Q: inlining

Q: inlining

[Hannes Haug]

Hi,

I'm new to Ada and have a question on inlining. Does
inlining work across compilation units ? In C I can
write the function definition in a header file. How
can I do this in Ada ?

Thanks

 - hannes

Re: Q: inlining

[Jon S Anthony]

In article <uvv20jh9j7p.fsf@chaq.informatik.uni-tuebingen.de> Hannes Haug <Hannes.Haug@Student.Uni-Tuebingen.de> writes:

> Hi,
> 
> I'm new to Ada and have a question on inlining. Does
> inlining work across compilation units ? In C I can

It can, yes.  I'm not sure if this is working in GNAT 3.05 yet or not.


> write the function definition in a header file. How
> can I do this in Ada ?

In a package specification.  In fact, this is virtually a requirement
to getting any "real" work done.

You might want to check out the following for online tutorials,
references, a good Ada for C++ programmers paper, etc:

http://lglwww.epfl.ch/Ada/

/Jon
-- 
Jon Anthony
Organon Motives, Inc.
1 Williston Road, Suite 4
Belmont, MA 02178

617.484.3383
jsa@organon.com

Re: Q: inlining

[Hannes Haug]

>>>>> "Jon" == Jon S Anthony <jsa@organon.com> writes:

    Jon> In article <uvv20jh9j7p.fsf@chaq.informatik.uni-tuebingen.de>
    Jon> Hannes Haug <Hannes.Haug@Student.Uni-Tuebingen.de> writes:
    >> Hi,
    >> 
    >> I'm new to Ada and have a question on inlining. Does inlining
    >> work across compilation units ? In C I can

    Jon> It can, yes.  I'm not sure if this is working in GNAT 3.05
    Jon> yet or not.

I know that this is possible. I just wanted to know if compilers actually do
it. I finally found the option for GNAT: -gnatn.


    >> write the function definition in a header file. How can I do
    >> this in Ada ?

    Jon> In a package specification.  In fact, this is virtually a
    Jon> requirement to getting any "real" work done.

I know. This made me look at Ada. It's really ugly to put all the stuff
in header files.

 - hannes

Re: Q: inlining

[Robert Dewar]

Jon Anthony said (of inter-unit inlining):

"It can, yes.  I'm not sure if this is working in GNAT 3.05 yet or not."

Certainly it is! Once again Robert consults his secret source of GNAT
information (namely the standard documentation in gnatinfo.txt :-)
and provides the full answer:

Inlining of Subprograms
-----------------------

A call to a subprogram in the current unit is inlined if all the following
conditions are met:

    o  The optimization level is at least -O1

    o  The called subprogram is suitable for inlining. It must be
       small enough and not contain nested subprograms or anything
       else that GCC cannot support in inlined subprograms.

    o  The call occurs after the definition of the body of the subprogram.

    o  Either pragma Inline applies to the subprogram, or it is very small
       and automatic inlining (optimization level -O3) is specified.

Calls to subprograms in with'ed units are normally not inlined. To achieve
this level of inlining, the following conditions must be true.

    o  The optimization level is at least -O1

    o  The called subprogram is suitable for inlining. It must be
       small enough and not contain nested subprograms or anything
       else that GCC cannot support in inlined subprograms.

    o  The call appears in a body (not in a package spec).

    o  There is a pragma Inline for the subprogram

    o  The -gnatn switch is used in the GCC command line

Note that specifying the -gnatn switch causes additional compilation
dependencies. Consider the following:


   package R is
     procedure Q;
     pragma Inline Q;
   end R;

   package body R is
     ...
   end R;

   with R;
   procedure Main is
   begin
     ...
     R.Q;
   end Main;

With the default behavior (no -gnatn switch specified), the compilation of
the Main procedure depends only on its own source, main.adb, and the spec
of the package in file r.ads. This means that editing the body of R does
not require recompiling Main.

On the other hand, the call R.Q is not inlined under these circumstances. If
the -gnatn switch is present when Main is compiled, then the call will be
inlined if the body of Q is small enough, but now Main depends on the body
of R in r.adb as well as the spec. This means that if the body is edited,
then the main program must be recompiled. Note that this extra dependency
occurs whether or not the call is in fact inlined by GCC.

Note: the GCC switch -fno-inline can be used to prevent all inlining. This
switch overrides all other conditions, and ensures that no inlining occurs.
The extra dependencies resulting from -gnatn will still be active, even if
the -fno-inline switch is used.


------------------
All this is in the section:

PERFORMANCE CONSIDERATIONS

which contains lots of useful stuff. In fact there is all sorts of
useful stuff in gnatinfo.txt, recommended reading for all GNAT users!

