This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: rtx_equal_function_value_matters
Date: Tue, 24 Nov 1998 10:30:41 -0800
From: Richard Henderson <rth@cygnus.com>
On Tue, Nov 24, 1998 at 07:13:14PM +0100, Jakub Jelinek wrote:
> That's what I proposed in my first mail. I think simply replacing a bunch of
> (reload_in_progress || reload_completed) tests with
> (cannot_generate_new_rtxs) or whatever (you'll probably name it better), it
> would fix a lot of bugs that may show up randomly.
While we should have some simple no_new_pseudos predicate, I wonder
if you win creating new pseudos when cse_not_expected.
If you are doing instruction scheduling, yes indeed you do win.
It is rarely a lose, because these pseudos we are creating here, even
if CSE does not take "advantage" of them, are very well allocated by
local alloc to hard registers. Their lifetimes are always inside
a single basic block, and they typically live for no more than 3 or 4
instructions.
Later,
David S. Miller
davem@dm.cobaltmicro.com