Re: [PATCH] Fix PR rtl-optimization/71634

2016-07-12 Thread Martin Liška
On 07/12/2016 04:18 PM, Bernd Schmidt wrote: > > > On 07/08/2016 04:27 PM, Martin Liška wrote: >> On 07/08/2016 02:54 PM, Martin Liška wrote: >>> On 07/08/2016 01:59 PM, Bernd Schmidt wrote: Gah, that's not right, that'll swap the numbers of kept/removed loops. I think the

Re: [PATCH] Fix PR rtl-optimization/71634

2016-07-12 Thread Bernd Schmidt
On 07/08/2016 04:27 PM, Martin Liška wrote: On 07/08/2016 02:54 PM, Martin Liška wrote: On 07/08/2016 01:59 PM, Bernd Schmidt wrote: Gah, that's not right, that'll swap the numbers of kept/removed loops. I think the right answer is simply for (i = 0; i < n - IRA_MAX_LOOPS_NUM; i++)

Re: [PATCH] Fix PR rtl-optimization/71634

2016-07-08 Thread Martin Liška
On 07/08/2016 02:54 PM, Martin Liška wrote: > On 07/08/2016 01:59 PM, Bernd Schmidt wrote: >> >> Gah, that's not right, that'll swap the numbers of kept/removed loops. >> >> I think the right answer is simply >> for (i = 0; i < n - IRA_MAX_LOOPS_NUM; i++) >> >> >> Bernd > > Thank you for the

Re: [PATCH] Fix PR rtl-optimization/71634

2016-07-08 Thread Martin Liška
On 07/08/2016 01:59 PM, Bernd Schmidt wrote: > > Gah, that's not right, that'll swap the numbers of kept/removed loops. > > I think the right answer is simply > for (i = 0; i < n - IRA_MAX_LOOPS_NUM; i++) > > > Bernd Thank you for the help, I've been testing the suggested change. Martin

Re: [PATCH] Fix PR rtl-optimization/71634

2016-07-08 Thread Bernd Schmidt
On 07/08/2016 01:52 PM, Bernd Schmidt wrote: int maxidx = MIN (IRA_MAX_LOOPS_NUM, n); for (i = 0; i < maxidx; i++) { Gah, that's not right, that'll swap the numbers of kept/removed loops. I think the right answer is simply for (i = 0; i < n - IRA_MAX_LOOPS_NUM; i++) Bernd

Re: [PATCH] Fix PR rtl-optimization/71634

2016-07-08 Thread Bernd Schmidt
On 06/23/2016 12:56 PM, Martin Liška wrote: Following patch changes minimum of ira-max-loops-num to 1. Having the minimum equal to zero does not make much sense. Ready after it finishes reg on x86_64-linux? Hmm, why wouldn't a number of zero make sense if you want try to have all loops

Re: [PATCH] Fix PR rtl-optimization/71634

2016-07-08 Thread Martin Liška
PING^1 On 06/23/2016 12:56 PM, Martin Liška wrote: > Hello. > > Following patch changes minimum of ira-max-loops-num to 1. > Having the minimum equal to zero does not make much sense. > > Ready after it finishes reg on x86_64-linux? > > Thanks, > Martin >

[PATCH] Fix PR rtl-optimization/71634

2016-06-23 Thread Martin Liška
> Date: Thu, 23 Jun 2016 12:52:44 +0200 Subject: [PATCH] Fix PR rtl-optimization/71634 gcc/ChangeLog: 2016-06-23 Martin Liska <mli...@suse.cz> * params.def: Change min of ira-max-loops-num to 1. --- gcc/params.def | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/gcc/p