RFC: Backport architecture specific improvements to release branches

RFC: Backport architecture specific improvements to release branches

[H.J. Lu]

Hi,

We'd like to backport architecture specific improvements to release branches,
similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
branch:

https://gitlab.com/x86-glibc/glibc/-/commits/users/intel/x86/2.34

Any comments?

Thanks.

--
H.J.

Re: RFC: Backport architecture specific improvements to release branches

[Noah Goldstein]

On Tue, Mar 29, 2022 at 12:29 PM H.J. Lu via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> Hi,
>
> We'd like to backport architecture specific improvements to release branches,
> similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
> branch:
>
> https://gitlab.com/x86-glibc/glibc/-/commits/users/intel/x86/2.34

Given that the compiler does not emit `bzero` and we dropped the `__memsetzero`
optimization maybe best w.o:
https://gitlab.com/x86-glibc/glibc/-/commit/76e17e47789ef4faf1367b2e48c19763559c79a0
for the sake of code size.

>
> Any comments?
>
> Thanks.
>
> --
> H.J.

Re: RFC: Backport architecture specific improvements to release branches

[H.J. Lu]

On Tue, Mar 29, 2022 at 10:36 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>
> On Tue, Mar 29, 2022 at 12:29 PM H.J. Lu via Libc-alpha
> <libc-alpha@sourceware.org> wrote:
> >
> > Hi,
> >
> > We'd like to backport architecture specific improvements to release branches,
> > similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
> > branch:
> >
> > https://gitlab.com/x86-glibc/glibc/-/commits/users/intel/x86/2.34
>
> Given that the compiler does not emit `bzero` and we dropped the `__memsetzero`
> optimization maybe best w.o:
> https://gitlab.com/x86-glibc/glibc/-/commit/76e17e47789ef4faf1367b2e48c19763559c79a0
> for the sake of code size.

I like to make the release branch as close to the master branch as possible
to make backporting easier.

> >
> > Any comments?
> >
> > Thanks.
> >
> > --
> > H.J.



-- 
H.J.

Re: RFC: Backport architecture specific improvements to release branches

[Noah Goldstein]

On Tue, Mar 29, 2022 at 2:02 PM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> On Tue, Mar 29, 2022 at 10:36 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
> >
> > On Tue, Mar 29, 2022 at 12:29 PM H.J. Lu via Libc-alpha
> > <libc-alpha@sourceware.org> wrote:
> > >
> > > Hi,
> > >
> > > We'd like to backport architecture specific improvements to release branches,
> > > similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
> > > branch:
> > >
> > > https://gitlab.com/x86-glibc/glibc/-/commits/users/intel/x86/2.34
> >
> > Given that the compiler does not emit `bzero` and we dropped the `__memsetzero`
> > optimization maybe best w.o:
> > https://gitlab.com/x86-glibc/glibc/-/commit/76e17e47789ef4faf1367b2e48c19763559c79a0
> > for the sake of code size.
>
> I like to make the release branch as close to the master branch as possible
> to make backporting easier.
>

Should we drop the optimized bzero on master? Or keep both?
> > >
> > > Any comments?
> > >
> > > Thanks.
> > >
> > > --
> > > H.J.
>
>
>
> --
> H.J.

Re: RFC: Backport architecture specific improvements to release branches

[Adhemerval Zanella]

On 29/03/2022 16:19, Noah Goldstein via Libc-alpha wrote:
> On Tue, Mar 29, 2022 at 2:02 PM H.J. Lu <hjl.tools@gmail.com> wrote:
>>
>> On Tue, Mar 29, 2022 at 10:36 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>>>
>>> On Tue, Mar 29, 2022 at 12:29 PM H.J. Lu via Libc-alpha
>>> <libc-alpha@sourceware.org> wrote:
>>>>
>>>> Hi,
>>>>
>>>> We'd like to backport architecture specific improvements to release branches,
>>>> similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
>>>> branch:
>>>>
>>>> https://gitlab.com/x86-glibc/glibc/-/commits/users/intel/x86/2.34
>>>
>>> Given that the compiler does not emit `bzero` and we dropped the `__memsetzero`
>>> optimization maybe best w.o:
>>> https://gitlab.com/x86-glibc/glibc/-/commit/76e17e47789ef4faf1367b2e48c19763559c79a0
>>> for the sake of code size.
>>
>> I like to make the release branch as close to the master branch as possible
>> to make backporting easier.
>>
> 
> Should we drop the optimized bzero on master? Or keep both?

