On Friday 19 August 2011 12:05:02 Janus Weil wrote: > Ping! (Maybe I should have posted the follow-up patch in a separate > thread to make it more visible.) I saw it, had a quick glance, thought that Thomas would jump on it, and forgot. Sorry.
> > 2011/8/13 Janus Weil <ja...@gcc.gnu.org>: > > Hi Thomas, hi all, > > > > 2011/8/7 Thomas Koenig <tkoe...@netcologne.de>: > >> When extending the values of gfc_dep_compare_expr, we will need to go > >> through all its uses (making sure we change == -2 to <= -2). > > > > attached is a patch which makes a start with this. > > > > For now, it changes the return value to "-3" for two cases: > > 1) different expr_types > > 2) non-identical variables > > > > I tried to take care of all places which are checking for a return > > value of "-2" and I hope I missed none. > > > > Any objections or ok for trunk? (Regtested successfully.) OK from my side for the code proper. I have one comment though about this: +/* Compare two expressions. Return values: + * +1 if e1 > e2 + * 0 if e1 == e2 + * -1 if e1 < e2 + * -2 if the relationship could not be determined + * -3 if e1 /= e2, but we cannot tell which one is larger. */ I think this is misleading, as the function does not always return -3 when e1/=e2. There is for example (currently) no special handling for operators. Here is an attempt at expressing it: * -3 in some cases where we could determine that e1 and e2 have different data dependencies (and thus are not guaranteed to have always the same value), but we cannot tell whether one is greater than the other. Mikael
