Monday Patch Queue Review update (2023-01-23)

Monday Patch Queue Review update (2023-01-23)

[Carlos O'Donell]

Most recent meeting status is always here:
https://sourceware.org/glibc/wiki/PatchworkReviewMeetings#Update

Meeting: 2023-01-23 @ 0900h EST5EDT

Video/Audio: https://bbb.linuxfoundation.org/room/adm-alk-1uu-7fu

IRC: #glibc on OFTC.

Review new patches and restart review at the top.

 * State NEW delegate NOBODY at 246 patches.
 * Carlos' SLI at 168 days.
 * Discussing glibc 2.37 release status.
 * Discussed "Remove --with-default-link option" (Adhemerval)
  * It is a major improvement for building with lld.
  * Improves security when building with lld.
 * Moved the "Remove --with-default-link option" to 2.38.
 * Marked "x86: Cache computation for AMD architecture" as committed.
 * Discussed review for "[v8] POSIX locale covers every byte [BZ# 29511]" and status.
  * Needs more strategic review about glibc direction.
 * This week starts machine testing for the release.
  * Gentoo is already starting to test.
  * Carlos to send out the request for machine testing.
 * Discussed RISC-V ABI break in the context of 2.37.
  * AArch64 uses the pointer-chain method and it works.
   * Szabolcs: No reported issues here.
 * Dicsussed PT_GNU_SFRAME constant in elf.h for 2.37
  * Needs coordination between bintuils, glibc, and gnu-gabi.
 * Discussed pthread cond var signal stealing.
  * Can be backported.
 * Discussed _r_debug and copy relocations.
  * Can be backported.
 * Discussed Math TS18661-2 for inclusion in C2x and impact.
  * No impact to the release, but can impact applications that define iszero().

-- 
Cheers,
Carlos.

Re: Monday Patch Queue Review update (2023-01-23)

[Joseph Myers]

[-- Attachment #1: Type: text/plain, Size: 730 bytes --]

On Mon, 23 Jan 2023, Carlos O'Donell via Libc-alpha wrote:

>  * Discussed Math TS18661-2 for inclusion in C2x and impact.
>   * No impact to the release, but can impact applications that define iszero().

18661-2 is decimal floating point - do we have someone interested in 
integrating it now?  
<https://sourceware.org/pipermail/libc-alpha/2019-September/106579.html> 
discusses relevant considerations.

18661-4 (supplementary functions) I hope to work on after 2.37 is out (at 
least to the extent that those functions were integrated in C2x), time 
permitting, given that we now have the MPFR 4.2 release that supports 
those functions and so can be used in gen-auto-libm-tests.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: Monday Patch Queue Review update (2023-01-23)

[Paul Zimmermann]

       Dear Joseph,

> 18661-4 (supplementary functions) I hope to work on after 2.37 is out (at 
> least to the extent that those functions were integrated in C2x), time 
> permitting, given that we now have the MPFR 4.2 release that supports 
> those functions and so can be used in gen-auto-libm-tests.

this is great news! For binary32, all these functions are already available in
CORE-MATH [1]. Feel free to reuse.

Paul

[1] https://gitlab.inria.fr/core-math/core-math/-/tree/master/src/binary32

Re: Monday Patch Queue Review update (2023-01-23)

[Adhemerval Zanella Netto]

On 23/01/23 17:00, Joseph Myers wrote:
> On Mon, 23 Jan 2023, Carlos O'Donell via Libc-alpha wrote:
> 
>>  * Discussed Math TS18661-2 for inclusion in C2x and impact.
>>   * No impact to the release, but can impact applications that define iszero().
> 
> 18661-2 is decimal floating point - do we have someone interested in 
> integrating it now?  
> <https://sourceware.org/pipermail/libc-alpha/2019-September/106579.html> 
> discusses relevant considerations.

I don't recall Carlos remark about decimal floating point on the call, so
maybe he was referring to something else.

Re: Monday Patch Queue Review update (2023-01-23)

[Joseph Myers]

On Tue, 24 Jan 2023, Paul Zimmermann via Libc-alpha wrote:

>        Dear Joseph,
> 
> > 18661-4 (supplementary functions) I hope to work on after 2.37 is out (at 
> > least to the extent that those functions were integrated in C2x), time 
> > permitting, given that we now have the MPFR 4.2 release that supports 
> > those functions and so can be used in gen-auto-libm-tests.
> 
> this is great news! For binary32, all these functions are already available in
> CORE-MATH [1]. Feel free to reuse.
> 
> Paul
> 
> [1] https://gitlab.inria.fr/core-math/core-math/-/tree/master/src/binary32

I expect each function implementation would take substantial work to adapt 
to use in glibc.  My plan is that as far as possible I'll do type-generic 
template implementations in terms of existing functions, to avoid needing 
separate work for each of five floating-point formats for most functions - 
and then it will be possible for people wishing to override the 
type-generic implementations with ones specific to a given format that are 
faster and more accurate to do so incrementally.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: Monday Patch Queue Review update (2023-01-23)

[Carlos O'Donell]

On 1/24/23 14:12, Adhemerval Zanella Netto wrote:
> 
> 
> On 23/01/23 17:00, Joseph Myers wrote:
>> On Mon, 23 Jan 2023, Carlos O'Donell via Libc-alpha wrote:
>>
>>>  * Discussed Math TS18661-2 for inclusion in C2x and impact.
>>>   * No impact to the release, but can impact applications that define iszero().
>>
>> 18661-2 is decimal floating point - do we have someone interested in 
>> integrating it now?  
>> <https://sourceware.org/pipermail/libc-alpha/2019-September/106579.html> 
>> discusses relevant considerations.
> 
> I don't recall Carlos remark about decimal floating point on the call, so
> maybe he was referring to something else.
 
To be honest I don't know why I wrote TS18661-2 in the notes, maybe
a cut-and-paste error. I didn't mean to write "TS18661-2."

The iszero() function is directly defined in C2x, not in any TSs, 
and we were discussing the C2x functions for math.h.

-- 
Cheers,
Carlos.