Over in #guile on freenode we just hit a bump where string-for-each on a string was a lot slower than (for-each proc (string->list lst)) due to a costly jump in

Over in #guile on freenode we just hit a bump where string-for-each on a string was a lot slower than (for-each proc (string->list lst)) due to a costly jump into C from the jitted code. Rewriting string-for-each in scheme solved the problem. I don't know what I wanted to say with that, but I know pypy has spent a lot of time rewriting some core libraries in python/Rpython. I thought it was to make pypy more hackable, but maybe it has some jit benefits as well.