RE: Maintainer of tree-inline.c and expr.c

RE: Maintainer of tree-inline.c and expr.c

[S. Bosscher]

Andrew Haley wrote:
> Be aware that we are working on an all-new inliner on the tree-ssa
> branch.

Oh? Can you elaborate?

Gr.
Steven

RE: Maintainer of tree-inline.c and expr.c

[Andrew Haley]

S. Bosscher writes:
 > Andrew Haley wrote:
 > > Be aware that we are working on an all-new inliner on the tree-ssa
 > > branch.
 > 
 > Oh? Can you elaborate?

It's at /local/aph/tree-ssa-20020619-branch/gcc/gcc/tree-inline.c.

Okay, so it's not all new.  It's a hybrid of the old C++ tree inliner,
the Java tree inliner, and new work for GIMPLE.

Andrew.

RE: Maintainer of tree-inline.c and expr.c

[Andrew Haley]

Andrew Haley writes:
 > 
 > It's at /local/aph/tree-ssa-20020619-branch/gcc/gcc/tree-inline.c.

D'oh.  It's at gcc/gcc/tree-inline.c on the tree-ssa-20020619-branch.

Andrew.

Re: Maintainer of tree-inline.c and expr.c

[Andrew Haley]

Carlo Wood writes:
 > On Mon, Sep 22, 2003 at 09:41:24AM +0100, Andrew Haley wrote:
 > > Be aware that we are working on an all-new inliner on the tree-ssa
 > > branch.  Any work that you do on the old inliner will only be of
 > > transient interest.
 > 
 > I really want this to be fixed in gcc 3.4.
 > Can you have a look at the RFC that I posted yesterday?
 > 
 > Basically, the first instruction in the inlined function
 > will have correct line info already anyway.  What is there
 > to be expected in between the start of the (inlined) call
 > and that first instruction except code that is responsible
 > for initializing the function parameters?  Maybe I should
 > take C++ class member and/or constructors into account?
 > 
 > I am still not sure if I am not forgetting to take something
 > into account.

What you say looks reasonable.  But as far as I am aware we have no
regression tests for line numbers in debugging information, so you'll
have to be careful.  Check C++ constructors, yes.  Also, the gdb
list may be interested.

And whatever you do, please make sure you also submit a patch for
tree-ssa.

Andrew.

Re: Maintainer of tree-inline.c and expr.c

[Carlo Wood]

On Mon, Sep 22, 2003 at 09:41:24AM +0100, Andrew Haley wrote:
> Be aware that we are working on an all-new inliner on the tree-ssa
> branch.  Any work that you do on the old inliner will only be of
> transient interest.

I really want this to be fixed in gcc 3.4.
Can you have a look at the RFC that I posted yesterday?

Basically, the first instruction in the inlined function
will have correct line info already anyway.  What is there
to be expected in between the start of the (inlined) call
and that first instruction except code that is responsible
for initializing the function parameters?  Maybe I should
take C++ class member and/or constructors into account?

I am still not sure if I am not forgetting to take something
into account.

-- 
Carlo Wood <carlo@alinoe.com>

Re: Maintainer of tree-inline.c and expr.c

[Andrew Haley]

Carlo Wood writes:
 > On Mon, Sep 22, 2003 at 01:10:24AM +0200, Gabriel Dos Reis wrote:
 > > If it is "all of us", then it is all of us.  At least all of us who
 > > care about inlining :-/  
 > 
 > Ok, I'll write up a description of the problem, my
 > proposal - the test cases that I think will be needed
 > and last but not least the questions I have in order to
 > be able to continue.  And post that here.

Be aware that we are working on an all-new inliner on the tree-ssa
branch.  Any work that you do on the old inliner will only be of
transient interest.

Andrew,

Re: Maintainer of tree-inline.c and expr.c

[Carlo Wood]

On Mon, Sep 22, 2003 at 01:10:24AM +0200, Gabriel Dos Reis wrote:
> If it is "all of us", then it is all of us.  At least all of us who
> care about inlining :-/  

Ok, I'll write up a description of the problem, my
proposal - the test cases that I think will be needed
and last but not least the questions I have in order to
be able to continue.  And post that here.

