[PATCH] PR29145 - be more restrict about restrict qualifier

Ian Lance Taylor iant@google.com
Tue Jan 16 15:27:00 GMT 2007


Dorit Nuzman <DORIT@il.ibm.com> writes:

> I tried that - it gave the same outcome - testcases vect-44.c, vect-50.c,
> vect-95.c and pr16105.c (all using restrict pointers) do not get vectorized
> if I drop the special handling of restrict and add a call to
> alias_sets_conflict_p instead.  On the other hand - testcases
> vect-[43,49,53].c do get vectorized although there's no restrict there and
> the pointers actually escape.

There must be something I don't understand.  Can you send me the patch
you are testing against mainline?  Thanks.

Ian



More information about the Gcc-patches mailing list