[gentoo-dev] New global USE flag "neon" for ARM NEON optimization(s)

[gentoo-dev] New global USE flag "neon" for ARM NEON optimization(s)

[Samuli Suominen]

i'll add USE=neon to use.desc and punt the local descriptions if nobody 
objects

media-libs/libpng: support ARM NEON cpu instruction set
media-libs/vo-aacenc: Enable neon cpu instructions
media-video/ffmpeg: Enables NEON optimizations for arm processors.
media-video/libav: Enable NEON optimizations for arm processors.
media-video/vlc: Enables NEON optimizations for arm processors.
x11-libs/pixman: Enables NEON optimizations for ARM processors.

[gentoo-dev] Re: New global USE flag "neon" for ARM NEON optimization(s)

[Duncan]

Samuli Suominen posted on Thu, 22 Dec 2011 21:03:41 +0200 as excerpted:

> i'll add USE=neon to use.desc and punt the local descriptions if nobody
> objects
> 
> media-libs/libpng: support ARM NEON cpu instruction set
> media-libs/vo-aacenc: Enable neon cpu instructions
> media-video/ffmpeg: Enables NEON optimizations for arm processors.
> media-video/libav: Enable NEON optimizations for arm processors.
> media-video/vlc: Enables NEON optimizations for arm processors.
> x11-libs/pixman: Enables NEON optimizations for ARM processors.

For users not familiar with the ARM arch, neon is arguably more likely to 
be associated with net-libs/neon.  Certainly that's what I would have 
immediately thought, tho my only familiarity with the library is as a 
dependency that IIRC used to trigger lots of subversion rebuilds, etc 
(now days I don't even have subversion on the system, it's all git, but 
according to equery d, neon is still required by musicbrainz).

As such, to avoid confusion I'd suggest arch-neon or arm-neon (or armneon/
archneon) if it's to be a global flag.

YMMV...

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

Re: [gentoo-dev] Re: New global USE flag "neon" for ARM NEON optimization(s)

[Matt Turner]

On Fri, Dec 23, 2011 at 4:26 AM, Duncan <1i5t5.duncan@cox.net> wrote:
> Samuli Suominen posted on Thu, 22 Dec 2011 21:03:41 +0200 as excerpted:
>
>> i'll add USE=neon to use.desc and punt the local descriptions if nobody
>> objects
>>
>> media-libs/libpng: support ARM NEON cpu instruction set
>> media-libs/vo-aacenc: Enable neon cpu instructions
>> media-video/ffmpeg: Enables NEON optimizations for arm processors.
>> media-video/libav: Enable NEON optimizations for arm processors.
>> media-video/vlc: Enables NEON optimizations for arm processors.
>> x11-libs/pixman: Enables NEON optimizations for ARM processors.
>
> For users not familiar with the ARM arch, neon is arguably more likely to
> be associated with net-libs/neon.  Certainly that's what I would have
> immediately thought, tho my only familiarity with the library is as a
> dependency that IIRC used to trigger lots of subversion rebuilds, etc
> (now days I don't even have subversion on the system, it's all git, but
> according to equery d, neon is still required by musicbrainz).
>
> As such, to avoid confusion I'd suggest arch-neon or arm-neon (or armneon/
> archneon) if it's to be a global flag.

Ugh.

NEON (the SIMD extensions) are turned on by the neon flag much more
often than support for net-libs/neon is. Let's not rename USE flags
like this.

[gentoo-dev] Re: New global USE flag "neon" for ARM NEON optimization(s)

[Duncan]

Matt Turner posted on Fri, 23 Dec 2011 08:09:30 -0500 as excerpted:

> to avoid confusion I'd suggest arch-neon or arm-neon (or armneon/
>> archneon) if it's to be a global flag.

> NEON (the SIMD extensions) are turned on by the neon flag much more
> often than support for net-libs/neon is. Let's not rename USE flags like
> this.

I'd argue that the library is far more frequently used, given that it's 
used across archs and arm is (for the time being, that seems to be 
gradually changing) still a rather obscure arch (for end-user installed 
distros, anyway), so the neon library is likely the most frequently used.

However you are probably correct about the USE flag, as the library usage 
seems to be required in many cases and thus not USE-flaggable.

