On Parameterized Tiled Loop Generation and Its Parallelization

Tiling is a loop transformation that decomposes computations into a set of smaller computation blocks. The transformation has proved to be useful for many high-level program optimizations, such as data locality optimization and exploiting coarse-grained parallelism, and crucial for architecture with limited resources, such as embedded systems, GPUs, and the Cell. Data locality and parallelism will continue to serve as major vehicles for achieving high performance on modern architectures. Parameterized tiling is tiling where the size of blocks is not fixed at compile time but remains a symbolic constant that can be selected/changed even at runtime. Parameterized tiled loops facilitate iterative and runtime optimizations, such as iterative compilation, auto-tuning and dynamic program adaption. Existing solutions to parameterized tiled loop generation are either restricted to perfectly nested loops or difficult to parallelize on distributed memory systems and even on shared memory systems when a program does not have synchronization free parallelism. We present an approach for parameterized tiled loop generation for imperfectly nested loops. We empoly a direct extension of the tiled code generation technique for perfectly nested loops and three simple optimizations on the resulting parameterized tiled loops. The generation as well as the optimizations are achieved purely syntactic processing, hence loop generation time remains negligible. Our code generation technique provides comparable efficiency of generated code to the existing code generation techniques while our generated code remains simple and suitable for parallelization. We also provide a technique for statically restructuring parameterized tiled loops to the wavefront scheduling on shared memory system. Because the formulation of parameterized tiling does not fit into the well established polyhedral framework, such static restructuring has been a great challenge. However, we achieve this limited restructuring through a syntactic processing without any sophisticated machinery.