A specializer, inverter, (abstract) interpreter, and normalizer for ARL. Based on previous work.

The project uses Stack to manage its dependencies.

The program can be built using the command stack build. It can be ran using stack run -- MODE, where MODE is either spec, interpret, invert, or bench.

Command: spec. Performs partial evaluation of a program with respect to a given specialization input. The specialization is given as a sequence of variable = 'constant declarations in a file. For examples see the "progs" folder.

Usage: PERevFlow-exe spec <Input RL file> <Output path> <Spec file>
                          [-u|--uniformBTA] [-s|--skipSpec] [-p|--skipPost]
                          [-v|--verbose] [-t|--trace]

• -u | --uniformBTA: Use the uniform binding-time analysis
• -s | --skipSpec: Disable post-processing of residual program
• -v | --verbose: Enable/Disable the default logging to stdout. Default: Enabled
• -t | --trace: Enable tracing of the specialization process. May be very long.

Invert the input program.

Usage: PERevFlow-exe invert <Input RL file> <Output path> [-v|--verbose]

• -v | --verbose: Enable/Disable the default logging to stdout. Default: Enabled

Interpret the input program with the given input. Shows final output and statistics. The input follows the same format as in the specializer.

Usage: PERevFlow-exe interpret <Input RL file> <Spec file> [-v|--verbose]

• -v | --verbose: Enable/Disable the default logging to stdout. Default: Enabled

Compare the size and speed of a source program for a given input with the size and speed of the residual programs from specializing it with the different BTA's.

The variables that should remain dynamic during specialization are specified in a space-separated list, e.g. "x y z".

Usage: PERevFlow-exe bench <Input RL file> <Spec file>
                           <List of dynamic input vars> [-v|--verbose]

• -v | --verbose: Enable/Disable the default logging to stdout. Default: Enabled