blob: 9eb89a94766202ad32962c8b1eaec1bb712a8a64 [file] [log] [blame]
sewardj034dc332004-11-15 13:30:26 +00001
2Last updated 15 Nov 04
3~~~~~~~~~~~~~~~~~~~~~~
4
5Critical (correctness)
6~~~~~~~~~~~~~~~~~~~~~~
sewardj034dc332004-11-15 13:30:26 +00007x86 isel: should free up all fp reg tags when calling a helper.
8Also check re saving of FP registers across calls (all caller-save),
9and also FPU control word.
10
11iropt: reconsider precise exceptions
12
sewardj14d3ddf2004-11-16 02:07:18 +000013x86 guest: look at FP accuracy
14
sewardj034dc332004-11-15 13:30:26 +000015
16Test
17~~~~
18Test adc/sbb carry dependency in memcheck
19
20
21Optimisation opportunities
22~~~~~~~~~~~~~~~~~~~~~~~~~~
23Improved isel for memcheck artefacts on x86 (generate neg ; sbbl)
24
25Assess tt_fast miss rates
26
27Better register allocation - spilling reduction
28
29iropt less conservative re precise exns
30
31improve stack-update pass
32
33proper profiling machinery
34
35do not CSE exprs :: Ity_Bit
36
37
38JIT speedups
39~~~~~~~~~~~~
40Ensure incremental flatness throughout
41
42Make sanity checker check flatness; also pass name of caller to
43sanity check so as to know after which pass it fails
44
45Profile again with memcheck
46
47change IRTemp to 16 bits?
48
49
50Integration
51~~~~~~~~~~~
52Get rid of sloppy-malloc
53
54Get rid of partial-loads-ok
55
56Optimisation after first instrumentation rather than 2nd ?
57
58disallow dirty helpers from writing SP/IP
59
60make instrumentation work at no optimisation
61
62write API doc, clarify IR semantics
63
64make IR utils module
65
66tt/tc simplification, also profiling
67
68generic stack pointer identification at startup?
69
70New memstack_k: old or new sp?
71
72Set host FPU mode before starting. Also set Dflag = 0
73
74skin_panic -> tool_panic
75