Comprehensive optimization of parametric kernels for graphics processing units

该提案涉及针对图形处理单元（GPU）的计算机程序 ** 的优化。 目标是从 ** 程序员到优化编译器，减轻指定 ** 程序和机器特定细节的沉重负担。 预期的好处是 ** 确保生成的 ** 计算机程序的健壮性、可移植性和效率。*我们通过在代码生成过程中将机器和程序 ** 参数视为未知符号，并基于机器和程序参数的 ** 可能值，以案例讨论的形式生成 ** 优化程序，来满足这些要求。这是优化编译程序设计的一种新方法。 通过利用计算机代数领域的最新进展，初步的实验已经证明了上述想法的优点。 通过这个提议，我们希望将它们转化为成熟的 ** 技术，以支持商业编译器，如IBM XL** 编译器。

This proposal deals with the optimization of computer programs**targeting Graphics Processing Units (GPUs). The goal is to lift, from**programmers to optimizing compilers, the heavy burden of specifying**program and machine-specific details. The expected benefit is to**ensure robustness, portability and efficiency of the generated**computer programs.****We address these requirements by treating machine and program**parameters as unknown symbols during code generation, and generating**optimized programs in the form of a case discussion, based on the**possible values of the machine and program parameters.****This is a new approach to the design of optimizing compilers. By**taking advantage of recent advances in the area of computer algebra,**preliminary experimentation has demonstrated the merit of the above**ideas. With this proposal, we want to turn them into mature**technology in support of commercial compilers such as the IBM XL**compiler.**