opt compiler parallelization

Details

  • Type: New Feature New Feature
  • Status: Open Open
  • Priority: Minor Minor
  • Resolution: Unresolved
  • Affects Version/s: None
  • Fix Version/s: 3.2
  • Component/s: Compiler: Optimizing
  • Labels:
    None
  • Environment:
    multicore host platforms

Description

An investigation of the parallelization opportunities in the optimizing compiler of Jikes RVM. This is a Google Summer of Code project suggestion.

Activity

Ascending order - Click to sort in descending order
Hide
Permalink
Erik Brangs added a comment -

AFAIK the corresponding GSoC project has been accepted. Is it already decided where the code is going to be located during development?

Show
Erik Brangs added a comment - AFAIK the corresponding GSoC project has been accepted. Is it already decided where the code is going to be located during development?
Hide
Permalink
Michael Taylor added a comment -

Here Is my GSOC blog covering the progress of this issue http://parallelcompiler.blogspot.com/

Show
Michael Taylor added a comment - Here Is my GSOC blog covering the progress of this issue http://parallelcompiler.blogspot.com/
Hide
Permalink
Jeremy Singer added a comment -

Hi Dave,
we spoke about parallel opt compilation at OOPSLA/SPLASH recently. It seems that having multiple opt compilation threads at runtime is a good idea in principle (as Kulkarni's OOPSLA '11 paper demonstrates for HotSpot - http://www.ittc.ku.edu/~kulkarni/CARS/papers/oopsla11.pdf )
Mike Taylor (GSoC student) played around with the infrastructure. He got a prototype of multiple compilation threads working, but we didn't have time to update the analytical cost/benefit model in a way that would give the multiple compiler threads lots of work to do. I guess a simple tweak to CompilerDNA might help?
I might dust down Mike's code and see whether it is feasible to use it in anger.
Cheers,
Jeremy

Show
Jeremy Singer added a comment - Hi Dave, we spoke about parallel opt compilation at OOPSLA/SPLASH recently. It seems that having multiple opt compilation threads at runtime is a good idea in principle (as Kulkarni's OOPSLA '11 paper demonstrates for HotSpot - http://www.ittc.ku.edu/~kulkarni/CARS/papers/oopsla11.pdf ) Mike Taylor (GSoC student) played around with the infrastructure. He got a prototype of multiple compilation threads working, but we didn't have time to update the analytical cost/benefit model in a way that would give the multiple compiler threads lots of work to do. I guess a simple tweak to CompilerDNA might help? I might dust down Mike's code and see whether it is feasible to use it in anger. Cheers, Jeremy
Hide
Permalink
David Grove added a comment -

Thanks Jeremy, that would be great.

Show
David Grove added a comment - Thanks Jeremy, that would be great.

People

  • Assignee:
    Unassigned
    Reporter:
    Jeremy Singer
Vote (1)
Watch (0)

Dates

  • Created:
    Updated: