Difference between revisions of "Paris 2010 Meeting"

From Eigen
Jump to: navigation, search
(Participants)
 
(29 intermediate revisions by 2 users not shown)
Line 1: Line 1:
= Participants =
+
The meeting was held in Paris, from Friday, February 19, to Sunday, February 21, 2010.
  
Confirmed participants:
+
This was the first real meeting of the Eigen team (except of course for the huge [http://listengine.tuxfamily.org/lists.tuxfamily.org/eigen/2009/08/msg00006.html 2009 World Conference]).
  
* Thomas Capricelli, [http://www.sylphide-consulting.com Sylphide Consulting]
+
Lots of discussion and development happened to prepare the upcoming 3.0 release, as well as more long-term items such as the beginning of SMP support within Eigen.
* Gael Guennebaud, INRIA Bordeaux
+
* Hauke Heibel, Technische Universität München
+
* Benoit Jacob, University of Toronto
+
* Dodji Seketeli, Red Hat (Friday morning)
+
* Joel Falcou, LRI, Université Paris Sud (Friday morning)
+
  
= Location =
+
On Friday, we got the visit of Dodji Seketli from Red Hat, who introduced us to some GCC internals, and of Joel Falcou from Université Paris Sud who presented NT2.
  
The meeting will be held in Paris.
+
The slides for most of the talks are available below.
* '''Friday:''' INRIA, 23 avenue d'Italie, 75013 Paris, [http://maps.google.com/maps?f=q&source=s_q&hl=en&geocode=&q=23+avenue+d%27italie+paris&sll=37.0625,-95.677068&sspn=48.374125,68.642578&ie=UTF8&hq=&hnear=23+Avenue+d%27Italie,+75013+Paris,+Ile-de-France,+France&t=h&z=16 Google Map], nearest subway station: Place d'Italie (metro lines 5,6,7). '''ID required to enter the building.''' Say that you're going to the Eigen meeting. The INRIA suites are on the 5th and 6th floors, and their entrance is on the 5th floor. [http://www.inria.fr/rocquencourt/ur/comment-venir/livret-accueil-visiteurs-externes_.pdf More information (in French).]
+
* '''Saturday and Sunday:''' (Thomas, what's the name of the company? What floor, etc?), 2 rue des quatre fils, 75003 Paris, [http://maps.google.it/maps?f=q&source=s_q&hl=it&geocode=&q=142+rue+du+faubourg+saint+denis,+paris,+france&sll=41.442726,12.392578&sspn=27.540387,66.005859&ie=UTF8&hq=&hnear=142+Rue+du+Faubourg+Saint-Denis,+75010+Paris,+Ile-de-France,+Francia&z=16 Google Map].
+
  
= Schedule =
+
[[Image:Meeting-paris-2010-rescaled.jpeg |link=http://download.tuxfamily.org/eigen/meetings/paris2010/meeting-paris-2010.jpeg |frame |alt=Aron, Gael, Hauke, Mathieu, Thomas, Benoit |Aron, Gael, Hauke, Mathieu, Thomas, Benoit. [http://download.tuxfamily.org/eigen/meetings/paris2010/meeting-paris-2010.jpeg Zoom] ]]
  
From Friday, February 19, to Sunday, February 21, 2010.
 
  
Working hours: from 9:00 AM to 6:00 PM.
 
  
Friday should be focused on:
+
= Participants =
* planning for the 3.0 release
+
* Aron Ahmadia, King Abdullah University of Science and Technology
* discussion of difficult technical points
+
* Thomas Capricelli, [http://www.sylphide-consulting.com Sylphide Consulting]
* master classes (see below).
+
* Joel Falcou, LRI, Université Paris Sud (Friday morning)
 
+
* Mathieu Gautier, CEA (Sunday)
Saturday and Sunday should be split between continuing the above, and coding, as will be discussed on Friday.
+
* Gael Guennebaud, INRIA Bordeaux
 
+
* Hauke Heibel, Technische Universität München
Also, at some time Dodji will be visiting us, so we must leverage that time to learn as much as possible about G++, see below.
+
* Benoit Jacob, University of Toronto
 
+
* Dodji Seketeli, Red Hat (Friday morning)
= Topics for the presentations/discussions =
+
* Manuel Yguel, INRIA (Friday)
 
+
* Design of the Eigen's novel expression tree (Gael)
+
* Non linear optimization module (Thomas)
+
* Block Householder transformations (Benoit)
+
 
+
= Accomodation =
+
 
+
Suggested hotels:
+
* http://www.hotelplacedesalpes.com/frameuk.htm
+
* [http://www.ibishotel.com/gb/hotel-5543-ibis-paris-avenue-d-italie-13eme/index.shtml, Ibis Paris Avenue d'Italie]
+
 
