Differences between revisions 7 and 9 (spanning 2 versions)
Revision 7 as of 2015-02-08 15:11:51
Size: 1386
Editor: vdelecroix
Comment:
Revision 9 as of 2015-02-08 17:16:53
Size: 2147
Editor: vdelecroix
Comment:
Deletions are marked like this. Additions are marked like this.
Line 20: Line 20:
In Sage there are various places where we have several possibilities to execute a task (e.g. calling pari or gap or a native sage function). It would be interesting to have a way of choosing the default parameters by performing benchmarkings at build time. That would also allow to check coherency between various implementations. In Sage there are various places where we can choose between several algorithms or underlying softwares to solve a problem. In Sage, this is often related to the presence of the keyword ''algorithm'' or ''method'' in methods and functions. The aim of this task is to build a generic dispatcher that would choose depending on the parameters the fastest solution available. The solution must be very light and not affect performance. The dispatch threshold must be static and decided at build time. This generic dispatcher could also be used to check coherency between the various implementations.

Note that it is different from what is called multimethods where the dispatch depends only on the input type. Here we consider a dispatcher that might also depend on the input values.

 * competence: good knowledge of Python and notions of Cython and C
 * (draft) timeline:
  1. write a simple prototype of generic dispatcher
  2. identify Sage functions/methods that could benefit from the dispatcher and test it
  3. release a first candidate for the dispatcher
  4. Sage integration

Project ideas for GSoC 2015 for Sage

Notebook mode with execution from top to bottom

In the current notebook (both Sage notebook and IPython notebook) the cells can be executed in any order. From a teaching point of view this is terrible and from a scientific point of view this leads to highly non reproducible computations.

The purpose of this task is to have a new mode for the IPython notebook that would force computations from top to bottom. If a cell is executed, then the state in which it is executed must be the one you obtain by executing all the cells above it. In order to make it work, one needs to save the Python state after each cell.

note: this is note completely Sage oriented... (see with IPython people)

Native GUI

Adapt Spyder to work with Sage.

See also this thread on sage-devel.

generic dispatcher

In Sage there are various places where we can choose between several algorithms or underlying softwares to solve a problem. In Sage, this is often related to the presence of the keyword algorithm or method in methods and functions. The aim of this task is to build a generic dispatcher that would choose depending on the parameters the fastest solution available. The solution must be very light and not affect performance. The dispatch threshold must be static and decided at build time. This generic dispatcher could also be used to check coherency between the various implementations.

Note that it is different from what is called multimethods where the dispatch depends only on the input type. Here we consider a dispatcher that might also depend on the input values.

  • competence: good knowledge of Python and notions of Cython and C
  • (draft) timeline:
    1. write a simple prototype of generic dispatcher
    2. identify Sage functions/methods that could benefit from the dispatcher and test it
    3. release a first candidate for the dispatcher
    4. Sage integration

Android App

iOS App

SageMathCloud

T.B.A.

GSoC/2015 (last edited 2015-03-16 20:44:44 by schilly)