On Thu, 17 Aug 2017, Richard Biener wrote: > > Without global analysis it seems best to move constants/negates to the > > toplevel if they can't be trivially removed in a subexpression. Eg. -x > > / (a * b * -c). > > Sure. So both patterns are canonicalization which is fine for match.pd. > Those followup transforms should be done at a place that can look at > more complicated patterns. We have the reassoc pass, then backprop (not > exactly matching), and the recip pattern matching / cse pass.
Watch out for -frounding-math issues, though. -a * b is always the same as a * -b, but not the same as -(a * b) with -frounding-math, and similarly, -frounding-math should prevent eliminating the negations from the -x / (a * b * -c) expression (whereas elimination from -x / -(a * b * c) would be fine). -- Joseph S. Myers jos...@codesourcery.com
