On Tue, 12 Mar 2013, Steven Bosscher wrote: > On Tue, Mar 12, 2013 at 4:25 PM, Richard Biener wrote: > > > > (Un-?)surprisingly the most effective compile-time reduction for > > the testcase in PR39326 is to employ ao_ref caching for > > alias oracle queries and caching of expanded affine-combinations > > for affine disambiguations. > > > > This reduces compile-time to a manageable amount in the first place > > for me (so I'm sending it "late" in the series). > > I suppose this renders my LIM patch obsolete.
Not really - it's still tree loop invariant motion: 588.31 (78%) usr so limiting the O(n^2) dependence testing is a good thing. But I can take it over from here and implement that ontop of my patches if you like. > Did you also look at the memory foot print? Yeah, unfortunately processing outermost loops separately doesn't reduce peak memory consumption. I'll look into getting rid of the all-refs bitmaps, but I'm not there yet. Currently the testcase peaks at 1.7GB for me (after LIM, then it gets worse with DSE and IRA). And I only tested -O1 sofar. Thanks, Richard.
