Re: [Sablevm-developer] RE: tail-recursion

[Etienne M. G. <eg...@j-...>]

Brent Fulgham wrote:
> Perhaps it would be sufficient to incorporate a "(rec)" or similar
> in the stacktrace output, to indicate that this function was
> handled through the tail-method-invocation handler.  :-)

That's a simple and elegant solution.  I would probably make that
"(tail)", but this only costs one bit of information. I'll have to see
if this could be stolen somewhere on the current frame design. In the
worst case, it would cost on additional byte per frame.

> It will be interesting to see the difference in performance with this
> feature turned on and off... :-) 

There will be some overhead cause by the displacement of method
arguments.  But, the effect on the memory cache should be positive, as
we are creating additional locality.  Reducing cache misses could very
well offest completely any overhead;-)  It's really exciting!

> Oh -- and now we are down to two new
> prospective bytecodes, which is great as well!  Now if we can just
> figure out a good way to work around that...

Let's wait for "Hello World" before we attack this more complex one.

Thanks for brigning this FP subject up!  I didn't realize it would be
that exciting.

Etienne
-- 
----------------------------------------------------------------------
Etienne M. Gagnon, M.Sc.                     e-mail: eg...@j-...
Author of SableVM:                             http://www.sablevm.org/
----------------------------------------------------------------------