On Tue, Apr 26, 2016 at 1:57 PM, Marc Glisse <marc.gli...@inria.fr> wrote: > On Tue, 26 Apr 2016, Richard Biener wrote: > >>>> By the way, it would be cool to be able to write: >>>> (lt:c @0 @1) >>>> >>>> which would expand to both >>>> (lt @0 @1) >>>> (gt @1 @0) >>>> >>>> (as per swap_tree_comparison or swapped_tcc_comparison) >>> >>> >>> Yeah, I know... I was hesitant to overload :c with "slightly" different >>> semantics though. >>> >>> I can give it a shot though - it would avoid quite some repetition I >>> guess. >> >> >> Being able to write (lt:c @0 @1) is easy, see attached (didn't check >> if it works), >> but being able to write >> >> (for cmp (lt gt) >> (cmp:c @0 @1) >> >> is harder (see FIXME), you'd have to create a new for at the nesting >> level of the >> old with the operator list adjusted. Not impossible, of course. >> >> Includes some verification I added locally at some point (which also >> exposed we >> use :c on non-commutative tree codes, thus the new :C ...). > > > Ah, I was hoping it would be as simple as adding op=commuted_op(op) at the > place where the regular commutation gets lowered. If it is significantly > more complicated, I guess it isn't that urgent...
Yeah, the issue is that lowering of (for ...) has to happen last and thus we have to commutate the (for...)s themselves. Possible, but some work. Richard. > -- > Marc Glisse
