Design doc: Minor GC for ExternalPointerTable

[Andy Wingo]

Hello all,

Please find below a link to a short design doc for adding a young
generation to ExternalPointerTable.

This work is intended to fix
https://bugs.chromium.org/p/v8/issues/detail?id=13640#c31.

https://docs.google.com/document/d/1Pvr4RbG_ZiaYZ-VkE22bSyCOiOntpbSBNJIhv45JTKk/edit

Feedback, comments, and corrections are very welcome. I have some
WIP patches not yet posted to add the space and arrange to allocate into
it, but have not yet done the actual promotion part yet. Happy to
change tack if appropriate.

Happy holidays,

Andy

[Jakob Kummerow]

Hi Andy,

Thanks for working on this!

I have no strong opinion on the general approach. Making the EPT trigger major GCs after a while (i.e. when the EPT has grown significantly without a major GC having happened due to other triggers) might also be acceptable, and might be (much) simpler.

Speaking of simplification, if you go with the suggested route of adding a Minor GC for the EPT, it might be fine to ignore the Scavenger and only support the MinorMS (which, IIUC, is intended to become the only supported young-generation GC at some point).

+Michael for a GC perspective.

--