This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[tree-ssa] Merge status 2004-04-05


The branch is in sync with mainline as of 2003-03-31.  It bootstraps
daily alphaev67-unknown-linux-gnu, i686-pc-linux-gnu,
x86_64-unknown-linux-gnu, powerpc-unknown-linux-gnu and
ia64-unknown-linux-gnu.

In terms of codegen quality, Jeff Law found a 97% pass rate for Fedora
Core builds.

Bootstrap times are roughly 14-15% behind mainline, however the branch
has 8-9% more C code than mainline and we currently run about 50 passes
in the tree optimizers.  Recently, Steven Bosscher and Paolo Bonzini
have started experimenting with the removal of RTL passes that should be
eventually replaced by tree-ssa counterparts.  There is also a new pass
by Andrew Pinski to remove superfluous cast operations that has a
positive effect on C++ compile times and tends to generate smaller code.

However, I believe that we are rapidly getting into diminishing returns
territory.  Some of the changes that we have planned do not jive well
with the current freeze.  Mainline is also changing quite frequently,
making the branch more difficult to keep stable.

We should not delay the merge much longer.  Improving compile and run
time issues over a broader range of applications and targets will only
happen if the branch is in mainline because (a) the code will be much
more widely exposed, (b) we'll be able to do major work on new passes.

IMO, we should spend a few more days finishing up the remaining
documentation bits, fix the testsuite regressions against mainline and
then merge.  Note that there are some regressions that may need
relatively major work which can only happen after the branch is merged
into mainline (I'm thinking mostly about the builtin optimization
tests).

I'd like to request opinions regarding a merge into mainline in the next
couple of weeks.  We should not rush into merging unnecessarily, but
keeping the branch frozen much longer will just cause us to spin our
wheels on ice.


Thanks.  Diego.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]