Created attachment 502110 [details] [diff] [review] Initial Patch. MQ rev 198 For the moment, this is one patch that is difficult to refactor into anything more much fine-grained. In is necessarily entangled. (I take it we cannot land patches that do not build?) It is also unclean, not-house style, and is polluted with thread-safe weakref implementation that should live in another patch. I'm adding it now to complete the AsymmGC patch stack. More docs and a clean-up to follow.
Created attachment 507350 [details] [diff] [review] Latest patch. MQ rev 228 This is a much cleaned-up version of the patch. It is close to being reviewable, mostly just documentation is required now. There is a possible refactoring into more fine-grained patches if judged necessary.
Attachment #502110 - Attachment is obsolete: true
Created attachment 522002 [details] [diff] [review] Rebased. TR rev 6090. Patch queue rev 261
Attachment #507350 - Attachment is obsolete: true
Who should review this one?
(In reply to comment #3) > Who should review this one? This is the big one. So Felix, Lars, Ruchi and Tommy. There's more docs to add first though.
(note from mtg): Does choosing only one victimThreadLocalAlloc suffice in GC::DoStealQuickListBudget? Not sure; don't see why it would. So we might need to generalize the loop structure.
Created attachment 525937 [details] [diff] [review] Latest. TR rev 6090. Patch queue rev 276
Attachment #522002 - Attachment is obsolete: true
Created attachment 525941 [details] [diff] [review] Latest. TR rev 6090. Patch queue rev 277
Attachment #525937 - Attachment is obsolete: true
Retargeting to Dolores.
Target Milestone: Q1 12 - Brannan → Q3 12 - Dolores
You need to log in before you can comment on or make changes to this bug.