Abstracto

Survey on Need of Loop Transformations for Automatic Parallelization

Nisha, Rafiya Shahana ,Mustafa B

With the increasing proliferation of multicore processors, parallelization of applications has become a priority task. One considers parallelism while writing new applications, to exploit the available computing power. Similarly, parallelization of legacy applications for performance benefits is also important. In modern computersystems loops present a great deal of opportunities for increasing Instruction Level andThread Level Parallelism. Techniques are needed to avoid unnecessary checks to assure that only the correct number of iterations are executed. In this paper we present a survey of basic loop transformation techniquesthat can improve the performance by eliminating some unnecessary conditional instructions checking for iteration bounds. The objective of this work is to come up with a nice survey of loop transformations. We present information on the number of instructionseliminated as well as on thegeneral transformations, mostly at the source level. Depending on the target architecture, the goal of loops transformations are: improve data reuse and data locality, efficient use of memory hierarchy, reducing overheads associated with executing loops, instructions pipeline, maximize parallelism.Loop transformations can be performed at different levels by the programmer, the compileror specialized tools. Loop optimization is the process of the increasing execution speed and reducing the overheads associated of loops. Thus, loops optimization is critical in high performance computing. Our techniques are applicable to mostmodern architecture including superscalar, multithreaded, VLIW or EPIC systems.