Statistics
| Branch: | Revision:

root / target-i386 / TODO @ 12e26b75

History | View | Annotate | Download (1.2 kB)

1
Correctness issues:
2

    
3
- some eflags manipulation incorrectly reset the bit 0x2.
4
- SVM: rework the implementation: simplify code, move most intercept
5
  tests as dynamic, correct segment access, verify exception safety,
6
  cpu save/restore, SMM save/restore. 
7
- x86_64: lcall/ljmp intel/amd differences ?
8
- better code fetch (different exception handling + CS.limit support)
9
- user/kernel PUSHL/POPL in helper.c
10
- add missing cpuid tests
11
- return UD exception if LOCK prefix incorrectly used
12
- test ldt limit < 7 ?
13
- fix some 16 bit sp push/pop overflow (pusha/popa, lcall lret)
14
- full support of segment limit/rights 
15
- full x87 exception support
16
- improve x87 bit exactness (use bochs code ?)
17
- DRx register support
18
- CR0.AC emulation
19
- SSE alignment checks
20
- fix SSE min/max with nans
21

    
22
Optimizations/Features:
23

    
24
- add SVM nested paging support
25
- add VMX support
26
- add AVX support
27
- add SSE5 support
28
- fxsave/fxrstor AMD extensions
29
- improve monitor/mwait support
30
- faster EFLAGS update: consider SZAP, C, O can be updated separately
31
  with a bit field in CC_OP and more state variables.
32
- evaluate x87 stack pointer statically
33
- find a way to avoid translating several time the same TB if CR0.TS
34
  is set or not.
35
- move kqemu support outside target-i386.