Re: Initial state and check_invariant

From: Fabio Somenzi (Fabio@colorado.edu)
Date: Mon Jul 08 2002 - 15:03:59 MDT

  • Next message: Srikanth Jujjuru: "help to convert C code to verilog or Vhdl"

    Jaen-Luc,

    If your .mv file does not specify the intial state(s), then all
    possible states are assumed initial.

    You want to add the following lines:

    .r xx35_out
    1

    to your model. This will force the initial state to 1.

    Fabio

    >>>>> "JM" == Jean-Luc Mercadier <jean-luc.mercadier@tni-valiosys.com> writes:

     JM> Hello,
     JM> I would like to check the following design (W10.trivial.blif):

     JM> .model w10l.s.pre_blif
     JM> .outputs xx35_out
     JM> .latch xx35 xx35_out 1;
     JM> .names xx35
     JM> 1
     JM> .end

     JM> with the following invariant (W10.B.invar):

     JM> xx35_out = 1 ;
     JM> and the following script :
     JM> set vis_stdout ci_trivial_1.vis.output
     JM> set vis_stderr ci_trivial_1.vis.error
     JM> read_blif W10.trivial.blif
     JM> flatten_hierarchy
     JM> static_order
     JM> build_partition_mdds
     JM> check_invariant -d 1 -i -v 2 -F ci_trivial_1.vis.debug W10.B.invar
     JM> time

     JM> I've obtained the following output :

     JM> ********************************
     JM> Reachability analysis results:
     JM> FSM depth = 1
     JM> reachable states = 2
     JM> BDD size = 1
     JM> analysis time = 0
     JM> reachability analysis = complete
     JM> # INV: Summary of invariant pass/fail
     JM> # INV: formula failed --- xx35_out=1
     JM> elapse: 0.0 seconds, total: 0.0 seconds

     JM> and the following debug trace :

     JM> # INV: formula 1 failed --- xx35_out=1
     JM> # INV: calling debugger
     JM> # INV: invariant fails at the following initial state
     JM> --State 0:
     JM> xx35_out:0

     JM> The built FSM accept value 0 for variable xx35_out even with the
     JM> .latch xx35 xx35_out 1;
     JM> initialising xx35_out with value 1.

     JM> How can I force the initial state of the FSM with chosen values ?
     JM> Could this be a bug ?
     JM> Thanks in advance for your response.
     JM> Jean-Luc Mercadier



    This archive was generated by hypermail 2b30 : Mon Jul 08 2002 - 15:07:11 MDT