The evolution of RISC technology at IBM

A 2000 retrospective by John Cocke and V. Markstein on the 1970s RISC work at IBM. I was particularly interested at the relationship between chip design, compilers, and programming languages.

There was still the question of higher-level languages. With limited resources, it was decided to concentrate on just one. PL/1 seemed to be a desirable language, since it supported many applications. Its very richness, however, also made it difficult for compilers to produce good code over much of its capability. The construct of PL/1 was therefore reduced to a subset useful to system programmers, and those features which seemed to defy reasonable translation were discarded... The result was the PL.8 language, the ".8" implying that it had about 80 percent of the richness of PL/1.

Originally published in the IBM Journal of Research and Development. The only place I could find to read it without a paywall was, very unhappily, scribd.

[Edit: Mattias Engdegård found a proper pdf.]

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

non-scribd location

Also found here.

Interesting article - thanks. Apparently it was taken for granted that compilers would be responsible for memory protection (which I suppose mainly means array indices in this context), rather than the programmer, as has become the case a little too often ever since.

It was also interesting to see how a lot of techniques now taken for granted actually came about, both in the architecture and in the compiler. I like how the improved static dataflow analysis suddenly found a lot of errors in existing code.

updated story

Thanks for the link, I've updated the story.

Thanks. Now I know where the

Thanks. Now I know where the .8 came from!

They should have called it PL/1.25