Wouldn't turning of memory function inlining of the normal core have the
same effect, which i recommended some time ago?

1return LoadMw(core.cseip+=2); 

Maybe =>

1return LoadMw((core.cseip++)++); 

non-inlining:
It would make it shorter, but would add a call right ?

Given the amount of code bloat the inlining introduces, i'd suppose it's faster
with the call than having the cache filled up with same code blocks.

+=2/+=4 breaks things, with it adding before use.

>return LoadMw((core.cseip++)++);

Complains about non-lvalue in increment.