-- 
Carlo Wood <carlo@alinoe.com>

Re: Maintainer of tree-inline.c and expr.c

[Gabriel Dos Reis]

Carlo Wood <carlo@alinoe.com> writes:

| On Sun, Sep 21, 2003 at 10:53:17PM +0100, Andrew Haley wrote:
| > Yeah, and I'm responsible for the Java part.  So the answer to "who is
| > in charge?" is "all of us!"

[...]

| I can post my proposal here instead, but I'd really like
| the person who will decide whether or not to accept the
| patch to read it (and give feedback when needed)!
| Without feedback, I'll never know if it was even read at
| all :/

If it is "all of us", then it is all of us.  At least all of us who
care about inlining :-/  

(No, sadely I did not contribute to the tree inliner).

-- Gaby

Re: Maintainer of tree-inline.c and expr.c

[Carlo Wood]

On Sun, Sep 21, 2003 at 10:53:17PM +0100, Andrew Haley wrote:
> Yeah, and I'm responsible for the Java part.  So the answer to "who is
> in charge?" is "all of us!"

Ok, sorry for the confusion.  The reason I ask is because
I'd like to discuss certain implementation details of a
patch that I intend to write for gcc up front: I believe
that it is better to first agree on what a patch should
do and _then_ write and submit it - instead of just writing
a patch, submitting it and hoping someone will take the
time and effort to convince themselfs it is what they
want.

About 1.5 years ago I reported a bug that never was
fixed.  I want to fix it myself now, with as release
target 3.4.  In order to accomplish that, I persue a quick
acceptance of the patch; I think the best way for that is
by discussing up front the changes that I intend.

I can post my proposal here instead, but I'd really like
the person who will decide whether or not to accept the
patch to read it (and give feedback when needed)!
Without feedback, I'll never know if it was even read at
all :/

-- 
Carlo Wood <carlo@alinoe.com>

Re: Maintainer of tree-inline.c and expr.c

[Andrew Haley]

Gabriel Dos Reis writes:
 > Andreas Jaeger <aj@suse.de> writes:
 > 
 > | Carlo Wood <carlo@alinoe.com> writes:
 > | 
 > | > Who is in charge of tree-inline.c and expr.c ?
 > | 
 > | If nobody is mentioned in the top-level MAINTAINERS file: Everybody -
 > | or nobody ;-)
 > | 
 > | tree-inline was initially written by Alexandre.
 > 
 > tree-inline.c was indeed contributed by Alexandre Oliva, but I believe
 > the core was already contributed by Mark Mitchell as part of
 > cp/optimize.c, and Alexandre pulled out most bits from there to make
 > them available for cc1 and latter for other front-ends. 

Yeah, and I'm responsible for the Java part.  So the answer to "who is
in charge?" is "all of us!"

Andrew.

Re: Maintainer of tree-inline.c and expr.c

[Gabriel Dos Reis]

Andreas Jaeger <aj@suse.de> writes:

| Carlo Wood <carlo@alinoe.com> writes:
| 
| > Who is in charge of tree-inline.c and expr.c ?
| 
| If nobody is mentioned in the top-level MAINTAINERS file: Everybody -
| or nobody ;-)
| 
| tree-inline was initially written by Alexandre.

tree-inline.c was indeed contributed by Alexandre Oliva, but I believe
the core was already contributed by Mark Mitchell as part of
cp/optimize.c, and Alexandre pulled out most bits from there to make
them available for cc1 and latter for other front-ends. 

-- Gaby

Re: Maintainer of tree-inline.c and expr.c

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 585 bytes --]

Carlo Wood <carlo@alinoe.com> writes:

> Who is in charge of tree-inline.c and expr.c ?

If nobody is mentioned in the top-level MAINTAINERS file: Everybody -
or nobody ;-)

tree-inline was initially written by Alexandre.

Why are you asking?

Andreas

P.S. Please fix your Reply-To, line, it come in with <carlo> - and got
completed by our mailserver to carlo@suse.de :-(
-- 
 Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj
  SuSE Linux AG, Deutschherrnstr. 15-19, 90429 Nürnberg, Germany
   GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Maintainer of tree-inline.c and expr.c

[Carlo Wood]

Who is in charge of tree-inline.c and expr.c ?

-- 
Carlo Wood <carlo@alinoe.com>