Maclisp

From Wikipedia, the free encyclopedia
  (Redirected from MacLisp)
Jump to: navigation, search

MACLISP (or Maclisp, sometimes styled MacLisp or MacLISP) is a dialect of the Lisp programming language. It originated at MIT's Project MAC[1] (from which it derived its prefix) in the late 1960s and was based on Lisp 1.5. Richard Greenblatt was the main developer of the original codebase for the PDP-6;[1] Jon L. White was responsible for its later maintenance and development. The name 'Maclisp' started being used in the early 1970s to distinguish it from other forks of PDP-6 Lisp, notably BBN Lisp.

History[edit]

Maclisp ran on DEC PDP-6/10 computers, initially only on ITS, but later under all the other PDP-10 operating systems. Its original implementation was in PDP-10 assembly language. It was later implemented on Multics using PL/I. Maclisp developed considerably in its lifetime, adding major features along the way which in other language systems would typically correspond to major release numbers.

Maclisp was used to implement the Macsyma symbolic algebra program; Macsyma's development also drove a number of features in Maclisp. The SHRDLU blocks-world program was written in Maclisp, and so the language was in widespread use in the artificial intelligence research community through the early 1980s. It was also used to implement other programming languages, such as Planner and Scheme. Multics Maclisp was used to implement the first Lisp-based Emacs.

Maclisp was very influential, but is no longer actively maintained. Nonetheless, it now runs on PDP-10 emulators and can be used for experimenting with early AI programs.

Characteristics[edit]

Maclisp started with a small, fixed number of data types: cons cell, atom (later called "symbol"), integer, and floating-point number. Later additions included: arrays, which were however never first-class data-types; arbitrary-precision integers (bignums); strings; and tuples. All objects (except inums) were implemented as pointers, and their data type was determined by the block of memory into which it pointed, with a special case for small numbers (inums).

Programs could be interpreted or compiled. Compiled behavior was the same as interpreted except that local variables were lexical by default in compiled code, and no error checking was done for inline operations such as CAR and CDR. The Ncomplr compiler (mid-1970s) introduced fast numeric support to the Lisp world, generating machine instructions for arithmetic rather than calling interpretive routines which dispatched on data type. This made Lisp arithmetic comparable in speed to Fortran for scalar operations (though Fortran array and loop implementation remained much better).

The original version was limited by the 18-bit word address of the PDP-10, and considerable effort was expended in keeping the implementation lean and simple. Multics Maclisp had a far larger address space, but was expensive to use. When the memory and processing power of the PDP-10 were exceeded, the Lisp Machine was invented: Lisp Machine Lisp is the direct descendant of Maclisp. Several other Lisp dialects were also in use, and the need to unify the community resulted in the modern Common Lisp language.

Name[edit]

MACLISP was named for Project MAC, and is unrelated to Apple's Macintosh ("Mac") computer, which it predates by many years. The Lisp systems for the Macintosh, MCL and Clozure CL (formerly OpenMCL), have no particular similarity to Maclisp. Texas Instruments and Symbolics both marketed NuBus-based LISP machine coprocessor cards for the Macintosh, but these do not run MACLISP.

References[edit]

External links[edit]