1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
|
This is a list of things that need to be worked on. It will hopefully
be complete soon.
Correctness
-----------
- Debugging info (check & fix as much as we can)
Optimizations (better compiled code)
------------------------------------
- Reduce register pressure in scheduler
- More strength reduction: multiply -> shift/add combos (Worth doing?)
- Add a value range propagation pass (for bounds elim & bitwidth reduction)
- Make dead store pass inter-block
- If there are a lot of MOVQ $0, ..., then load
0 into a register and use the register as the source instead.
- Allow arrays of length 1 (or longer, with all constant indexes?) to be SSAable.
- If strings are being passed around without being interpreted (ptr
and len fields being accessed) pass them in xmm registers?
Same for interfaces?
- Non-constant rotate detection.
- Do 0 <= x && x < n with one unsigned compare
- nil-check removal in indexed load/store case:
lea (%rdx,%rax,1),%rcx
test %al,(%rcx) // nil check
mov (%rdx,%rax,1),%cl // load to same address
- any pointer generated by unsafe arithmetic must be non-nil?
(Of course that may not be true in general, but it is for all uses
in the runtime, and we can play games with unsafe.)
Optimizations (better compiler)
-------------------------------
- OpStore uses 3 args. Increase the size of Value.argstorage to 3?
- Use a constant cache for OpConstNil, OpConstInterface, OpConstSlice, maybe OpConstString
- Handle signed division overflow and sign extension earlier
Regalloc
--------
- Make less arch-dependent
- Handle 2-address instructions
- Make liveness analysis non-quadratic
Future/other
------------
- Start another architecture (arm?)
- 64-bit ops on 32-bit machines
- Investigate type equality. During SSA generation, should we use n.Type or (say) TypeBool?
- Should we get rid of named types in favor of underlying types during SSA generation?
- Should we introduce a new type equality routine that is less strict than the frontend's?
- Infrastructure for enabling/disabling/configuring passes
|
