Re: [PATCH] Fix some EVRP stupidness

2018-10-23 Thread Aldy Hernandez
Thanks! On Tue, Oct 23, 2018, 13:37 Richard Biener wrote: > On Tue, 23 Oct 2018, Richard Biener wrote: > > > On Tue, 23 Oct 2018, Aldy Hernandez wrote: > > > > > > > > > + if (tem.kind () == old_vr->kind () > > > > + && tem.min () == old_vr->min () > > > > + && tem.max

Re: [PATCH] Fix some EVRP stupidness

2018-10-23 Thread Richard Biener
On Tue, 23 Oct 2018, Richard Biener wrote: > On Tue, 23 Oct 2018, Aldy Hernandez wrote: > > > > > > + if (tem.kind () == old_vr->kind () > > > + && tem.min () == old_vr->min () > > > + && tem.max () == old_vr->max ()) > > > + continue; > > > > I think it would

Re: [PATCH] Fix some EVRP stupidness

2018-10-23 Thread Richard Biener
On Tue, 23 Oct 2018, Aldy Hernandez wrote: > > > + if (tem.kind () == old_vr->kind () > > + && tem.min () == old_vr->min () > > + && tem.max () == old_vr->max ()) > > + continue; > > I think it would be cleaner to use tem.ignore_equivs_equal_p (*old_vr).

Re: [PATCH] Fix some EVRP stupidness

2018-10-23 Thread Aldy Hernandez
+ if (tem.kind () == old_vr->kind () + && tem.min () == old_vr->min () + && tem.max () == old_vr->max ()) + continue; I think it would be cleaner to use tem.ignore_equivs_equal_p (*old_vr). The goal was to use == when the equivalence

Re: [PATCH] Fix some EVRP stupidness

2018-10-22 Thread Richard Biener
On Mon, 22 Oct 2018, David Malcolm wrote: > On Mon, 2018-10-22 at 15:56 +0200, Richard Biener wrote: > [...snip...] > > > This is what I finally applied for the original patch after fixing > > the above issue. > > > > Bootstrapped and tested on x86_64-unknown-linux-gnu, applied. > > > > Richard

Re: [PATCH] Fix some EVRP stupidness

2018-10-22 Thread David Malcolm
On Mon, 2018-10-22 at 15:56 +0200, Richard Biener wrote: [...snip...] > This is what I finally applied for the original patch after fixing > the above issue. > > Bootstrapped and tested on x86_64-unknown-linux-gnu, applied. > > Richard. > > 2018-10-22 Richard Biener > > * gimple-ssa-e

Re: [PATCH] Fix some EVRP stupidness

2018-10-22 Thread Richard Biener
On Thu, 18 Oct 2018, Richard Biener wrote: > On October 18, 2018 5:42:56 PM GMT+02:00, Aldy Hernandez > wrote: > > > > > >On 10/18/18 8:11 AM, Richard Biener wrote: > >> On Thu, 18 Oct 2018, Richard Biener wrote: > >> > >>> > >>> At some point we decided to not simply intersect all ranges we ge

Re: [PATCH] Fix some EVRP stupidness

2018-10-18 Thread Richard Biener
On October 18, 2018 5:42:56 PM GMT+02:00, Aldy Hernandez wrote: > > >On 10/18/18 8:11 AM, Richard Biener wrote: >> On Thu, 18 Oct 2018, Richard Biener wrote: >> >>> >>> At some point we decided to not simply intersect all ranges we get >>> via register_edge_assert_for. Instead we simply registe

Re: [PATCH] Fix some EVRP stupidness

2018-10-18 Thread Aldy Hernandez
On 10/18/18 8:11 AM, Richard Biener wrote: On Thu, 18 Oct 2018, Richard Biener wrote: At some point we decided to not simply intersect all ranges we get via register_edge_assert_for. Instead we simply register them in-order. That causes things like replacing [64, +INF] with ~[0, 0]. The

Re: [PATCH] Fix some EVRP stupidness

2018-10-18 Thread Richard Biener
On Thu, 18 Oct 2018, Richard Biener wrote: > > At some point we decided to not simply intersect all ranges we get > via register_edge_assert_for. Instead we simply register them > in-order. That causes things like replacing [64, +INF] with ~[0, 0]. > > The following patch avoids replacing a ra

[PATCH] Fix some EVRP stupidness

2018-10-18 Thread Richard Biener
At some point we decided to not simply intersect all ranges we get via register_edge_assert_for. Instead we simply register them in-order. That causes things like replacing [64, +INF] with ~[0, 0]. The following patch avoids replacing a range with a larger one as obvious improvement. Compared