Fw: [patch] [4.2 projects] vectorize strided accesses (4 patches)

Roger Sayle roger@eyesopen.com
Tue Sep 5 13:48:00 GMT 2006


Hi Ira (and the other autovect folks),

On Tue, 5 Sep 2006, Ira Rosen wrote:
> Is it O.K. for mainline 4.3 with those changes?

I recently tried doing a full top-level bootstrap of the autovect branch
on i686-pc-linux-gnu where it currently fails to build due to an ICE
whilst building libjava.  After waiting a few weeks I tried again, but
with the same result.  It's not clear whether this is a pre-existing
java condition from the branch's last merge with mainline, or a problem
associated with the autovect changes.

Any chance I could ask one of the autovect folks to look into the issue
and keep the autovect branch buildable?  This helps inspire confidence
that integrating changes to mainline should be safe.  If a branch has
problems even for common targets the new code probably hasn't stablized
enough to be integrated.  This applies not just to autovect but all of
GCC's other development branches.

Roger
--



More information about the Gcc-patches mailing list