[Bug tree-optimization/38747] [4.4 Regression] Wrong code due to VIEW_CONVERT_EXPR

rguenth at gcc dot gnu dot org gcc-bugzilla@gcc.gnu.org
Fri Jan 16 09:23:00 GMT 2009



------- Comment #3 from rguenth at gcc dot gnu dot org  2009-01-16 09:23 -------
I think you should simply do the check for an indirect base only if the folding
of the view-convert-expr ended up in a view-convert-expr.  Playing with
alias-sets
here will still cause these nested v-c-es with -fno-strict-aliasing and loads
of unwanted (and likely unprofitable) v-c-es.  Your testcase should be still
optimized then.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38747



More information about the Gcc-bugs mailing list