garbage collectable contexts

[whiteknight]

Okay, talking to Allison++ on IRC has given me a good idea about how to get this all done. Here's the general roadmap for this task. Note that not all of these will need to be completed to close this particular ticket, so once we get the first two items done we can open tickets to support the others:

1) Create a Context PMC that mirrors the functionality of the current Parrot_Context structure 2) Replace all uses of Interp_Context and Parrot_Context structures with the new Context PMC 3) Combine the RetContinuation PMC into the Context PMC 4) Combine the Call_State structure, which is used extensively in the /src/call/* and the /src/jit/* subsystems, into the Context PMC. 5) Combine the CallSignature PMC into the Context PMC. Optimize the calling paths heavily to reduce the amount of data shuffling that needs to be done to make a call.

Lots of work to do, but again only #1 and #2 really need to be done to complete this ticket. The remainder are wishlist items for the distant future (Possibly post-2.0)

[coke]

This was originally mentioned in " http://rt.perl.org/rt3/Ticket/Display.html?id=60564", which is superseded by this ticket.

[whiteknight]

an example file for how a new Context abstraction API could look. Does not build and is not pretty

[whiteknight]

I've attached an example file here for the start of what could become a context API. Here is a rough path to complete this ticket:

• Implement an API (like the one attached, but prettier) that does everything we need to do with contexts without having to poke into the guts of the data structure directly
• Begin moving context accesses behind this abstraction layer
• Once contexts are completely abstracted behind the API, we can convert them to a new storage format more easily.

[bacek]

context_pmc3 branch with GCable Contexts merged into trunk.