Re: Q: inlining

[Tucker Taft]

Jon S Anthony (jsa@organon.com) wrote:
: In article <uvv20jh9j7p.fsf@chaq.informatik.uni-tuebingen.de> 
: Hannes Haug <Hannes.Haug@Student.Uni-Tuebingen.de> writes:

: > Hi,
: > 
: > I'm new to Ada and have a question on inlining. Does
: > inlining work across compilation units ? In C I can

: It can, yes.  I'm not sure if this is working in GNAT 3.05 yet or not.


: > write the function definition in a header file. How
: > can I do this in Ada ?

: In a package specification.  In fact, this is virtually a requirement
: to getting any "real" work done.

This answer could be confusing.  In C++, the function *definition*
usually needs to be in the header file for inlining to work.  In Ada,
only the function *declaration* is allowed in the package spec.  The
function body ("definition" in C parlance) goes in the package body, even
if there is a pragma inline on the function declaration.  The Ada compiler
needs to be smart enough to find the function body when it is needed.
Most Ada compilers do this (I don't know of any that don't, other
than older versions of GNAT).

: You might want to check out the following for online tutorials,
: references, a good Ada for C++ programmers paper, etc:

: http://lglwww.epfl.ch/Ada/

: Jon Anthony
: Organon Motives, Inc.
: 1 Williston Road, Suite 4
: Belmont, MA 02178

: 617.484.3383
: jsa@organon.com

-Tucker Taft   stt@inmet.com   http://www.inmet.com/~stt/
Intermetrics, Inc.  Cambridge, MA  USA

Re: Q: inlining

[Jon S Anthony]

In article <uvvd92zuczi.fsf@chaq.informatik.uni-tuebingen.de> Hannes Haug <Hannes.Haug@Student.Uni-Tuebingen.de> writes:

>     >> I'm new to Ada and have a question on inlining. Does inlining
>     >> work across compilation units ? In C I can
> 
>     Jon> It can, yes.  I'm not sure if this is working in GNAT 3.05
>     Jon> yet or not.
> 
> I know that this is possible. I just wanted to know if compilers actually do
> it. I finally found the option for GNAT: -gnatn.

Great.  There were Ada83 compilers that did this also.  Does anyone
know if ObjectAda/AdaMagic does this?


>     >> write the function definition in a header file. How can I do
>     >> this in Ada ?
> 
>     Jon> In a package specification.  In fact, this is virtually a
>     Jon> requirement to getting any "real" work done.
> 
> I know. This made me look at Ada. It's really ugly to put all the stuff
> in header files.

Oh.  I guess I misunderstood that you already knew this...

/Jon
-- 
Jon Anthony
Organon Motives, Inc.
1 Williston Road, Suite 4
Belmont, MA 02178

617.484.3383
jsa@organon.com

Re: Q: inlining

[Jon S Anthony]

In article <Dt4F36.1pG.0.-s@inmet.camb.inmet.com> stt@henning.camb.inmet.com (Tucker Taft) writes:

> Jon S Anthony (jsa@organon.com) wrote:
> : In article <uvv20jh9j7p.fsf@chaq.informatik.uni-tuebingen.de> 
> : Hannes Haug <Hannes.Haug@Student.Uni-Tuebingen.de> writes:
> 
> : > Hi,
> : > 
> : > I'm new to Ada and have a question on inlining. Does
> : > inlining work across compilation units ? In C I can
> 
> : It can, yes.  I'm not sure if this is working in GNAT 3.05 yet or not.
> 
> 
> : > write the function definition in a header file. How
> : > can I do this in Ada ?
> 
> : In a package specification.  In fact, this is virtually a requirement
> : to getting any "real" work done.
> 
> This answer could be confusing.  In C++, the function *definition*
> usually needs to be in the header file for inlining to work.  In Ada,
> only the function *declaration* is allowed in the package spec.  The

I see that I have _COMPLETELY_ misunderstood the question.  Duh!
I apologize to all...

/Jon
-- 
Jon Anthony
Organon Motives, Inc.
1 Williston Road, Suite 4
Belmont, MA 02178

617.484.3383
jsa@organon.com

Re: Q: inlining

[Tucker Taft]

Jon S Anthony (jsa@organon.com) wrote:
: In article <uvvd92zuczi.fsf@chaq.informatik.uni-tuebingen.de> 
: Hannes Haug <Hannes.Haug@Student.Uni-Tuebingen.de> writes:

: >     >> I'm new to Ada and have a question on inlining. Does inlining
: >     >> work across compilation units ? In C I can
: > 
: >     Jon> It can, yes.  I'm not sure if this is working in GNAT 3.05
: >     Jon> yet or not.
: > 
: > I know that this is possible. I just wanted to know if compilers actually do
: > it. I finally found the option for GNAT: -gnatn.

