Section: New Results

Rewriting System for Profile-Guided Data Layout Transformations on Binaries

Careful data layout design is crucial for achieving high performance. However exploring data layouts is time-consuming and error-prone, and assessing the impact of a layout transformation on performance is difficult without performing it. We proposed [7] a method and implemented a prototype to guide application programmers through data layout restructuring for improving kernel performance and SIMDizability, by providing a comprehensive multidimensional description of the initial layout, built from trace analysis, and then by giving a performance evaluation of the transformations tested and an expression of each transformed layout. The programmer can limit the exploration to layouts matching some patterns. We apply this method to two multithreaded applications. The performance prediction of multiple transformations matches within 5% the performance of hand-transformed layout code.