llvm dev - Aug 2018 - [RFC] Remove Darwin support from POWER Backends

Hi all,
I just posted two patches on phabricator to remove Darwin support from the
POWER backend.

I thought I had posted this on the dev mailing list some time ago, but I
cannot find a reference to it now.

The intention here is to cleanup the POWER backend. Darwin has not been
supported on POWER for quite some time now, and we'd like to use this as an
opportunity to clean up several aspects of the backend code. If there is
any concerns with this, please let me know.

As for the approach, the patch that I posted will issue an error message
from the backend if you attempt to use the Darwin ABI on PPC. The intention
is that once this patch lands, all of the Darwin-specific code paths will
essentially be dead-code, and we can remove them on demand as we're
touching various pieces of code. This cleanup should be done in stand-alone
patches, but can be done as NFC changes and thus reviewed post-commit,
whenever possible.

I've cleaned up all of the test cases effected by this, mostly by
converting them to a different powerpc target so as to not lose test
coverage. There were a few tests that were ppc-darwin specific, which were
removed.

There is a companion patch to clean up test cases in Clang as well. I think
we can use a similar approach for cleaning up ppc-specific darwin code in
Clang, although I don't have a good idea right now for how much cleanup
that will be.

Thanks.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180820/5434f6fe/attachment.html>

I know we spoke on IRC and at the dev meeting, but a good email here helps
too. I've added Hal and Iain explicitly here since we've had talks with
both of them about this.

For the record: I'm in favor of this.

Thanks!

-eric

On Mon, Aug 20, 2018 at 12:42 PM Kit Barton via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> Hi all,
> I just posted two patches on phabricator to remove Darwin support from the
> POWER backend.
>
> I thought I had posted this on the dev mailing list some time ago, but I
> cannot find a reference to it now.
>
> The intention here is to cleanup the POWER backend. Darwin has not been
> supported on POWER for quite some time now, and we'd like to use this
as an
> opportunity to clean up several aspects of the backend code. If there is
> any concerns with this, please let me know.
>
> As for the approach, the patch that I posted will issue an error message
> from the backend if you attempt to use the Darwin ABI on PPC. The intention
> is that once this patch lands, all of the Darwin-specific code paths will
> essentially be dead-code, and we can remove them on demand as we're
> touching various pieces of code. This cleanup should be done in stand-alone
> patches, but can be done as NFC changes and thus reviewed post-commit,
> whenever possible.
>
> I've cleaned up all of the test cases effected by this, mostly by
> converting them to a different powerpc target so as to not lose test
> coverage. There were a few tests that were ppc-darwin specific, which were
> removed.
>
> There is a companion patch to clean up test cases in Clang as well. I
> think we can use a similar approach for cleaning up ppc-specific darwin
> code in Clang, although I don't have a good idea right now for how much
> cleanup that will be.
>
> Thanks.
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180820/5b4c1016/attachment.html>

Added data points. Much of the support for targeting PowerPC has been removed
from Apple's supported tools for years now, and we've more or less
officially stopped supporting building for 10.5 at this point. We even recently
dropped support for 10.4 x86. I recently cleaned up some of the Darwin driver
code to remove functionality that is broken with Xcode 10 (currently in beta)
see r339807, which means clang no longer supports 10.5 and earlier.

-Chris
> On Aug 20, 2018, at 12:59 PM, Eric Christopher via llvm-dev <llvm-dev at
lists.llvm.org> wrote:
> 
> I know we spoke on IRC and at the dev meeting, but a good email here helps
too. I've added Hal and Iain explicitly here since we've had talks with
both of them about this.
> 
> For the record: I'm in favor of this. 
> 
> Thanks!
> 
> -eric
> 
> On Mon, Aug 20, 2018 at 12:42 PM Kit Barton via llvm-dev <llvm-dev at
lists.llvm.org <mailto:llvm-dev at lists.llvm.org>> wrote:
> Hi all,
> I just posted two patches on phabricator to remove Darwin support from the
POWER backend.
> 
> I thought I had posted this on the dev mailing list some time ago, but I
cannot find a reference to it now.
> 
> The intention here is to cleanup the POWER backend. Darwin has not been
supported on POWER for quite some time now, and we'd like to use this as an
opportunity to clean up several aspects of the backend code. If there is any
concerns with this, please let me know.
> 
> As for the approach, the patch that I posted will issue an error message
from the backend if you attempt to use the Darwin ABI on PPC. The intention is
that once this patch lands, all of the Darwin-specific code paths will
essentially be dead-code, and we can remove them on demand as we're touching
various pieces of code. This cleanup should be done in stand-alone patches, but
can be done as NFC changes and thus reviewed post-commit, whenever possible.
> 
> I've cleaned up all of the test cases effected by this, mostly by
converting them to a different powerpc target so as to not lose test coverage.
There were a few tests that were ppc-darwin specific, which were removed.
> 
> There is a companion patch to clean up test cases in Clang as well. I think
we can use a similar approach for cleaning up ppc-specific darwin code in Clang,
although I don't have a good idea right now for how much cleanup that will
be.
> 
> Thanks.
> 
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
<http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180820/abc3d1e2/attachment.html>

😥 it was nice knowing you darwin/ppc, but it is now time to say goodbye.  +1
from me.

