-
Notifications
You must be signed in to change notification settings - Fork 3
/
TODO
39 lines (25 loc) · 1.16 KB
/
TODO
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
TODO
DONE Native while loops
DONE Integrated prep-oce-eqs to preprocessor
DONE verified bin-search
DONE: Heap and Heapmap
Instantiate higher-order (template) parameters in preprocessor
(could also help for locale instantiations, e.g., in the sorting examples)
Elaborate memory model: No confusion for better verifiability!
Refinement framework: more flexible matching,
e.g., only on combinator structure/loop structure.
Use pw inside.
Splitting on refinement goals: Split conc side, then simplify,
and only split abs side when necessary.
more verified algos
Ideas:
* Heaps, Dijkstra [we have some performance data there]
* gratchk: Exception compilation, memmap of certificate. Actually fast enough.
* drat-checker: Quite complex, but would be cool ... own paper?
* Push-Relabel: Performance data available, compare with ProgComp C++ algo!
Motivation for paper:
Efficient verified code, still small TCB
How big is TCB of Isabelle's standard codegen? How big is ours?
Approach:
* Code generation from shallowly embedded IR, which is easy to compile to LLVM.
* Outside TCB: Preprocessing, VCG, integration of Sepref-Tool