-
Notifications
You must be signed in to change notification settings - Fork 9
/
TODO
17 lines (17 loc) · 1.09 KB
/
TODO
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
- {A,B} = {[],[]} does not compile
- when binary is matched it is first copied to a new register
- pieces of matched binary are cleared out prematurely
- on binary allocation the ultimate parent should be traced
- examine whether it is worthwhile to have apr_pollset_t allocated once; also apr_pollcb_t
- binary_to_term fails on nested oddball binaries
- unused registers can be sorted on clear_dead; or may be not; check compilation times etc
- one register may be shaved in code for make_fun; in some cases
- register touched by a local function can be tracked; probably not worthwhile
- implement label-Op-label-Op optimization; happens many a time; only terminal ops
- mp_set_int takes long argument and make assumption that sizeof(long) > sizeof(int)
- compile APR without threads; no APR_HAS_THREADS
- exceptions leak from guards
- update optimization taking into account possibility of {source,_,_}
- named tuple update not implemented, e.g. {A.car#model = ""}
- named tuple access, index, etc does not check that argument is of proper type (no is_record())
- named tuple patterns not implemented (tough one)