Richard Biener writes:
> On Mon, 28 Oct 2019, Jiufu Guo wrote:
>
>> Richard Biener writes:
>>
>> > On Fri, 25 Oct 2019, Jiufu Guo wrote:
>> >
>> >> Hi,
>> >>
>> >> In PR88760, there are a few disscussion about improve or tune unroller for
>> >> targets. And we would agree to enable unroller fo
On October 28, 2019 6:18:26 PM GMT+01:00, Segher Boessenkool
wrote:
>On Mon, Oct 28, 2019 at 09:08:04AM +0100, Richard Biener wrote:
>> On Fri, 25 Oct 2019, Jiufu Guo wrote:
>> > On powerpc64le, for O2 , enable -funroll-loops and limit
>> > PARAM_MAX_UNROLL_TIMES=2 and PARAM_MAX_UNROLLED_INSNS=20
On Mon, Oct 28, 2019 at 09:08:04AM +0100, Richard Biener wrote:
> On Fri, 25 Oct 2019, Jiufu Guo wrote:
> > On powerpc64le, for O2 , enable -funroll-loops and limit
> > PARAM_MAX_UNROLL_TIMES=2 and PARAM_MAX_UNROLLED_INSNS=20, we can see >2%
> > overall improvement for SPEC2017.
>
> Note the behav
On Mon, 28 Oct 2019, Jiufu Guo wrote:
> Richard Biener writes:
>
> > On Fri, 25 Oct 2019, Jiufu Guo wrote:
> >
> >> Hi,
> >>
> >> In PR88760, there are a few disscussion about improve or tune unroller for
> >> targets. And we would agree to enable unroller for small loops at O2 first.
> >> And
Richard Biener writes:
> On Fri, 25 Oct 2019, Jiufu Guo wrote:
>
>> Hi,
>>
>> In PR88760, there are a few disscussion about improve or tune unroller for
>> targets. And we would agree to enable unroller for small loops at O2 first.
>> And we could see performance improvement(~10%) for below code
On Fri, 25 Oct 2019, Jiufu Guo wrote:
> Hi,
>
> In PR88760, there are a few disscussion about improve or tune unroller for
> targets. And we would agree to enable unroller for small loops at O2 first.
> And we could see performance improvement(~10%) for below code:
> ```
> subroutine foo (i, i1
Hi Jiufu,
On Fri, Oct 25, 2019 at 10:44:39PM +0800, Jiufu Guo wrote:
> In PR88760, there are a few disscussion about improve or tune unroller for
> targets. And we would agree to enable unroller for small loops at O2 first.
[ snip ]
> PR tree-optimization/88760
> * config/rs6000/rs60
Hi,
In PR88760, there are a few disscussion about improve or tune unroller for
targets. And we would agree to enable unroller for small loops at O2 first.
And we could see performance improvement(~10%) for below code:
```
subroutine foo (i, i1, block)
integer :: i, i1
integer :: block(9,