[commit: ghc] ghc-lwc2: Added support for handling cases where SConts migrate capabilities. However, Sieve/ConcurrentListStealing.hs doesn't perform well when compared to Sieve/ConcurrentList.hs. Hmmm.. Perhaps it has to do with bound threads being stolen? (061e541)

Sivaramakrishnan Krishnamoorthy Chandrasekaran t-sichan at microsoft.com
Sun May 5 07:05:09 CEST 2013


Repository : http://darcs.haskell.org/ghc.git/

On branch  : ghc-lwc2

https://github.com/ghc/ghc/commit/061e5415e9d0089ebe0f28733876d858fe8e3dc4

>---------------------------------------------------------------

commit 061e5415e9d0089ebe0f28733876d858fe8e3dc4
Author: KC Sivaramakrishnan <chandras at cs.purdue.edu>
Date:   Sun May 5 00:49:37 2013 -0400

    Added support for handling cases where SConts migrate capabilities.
    However, Sieve/ConcurrentListStealing.hs doesn't perform well when
    compared to Sieve/ConcurrentList.hs. Hmmm.. Perhaps it has to do with
    bound threads being stolen?
    
    Avoided migrating tso's hosting user-level schedulers (ULS) since there
    is a possibility that naive implementation of user-level schedulers will
    get stuck if it were not to account for SCont's (tso's) migrating
    between capabilities. Migration of ULS is safe, however.

 libraries/base/LwConc/Substrate.hs                 |  1 +
 rts/Schedule.c                                     | 21 +++++++++--
 ...ConcurrentList.hs => ConcurrentListStealing.hs} | 42 ++++++++++++++--------
 3 files changed, 48 insertions(+), 16 deletions(-)


Diff suppressed because of size. To see it, use:

    git show 061e5415e9d0089ebe0f28733876d858fe8e3dc4



More information about the ghc-commits mailing list