Jeff Law <l...@redhat.com> writes: > On 09/03/2015 12:53 PM, Wilco Dijkstra wrote: >>> Segher Boessenkool wrote: >>> On Thu, Sep 03, 2015 at 10:09:36AM -0600, Jeff Law wrote: >>>>>> You will end up with a *lot* of target hooks like this. It will also >>>>>> make testing harder (less coverage). I am not sure that is a good idea. >>>>> >>>>> We certainly need a lot more target hooks in general so GCC can do the >>>>> right thing >>>>> (rather than using costs inconsistently all over the place). But that's a >>>>> different >>>>> discussion... >>>> Let's be very careful here, target hooks aren't always the solution. >>>> I'd rather see the costing models fixed and use those across the board. >>>> But frankly, I don't know how to fix the costing models. >>> >>> Combine doesn't currently use costs to decide how to simplify and >>> canonicalise things. Simplifications are what is simpler RTL; combine's >>> job is to make fewer RTL instructions (which is not the same thing as >>> fewer machine instructions, or cheaper instructions). Changing what is >>> canonical based on target hooks would be, uh, interesting. >> >> Would it be reasonable to query the rtx_cost of a compare+and and if the cost >> is the same as an AND assume that that instruction does not need to be >> "improved" >> into the canonical form? That way it will use the compare+and pattern >> if it exists >> and still try the zero_extract/shift+and forms for targets that don't have a >> compare+and instruction. > Perhaps -- but you also have to make sure that you don't regress cases > where canonicalization in turn exposes simplifications due to related > insns in the chain.
I agree with Segher that the canonical form really shouldn't depend on costs or target hooks. It's just going to be a can of worms. And patterns shouldn't match non-canonical rtl. If the (and ...) form is a better canonical form (IMO yes) then I think it would be better to make it the canonical form across the baord and update the existing ports to use it. The criteria could be something like no unjustifiable differences in gcc.dg, g++.dg and gcc.c-torture .s output for -O2, which is relatively easy to test. Thanks, Richard
