* * * * *

               On tail call optimization in certain C compilers

> From: Mark Grosberg <XXXXXXXXXXXXXXXXX>
> To: Sean Conner <[email protected]>
> Subject: Tail calls.
> Date: Sun, 24 Mar 2013 12:35:18 PM -0500
>
> > But here, I can't rely on the C compiler to optimize tail calls (GCC can,
> > but only with certain options; I don't know about the Solaris C
> > compiler). I could have the routines return the next function to call and
> > use a loop:
> >
>
> I haven't verified it but it probably does. It's a pretty easy optimization
> (compared to what compilers do today) so I'd be surprised if the Sun C
> compiler doesn't handle this. At least for C code (C++ exceptions can throw
> some wrinkles in this in some cases).
>
> -MYG
>

I decided to check. And yes, the Solaris C compiler does support tail call
optimizations [1]. So I figured I would play around with this a bit, both
under gcc and the Solaris C compiler.

Final results: gcc and the Solaris C compiler both support tail call
optimizations (some restrictions apply; void where prohibited; your mileage
may vary; results presented are not typical and yours might vary; use only as
directed; this program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE; do not taunt Happy Fun Ball).

First off, the number of parameters must match among the functions; the types
don't appear to matter that much, just the number. Second, the number (or
size) of locally defined variables also matters. I'm not sure what the upper
size (or number) for variables is (and it may differ between the two
compilers) but it does appear to be a factor. Third, the only safe way to
determine if tail call optimizations are being performed is to check the
assembly code and check for calls (or, just run it and see if it crashes
after a period of time).

So I can, kind of, assume tail call optimization in C code. It'll be
something to keep in mind.

[1] http://docs.oracle.com/cd/E18659_01/html/821-1379/afamv.html#afana

Email author at [email protected]