History: How did Forth get its stacks

On 18-7-06 13:20, in article

See:

Turing, A. M. "Proposals for the development in the Mathematics Division of an Automatic Computing Engine (ACE)." Report E882, Executive Committee, NPL February 1946. (Reprinted April 1972 as NPL Report Com. Sci 57.)

Turing describes the use of a subroutine RA stack, with PUSH RA called 'BURY' and POP RA called 'UNBURY", implemented in his proposed ACE instruction set (which was effectively user microprogramming in today's terms).

True with exceptions. The English Electric KDF9 ca.1960-61 had a 16-deep return address stack (along with a a 16-deep arithmetic stack and a bank of 16 counting-indexing registers) for each of 4 processes, driven by blocking and unblocking on I-O transfer termination status, and implemented entirely in hardware.

The Burroughs B5000 from the same era had an unbounded activation stack managed in core by the hardware.

"Modern-textbook-style recursion" based on an activation stack (usually software-managed) was a feature of Algol 60 and its many successors and derivatives implemented during the 1960s, including PL-1, Algol 68, Algol W, Pascal, CPL, BCPL, ... (we can all see where that is heading, I guess 8-).

-- Bill Findlay