This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Calls.c code permutation patch take2
- To: Jan Hubicka <hubicka at atrey dot karlin dot mff dot cuni dot cz>
- Subject: Re: Calls.c code permutation patch take2
- From: Richard Henderson <rth at cygnus dot com>
- Date: Fri, 5 May 2000 14:44:44 -0700
- Cc: egcs-patches at egcs dot cygnus dot com
- References: <20000505140220.D27697@atrey.karlin.mff.cuni.cz>
On Fri, May 05, 2000 at 02:02:20PM +0200, Jan Hubicka wrote:
> * calls.c (expand_call): Reorganize somewhat in order
> to avoid unnecesary recalculation inside loop and make
> tail call failure code cleaner.
It appears as if currently_expanding_call is now incremented too late.
In particular, after we expand the libcall to __bb_fork_func.
> + Also do not make a sibling call. */
> +
> + flags &= ~(ECF_CONST | ECF_PURE);
> + must_preallocate = 1;
Should move the "do not make a sibling call" comment, as it's now
confusing here.
> + /* We need to build actparms for optimize_tail_recursion. We can
> + safely trash away TREE_PURPOSE, since it is unused by this
> + function. */
> + if (try_tail_call)
> + actparms = tree_cons (NULL_TREE, args[i].tree_value, actparms);
I don't see actparms cleared first. Plus, why only tail_call
and not tail_recursion as well?
> /* Various reasons we can not use a sibling call. */
> ! if (! try_tail_call)
> continue;
The comment is somewhat out of date here. ;-)
r~