Re: [Fwd: Register stacks again]

Leopold Toetsch <lt@xxxxxxxxxx> wrote:
> Didn't get much response

... so I'm responding to myself ;)

I've checked in docs/nanoparrot.c. While it should be a document
showing, how the interpreter actually performs its work, it also allows
comparison of the additional performance penalty of the indirect
register access.

Its around 1% (OS X) up to 3% (athlon/linux).

This impact would be on function-based cores only and is worst case.
Other opcode functions burn much more cycles, so that the one additional
assembler instruction does harm less.

leo