-Chris

> On Aug 20, 2018, at 12:42 PM, Kit Barton via llvm-dev <llvm-dev at
lists.llvm.org> wrote:
> 
> Hi all,
> I just posted two patches on phabricator to remove Darwin support from the
POWER backend.
> 
> I thought I had posted this on the dev mailing list some time ago, but I
cannot find a reference to it now.
> 
> The intention here is to cleanup the POWER backend. Darwin has not been
supported on POWER for quite some time now, and we'd like to use this as an
opportunity to clean up several aspects of the backend code. If there is any
concerns with this, please let me know.
> 
> As for the approach, the patch that I posted will issue an error message
from the backend if you attempt to use the Darwin ABI on PPC. The intention is
that once this patch lands, all of the Darwin-specific code paths will
essentially be dead-code, and we can remove them on demand as we're touching
various pieces of code. This cleanup should be done in stand-alone patches, but
can be done as NFC changes and thus reviewed post-commit, whenever possible.
> 
> I've cleaned up all of the test cases effected by this, mostly by
converting them to a different powerpc target so as to not lose test coverage.
There were a few tests that were ppc-darwin specific, which were removed.
> 
> There is a companion patch to clean up test cases in Clang as well. I think
we can use a similar approach for cleaning up ppc-specific darwin code in Clang,
although I don't have a good idea right now for how much cleanup that will
be.
> 
> Thanks.
> 
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

Thanks for the feedback everyone.

Does anyone know if there are any remaining darwin/ppc build bots? I looked
but could not find any yesterday. I think that is the last thing to take
care of before committing the patches.

Thanks,

Kit

On Thu, Aug 23, 2018 at 12:28 AM Chris Lattner <clattner at nondot.org>
wrote:
> 😥 it was nice knowing you darwin/ppc, but it is now time to say goodbye.
> +1 from me.
>
> -Chris
>
>
> > On Aug 20, 2018, at 12:42 PM, Kit Barton via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
> >
> > Hi all,
> > I just posted two patches on phabricator to remove Darwin support from
> the POWER backend.
> >
> > I thought I had posted this on the dev mailing list some time ago, but
I
> cannot find a reference to it now.
> >
> > The intention here is to cleanup the POWER backend. Darwin has not
been
> supported on POWER for quite some time now, and we'd like to use this
as an
> opportunity to clean up several aspects of the backend code. If there is
> any concerns with this, please let me know.
> >
> > As for the approach, the patch that I posted will issue an error
message
> from the backend if you attempt to use the Darwin ABI on PPC. The intention
> is that once this patch lands, all of the Darwin-specific code paths will
> essentially be dead-code, and we can remove them on demand as we're
> touching various pieces of code. This cleanup should be done in stand-alone
> patches, but can be done as NFC changes and thus reviewed post-commit,
> whenever possible.
> >
> > I've cleaned up all of the test cases effected by this, mostly by
> converting them to a different powerpc target so as to not lose test
> coverage. There were a few tests that were ppc-darwin specific, which were
> removed.
> >
> > There is a companion patch to clean up test cases in Clang as well. I
> think we can use a similar approach for cleaning up ppc-specific darwin
> code in Clang, although I don't have a good idea right now for how much
> cleanup that will be.
> >
> > Thanks.
> >
> > _______________________________________________
> > LLVM Developers mailing list
> > llvm-dev at lists.llvm.org
> > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180824/31826d3f/attachment.html>

On 08/20/2018 02:42 PM, Kit Barton via llvm-dev wrote:
> Hi all,
> I just posted two patches on phabricator to remove Darwin support from
> the POWER backend.
Thanks, Kit.

To connect the two threads, the last RFC we had on this subject was
here: http://lists.llvm.org/pipermail/llvm-dev/2016-October/106359.html
- it looks like we're all now in agreement to move forward.
>
> I thought I had posted this on the dev mailing list some time ago, but
> I cannot find a reference to it now. 
>
> The intention here is to cleanup the POWER backend. Darwin has not
> been supported on POWER for quite some time now, and we'd like to use
> this as an opportunity to clean up several aspects of the backend
> code. If there is any concerns with this, please let me know.
>
> As for the approach, the patch that I posted will issue an error
> message from the backend if you attempt to use the Darwin ABI on PPC.
> The intention is that once this patch lands, all of the
> Darwin-specific code paths will essentially be dead-code, and we can
> remove them on demand as we're touching various pieces of code. This
> cleanup should be done in stand-alone patches, but can be done as NFC
> changes and thus reviewed post-commit, whenever possible.
This is reasonable. Please proceed.

 -Hal
>
> I've cleaned up all of the test cases effected by this, mostly by
> converting them to a different powerpc target so as to not lose test
> coverage. There were a few tests that were ppc-darwin specific, which
> were removed.
>
> There is a companion patch to clean up test cases in Clang as well. I
> think we can use a similar approach for cleaning up ppc-specific
> darwin code in Clang, although I don't have a good idea right now for
> how much cleanup that will be.
>
> Thanks.
>
>
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-- 
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180825/e7c7e4b4/attachment.html>