Re: [stree] Committed: initial stree infrastructure, and stree version of enumerators

Re: [stree] Committed: initial stree infrastructure, and stree version of enumerators

[Matt Austern]

[-- Attachment #1: Type: text/plain, Size: 629 bytes --]

On Mar 9, 2004, at 12:19 PM, Mark Mitchell wrote:

>
>>
>> I'm definitely sensitive to the problem of gcc "improvements" that 
>> make the compiler even more complicated and harder to understand, and 
>> I certainly agree that there are algorithmic improvements to be had 
>> as well.  But I also think the stree approach is promising enough 
>> that we ought at least to explore it.  We need to make major 
>> speedups, not just avoid further slowdowns.
>
> Agreed.  I look forward to the design document!

Here's the first draft!  Sorry it took a little longer than I'd 
expected.

As usual: comments are welcome.

			--Matt



[-- Attachment #2: stree-notes.pdf --]
[-- Type: application/pdf, Size: 61967 bytes --]

Re: [stree] Committed: initial stree infrastructure, and stree version of enumerators

[Mark Mitchell]

Matt Austern wrote:

> On Mar 9, 2004, at 12:19 PM, Mark Mitchell wrote:
>
>>
>>>
>>> I'm definitely sensitive to the problem of gcc "improvements" that 
>>> make the compiler even more complicated and harder to understand, 
>>> and I certainly agree that there are algorithmic improvements to be 
>>> had as well.  But I also think the stree approach is promising 
>>> enough that we ought at least to explore it.  We need to make major 
>>> speedups, not just avoid further slowdowns.
>>
>>
>> Agreed.  I look forward to the design document!
>
>
> Here's the first draft!  Sorry it took a little longer than I'd expected.

Thanks for writing this up.  This is a great example of what such a 
document should look like.  This document makes it possible to discuss 
your idea without having to wade through code.

(Gerald, would you care to put this somewhere on our web site as an 
example of how to write an informal design?)

I'm still digesting the idea and deciding what my opinion is.

Thanks,

-- 
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com

Re: [stree] Committed: initial stree infrastructure, and stree version of enumerators

[Matt Austern]

On Mar 16, 2004, at 7:39 PM, Mark Mitchell wrote:

> Matt Austern wrote:
>
>> On Mar 9, 2004, at 12:19 PM, Mark Mitchell wrote:
>>
>>>
>>>>
>>>> I'm definitely sensitive to the problem of gcc "improvements" that 
>>>> make the compiler even more complicated and harder to understand, 
>>>> and I certainly agree that there are algorithmic improvements to be 
>>>> had as well.  But I also think the stree approach is promising 
>>>> enough that we ought at least to explore it.  We need to make major 
>>>> speedups, not just avoid further slowdowns.
>>>
>>>
>>> Agreed.  I look forward to the design document!
>>
>>
>> Here's the first draft!  Sorry it took a little longer than I'd 
>> expected.
>
> Thanks for writing this up.  This is a great example of what such a 
> document should look like.  This document makes it possible to discuss 
> your idea without having to wade through code.

Thank you!

> I'm still digesting the idea and deciding what my opinion is.

Can you think of any information (measurements, experiments, whatever) 
that would help you form an opinion?

Obviously there's some self-interest here: all of us would like to have 
some more assurance that this is the right way to go before we do lots 
more work on it.

			--Matt

Re: [stree] Committed: initial stree infrastructure, and stree version of enumerators

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 1010 bytes --]

Matt Austern <austern@apple.com> writes:

> On Mar 9, 2004, at 12:19 PM, Mark Mitchell wrote:
>
>>
>>>
>>> I'm definitely sensitive to the problem of gcc "improvements" that
>>> make the compiler even more complicated and harder to understand,
>>> and I certainly agree that there are algorithmic improvements to be
>>> had as well.  But I also think the stree approach is promising
>>> enough that we ought at least to explore it.  We need to make major
>>> speedups, not just avoid further slowdowns.
>>
>> Agreed.  I look forward to the design document!
>
> Here's the first draft!  Sorry it took a little longer than I'd
> expected.
>
> As usual: comments are welcome.

Can you convert this into a webpage and put it on gcc.gnu.org together
as an explanation what the branch is about?

Andreas
-- 
 Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj
  SuSE Linux AG, Maxfeldstr. 5, 90409 Nürnberg, Germany
   GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: [stree] Committed: initial stree infrastructure, and stree version of enumerators

[Gerald Pfeifer]

On Wed, 17 Mar 2004, Andreas Jaeger wrote:
> Can you convert this into a webpage and put it on gcc.gnu.org together
> as an explanation what the branch is about?

Yup, I'm all for that!  It doesn't need to be very elaborate in terms
of markup.  Heck, even pdf2txt or similar would already be useful.

You could put this somewhere under /projects, for example.

Gerald
-- 
Gerald Pfeifer (Jerry)   gerald@pfeifer.com   http://www.pfeifer.com/gerald/

Re: [stree] Committed: initial stree infrastructure, and stree version of enumerators

[Matt Austern]

On Mar 24, 2004, at 10:40 AM, Gerald Pfeifer wrote:

> On Wed, 17 Mar 2004, Andreas Jaeger wrote:
>> Can you convert this into a webpage and put it on gcc.gnu.org together
>> as an explanation what the branch is about?
>
> Yup, I'm all for that!  It doesn't need to be very elaborate in terms
> of markup.  Heck, even pdf2txt or similar would already be useful.
>
> You could put this somewhere under /projects, for example.

Done!

http://gcc.gnu.org/projects/strees/index.html

			--Matt

Re: [stree] Committed: initial stree infrastructure, and stree version of enumerators

[Ranjit Mathew]

Matt Austern wrote:
> Here's the first draft!  Sorry it took a little longer than I'd 
> expected.

It would have been much better if you'd used
a plain text file - among other things, it makes
it easier to quote from it while replying.

Anyways:

1. "We can add incrementally add cases where we are
   able to generate strees."

Extra "add".

2. "GCC 3.4 does not generated DWARF debug info for function
   declarations, and does not generate debug info for used
   types unless -fno-eliminate-unused-debugtypes is specified."

I guess you mean "does not generate debug info for *unused*
types".

Of course, these matter only if you publish this, say,
on the GCC web pages.

Ranjit.

-- 
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/

Re: [stree] Committed: initial stree infrastructure, and stree version of enumerators

[Paolo Carlini]

Ranjit Mathew wrote:

>Matt Austern wrote:
>  
>
>>Here's the first draft!  Sorry it took a little longer than I'd 
>>expected.
>>    
>>
>It would have been much better if you'd used
>a plain text file - among other things, it makes
>it easier to quote from it while replying.
>  
>
ps2ascii, part of the GNU Ghostscript package, works well.

Paolo.