Ticket #13 (new proposed-project)

Opened 3 years ago

Last modified 1 year ago

Improve performance of numerical code

Reported by: paolo Assigned to: none
Priority: bad Keywords:
Cc: Topic: GHC
Difficulty: unknown Mentor: not-accepted

Description (Last modified by chak@cse.unsw.edu.au)

GHC's performance for double and float intensive code is not as good as it could be. Find out why and improve it. Requires GHC backend hacking. Must be very Haskell literate or have knowledge of code generators.

Interested Mentors

Interested Students

Change History

04/26/06 14:54:56 changed by paolo

  • description changed.

05/02/06 12:56:24 changed by wchogg

  • description changed.

03/18/07 19:20:24 changed by chak@cse.unsw.edu.au

  • description changed.

03/18/08 14:18:51 changed by nominolo

  • priority set to OK.

03/18/08 21:52:58 changed by dons

Needs close discussion with SimonM about related issues (e.g. SSE support in the native code gen). Could be integrated into the general work on the code gen appearing in the next few months.

If you're interested in doing code generator hacking, this is the ticket, and we can discuss more with Simon about what to do.

03/19/08 09:27:42 changed by simonmar

  • priority changed from OK to bad.

I think the timing isn't right for this project, as we have significant upheavals in the pipeline for the back end this summer, so coordination will be difficult. For a SoC project we need something with more clearly-defined goals, and less interaction with other ongoing development work.

An example of a good project would be "implement optimisation X in the backend" - but that won't be doable until we have the architectural changes in place. So I'm going to mark this as "bad" for now.