Changes between Version 44 and Version 45 of WikiStart

Show
Ignore:
Timestamp:
04/13/10 04:44:10 (4 years ago)
Author:
benl (IP: 129.94.242.47)
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • WikiStart

    v44 v45  
    99This section records what the DDC developers are actively working on. If you're hacking on something DDC related then by all means add yourself here. Records are dropped off if they haven't been updated for more than a month. 
    1010 
    11 ==== Ben (benl23) -- updated 10th March 2010  ==== 
    12  * I'm currently refactoring the representation of types (in Type.Exp) to use a Data.Map to hold type constraints instead of a flat list. Along the way I'm redoing some of the more naive parts of the type inferencer, which should result in a major performance improvement. Once the representation of types has stabilised I'll be able to see about fixing #2 (checking the types of instance functions against their definitions) which is a major cause of confusing panic messages. 
    13  * As part of that, I'm redoing the code that manages interface files. The previous setup was very hacky, and compiling a module meant totally loading all interface files for all imported modules, even in only a tiny proportion of the information was used. The new module code should be significantly cleaner and faster. 
     11==== Ben (benl23) -- updated 13th April 2010  ==== 
     12 * Over the last couple of weeks I've been doing thesis corrections, so not as much coding of late. 
     13 * When coding I'm mostly refactoring and cleaning up the code base, getting it more stable before moving on to bug fixing. 
     14 * Refactoring the representation of types to use a Data.Map to hold type constraints instead of a flat list. 
     15 * Refactoring the top-level driver to use a better structure to hold top-level-things instead of a flat list. 
     16 * Refactoring the representation of type and kind constructors to match last year's APLAS paper. 
     17 * Also implementing a new module interface file format. We need this to propagate constraints on top-level regions between modules, and also to move towards being able to do cross-module inlining. 
    1418 
    1519==== Erik de Castro Lopo (m3ga) -- updated 6th Apr 2010  ====