RVM

Make lock allocation thread-model-agnostic.

Details

  • Patch Submitted:
    Yes
  • Number of attachments :
    2

Description

The current lock allocation strategy is M:N-specific. As well, it requires a complex balancing scheme that is disabled by default. This patch introduces a new lock allocation strategy that: 1) scales just as well (differences are in the noise), 2) has a simpler balancing strategy, and 3) is thread model agnostic.

  1. simple_lock_alloc.patch
    06/May/08 10:18 AM
    13 kB
    Filip Pizlo
  2. soo.txt
    14/May/08 4:52 PM
    0.5 kB
    Filip Pizlo

Activity

Hide
David Grove added a comment -

pushing to 3.0.1 as 3.0 is close enough we should be minimizing changes to avoid delaying release.

Show
David Grove added a comment - pushing to 3.0.1 as 3.0 is close enough we should be minimizing changes to avoid delaying release.
Hide
David Grove added a comment -

committed in r14965.

Show
David Grove added a comment - committed in r14965.
Hide
David Grove added a comment -
Show
David Grove added a comment - No obvious performance problems: For example see: http://jikesrvm.anu.edu.au/cattrack/results/habanero.anu.edu.au/perf/5238/performance_report

People

Vote (0)
Watch (0)

Dates

  • Created:
    Updated:
    Resolved: