Re: Bug in ra-colorize.c:merge_moves?

[Daniel Berlin <dberlin at dberlin dot org>]

On Thursday, March 6, 2003, at 06:08  AM, Michael Matz wrote:

> Hi,
>
> On Wed, 5 Mar 2003, Daniel Berlin wrote:
>
> > Yes, but unless you've made symbol_refs of some new symbol to 
> > represent
> > the stack space you are using, they are still based off a register (in
> > this case, the stack register).
>
> Which is fixed, and therefore not even a condidate for coloring.


But still causes conflicts, regardless, since the register is no longer 
available.

>  Ergo
> there's no conflict between webs and the stack pointer.

Of course there is, since they can't use that register as a color now, 
and could before (since the pseudo spill slots could get different 
colors).

>  Even if it were a
> candidate the webs would most probably already have conflicted with 
> it, so
> it doesn't really change the conflicts for the webs.
This is pure conjecture, i have stats that say otherwise (that they 
didn't conflict with it before, and do now).

Theory is nice, but the fact that it takes 15 passes to color something 
that used to take 3, the only difference is stack slots, and the number 
of conflicts on the newly spilled webs has gone up, tells me you are 
wrong.

All of this is academic anyway, since my solution is to not even 
consider merging the stack slot code until it doesn't cause 15 passes 
of coloring, *whatever* the reason.
Not that this stops you or anyone else, of course.

--Dan