Jeff Law wrote:
Joseph S. Myers wrote:
Vladimir, is there any news on the patch to remove the old allocator
since <http://gcc.gnu.org/ml/gcc/2008-10/msg00428.html>, or on getting
m32c and IRA working together?
Vlad is relatively busy dealing with PRs as they're reported, the
optimization regressions in particular take considerable time due to
benchmarking.
That is right, Jeff. The most time takes optimization regression PRs.
It's my understanding Vlad has a patch to remove the old allocator,
but it's become somewhat dated and probably doesn't apply to the
trunk. If Vlad will pass that patch along, I'll pick it up.
Yes, that is right. I have the patch to remove old RA ready. The last
time it was synchronized a week ago. So it is pretty fresh. I need a
day to check it again.
For the m32c, Vlad has a patch which was supposed to help deal with
specific issues that arise on the m32c; that patch needs updating so
that DJ & I can look at the m32c again.
I am working on m32c and mostly done with that. I am going to submit
the patch tomorrow. It makes small changes to IRA and implements Chow's
priority coloring (for regional or all function RA) which permits to use
intersected register classes. In fact, it permits not to describe
IRA_COVER_CLASESS for some weird targets (like m32c). But without
description of this macro, target can not use a better Chaitin-Briggs
RA. This patch will permit to use IRA for all targets even without
porting them to IRA.
So if nobody objects, I could submit the patch removing the old RA on
next week for review and commit it after getting an approval and after
committing patch for priority coloring.
