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]

Re: [lno] copy header and vectorization



On Feb 18, 2004, at 1:44 PM, Devang Patel wrote:



On Feb 17, 2004, at 8:32 PM, Daniel Berlin wrote:


In the meantime, why not clean up the DDG GTY'ing mess, or use loop versioning somewhere?

Because before completing new implementation or starting to use new features in vectorizer it is important to fix regressions in vectorizer.
Then shouldn't this have been done before committing more code that isn't used yet?
Why commit the messy version of the DDG, for example, instead of fixing regressions, if you are just going to end up undoing a lot of the DDG changes?


I'm all for fixing bugs and regressions, but it seems a bit silly to introduce unused code, then leave it sit for weeks while fixing other bugs. It just makes merges harder.

Right now all tree-ssa-vect tests fail on powerpc-darwin.



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