The FORTRAN automatic coding system

Staged Computation with Staged Lexical Scope

“…Languages that evaluate under future-stage binders demand a careful treatment in the type system of potentially open code (that is, code that contains free variables). This is particularly challenging in the presence of an eval function and of assignments to mutable state since (1) an eval function must be passed closed code values only and (2) assignments enable code values to escape the scope in which they are generated, which in turn enables future-stage variables to escape their binder. Hence type systems for staged computation must distinguish between open and closed code values and must prevent future-stage variables from being captured by any binder but their own.…”

“…Code manipulation as a programming discipline dates back to the development of the first Fortran compiler in the late 1950s [1]. In the early 1960s, McCarthy proposed S-expressions as a uniform representation of code (and other data) in Lisp [15].…”

Essentials of programming languages

“…Fortran (Backus et al, 1957) was the first language to use call-by-reference, Algol 60 (Naur et al, 1963) was the first language to use call-by-name, and Haskell (Hudak et al, 1990) was the first practical language to use call-by-need. (Plotkin, 1975) …”

Tera-OP Reliable Intelligently Adaptive Processing System (TRIPS) Implementation

“…Figure 14 shows that the TRIPS backend compiler phase iterates in several places, whereas compilers for conventional architectures do not. These phases break separation-of-concerns, a key tenet of compiler design that helps simplify the compiler [3]. The first compiler for Fortran and every compiler since then divides the compiler into distinct phases that operate on one or more well-specified intermediate forms.…”