Fix bug in Pmc2c which always generate WriteBarrier even :manual_wb w…
Few more optimisations
Mark shift_foo as manual_wb. It's safe because we are not chaning GCa…
Remove unused code
Made CallContext 'final' by poking directly into attributes. It gives…
… significant speedup
Rework CallContext slightly to avoid VTABLE calls and allocations
Check that all named are after positionals
Optimize creation of hash for named params
Factor out cleanup function for fill_params
Fix extracting named param
Finish FSM early if needed.
Implement dissecting of hashes
Don't immidiately restart after dissecting of array.
Rework dissecting of arrays to properly stop at the end
Collect named params and handle them. In very naive way for now
First step to handle named params
Use proper GET_ATTR macros
More fixes of CC.get_pointer
* Stop if needed after dissecting of array.
* Return NULL if ret wasn't updated.
* Use STATICSELF to avoid VTABLE call.
Down to 35 failures in t/op/calling.t
Make compiler happy
Fix copy-paste error in CC.shift_foo VTABLEs
Add NYI assert to enable coretest
First pass on fill_params to avoid indexed access to params
Fix CC.morph to actually reset state
Bit of removal of old code from CC
Add CCFC.shift_integer and .shift_number
Replace few accessors with new code
Start using CallContextFromC
Remove now wrong asserts.
Add (unused) new boxing functions
Move logic about parameter processing into CallContext.get_pointer
This will allow to use it between all accessors internally and
external (think nqp/rakudo) parties.
Remove unused state
First brain dump of fast PCC call processing
Remove unused accessors in fill_params
Copy CallContext to CallContextFromC.
Inherit CallContext from CallContextBase. Will rename it later