Pointer ownership semantics, Attaching debug info and "unsigned" usage in LLVM -

i've started using llvm ir generation apis project. documentation , llc tool pretty helpful, haven't been able find answer following -

question-1 - pointer ownership semantics

all code creates llvm ir instructions using apis seems "new" instructions instead of creating them stack variables.

auto x = new alloca(...) vs. alloca(...) x; 

i wondering ownership semantics pointers created? have call delete on these instruction objects. code i've seen calls "delete engine;".

i guessing memory owned module object , when module destroyed, memory occupied these instructions destroyed. looking @ of code, seems these instruction objects created using "placement new" ... understanding correct?

question-2 - why doe llvm ir apis take "unsigned" data-type args. (e.g. unsigned addresspace 1 common argument. why not sized type uint32_t ?)

question-3 - how attach debug information ir instruction? pointers llvm apis attach debug info useful.

it's best split 3 questions, answer 3rd question did talk @ year's llvm developer conference on this:

http://llvm.org/devmtg/2014-10/#tutorial3

and can use documentation wrote after guideline:

http://llvm.org/docs/tutorial/langimpl8.html