: Great.  There were Ada83 compilers that did this also.  Does anyone
: know if ObjectAda/AdaMagic does this?

Yes, ObjectAda/AdaMagic does inlining across compilation units.
It requires that the source file containing the body already be
"registered" in the program library when you compile a call to the 
subprogram.

: Jon Anthony
: Organon Motives, Inc.
: 1 Williston Road, Suite 4
: Belmont, MA 02178
: 617.484.3383
: jsa@organon.com

-Tucker Taft   stt@inmet.com   http://www.inmet.com/~stt/
Intermetrics, Inc.  Cambridge, MA  USA

Re: Q: inlining

[Hannes Haug]

>>>>> "Robert" == Robert Dewar <dewar@cs.nyu.edu> writes:

    Robert> [...]

    Robert> In fact there is all sorts of useful stuff in gnatinfo.txt,
    Robert> recommended reading for all GNAT users!

Gnat.info would be even better ;-) But there is a lot of stuff for which
you have to read the sources of the lib. How to mix Ada and C I/O ?  That's
implementation specific and should be in the docs. It would be nice if all
that goes beyond ARM would be documented. And a TODO would be fine. Will
support for machine intrinsic subprograms be available soon ? Perhaps I need
"ldstub" and "ta" on Sparc.

 -hannes

Re: Q: inlining

[Robert Dewar]

Tucker said

"Yes, ObjectAda/AdaMagic does inlining across compilation units.
It requires that the source file containing the body already be
"registered" in the program library when you compile a call to the
subprogram."

Tuck, you might want to emphasize that this registration has nothing
to do with compilation, so, as is the case with GNAT, Ada Magic allows
completely general inlining without having to worry about the order
in which things are compiled (at least I assume this is the case!)

Note that the registration in Ada Magic is essentially similar to placing
Source_File_Name pragmas in gnat.adc in GNAT, the only difference is that
in GNAT you can if you like avoid this step by using the default file
names (which is what most, but not all, GNAT users prefer to do).

Re: Q: inlining

[Robert Dewar]

Hannes asks

"Gnat.info would be even better ;-) But there is a lot of stuff for which
you have to read the sources of the lib. How to mix Ada and C I/O ?  That's
implementation specific and should be in the docs. It would be nice if all
that goes beyond ARM would be documented. And a TODO would be fine. Will
support for machine intrinsic subprograms be available soon ? Perhaps I need
"ldstub" and "ta" on Sparc."

Yes, there is a lot more documentatoin needed, for some idea, see the
SGI GNAT manuals, which are the model for the full GNAT documenation
currently being written.

Intrinsics will be available soon.
(meanwhile you can always write a C function that uses the gcc instrinsic
feature).

Re: Q: inlining

[Hannes Haug]

>>>>> "Robert" == Robert Dewar <dewar@cs.nyu.edu> writes:

    Robert> Hannes asks "Gnat.info would be even better ;-) But there
    Robert> is a lot of stuff for which you have to read the sources
    Robert> of the lib. How to mix Ada and C I/O ?  That's
    Robert> implementation specific and should be in the docs. It
    Robert> would be nice if all that goes beyond ARM would be
    Robert> documented. And a TODO would be fine. Will support for
    Robert> machine intrinsic subprograms be available soon ? Perhaps
    Robert> I need "ldstub" and "ta" on Sparc."

    Robert> Yes, there is a lot more documentatoin needed, for some
    Robert> idea, see the SGI GNAT manuals, which are the model for
    Robert> the full GNAT documenation currently being written.

Where can I find it? I looked at the Ada FAQs, GNAT docs and www.gnat.com.
But I found no hint. 

    Robert> Intrinsics will be available soon.  (meanwhile you can
    Robert> always write a C function that uses the gcc instrinsic
    Robert> feature).

I'm looking forward to it. But C isn't an option. Ldstub has to be
instrinsic. I'd like to use it for fast synchronization. It think
gnat uses mutexes or semaphores for synchronization. That's quite
slow. It would slow down consing. I have to care about this at the
low level. Otherwise I'd have to care about this stuff at the high
level, i.e. algorithms. :-(

Perhaps gnat-4 will have faster synchronization? Tasks seem to be
quite expensive, too. But I think my demands (not only in this area)
are a little bit unusual for Ada.

 -hannes