Hi, > On 03/10/08 08:24, Richard Guenther wrote: > > >You could either do > > > >GIMPLE_ASSIGN <COND_EXPR, cond, x, y> > > But 'cond' would be an unflattened tree expression. I'm trying to avoid > that. > > >or invent COND_GT_EXPR, COND_GE_EXPR, etc. (at least in GIMPLE > >we always have a comparison in COND_EXPR_COND, never a plain > >boolean variable). > > Yeah, that would mean adding 5 more tree codes, though. Unless we gave > up and invented a new set of subcodes exclusively for gimple. That > seems like a waste, since tree.def neatly defines all the subcodes we > want already.
another possibility would be to represent a = b < c ? d : e as GIMPLE_ASSIGN (LT_EXPR, a, b, c, d, e) and a = (b < c) as GIMPLE_ASSIGN (LT_EXPR, a, b, c, true, false) Zdenek
