SPITBOL was (is?) a compiled implementation of SNOBOL4 for the IBM360/370. It was created by Robert Dewar and Ken Belcher (names coming from my memory--hope I have them right) at the Illinois Institute of Technology (IIT). (Dewar later went to NYU, if memory serves me, where I think he created an Algol68 compiler.).
Until SPITBOL came along, SNOBOL4 was thought to be (1) slow, (2) a memory hog, and (3) impossible to compile due to its dynamic nature. While the delayed binding prevented everything from being determined at compile time, SPITBOL was very clever about doing as much as early as possible. SNOBOL programs run under SPITBOL were indeed amazingly fast. Notable was the SPIBTOL garbage collector which ran with almost no spare memory.
A contribution of SPITBOL to computer science was to demonstrate the clear distinction between a language and its implementation. To say that a language is slow is to reveal a lack of imagination. It's better to say that writing a fast implementation is hard. We see this same issue now with Java, which is often erroneously called an interpreted language. Of course, it's an implementation that is an interpreter. The language is just the language.
--Marc Rochkind [Do we sign Wikipedia articles?]