As a side note, I have removed bzero optimization on all targets except x86.
It is on my plan to clean x86 as well.

Re: RFC: Backport architecture specific improvements to release branches

[Adhemerval Zanella]

On 29/03/2022 14:28, H.J. Lu via Libc-alpha wrote:
> Hi,
> 
> We'd like to backport architecture specific improvements to release branches,
> similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
> branch:
> 
> https://gitlab.com/x86-glibc/glibc/-/commits/users/intel/x86/2.34
> 
> Any comments?


I guess it should be ok, it might add some extra burden on performance bug
triage if consumer use the release tarball.  But it does make sense to make
it as close as possible to release.

Re: RFC: Backport architecture specific improvements to release branches

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 707 bytes --]

Hi, 

> We'd like to backport architecture specific improvements to release branches,
> similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
> branch:

please, if you do this, when possible make some well-defined patch series
(with a start and end) instead of a continuous trickle of fixes and improvements.

[When I tried to make a cutoff for the 2.34 patchlevel to be "stabilized" for Gentoo, 
I was never sure if the series is finished or if another important thing follows
in a few days...]

Thanks a lot for your work.

Cheers,
Andreas

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: RFC: Backport architecture specific improvements to release branches

[Adhemerval Zanella]

On 29/03/2022 16:33, Andreas K. Huettel via Libc-alpha wrote:
> Hi, 
> 
>> We'd like to backport architecture specific improvements to release branches,
>> similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
>> branch:
> 
> please, if you do this, when possible make some well-defined patch series
> (with a start and end) instead of a continuous trickle of fixes and improvements.
> 
> [When I tried to make a cutoff for the 2.34 patchlevel to be "stabilized" for Gentoo, 
> I was never sure if the series is finished or if another important thing follows
> in a few days...]
> 
> Thanks a lot for your work.
> 
> Cheers,
> Andreas
> 

Specially for release branches, each patch should be self-contained (meaning it builds
and do not trigger any regression).  Are you seeing something differently?

Re: RFC: Backport architecture specific improvements to release branches

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 355 bytes --]

> 
> Specially for release branches, each patch should be self-contained (meaning it builds
> and do not trigger any regression).  Are you seeing something differently?
> 

That's a great approach. I fully agree with this.


-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer
(council, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: RFC: Backport architecture specific improvements to release branches

[H.J. Lu]

On Tue, Mar 29, 2022 at 12:30 PM Adhemerval Zanella via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
>
>
> On 29/03/2022 16:19, Noah Goldstein via Libc-alpha wrote:
> > On Tue, Mar 29, 2022 at 2:02 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> >>
> >> On Tue, Mar 29, 2022 at 10:36 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
> >>>
> >>> On Tue, Mar 29, 2022 at 12:29 PM H.J. Lu via Libc-alpha
> >>> <libc-alpha@sourceware.org> wrote:
> >>>>
> >>>> Hi,
> >>>>
> >>>> We'd like to backport architecture specific improvements to release branches,
> >>>> similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
> >>>> branch:
> >>>>
> >>>> https://gitlab.com/x86-glibc/glibc/-/commits/users/intel/x86/2.34
> >>>
> >>> Given that the compiler does not emit `bzero` and we dropped the `__memsetzero`
> >>> optimization maybe best w.o:
> >>> https://gitlab.com/x86-glibc/glibc/-/commit/76e17e47789ef4faf1367b2e48c19763559c79a0
> >>> for the sake of code size.
> >>
> >> I like to make the release branch as close to the master branch as possible
> >> to make backporting easier.
> >>
> >
> > Should we drop the optimized bzero on master? Or keep both?
>
> As a side note, I have removed bzero optimization on all targets except x86.
> It is on my plan to clean x86 as well.

This will be included in backport.

-- 
H.J.

Re: RFC: Backport architecture specific improvements to release branches

[Florian Weimer]

* H. J. Lu via Libc-alpha:

> We'd like to backport architecture specific improvements to release branches,
> similar to backports under sysdeps/x86_64/multiarch on users/intel/x86/2.34
> branch:
>
> https://gitlab.com/x86-glibc/glibc/-/commits/users/intel/x86/2.34
>
> Any comments?

I would be reluctant to backport the new set of RTM string functions,
but those are already in 2.34, so no objection from me.

Thanks,
Florian