[tromp]

> Writing in continuation-passing style also helps the lambda calculus interpreter to prevent re-evaluating the same term multiple times. This is very, very important and critical when writing programs for the 521-byte binary lambda calculus interpreters Blc, tromp and uni, since it seems that they lack a memoization feature, although they readily have a garbage collection feature.

Note that Melvin Zhang added a memoization feature (call-by-need evaluation with result sharing) in his refactoring [1] of my BLC interepreter.

[1] https://github.com/melvinzhang/binary-lambda-calculus