So I'd still argue that to prevent confusion... but it's not something I 
feel strongly about, so given no one else objecting, use=neon for the simd 
extensions works as a global USE flag, and if it's ever used for the net-
lib, that one can change I guess.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

Re: [gentoo-dev] Re: New global USE flag "neon" for ARM NEON optimization(s)

[Mike Frysinger]

[-- Attachment #1: Type: Text/Plain, Size: 1769 bytes --]

On Friday 23 December 2011 11:44:32 Duncan wrote:
> Matt Turner posted on Fri, 23 Dec 2011 08:09:30 -0500 as excerpted:
> > to avoid confusion I'd suggest arch-neon or arm-neon (or armneon/
> >> archneon) if it's to be a global flag.
> > 
> > NEON (the SIMD extensions) are turned on by the neon flag much more
> > often than support for net-libs/neon is. Let's not rename USE flags like
> > this.
> 
> I'd argue that the library is far more frequently used, given that it's
> used across archs and arm is (for the time being, that seems to be
> gradually changing) still a rather obscure arch (for end-user installed
> distros, anyway), so the neon library is likely the most frequently used.
> 
> However you are probably correct about the USE flag, as the library usage
> seems to be required in many cases and thus not USE-flaggable.
> 
> So I'd still argue that to prevent confusion... but it's not something I
> feel strongly about, so given no one else objecting, use=neon for the simd
> extensions works as a global USE flag, and if it's ever used for the net-
> lib, that one can change I guess.

if we're going to arch namespace flags, we should do it for all of them.  i 
probably wouldn't complain about that (although i'd prefer more of an ISA 
prefix than Gentoo $ARCH), but doing it for one flag is not nice.

i agree that for some users, they've never heard of the the ARM NEON 
extensions, but they have heard of the neon library.  i'd counter that with a 
few points: (1) i don't think there are any packages in the tree that have 
optional neon (the library) support (2) people are good at reading `quse -D 
neon` (3) the flag is masked everywhere except for ARM ports, so people can't 
really bite themselves.
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: [gentoo-dev] Re: New global USE flag "neon" for ARM NEON optimization(s)

[Samuli Suominen]

> i agree that for some users, they've never heard of the the ARM NEON
> extensions, but they have heard of the neon library.  i'd counter that with a
> few points: (1) i don't think there are any packages in the tree that have
> optional neon (the library) support (2) people are good at reading `quse -D
> neon` (3) the flag is masked everywhere except for ARM ports, so people can't
> really bite themselves.
> -mike

i agree, and it is why i've already committed the global USE=neon few 
days ago (approx)

Re: [gentoo-dev] Re: New global USE flag "neon" for ARM NEON optimization(s)

[Arfrever Frehtes Taifersar Arahesis]

[-- Attachment #1: Type: Text/Plain, Size: 1899 bytes --]

2012-01-18 12:37:12 Mike Frysinger napisał(a):
> On Friday 23 December 2011 11:44:32 Duncan wrote:
> > Matt Turner posted on Fri, 23 Dec 2011 08:09:30 -0500 as excerpted:
> > > to avoid confusion I'd suggest arch-neon or arm-neon (or armneon/
> > >> archneon) if it's to be a global flag.
> > > 
> > > NEON (the SIMD extensions) are turned on by the neon flag much more
> > > often than support for net-libs/neon is. Let's not rename USE flags like
> > > this.
> > 
> > I'd argue that the library is far more frequently used, given that it's
> > used across archs and arm is (for the time being, that seems to be
> > gradually changing) still a rather obscure arch (for end-user installed
> > distros, anyway), so the neon library is likely the most frequently used.
> > 
> > However you are probably correct about the USE flag, as the library usage
> > seems to be required in many cases and thus not USE-flaggable.
> > 
> > So I'd still argue that to prevent confusion... but it's not something I
> > feel strongly about, so given no one else objecting, use=neon for the simd
> > extensions works as a global USE flag, and if it's ever used for the net-
> > lib, that one can change I guess.
> 
> if we're going to arch namespace flags, we should do it for all of them.  i 
> probably wouldn't complain about that (although i'd prefer more of an ISA 
> prefix than Gentoo $ARCH), but doing it for one flag is not nice.
> 
> i agree that for some users, they've never heard of the the ARM NEON 
> extensions, but they have heard of the neon library.  i'd counter that with a 
> few points: (1) i don't think there are any packages in the tree that have 
> optional neon (the library) support

http://qa-reports.gentoo.org/output/genrdeps/rindex/net-libs/neon lists 6 packages, which optionally
depend on net-libs/neon.

-- 
Arfrever Frehtes Taifersar Arahesis

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: [gentoo-dev] Re: New global USE flag "neon" for ARM NEON optimization(s)

[Mike Frysinger]

[-- Attachment #1: Type: Text/Plain, Size: 714 bytes --]

On Wednesday 18 January 2012 13:40:23 Arfrever wrote:
> 2012-01-18 12:37:12 Mike Frysinger napisał(a):
> > i agree that for some users, they've never heard of the the ARM NEON
> > extensions, but they have heard of the neon library.  i'd counter that
> > with a few points: (1) i don't think there are any packages in the tree
> > that have optional neon (the library) support
> 
> http://qa-reports.gentoo.org/output/genrdeps/rindex/net-libs/neon lists 6
> packages, which optionally depend on net-libs/neon.

fair enough.  some of those deps prob still wouldn't be USE=neon since they're 
webdev related, and others are xml related.

but too late now as the ship as sailed for everyone :).
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]