Re: loop-ch tweek

2016-06-07 Thread Richard Biener
On Mon, 6 Jun 2016, Jan Hubicka wrote: > Hi, > does this look better? Can you make the argument of gimple_inexpensive_call_p a gcall * please? Ok with that change. Thanks, Richard. > Honza > > * gimple.c: Include builtins.h > (gimple_inexpensive_call_p): New function. > * gi

Re: loop-ch tweek

2016-06-06 Thread Jan Hubicka
Hi, does this look better? Honza * gimple.c: Include builtins.h (gimple_inexpensive_call_p): New function. * gimple.h (gimple_inexpensive_call_p): Declare. * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Use it. * tree-ssa-loop-ivcanon.c (tree_estima

Re: loop-ch tweek

2016-06-06 Thread Jan Hubicka
> On Mon, 6 Jun 2016, Jan Hubicka wrote: > > > > On Sun, 5 Jun 2016, Jan Hubicka wrote: > > > > > > > Hi, > > > > both loop-ch and loop-ivcanon want to trottle down the heuristics on > > > > paths > > > > containing call. Testing for presence of GIMPLE_CALL is wrong for > > > > internal > > > >

Re: loop-ch tweek

2016-06-06 Thread Richard Biener
On Mon, 6 Jun 2016, Jan Hubicka wrote: > > On Sun, 5 Jun 2016, Jan Hubicka wrote: > > > > > Hi, > > > both loop-ch and loop-ivcanon want to trottle down the heuristics on paths > > > containing call. Testing for presence of GIMPLE_CALL is wrong for internal > > > call and cheap builtins that are

Re: loop-ch tweek

2016-06-06 Thread Jan Hubicka
> On Sun, 5 Jun 2016, Jan Hubicka wrote: > > > Hi, > > both loop-ch and loop-ivcanon want to trottle down the heuristics on paths > > containing call. Testing for presence of GIMPLE_CALL is wrong for internal > > call and cheap builtins that are expanded inline. > > > > Bootstrapped/regtested x86

Re: loop-ch tweek

2016-06-06 Thread Richard Biener
On Sun, 5 Jun 2016, Jan Hubicka wrote: > Hi, > both loop-ch and loop-ivcanon want to trottle down the heuristics on paths > containing call. Testing for presence of GIMPLE_CALL is wrong for internal > call and cheap builtins that are expanded inline. > > Bootstrapped/regtested x86_64-linux, OK?