This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Let's be clear about store motion
- From: Daniel Berlin <dberlin at dberlin dot org>
- To: gcc at gcc dot gnu dot org
- Date: Wed, 15 May 2002 13:14:37 -0400 (EDT)
- Subject: Let's be clear about store motion
The original discussion was whether or not it was broken.
Which led to the related question of whether it should be disabled on
the mainline, since it's disabled on the branch, which started the
discussion about it's usefulness.
These are starting to get mixed together.
Let's clearly seperate these issues. If you want to discuss a certain
issue, let's prefix the subject with the issue, or something, so that we
don't digress (probably my fault, ADHD and all that) onto other subjects.
1. Is it broken?
Possibly. My memory is fuzzy on whether the bugs i pointed out were
fixed. Giving the benefit of the doubt that they were (I can't remember
if the patches were committed), i'll make the assumption it's
not broken anymore, until I run across problems that show otherwise.
2. Is it currently useful in any general context?
Nope. I don't think anyone is seriously arguing it is anymore.
3. Should it be removed?
Probably not, until it's superceded by something else. Even if
it does nothing, it's not significantly intertwined with other
code such that leaving it bitrotting hurts.
I'm just being cautious here though, it might make sense to at least move
it into it's own file, to highlight the fact that it's really a separate
pass that just reuses a few of gcse's functions for simplicity.
That would probably make it
A. less intimidating to others who might think they have to get into the
dirty details of GCSE to improve it.
B. more noticeable to the casual eye so that someone might notice it
exists, and go about improving it.
4. Should it be disabled?
I believe it should, for the reasons given before.
--Dan