System z vector language extensions for 3.7 branch?

Hi Hans, hi Richard,

I've just checked support for the System z vector language extensions (and
associated intrinsics header file) into clang mainline as revisions r243642
and r243643.

This is the last missing part to complete support for the IBM z13 mainframe
architecture in LLVM and clang at a level compatible with current GCC
mainline. All the other parts (LLVM back-end support, and basic clang
support for -march=z13 except for the language extension) are already in
3.7 since they went in before the release branched. The language extension
however is not in the branch.

In order for LLVM 3.7 to be able to claim full z13 support, it would be
really good if that last missing feature could still make it into the
branch. I'm aware that we're already quite late in the cycle -- do you
still see a chance to get it in? If there's anything I can to do help with
this, please let me know. Thanks!

Mit freundlichen Gruessen / Best Regards

Ulrich Weigand

Hi Ulrich,

I'm happy to merge if Richard gives the owner's approval.

Cheers,
Hans

Hi Hans,

thanks! Just a quick update: it turns out the sanitizer build bots
uncovered an uninitialized variable access. I've now fixed this in
r243644, and the build bot is green again. So if this feature is
backported, please add that revision as well.

Sorry for the extra confusion ...

Bye,
Ulrich

Hi Ulrich,

I'm happy to merge if Richard gives the owner's approval.

Cheers,
Hans

>
> Hi Hans, hi Richard,
>
> I've just checked support for the System z vector language extensions

(and

> associated intrinsics header file) into clang mainline as revisions

r243642

> and r243643.
>
> This is the last missing part to complete support for the IBM z13

mainframe

> architecture in LLVM and clang at a level compatible with current GCC
> mainline. All the other parts (LLVM back-end support, and basic clang
> support for -march=z13 except for the language extension) are already

in

> 3.7 since they went in before the release branched. The language

extension

> however is not in the branch.
>
> In order for LLVM 3.7 to be able to claim full z13 support, it would be
> really good if that last missing feature could still make it into the
> branch. I'm aware that we're already quite late in the cycle -- do you
> still see a chance to get it in? If there's anything I can to do help

with

> this, please let me know. Thanks!
>
>
> Mit freundlichen Gruessen / Best Regards
>
> Ulrich Weigand
>
> --
> Dr. Ulrich Weigand | Phone: +49-7031/16-3727
> STSM, GNU/Linux compilers and toolchain
> IBM Deutschland Research & Development GmbH
> Vorsitzende des Aufsichtsrats: Martina Koederitz | Geschäftsführung:

Dirk

Hi Hans,

thanks! Just a quick update: it turns out the sanitizer build bots
uncovered an uninitialized variable access. I've now fixed this in
r243644, and the build bot is green again. So if this feature is
backported, please add that revision as well.

Sorry for the extra confusion ...

Bye,
Ulrich

> Hi Ulrich,
>
> I'm happy to merge if Richard gives the owner's approval.

It's a bit unusual to take a whole new feature at this stage in the
release, but if you're happy with it, I think it's a suitably low-risk
patch (the biggest risk is that the feature itself would not work in some
way), so go ahead.

Hi Hans,

thanks! Just a quick update: it turns out the sanitizer build bots
uncovered an uninitialized variable access. I've now fixed this in
r243644, and the build bot is green again. So if this feature is
backported, please add that revision as well.

Sorry for the extra confusion ...

Bye,
Ulrich

> Hi Ulrich,
>
> I'm happy to merge if Richard gives the owner's approval.

It's a bit unusual to take a whole new feature at this stage in the release,
but if you're happy with it, I think it's a suitably low-risk patch (the
biggest risk is that the feature itself would not work in some way), so go
ahead.

It's on the cusp, but I figured it could go in under the "finish up
existing work" wording for Phase I.

Merged r243642, r243643, and r243644 to 3.7 in r243699.

Thanks,
Hans

>>
>> > Hi Ulrich,
>> >
>> > I'm happy to merge if Richard gives the owner's approval.
>
>
> It's a bit unusual to take a whole new feature at this stage in the

release,

> but if you're happy with it, I think it's a suitably low-risk patch

(the

> biggest risk is that the feature itself would not work in some way), so

go

> ahead.

It's on the cusp, but I figured it could go in under the "finish up
existing work" wording for Phase I.

Merged r243642, r243643, and r243644 to 3.7 in r243699.

Thanks Hans and Richard for your support on this!

I've now tested the 3.7 branch, and everything looks good to me
(clang bootstrap, test-suite, GCC/clang ABI compatibility tests).

Bye,
Ulrich