llvm dev - Nov 2015 - [cfe-dev] [RFC] __attribute__((internal_linkage))

I haven't been able to figure out from this thread why this attribute is
necessary at all.

Anonymous or unnamed namespaces were added to C++ for this very purpose, but the
ISO C++ Standard does not discuss "linkage" per-se because it is
outside the scope of the language specification.  But it does describes it in
terms of having a hidden name which is "unique" to the translation
unit, and under the "as if" rule, internal linkage is a common
convention for achieving this.

This is just Standardese for dealing with what compiler implementers typically
handle as "internal linkage" anyway; at the same time, the use of
'static' for this purpose was deprecated.  This is specifically stated
in C++98 section 7.3.1.1 and unnamed namespaces are still similarly defined in
C++17 (Working Paper) although the specific reference to the deprecation of
'static' for this purpose is now gone.

The closest the Standard gets to talking about linkage is with Linkage
Specifications, and even here it tries to avoid to avoid treading on the
definitions things like internal versus external linkage.

So I am curious, what does this proposed attribute on namespaces achieve that
cannot already be achieved with an anonymous or unnamed namespace?

Thanks,

	Martin O'Riordan - Movidius Ltd.

-----Original Message-----
From: cfe-dev [mailto:cfe-dev-bounces at lists.llvm.org] On Behalf Of Evgenii
Stepanov via cfe-dev
Sent: 23 October 2015 22:41
To: Steven Stewart-Gallus <sstewartgallus00 at mylangara.bc.ca>
Cc: Clang Dev <cfe-dev at lists.llvm.org>
Subject: Re: [cfe-dev] [RFC] __attribute__((internal_linkage))

Sounds right. The proposed attribute is a rough equivalent of

static void foo::do_method(void)
{
  // implementation here
}

where "static" does not mean a static class member, but makes the
declaration local to the translation unit. C-style "static".

The patch in http://reviews.llvm.org/D13925 (waiting for review, btw!) supports
this attribute on classes and namespaces, with this syntax:

namespace __attribute__((internal_linkage)) { }


On Fri, Oct 23, 2015 at 9:03 AM, Steven Stewart-Gallus via cfe-dev <cfe-dev
at lists.llvm.org> wrote:
> Hello,
>
> Can I ask for some clarification?
>
> Apparently, C++ doesn't allow to static class methods?
>
> While in C one might write inside a header file:
>
> static inline void mylib_foo_do_method(struct mylib_foo *foo) {
>    // implementation here
> }
>
> In C++ the typical style would be to write something like:
>
> namespace mylib {
>
> void foo::do_method(void)
> {
>   // implementation here
> }
> }
>
> Unfortunately, the C++ case then implies some linkage behaviour that 
> some might not want.
>
> Apparently, one can't do things like:
>
> namespace mylib {
>
> static void foo::do_method(void)
> {
>   // implementation here
> }
> }
>
> Or,
>
> namespace {
> namespace mylib {
>
> static void foo::do_method(void)
> {
>   // implementation here
> }
> }
> }
>
> Also, if you wanted to an attribute to a whole namespace you should do 
> something like the following I think:
>
> namespace mylib {
> [[clang::internal_linkage]];
>
> static void foo::do_method(void)
> {
>   // implementation here
> }
> }
>
>
> Thank you,
> Steven Stewart-Gallus
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
_______________________________________________
cfe-dev mailing list
cfe-dev at lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev

Hi Stewart,

I saw this get brought up at the LLVM devmtg in the context of improving the
stack size of libc++ frames.

Have you guys considered a different approach to solving this problem?  In the
case of libc++, the _LIBCPP_INLINE_VISIBILITY macro currently expands out to
"__attribute__ ((__always_inline__))”.  It seems reasonable to me to have
it also add the “nodebug” attribute as well.  This should allow the allocas
generated by inlining lots of frames to be promoted to registers (because there
is no debug info to pessimize).

This would dramatically shrink the stack frames of code using libc++ at -O0, and
would also make it go a lot faster.

