I recently bought a copy of "Computer Engineering: A DEC View of Hardware Systems Design". I still prefer to read real books rather than web pages or PDFs.

I recently finished reading the book's Chapter 15, "Turning Cousins into Sisters: An Example of Software Smoothing of Hardware Differences", by Ronald Brender. I found the topic particularly interesting.

In the early 1970s, DEC required a FORTRAN compiler in order for the PDP-11 minicomputers to compete against the offerings of their competitors, especially IBM. The limited amount of memory in their systems drove them to implement the compiler so that it produced the application as threaded code. The resulting compiler produced very efficient code that allowed the compiler, the linker, and the application to run in an 8kW (16kB) environment along with whatever operating system was present on the computer. The instruction set of the PDP-11 easily realized the threaded code using its complex addressing modes without implementing an interpreter like that used for BASIC or most FORTH implementations.

I found the article particularly enlightening and thought provoking in relation to my current effort to add some instructions to the M65C02 FPGA core to support direct threaded code (DTC) FORTH. I think others here might also find the article interesting and thought provoking. Therefore, I found and downloaded a copy of "Computer Engineering" from the archive.org, and printed out Chapter 15. It is attached below for any who have an interest in threaded code compilers.

Edit: 21 July 2014. Added the following paper by James Bell, Digital Equipment Corporation, that is referenced by Ronald Brender. (C. Gordon Bell is one of three editors of the book for which Ronald Brender wrote Chapter 15.)

Edit: 16 Aug 2014. Added the following paper. It presents a slightly different approach than that presented in the the original paper by James Bell. It appears that the approach described above by Brender is an improvement on the basic technique the Bell described, but DEC's approach does not appear to be the same as that described by Dewar in the following paper which proposes indirect threading rather than direct threading.

_________________
Michael A.

Thanks!

I like the frequent use of the term "ad hoc". I don't have enough imagination to anticipate what the problem spots might be when starting from a top-down code-last approach. I need to start fiddling around with code before I can even begin to see what the whole solution might be.

Thanks for that! I knew that DEC had used threaded code in a FORTRAN compiler, but I didn't have a reference for it. Until now.

_________________
Because there are never enough Forth implementations: http://www.camelforth.com

Okay, I give up. Where did you find the complete "Computer Engineering" for download? I looked at archive.org and all I can find is an image scan of the first 16 pages or so.

Edited to add: never mind, I think I found it here: http://www.computerhistory.org/collecti ... /102630383

_________________
Because there are never enough Forth implementations: http://www.camelforth.com

See also this page-by-page version online (in text):
http://research.microsoft.com/en-us/um/ ... 000387.htm

Brad:

The one at the Computer History Museum is the one that I downloaded.

_________________
Michael A.