Compiler Code Generators That Generate Complex Instructions

当前一代的可重目标编译器可以很容易地生成包含大多数指令和目标机器架构实现的所有寻址的最佳本地代码。然而，这些编译器不能利用在许多CISC目标机器体系结构中发现的复杂指令。这些指令通常操作字段和字符串，或者实现复杂的循环控制机制。该项目将寻求构建一个可重目标编译器，以便在编译用Algol家族的高级语言编写的程序时利用复杂的指令。这个系统将分为两部分。代码生成器将是一个受属性影响的图形匹配器，它可以匹配中间表示并将其转换为由复杂指令直接匹配的形式。将使用高级语言来指定由复杂指令执行的操作。规范将在代码生成时被分析并转换为代码生成器使用的计划。

The current generation of retargetable compilers can easily generate optimal local code incorporating most instructions and all addressing implemented by the target machine architecture. However, these compilers can not utilize the complex instructions found on many CISC target machine architectures. These instructions typically manipulate fields and strings, or implement complex loop control mechanisms. The project will seek to construct a retargetable compiler than can exploit complex instructions when compiling programs written in a high level language in the Algol family. The system will have two pieces. The code generator will be an attributed influenced graph matcher that can match and transform an intermediate representation into a form that is directly matched by a complex instruction. A high level language will be used to specify the actions performed by a complex instruction. The specification will be analyzed and transformed at code generation time into a plan that the code generator uses.