The Analysis, Measurement and Exploitation of Eager Evaluation in Imperative Instruction Streams

本研究关注的是通过动态求值来增加输入代码的并发性或并行性，从而显著提高传统语言计算机器的性能。即时求值是在条件分支执行之前，通过条件分支执行代码的条件执行。研究的具体目标是通过建立一个描述渴望评价的图模型来更充分地理解和表征渴望评价；测量典型代码执行时的性能增益和使用的资源，在理想意义上给出边界，并假设不同的体系结构或执行引擎。同时，开发新的机器，以有效地利用渴望评价。要遵循的方法的一个新颖方面是考虑执行的静态指令流模型，其中代码按照其存在于内存中的顺序排列，而不是仅仅动态或经典指令流，其中程序计数器遵循控制流。要进行的测量和评估将通过计算机模拟代码的执行，假设特定的硬件是适当的。

This research is concerned with dramatically improving the performance of computing machines for traditional languages by increasing the concurrency or parallelism apparent in the input code via eager evaluation. Eager evaluation is the conditional execution of code past conditional branches, before the branches have executed. The specific objectives of the research are to more fully understand and characterize eager evaluation by developing a graph model for its description; measuring the performance gain and resources used when typical codes are executed with eager evaluation, both in an ideal sense to give bounds, and assuming different architectures or execution engines. Also, the development of novel machines to exploit eager evaluation in an efficient manner is to be pursued. One novel aspect of an approach to be followed is to consider the static instruction stream model of execution, in which the code is ordered as it exists in memory, as opposed to just the dynamic, or classic, instruction stream, in which the program counter follows the control flow. The measurements and evaluations to be made will be via computer simulations of the execution of the codes, assuming specific hardware as appropriate.