Donald E. Knuth is in the process of updating his classic, The art of computer programming. Rather than revising entire volumes at once, he is targeting the most relevant kernels of each volume as fascicles. This is the first fascicle for volume 1, “Fundamental Algorithms.” It addresses sections of two chapters from the original volume: “MIX” (1.3.1-1.3.3) and “Some Fundamental Programming Techniques” (1.4.1-1.4.3). To distinguish the rewritten sections from the originals, he adds a single quote (’) to the section number in the fascicle.

A core component for the update is the specification of the old MIX computer architecture. Section 1.3.1’ introduces MMIX, the modern 64-bit pipelined reduced instruction set computer (RISC) successor to the now obsolete MIX machine. MMIX assembly language is the focus of section 1.3.2’, while 1.3.3’ is left as a challenge for interested readers to convert old MIX programs to the new MMIX notation. Chapter 1.4.1’ discusses subroutines as an instance of fundamental programming techniques, expressed in MMIX, while 1.4.2’ argues convincingly that subroutines may be viewed as a special case of coroutines. Chapter 1.4.3’ discusses simulators, including an MMIX simulator, naturally expressed in MMIX itself.

Lamentably, there is little to improve about this fascicle, especially since the author promises monetary rewards to the tune of $2.56 for any first-detected error. Even though some of the subjective preferences proposed by Knuth are debatable, even they would stimulate constructive architecture discussions. Subjective choices include the memory width nomenclature, such as wyde, tetra, and octa, and the unusual permission to let addresses be unaligned for memory units longer than one byte. And in light of the phenomenal accuracy of dynamic branch predictors in modern computers, it is debatable whether there is real value in giving programmers the power and burden of statically biasing conditional branches with a probable direction.

Part of the beauty of MMIX lies in the fact that its inventor was able to devise an architecture that incorporates all interesting high-performance, clean machine attributes. Knuth selected RISC over complex instruction set computer (CISC), but failed to tell us why. The fascicle is a must-read for anybody wishing to design a new computer architecture, real or hypothetical, or anybody wishing to see all performance-enhancing attributes in one machine, including a large register file, conditional branches, a circular register stack, and hints about cache-line fill. Proposing an MMIX assembly language that permits reusable local names as well as global symbols offers a refreshing convenience to low-level language programmers.

The general computer science audience anxiously awaits the remaining revisions, as well as the first edition of the promised, but as yet unpublished, volumes in the seven-volume series. Many of us wish to benefit from Knuth’s view and wit before we retire, and would be envious if only our successor generation can start studying the art of computer programming with a fresh and complete set of classics.