made by tailoring compilers of logic programs to account for what ILP systems do. In concluding his empirical study, Knuth states: the ....

D. B. Loveman. Program improvement by source-to-source transformation. JACM, 24(1):121-145, 1977.

....stream operations are applied to an entire input batch at the same time. In a real system, the application or driver would assemble the input into batches, but in our tests we batch the input before beginning processing. This method is modeled after the vector programming technique of strip mining [14, 17]. 3.1 Implementation Challenges Load balancing Stages of the rendering pipeline such as scan conversion and span rasterization require unequal amounts of work for each input element. Large triangles, for instance, require more work to scan convert than do smaller ones. The challenge of load ....

David B. Loveman. Program improvement by source-to-source transformation. Journal of the ACM, 24(1):121--145, January 1977.

.... of a loop nest to a single loop with the index taking one value for each point in the original iteration space [52, 55] ffl loop unrolling the replication of the loop body and shrinking of the iteration space [19] ffl strip mining the partitioning of the index range of a loop into strips [45], and tiling the generalisation of strip mining to multiple loops [35, 57, 73] ffl loop skewing a very useful transformation which skews the iteration space such as to align independent points in it [35, 40, 57, 72] ffl loop concurrentisation the specification that a loop may be ....

D.B. Loveman. Program improvement by source-to-source transformation. Journal of the ACM, 24(1):121--145, January 1977.

....of a program. Program transformation has been extensively studied in the fields of compiler design [16] automatic logic and functional program generation [18] and parallel program optimization. However, transformations used in SIPLeS II are only those that are applicable at source code level [14], not at intermediate code level, and are necessary for handling the variations that appear. There are three levels of program comparisons: simple text level comparison, syntax level comparison [9] and semantic level comparison [10, 11, 25] The comparison algorithm used in our approach is a ....

Loveman, D. (1977). Program improvement by source to source transformation, Journal of ACM, 24(1), 121-145.

....has been extensively studied in the fields of compiler design (Muchnick, 1997) automatic logic and functional program generation (Pettorossi Proietti, 1994) and parallel program optimization. However, transformations used here are only those that are applicable at source code level (Loveman, 1977), not at intermediate code level, and are necessary for handling the variations that appear. We define five kinds of standardization transformations here. Due to space limitation, implementation details of these transformations are not presented here. 3.1 Statement separation When cascaded ....

Loveman, D. (1977). Program improvement by source to source transformation, Journal of ACM, 24(1), 121-145.

....Pipelining Pipelining computations can provide an efficient solution to cross node dependencies in loops. As a result, a number of parallelizing compilers and systems use pipelining to exploit parallelism. There is a wealth of related work on pipelining [AK87, BCG 95, HLKL97, KTG95, Lov77, VSM96] this section presents some of it. Before support for pipelining was built into parallelizing compilers, a number of researchers wrote programs that were pipelined by hand . The computations in these application programs were pipelined by explicitly passing messages. An optimizing ....

D.B. Loveman. Program improvement by source-to-source transformation. Journal of the ACM, 24(1):129--138, January 1977.

....is not as necessary in optimization as it is in restructuring. Also, optimization is typically not required to produce a valid or readable source language representation, working instead on intermediate code or assembler, a simpler task. Exceptions are the transformations described by Loveman [Loveman 77] and recent results in optimizing register accesses in loops [Callahan et al. 90] Still, neither of these require a readable source, nor an interactive user interface. Another difference is that most optimizations are, in practice, local in character manipulations on loops and basic blocks. ....

....errors during restructuring, and still allows changing structure. Using a transformational approach, Chapter 2 defined a set of structural changes exploiting this constraint. The transformations as a group can manipulate structures spanning several common types. The transformations are not new [Loveman 77] Burstall Darlington 77] Stankovic 82] but their style of application and purpose are. A transformation is applied to a single syntactic construct, and the tool can make the compensating changes in the rest of the program to preserve its original meaning. This style of transformation removes ....

D. B. Loveman. Program improvement by source-to-source transformation. Journal of the ACM, 24(1):121--145, January 1977.

....the transformed source. Ped supports a large set of transformations that have proven useful for introducing, discovering, and exploiting parallelism. Ped also supports transformations for enhancing the use of the memory hierarchy. These transformations are described in detail in the literature [2, 4, 12, 32, 33, 36, 41, 56]. We classify the transformations in Ped as follows. Reordering Transformations Loop Distribution Loop Interchange Loop Skewing Loop Reversal Statement Interchange Loop Fusion Dependence Breaking Transformations Privatization Scalar Expansion Array Renaming Loop Peeling Loop Splitting ....

D. Loveman. Program improvement by source-to-source transformations. Journal of the ACM, 17(2):121--145, January 1977.

....section 7 draws some conclusions based on the results obtained from the example. 2. SynGuide Environment High level transformations can have a important impact on speed and hardware used in the final implementation. Study on transformations has resulted in classification of transformations [9] but when to apply a transformation is still a difficult and open problem. Because of this, an interactive design style is at this level very important. The most important reason to have an interactive design style is to enable the designer to realize his own solution or to give him the ....

D.B. Loveman. Program improvement by source-to-source transformation. Journal of the ACM , vol. 24, no. 1, pp. 121-145, 1977.

No context found.

Loveman, D. B., "Program Improvement by Source-to-Source Transformation," JACM 20(1) pp. 121-145 (January 1977).

No context found.

D. B. Loveman, "Program improvement by source-to-source transformation," Journal of the ACM, vol. 24, no. 1, pp. 121-145, January 1977.

No context found.

Loveman, D., "Program Improvement by Source-to-Source Transformation", Journal of the Association for Computing Machinery, Vol. 24, No. 1, January 1977, pp.121-145

No context found.

LOVEMAN, D. B. 1977. Program improvement by source-to-source transformation. J. ACM 24, 1 (Jan.), 121--145.

No context found.

D.B. Loveman, Program improvement by source-to-source transformation, Journal of the ACM, 24(1):121-145, 1977.

No context found.

D.B. Loveman, "Program improvement by source-to-source transformation", Journal of the ACM, Vol.24, No.1, pp.121145, 1977.

No context found.

D.B. Loveman. Program Improvement by Source-to-Source Transformation. In Jour. of the ACM 24, 1, pp. 121-145, Jan. 1977.

No context found.

D. B. Loveman. Program improvement by source-to-source transformation. Journal of the Association for Computing Machinery, 24(1):121-145, 1977.

No context found.

D. Loveman. Program Improvement by source-to-source transformation. JACM, 24(1), 1977.

No context found.

D.B. Loveman, Program improvements by source to source transformations, in: [94]

No context found.

D. Loveman, Program Improvement by Source-toSource Transformations, J. of the ACM, V. 20, n. 1, pp. 121-145

First 50 documents

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