[aarch64-port-dev ] RFR: 8179933: aarch64: incorrect match rule for immL_63 and immL_255

[Zhongwei Yao]

Hi, Edward, Thanks for pointing out.

And I've checked the generated nodes of dst = src / 2, (y and x are both
long) in current c2 is:
  RShiftL (AddL src (URShiftL (RShiftL src immI_63) immI_63))

On 13 May 2017 at 02:11, Edward Nevill <edward.nevill at gmail.com> wrote:

> On Fri, 2017-05-12 at 10:45 +0100, Andrew Dinn wrote:
> > On 12/05/17 10:25, Andrew Haley wrote:
> > >
> > > On 12/05/17 10:14, Zhongwei Yao wrote:
> > > >
> > > > On 12 May 2017 at 16:14, Andrew Haley <aph at redhat.com> wrote:
> > > > >
> > > > > So it's
> > > > >
> > > > >    ((aLong >> 63) >>> 63)
> > > > >
> > > > > This is so weird I don't know what to think about it.
> > > > Yeah, it is weird. I was thinking (aLong >>> 63) should be enough.
> Shall we
> > > > just remove this pattern?
> > > Well, maybe there was a reason for it.
> > Ok, these rules were first introduced into jdk8 by Ed Nevill
> >
> >   http://hg.openjdk.java.net/aarch64-port/jdk8u/hotspot/rev/1fcabae0e46f
> >
> > They were then imported into jdk9 as part of the wholesale upload.
> >
> > So, I think there is probably a good reason why they are there :-)
> >
>
> Here is the original posting for the int case on aarch64-dev
>
> http://mail.openjdk.java.net/pipermail/aarch64-port-dev/
> 2014-April/001047.html
>
> On Mon, 2014-05-12 at 13:10 +0100, Andrew Haley wrote:
>
> > Why is there no long version of this?
>
> So I added a long version
>
> http://mail.openjdk.java.net/pipermail/aarch64-port-dev/
> 2014-May/001070.html
>
> Ed.
>
>


-- 
Best regards,
Zhongwei