[PATCH] PR 10011: Introduce gcse_constant_p

[Richard Henderson]

On Sun, Mar 30, 2003 at 04:21:33PM -0700, Roger Sayle wrote:
> Ok, after giving more thought to your solution I'm not opposed
> to the new location, provided the OK above counts as approval
> for the gcse_constant_p fix.

Yes, that's approval.

> This is the comment that led me astray.  It turns out that the only
> thing run between the new location and the old location are CSE and
> CFG clean-ups.  So it isn't "loop" that's responsible for moving
> the assignment to a different basic block, but CSE proving that an
> earlier assignment of the condition codes are still live.

Absolutely not.  Remove my patch and look at the .loop dump.  There's
an instruction being moved by move_movables.  I think you are being led
astray by the fact that purge_constant_p was run *between* dump files.
So its effects appear nowhere.  Which I also considered a bug.


r~