On Fri, Apr 27, 2012 at 11:00:19AM +0200, Richard Guenther wrote:
> But won't it re-introduce bugs like PR52080, 52097 or 48124? Also the
No. All those are about bitfield stores, not reads. All extract_bit*
functions currently pass 0, 0 as bitrange_{start,end}.
> proper place for this optimization is lowering and CSEing of bit-field loads
> (not that this lowering already exists).
I agree, but we don't have that yet. When it is added,
optimize_bitfield* should be certainly moved there, but if we do that
before, we regress generated code quality.
Jakub
