X10

Optimize atomic regions that are simple instance field updates

Details

  • Number of attachments :
    0

Description

Given a simple x10 atomic:

atomic o.f = v;

we might be able to implement this with an AtomicFieldUpdater (Java) or an atomic memory operation (C++) instead of a full blown X10 atomic.

Activity

Ascending order - Click to sort in descending order
Hide
Permalink
David Grove added a comment -

Unassign issues assigned to Nate.

Show
David Grove added a comment - Unassign issues assigned to Nate.
Hide
Permalink
Josh Milthorpe added a comment -

Given that in the general case atomic is currently implemented as "stop the world" – which is antithetical to multithreaded performance – the compiler should generate a warning for any non-optimized section.

I would also vote +1 on priority and -998 on fix version

Show
Josh Milthorpe added a comment - Given that in the general case atomic is currently implemented as "stop the world" – which is antithetical to multithreaded performance – the compiler should generate a warning for any non-optimized section. I would also vote +1 on priority and -998 on fix version
Hide
Permalink
Igor Peshansky added a comment -

It would have to also notify the runtime to process any conditional atomics waiting on the value (see x10.lang.Runtime.Latch.release()).

Show
Igor Peshansky added a comment - It would have to also notify the runtime to process any conditional atomics waiting on the value (see x10.lang.Runtime.Latch.release() ).
Hide
Permalink
Vijay Saraswat added a comment -

+1

Show
Vijay Saraswat added a comment - +1

People

  • Assignee:
    Unassigned
    Reporter:
    David Grove
Vote (2)
Watch (1)

Dates

  • Created:
    Updated: