Automatic storage optimization

Abstract

Most studies of optimization techniques for higher level languages have focused on improving execution time of generated programs, often at the expense of increased storage. When storage optimization has been addressed, it is usually in conjunction with time optimization, such as in instruction-reducing code transformations. In the Bliss Compiler (WJWHG75), a storage-optimizing compiler, transformations that reduce register temporary storage are also performed, but automatic overlay of program variables is not addressed. The rising popularity of mini-computers and micro-processors suggests that the time has come to examine the problem of automatic storage optimization in its totality. Because lack of space has always been a problem in the small systems environment, the proliferation of small machines implies the growing importance of the problem. Although the decreasing cost of memory may mitigate this trend, a variant of Murphy's Law ensures that program size will always increase faster than the available storage. In other words, programmers always write programs that don't fit, and, as time goes on, more of them will be doing it.