[GHC] #13164: idle time full GCs (idle cpu usage)

[GHC]

#13164: idle time full GCs (idle cpu usage)
-------------------------------------+-------------------------------------
        Reporter:  lspitzner         |                Owner:
            Type:  feature request   |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Runtime System    |              Version:  8.0.1
      Resolution:                    |             Keywords:  idle GC
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by lspitzner):

 Firstly: I said "I just noticed that this means", i.e. this was a pure
 deduction, not an observation.

 Apart from that, I never meant to suggest making -I0 the default, and I
 think the two goals (deadlock detection and latency reduction) do align
 here. So between options a) status quo b) default -I0 c) something more
 clever that both does deadlock detection and uses, lets say.. 3% idle cpu
 time, I'd choose c) over a) over b).

 I am not sure how the budget would be enforced, but it sounds like a good
 idea, especially if that is easier to implement than the heuristics I
 mentioned above.

 Another thought: If there was an upper bound on the interval between full
 GCs (idle or not) you'd get reliable deadlock detection even with "forever
 threaddelay" cases.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13164#comment:7>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler