Re: indirect threading for bytecode interpreter

[Helmut Eller <eller.helmut@gmail.com>]

* Tom Tromey [2009-09-17 21:38+0200] writes:

>>>>>> "Helmut" == Helmut Eller <eller.helmut@gmail.com> writes:
>
> Helmut> Despite that, a few years ago somebody has already proposed that 
> Helmut> http://lists.gnu.org/archive/html/emacs-devel/2004-05/msg00095.html
> Helmut> and it wasn't deemed worthwhile back then.
>
> That is not my take on that thread.  It just sort of petered out with no
> results.  I didn't see any actual objection.

Hmm.. maybe I'm misremembering something.  But then such a change
introduces quite a bit of clutter and it's not so clear how much speedup
is required to justify that.  5% doesn't sound like a lot to some
people.  Well, at least not to me. :-) 

vmgen sounds like a good idea, but I fear that it makes the build
process quite a bit more complicated.

I'm wondering why gcc can't perform this transformation from the switch
based code.  Is there no compiler setting to skip the range check in the
switch statement?  Quite bizarre, considering C's philosophy of skipping
safety checks.

> Also, the Emacs project has a problem with losing patches.
> This is the second time I've reimplemented something only to find that a
> nearly identical patch has been sitting in the archives for years.

Looks like it.  Maybe things are better now with the bug tracker.

Helmut