Yu-zh / minidafny

CS545 project

Usage

We developed a perl script to automatically compile & run the test.

Example：

./run.pl -clean order.imp

-clean indicates cleaning compiliation garbage and results from previous run order.imp is the benchmark that applied to minidafny, both relative and absulote path are welcome.

./run.pl -all

-all indicates running all the benchmarks under the Benchmarks director. The script will enumerate them and run them one by one.

./run.pl -info=5 -all

info=5 indicates printing all the compilation, execution and debug information.

Running the perl script will re-compile the scala source file every time. In case you have compiled it and do not want to waste time waiting compiling, you can run the bytecode directly:

scala -cp _Objs VCGen path/to/imp/file

Collaborators

Yu Zhang and Bowen Huang

Benchmarks

We supplied new benchmarks and they can be found in the directory Benchmark/valid

• arr_order.imp: Given two arrays a and b, swap a[i] and b[i] if a[i] > b[i]. Thus, for all index i, a[i] <= b[i]. Simple and trivial.
• arr_order_par.imp: An implementation of arr_order.imp using parallel array write.
• bsearch.imp: Binary search.
• bubble_improved.imp: A slightly optimized version of bubble sort so the invariants are slightly more complicated. Parallel array write is used.
• bubble_parwrite.imp: The same as bubble.imp except parallel array write is used to swap two elements.
• insertion.imp: Insertion sort.
• max.imp: Find the maximum element in an array.
• order_corner_case.imp: Test of the corner cases, e.g. a while loop with empty body, etc.
• order_empty_forall.imp: Test of the corner cases, e.g. a forall formula with no quantified variables, etc.
• partition.imp: Array Partition. Parallel array write is used.
• rev_rev.imp: Reverse an array twice.
• selection.imp: Selection sort.

All the benchmarks above can be verified. There is one more interesting benchmark in the directory Benchmark/valid

• gcd_euclidean.imp: An euclidean algorithm implementation for GCD problem. I believe the program is valid. However, z3 cannot figure it out because of its insuffient support for modular linear arithmetic. So it will run for a long while and return unknown. Thus, the output of our verifier for this test case is Unknown.

Implementation of VCG

• Object ImpParser: A parser that parses the program with its pre- and post-conditions.
• Object VCGenerator: A generator that generate the weakest pre-condition directly from the program and its post-conditions. The intermidiate guard condition language is bypassed and the problem of generating fresh variable names is also avoided.
• Object SMTFormat: A translator that translates the assertions to SMT2 format formulars.
• Object Helper: Auxilary functions.
• Object PrettyPrinter: A simple pretty printer. This part is not working very well and the output is still a little messy. It's only used for the purpose of debug.