Timing Analysis and Scheduling
Optimizing compilers in general, and parallelizing compilers in particular, try to improve the performance of programs by performing a series of transformations. A transformation is an optimization if a set (possibly empty) of conditions is satisfied. These conditions are derived from a comparison of the cost functions of the original and the transformed code. The ideal cost function is the real execution time of the program, which in the case of an arbitrary program is not computable even if its data are given. Therefore, it is essential to find heuristics for estimating the execution time of programs at various granularity levels, for example program, subprogram, loop, basic block, statement, and expression.
KeywordsExecution Time Assignment Statement Iteration Space Loop Nest Conditional Statement
Unable to display preview. Download preview PDF.