for review (M): 6863023: need non-perm oops in code cache for JSR 292
Thomas.Rodriguez at Sun.COM
Thu Aug 27 12:34:08 PDT 2009
On Aug 27, 2009, at 12:12 PM, John Rose wrote:
> On Aug 27, 2009, at 12:00 PM, Tom Rodriguez wrote:
>> Hmm. Having to scan the stacks an extra time seems bad,
>> particularly since it's only needed during a scavenge because we're
>> treating some nmethods as strong roots. Can't you just mark the
>> nmethods during the stack walk and clear it during the gc_epilogue?
> Yes, that's what I was intending. In the epilogue either a stack
> walk or a code-cache walk will suffice to clear the bits, and I
> think there's already a code-cache walk in all the relevant cases.
Oh, I thought you meant an additional walk in gc_prologue.
gc_epiloque already includes a code cache walk so in the end it should
require no extra walks, just some extra marking.
> -- John
More information about the hotspot-compiler-dev