-Chris

> On Oct 29, 2015, at 6:35 AM, Martin J. O'Riordan via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> 
> I haven't been able to figure out from this thread why this attribute
is necessary at all.
> 
> Anonymous or unnamed namespaces were added to C++ for this very purpose,
but the ISO C++ Standard does not discuss "linkage" per-se because it
is outside the scope of the language specification.  But it does describes it in
terms of having a hidden name which is "unique" to the translation
unit, and under the "as if" rule, internal linkage is a common
convention for achieving this.
> 
> This is just Standardese for dealing with what compiler implementers
typically handle as "internal linkage" anyway; at the same time, the
use of 'static' for this purpose was deprecated.  This is specifically
stated in C++98 section 7.3.1.1 and unnamed namespaces are still similarly
defined in C++17 (Working Paper) although the specific reference to the
deprecation of 'static' for this purpose is now gone.
> 
> The closest the Standard gets to talking about linkage is with Linkage
Specifications, and even here it tries to avoid to avoid treading on the
definitions things like internal versus external linkage.
> 
> So I am curious, what does this proposed attribute on namespaces achieve
that cannot already be achieved with an anonymous or unnamed namespace?
> 
> Thanks,
> 
> 	Martin O'Riordan - Movidius Ltd.
> 
> -----Original Message-----
> From: cfe-dev [mailto:cfe-dev-bounces at lists.llvm.org] On Behalf Of
Evgenii Stepanov via cfe-dev
> Sent: 23 October 2015 22:41
> To: Steven Stewart-Gallus <sstewartgallus00 at mylangara.bc.ca>
> Cc: Clang Dev <cfe-dev at lists.llvm.org>
> Subject: Re: [cfe-dev] [RFC] __attribute__((internal_linkage))
> 
> Sounds right. The proposed attribute is a rough equivalent of
> 
> static void foo::do_method(void)
> {
>  // implementation here
> }
> 
> where "static" does not mean a static class member, but makes the
declaration local to the translation unit. C-style "static".
> 
> The patch in http://reviews.llvm.org/D13925 (waiting for review, btw!)
supports this attribute on classes and namespaces, with this syntax:
> 
> namespace __attribute__((internal_linkage)) { }
> 
> 
> On Fri, Oct 23, 2015 at 9:03 AM, Steven Stewart-Gallus via cfe-dev
<cfe-dev at lists.llvm.org> wrote:
>> Hello,
>> 
>> Can I ask for some clarification?
>> 
>> Apparently, C++ doesn't allow to static class methods?
>> 
>> While in C one might write inside a header file:
>> 
>> static inline void mylib_foo_do_method(struct mylib_foo *foo) {
>>   // implementation here
>> }
>> 
>> In C++ the typical style would be to write something like:
>> 
>> namespace mylib {
>> 
>> void foo::do_method(void)
>> {
>>  // implementation here
>> }
>> }
>> 
>> Unfortunately, the C++ case then implies some linkage behaviour that 
>> some might not want.
>> 
>> Apparently, one can't do things like:
>> 
>> namespace mylib {
>> 
>> static void foo::do_method(void)
>> {
>>  // implementation here
>> }
>> }
>> 
>> Or,
>> 
>> namespace {
>> namespace mylib {
>> 
>> static void foo::do_method(void)
>> {
>>  // implementation here
>> }
>> }
>> }
>> 
>> Also, if you wanted to an attribute to a whole namespace you should do 
>> something like the following I think:
>> 
>> namespace mylib {
>> [[clang::internal_linkage]];
>> 
>> static void foo::do_method(void)
>> {
>>  // implementation here
>> }
>> }
>> 
>> 
>> Thank you,
>> Steven Stewart-Gallus
>> _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
> 
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

