The amount of stack in the L1 cache is essentially that. A shifting fast access working memory area. Then think of registers as just part of the fetch & store

The amount of stack in the L1 cache is essentially that. A shifting fast access working memory area. Then think of registers as just part of the fetch & store pipelines for staging operations on stack values. Forth goes all in with this approach.