论文部分内容阅读
得克萨斯仪器公司的先进科学计算机(ASC)为了对各种科学计算问题提供更有效的解答,在它的中央处理机(CP)中使用了流水线概念。虽然通过利用向量指令达到了最大的效率,但通过优化CP标量代码可以有效地增加中央处理机的总效率。本文提出影响CP标量代码执行速度的因素,并考虑用来减少CP标量执行时间的技术。给了一个通过ASCNX FORTRAN编译程序使用其中一些技术的例子,还包括一项有关通过重排数据改进CP效率的最后注记。
Texas Instruments Advanced Scientific Computing (ASC) uses a pipeline concept in its central processing unit (CP) to provide a more efficient answer to a variety of scientific computing problems. Although the maximum efficiency is achieved by using vector instructions, the overall efficiency of the central processing unit can be effectively increased by optimizing the CP scalar code. This paper presents the factors that affect the execution speed of CP scalar code and considers techniques to reduce the execution time of CP scalars. An example was given of using some of these techniques with the ASCNX FORTRAN compiler, as well as a final note on improving CP efficiency by rearranging data.