Asymptus is a tool for automatic inference of function complexity. It uses a hibrid approach, mixing static analysis and dynamic profiling, in order to provide more accurate results.
Asymptus requires LLVM 3.4 to performe static analysis. To install LLVM 3.4 you can follow these tutorials: downloading a specific version of LLVM and installing LLVM. Note that its necessary to also install clang, the LLVM front-end.
Once you have a working installation of LLVM 3.4, you can download the Asymptus source code and install it.
Asymptus executes over either a C/C++ file or LLVM bytecodes. The simplest way to run our tool
is to execute ./asymptus input_file.(c|cpp|bc). It will identify how many command line arguments
the input program reads and generate numeric values for them. You may also use the following
options:
--args type .... Allows the user to specify the types of command line arguments.
It is necessary to specify as many argument as the program reads
in order to have a correct execution. The types may be one of the
following: int, long, float, double, num, char or string.
--mix ARG .. Allows the user to mix random generated arguments with concrete
data. When using this option, the random arguments types have to
come between {}. For instance, --mix myInput {int}.
--man This option will ask the user for only concrete inputs. Asymptus will
ask for the inputs of each desired execution. An empty line means all
data has been provided.
-v Verbose mode. Prints the polynomial for each loop inside a function.
-h or --help Prints this usage message.
If you want to run Asymptus with LLVM 3.7, go to llvm-3.7 branch (experimental).