Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-08-04 Thread Bill Schmidt
On Aug 1, 2017, at 10:56 AM, Bill Schmidt wrote: > > On Aug 1, 2017, at 8:50 AM, Bill Schmidt wrote: >> >> On Aug 1, 2017, at 7:44 AM, Bill Schmidt wrote: >>> On Aug 1, 2017, at 3:46 AM, Richard Biener wrote: On Mon, Jul 31, 2017 at 4:03 PM, Bill Schmidt wro

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-08-02 Thread Bill Schmidt
On Aug 2, 2017, at 3:09 AM, Richard Biener wrote: > > On Tue, Aug 1, 2017 at 5:56 PM, Bill Schmidt > wrote: >> On Aug 1, 2017, at 8:50 AM, Bill Schmidt wrote: >>> >>> On Aug 1, 2017, at 7:44 AM, Bill Schmidt >>> wrote: > > On Aug 1, 2017, at 3:46 AM, Richard Biener > wro

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-08-02 Thread Richard Biener
On Tue, Aug 1, 2017 at 5:56 PM, Bill Schmidt wrote: > On Aug 1, 2017, at 8:50 AM, Bill Schmidt wrote: >> >> On Aug 1, 2017, at 7:44 AM, Bill Schmidt wrote: >>> On Aug 1, 2017, at 3:46 AM, Richard Biener wrote: On Mon, Jul 31, 2017 at 4:03 PM, Bill Schmidt wrote:

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-08-01 Thread Bill Schmidt
On Aug 1, 2017, at 8:50 AM, Bill Schmidt wrote: > > On Aug 1, 2017, at 7:44 AM, Bill Schmidt wrote: >> >>> >>> On Aug 1, 2017, at 3:46 AM, Richard Biener >>> wrote: >>> >>> On Mon, Jul 31, 2017 at 4:03 PM, Bill Schmidt >>> wrote: > On Jul 31, 2017, at 8:19 AM, Bill Schmidt

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-08-01 Thread Richard Biener
On Tue, Aug 1, 2017 at 3:50 PM, Bill Schmidt wrote: > On Aug 1, 2017, at 7:44 AM, Bill Schmidt wrote: >> >>> >>> On Aug 1, 2017, at 3:46 AM, Richard Biener >>> wrote: >>> >>> On Mon, Jul 31, 2017 at 4:03 PM, Bill Schmidt >>> wrote: > On Jul 31, 2017, at 8:19 AM, Bill Schmidt > w

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-08-01 Thread Bill Schmidt
On Aug 1, 2017, at 7:44 AM, Bill Schmidt wrote: > >> >> On Aug 1, 2017, at 3:46 AM, Richard Biener >> wrote: >> >> On Mon, Jul 31, 2017 at 4:03 PM, Bill Schmidt >> wrote: >>> On Jul 31, 2017, at 8:19 AM, Bill Schmidt wrote: That would certainly be much simpler! I'll

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-08-01 Thread Bill Schmidt
> On Aug 1, 2017, at 3:46 AM, Richard Biener wrote: > > On Mon, Jul 31, 2017 at 4:03 PM, Bill Schmidt > wrote: >> >>> On Jul 31, 2017, at 8:19 AM, Bill Schmidt >>> wrote: >>> >>> That would certainly be much simpler! I'll regstrap it and test it on the >>> other >>> occurrence I've found

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-08-01 Thread Richard Biener
On Mon, Jul 31, 2017 at 4:03 PM, Bill Schmidt wrote: > >> On Jul 31, 2017, at 8:19 AM, Bill Schmidt >> wrote: >> >> That would certainly be much simpler! I'll regstrap it and test it on the >> other >> occurrence I've found to be certain. > > Unfortunately, this fails bootstrap: > > /home/wsch

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-07-31 Thread Bill Schmidt
> On Jul 31, 2017, at 8:19 AM, Bill Schmidt wrote: > > That would certainly be much simpler! I'll regstrap it and test it on the > other > occurrence I've found to be certain. Unfortunately, this fails bootstrap: /home/wschmidt/gcc/gcc-mainline-test/gcc/calls.c: In function 'rtx_def* emit

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-07-31 Thread Bill Schmidt
That would certainly be much simpler! I'll regstrap it and test it on the other occurrence I've found to be certain. -- Bill > On Jul 31, 2017, at 4:15 AM, Richard Biener > wrote: > > On Sun, Jul 30, 2017 at 8:04 PM, Bill Schmidt > wrote: >> Hi, >> >> PR81354 identifies a latent bug that ca

Re: [PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-07-31 Thread Richard Biener
On Sun, Jul 30, 2017 at 8:04 PM, Bill Schmidt wrote: > Hi, > > PR81354 identifies a latent bug that can happen in SLSR since the > conditional candidate support was first added. SLSR relies on the > address of a GIMPLE PHI remaining constant during the course of the > optimization pass, but it ne

[PATCH] Fix PR81354 (rewrite gimple_split_edge)

2017-07-30 Thread Bill Schmidt
Hi, PR81354 identifies a latent bug that can happen in SLSR since the conditional candidate support was first added. SLSR relies on the address of a GIMPLE PHI remaining constant during the course of the optimization pass, but it needs to split edges. The use of make_single_succ_edge and reinsta