+
= Master classes =
+
 
+
The goal is to take advantage of this meeting to share knowledge and have existing core developers teach technical points to other developers according to their interests. For that to work, it would be nice to know in advance if there is something specific that you would like to learn. Add that below.
+
 
+
= Dodji's visit =
+
  
* When? How long can you stay with us?
+
= Friday =
  
== Talk ==
+
'''Morning:'''
 +
* [http://download.tuxfamily.org/eigen/meetings/paris2010/Hauke-ShortIntroduction.pdf Presentation of Eigen], by Hauke
 +
* [http://download.tuxfamily.org/eigen/meetings/paris2010/Joel-nt2.pdf Presentation of NT2], by Joel
 +
* [http://download.tuxfamily.org/eigen/meetings/paris2010/Dodji-GCC.pdf Presentation of G++ internals], by Dodji
 +
'''Afternoon:'''
 +
* [http://download.tuxfamily.org/eigen/meetings/paris2010/Gael-NewInternals.pdf Presentation of Eigen's new class hierarchy (DenseBase...)], by Gael
 +
* Hacking
  
Dodji can do a talk for us on G++. What would you be the most interested in listening to? Add your ideas / comment below:
+
= Saturday =
* Go over a simple C++ example with templates, see how G++ handles it step by step.
+
* Same thing with a more involved example taken directly from Eigen.
+
* General presentation of how G++ handles things like: inheritance, template specialization, SFINAE... with the goal of understanding what makes code slower to compile, and what doesn't.
+
* ...add your ideas...
+
  
== Questions ==
+
'''Morning:'''
 +
* [http://download.tuxfamily.org/eigen/meetings/paris2010/Thomas-NonlinearOptimization.pdf Non-linear optimization in Eigen], by Thomas
 +
* Presentation of use cases of Eigen in supercomputing centres, by Aron
 +
* Hacking
  
What other questions would you like to ask Dodji?
+
'''Afternoon:'''
* When we want to understand why GCC misses an optimization on some code, what can we do? So far all we know how to do, is to generate assembly output... is there something else that we can generate and examine, something closer to the front-end?
+
* Proposal for SVD decomposition API and TODO, by Benoit
* General discussion of what's costly and what's cheap, regarding compilation times.
+
* Hacking
* ...add your ideas...
+
  
== Feedback ==
+
= Sunday =
  
We should also take this occasion to let Dodji know about our needs, our experiences...
+
(Mostly hacking)
* Prepare examples of code that GCC handles poorly (or is GCC 4.4 perfect already?)
+
* Prepare examples of workarounds / special care we had to use to help GCC generate good code
+
* Discuss the nest-xprs-by-value recent change, try to understand _why_ it is such an improvement with all compilers?
+
* ...add your ideas...
+
  
Dodji also expressed interest in listening to a talk on Eigen, so we should at least think about a quick explanation of what Eigen is doing from a C++ perspective, why it is doing it, etc...
+
Presentations:
 +
* Mathieu: about providing a general Lie Group framework within Eigen.
 +
* Thomas: InModelia software for non linear models: what we do with Eigen.

Latest revision as of 13:11, 7 May 2010

The meeting was held in Paris, from Friday, February 19, to Sunday, February 21, 2010.

This was the first real meeting of the Eigen team (except of course for the huge 2009 World Conference).

Lots of discussion and development happened to prepare the upcoming 3.0 release, as well as more long-term items such as the beginning of SMP support within Eigen.

On Friday, we got the visit of Dodji Seketli from Red Hat, who introduced us to some GCC internals, and of Joel Falcou from Université Paris Sud who presented NT2.

The slides for most of the talks are available below.

Aron, Gael, Hauke, Mathieu, Thomas, Benoit
Aron, Gael, Hauke, Mathieu, Thomas, Benoit. Zoom


Participants

  • Aron Ahmadia, King Abdullah University of Science and Technology
  • Thomas Capricelli, Sylphide Consulting
  • Joel Falcou, LRI, Université Paris Sud (Friday morning)
  • Mathieu Gautier, CEA (Sunday)
  • Gael Guennebaud, INRIA Bordeaux
  • Hauke Heibel, Technische Universität München
  • Benoit Jacob, University of Toronto
  • Dodji Seketeli, Red Hat (Friday morning)
  • Manuel Yguel, INRIA (Friday)

Friday

Morning:

Afternoon:

Saturday

Morning:

Afternoon:

  • Proposal for SVD decomposition API and TODO, by Benoit
  • Hacking

Sunday

(Mostly hacking)

Presentations:

  • Mathieu: about providing a general Lie Group framework within Eigen.
  • Thomas: InModelia software for non linear models: what we do with Eigen.