murphi2uclid1

Usage

Overview

The utility murphi2uclid is bundled with the model checker Rumur and can be used to translate a Murphi model into a Uclid5 model. See rumur(1) for more information about Rumur or Murphi.

Options

--help or -?

Display usage information.

--module NAME or -m NAME

Set the name to use for the generated Uclid5 module. If you do not supply this, it defaults to main.

--numeric-type TYPER or -n TYPE

Set the Uclid5 type used for ranges, scalarsets and numeric literals. This can be either integer or a bit-vector type, for example, bv8. If this option is omitted, either integer or bv64 is chosen automatically based on the content of the input model.

--output FILE or -o FILE

Set the path to write the output Uclid5 model to. If this argument is omitted, output is written to standard out.

--quiet or -q

Silence warning diagnostics.

--verbose or -v

Report extra debugging information.

--version

Display version information and exit.

Notes

Translation to Uclid5 is imprecise in the case where there is no direct Uclid5 equivalent for a Murphi concept. For example, an undefine statement in Murphi has no equivalent in Uclid5. This is translated to a havoc statement. Obviously this is not exactly equivalent. However the hope is that in real world models this achieves a similar effect. Namely, any read of an undefined variable results in the verifier exploring unintended behavior. To get reasonable utility out of translation of a model using undefine, you will likely need an assertion-heavy model. Or, put another way, defensive programming is advised.

Limitations

The following Murphi concepts have no translation to Uclid5 and are rejected by murphi2uclid:

Function calls within expressions are translated as if they were calls to uninterpreted functions. Uclid5 does not support calling interpreted functions (procedures, in Uclid5 terminology) this way. The desirable mapping (uninterpreted function vs rephrasing the call site) cannot be determined automatically. So it is left to the user to tweak or post-process the output, as it will not be accepted by Uclid5 as-is.