> -----Original Message-----
> From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of
Chris
> Lattner via llvm-dev
> Sent: Saturday, October 31, 2015 2:58 PM
> To: sstewartgallus00 at mylangara.bc.ca
> Cc: LLVM Developers
> Subject: Re: [llvm-dev] [cfe-dev] [RFC] __attribute__((internal_linkage))
> 
> Hi Stewart,
> 
> I saw this get brought up at the LLVM devmtg in the context of improving
> the stack size of libc++ frames.
> 
> Have you guys considered a different approach to solving this problem?  In
> the case of libc++, the _LIBCPP_INLINE_VISIBILITY macro currently expands
> out to "__attribute__ ((__always_inline__))”.  It seems reasonable to
me
> to have it also add the “nodebug” attribute as well.  This should allow
> the allocas generated by inlining lots of frames to be promoted to
> registers (because there is no debug info to pessimize).
Are you suggesting that 'nodebug' should affect code generation?
It most definitely should not...
--paulr
> 
> This would dramatically shrink the stack frames of code using libc++ at -
> O0, and would also make it go a lot faster.
> 
> -Chris
> 
> 
> > On Oct 29, 2015, at 6:35 AM, Martin J. O'Riordan via llvm-dev
<llvm-
> dev at lists.llvm.org> wrote:
> >
> > I haven't been able to figure out from this thread why this
attribute is
> necessary at all.
> >
> > Anonymous or unnamed namespaces were added to C++ for this very
purpose,
> but the ISO C++ Standard does not discuss "linkage" per-se
because it is
> outside the scope of the language specification.  But it does describes it
> in terms of having a hidden name which is "unique" to the
translation
> unit, and under the "as if" rule, internal linkage is a common
convention
> for achieving this.
> >
> > This is just Standardese for dealing with what compiler implementers
> typically handle as "internal linkage" anyway; at the same time,
the use
> of 'static' for this purpose was deprecated.  This is specifically
stated
> in C++98 section 7.3.1.1 and unnamed namespaces are still similarly
> defined in C++17 (Working Paper) although the specific reference to the
> deprecation of 'static' for this purpose is now gone.
> >
> > The closest the Standard gets to talking about linkage is with Linkage
> Specifications, and even here it tries to avoid to avoid treading on the
> definitions things like internal versus external linkage.
> >
> > So I am curious, what does this proposed attribute on namespaces
achieve
> that cannot already be achieved with an anonymous or unnamed namespace?
> >
> > Thanks,
> >
> > 	Martin O'Riordan - Movidius Ltd.
> >
> > -----Original Message-----
> > From: cfe-dev [mailto:cfe-dev-bounces at lists.llvm.org] On Behalf Of
> Evgenii Stepanov via cfe-dev
> > Sent: 23 October 2015 22:41
> > To: Steven Stewart-Gallus <sstewartgallus00 at mylangara.bc.ca>
> > Cc: Clang Dev <cfe-dev at lists.llvm.org>
> > Subject: Re: [cfe-dev] [RFC] __attribute__((internal_linkage))
> >
> > Sounds right. The proposed attribute is a rough equivalent of
> >
> > static void foo::do_method(void)
> > {
> >  // implementation here
> > }
> >
> > where "static" does not mean a static class member, but
makes the
> declaration local to the translation unit. C-style "static".
> >
> > The patch in http://reviews.llvm.org/D13925 (waiting for review, btw!)
> supports this attribute on classes and namespaces, with this syntax:
> >
> > namespace __attribute__((internal_linkage)) { }
> >
> >
> > On Fri, Oct 23, 2015 at 9:03 AM, Steven Stewart-Gallus via cfe-dev
<cfe-
> dev at lists.llvm.org> wrote:
> >> Hello,
> >>
> >> Can I ask for some clarification?
> >>
> >> Apparently, C++ doesn't allow to static class methods?
> >>
> >> While in C one might write inside a header file:
> >>
> >> static inline void mylib_foo_do_method(struct mylib_foo *foo) {
> >>   // implementation here
> >> }
> >>
> >> In C++ the typical style would be to write something like:
> >>
> >> namespace mylib {
> >>
> >> void foo::do_method(void)
> >> {
> >>  // implementation here
> >> }
> >> }
> >>
> >> Unfortunately, the C++ case then implies some linkage behaviour
that
> >> some might not want.
> >>
> >> Apparently, one can't do things like:
> >>
> >> namespace mylib {
> >>
> >> static void foo::do_method(void)
> >> {
> >>  // implementation here
> >> }
> >> }
> >>
> >> Or,
> >>
> >> namespace {
> >> namespace mylib {
> >>
> >> static void foo::do_method(void)
> >> {
> >>  // implementation here
> >> }
> >> }
> >> }
> >>
> >> Also, if you wanted to an attribute to a whole namespace you
should do
> >> something like the following I think:
> >>
> >> namespace mylib {
> >> [[clang::internal_linkage]];
> >>
> >> static void foo::do_method(void)
> >> {
> >>  // implementation here
> >> }
> >> }
> >>
> >>
> >> Thank you,
> >> Steven Stewart-Gallus
> >> _______________________________________________
> >> cfe-dev mailing list
> >> cfe-dev at lists.llvm.org
> >> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
> > _______________________________________________
> > cfe-dev mailing list
> > cfe-dev at lists.llvm.org
> > http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
> >
> > _______________________________________________
> > LLVM Developers mailing list
> > llvm-dev at lists.llvm.org
> > 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