src/base/io/ioWriteList.c

Go to the documentation of this file.

#include "io.h"

/*
-------- Original Message --------
Subject: Re: abc release and retiming
Date:    Sun, 13 Nov 2005 20:31:18 -0500 (EST)
From:    Luca Carloni <luca@cs.columbia.edu>
To:      Alan Mishchenko <alanmi@eecs.berkeley.edu>

Alan,

My graph-representation file format is based on an adjacency list
representation and is indeed quite simple, in fact maybe too simple... I
used it in order to reason on relatively small weighed direct graphs. I
simply list all vertices, one per line and for each vertex "V_source" I
list all vertices that are "sinks" with respect to it, i.e. such that
there is a distinct arc between "V_source" and each of them (in
paranthesis I list the name of the edge and its weight (number of latency
on that path). For instance, if you look at the following graph, you have
that vertex "v_5" is connected to vertex "v_6" through a directed arc
called "v_5_to_v_6" whose latency is equal to 3, i.e. there are three
flip-flops on this arc. Still, notice that I sometime interpret the graph
also as the representation of a  LIS where each node corresponds to a
shell encapsulating a sequential core module (i.e. a module which does not
contain any combinational path between its inputs and its outputs). With
this representation an arc of latency 3 is interpreted as a wire where two
relay stations have been inserted in addition to the flip-flop terminating
the output of the core module.

Finally, notice that the name of the arc does not necessarily have to be
"v_5_to_v_6", but it could have been something like "arc_222" or "xyz" as
long as it is a unique name in the graph.

Thanks,
Luca

Example of graph representation
-----------------------------------------------------------------------------
v_5     >       v_6 ([v_5_to_v_6] = 3), v_12 ([v_5_to_v_12] = 2).
v_2     >       v_4 ([v_2_to_v_4] = 1), v_10_s0 ([v_2_to_v_10_s0] = 6), v_12 ([v_2_to_v_12] = 3).
v_9     >       v_10_s0 ([v_9_to_v_10_s0] = 5), v_12 ([v_9_to_v_12] = 2).
v_12    >       v_13 ([v_12_to_v_13] = 5).
v_13    >       v_14 ([v_13_to_v_14] = 1).
v_6     >       v_7 ([v_6_to_v_7] = 2).
v_4     >       v_5 ([v_4_to_v_5] = 2).
v_1     >       v_2 ([v_1_to_v_2] = 1).
v_7     >       v_8 ([v_7_to_v_8] = 2).
t       >       .
v_14    >       t ([v_14_to_t] = 1), v_5 ([v_14_to_v_5] = 1).
v_8     >       v_9 ([v_8_to_v_9] = 2).
s       >       v_1 ([s_to_v_1] = 1).
v_10_s0 >       v_10_s1 ([v_10_s0_to_v_10_s1] = 1).
v_10_s1 >       v_4 ([v_10_s1__v_4] = 1), v_8 ([v_10_s1__v_8] = 1).
-----------------------------------------------------------------------------
*/


static void Io_WriteListEdge( FILE * pFile, Abc_Obj_t * pObj );
static void Io_WriteListHost( FILE * pFile, Abc_Ntk_t * pNtk );


void Io_WriteList( Abc_Ntk_t * pNtk, char * pFileName, int fUseHost )
{
    FILE * pFile;
    Abc_Obj_t * pObj;
    int i;

//    assert( Abc_NtkIsSeq(pNtk)  );

    // start the output stream
    pFile = fopen( pFileName, "w" );
    if ( pFile == NULL )
    {
        fprintf( stdout, "Io_WriteList(): Cannot open the output file \"%s\".\n", pFileName );
        return;
    }

    fprintf( pFile, "# Adjacency list for sequential AIG \"%s\"\n", pNtk->pName );
    fprintf( pFile, "# written by ABC on %s\n", Extra_TimeStamp() );

    // write the constant node
    if ( Abc_ObjFanoutNum( Abc_AigConst1(pNtk) ) > 0 )
        Io_WriteListEdge( pFile, Abc_AigConst1(pNtk) );

    // write the PI edges
    Abc_NtkForEachPi( pNtk, pObj, i )
        Io_WriteListEdge( pFile, pObj );

    // write the internal nodes
    Abc_AigForEachAnd( pNtk, pObj, i )
        Io_WriteListEdge( pFile, pObj );

    // write the host node
    if ( fUseHost )
        Io_WriteListHost( pFile, pNtk );
    else
        Abc_NtkForEachPo( pNtk, pObj, i )
            Io_WriteListEdge( pFile, pObj );

        fprintf( pFile, "\n" );
        fclose( pFile );
}

void Io_WriteListEdge( FILE * pFile, Abc_Obj_t * pObj )
{
    Abc_Obj_t * pFanout;
    int i;
    fprintf( pFile, "%-10s >    ", Abc_ObjName(pObj) );
    Abc_ObjForEachFanout( pObj, pFanout, i )
    {
        fprintf( pFile, " %s", Abc_ObjName(pFanout) );
        fprintf( pFile, " ([%s_to_", Abc_ObjName(pObj) );
//        fprintf( pFile, "%s] = %d)", Abc_ObjName(pFanout), Seq_ObjFanoutL(pObj, pFanout) );
        fprintf( pFile, "%s] = %d)", Abc_ObjName(pFanout), 0 );
        if ( i != Abc_ObjFanoutNum(pObj) - 1 )
            fprintf( pFile, "," );
    }
    fprintf( pFile, "." );
    fprintf( pFile, "\n" );
}

void Io_WriteListHost( FILE * pFile, Abc_Ntk_t * pNtk )
{
    Abc_Obj_t * pObj;
    int i;

    Abc_NtkForEachPo( pNtk, pObj, i )
    {
        fprintf( pFile, "%-10s >    ", Abc_ObjName(pObj) );
        fprintf( pFile, " %s ([%s_to_%s] = %d)", "HOST", Abc_ObjName(pObj), "HOST", 0 );
        fprintf( pFile, "." );
        fprintf( pFile, "\n" );
    }

    fprintf( pFile, "%-10s >    ", "HOST" );
    Abc_NtkForEachPi( pNtk, pObj, i )
    {
        fprintf( pFile, " %s", Abc_ObjName(pObj) );
        fprintf( pFile, " ([%s_to_%s] = %d)", "HOST", Abc_ObjName(pObj), 0 );
        if ( i != Abc_NtkPiNum(pNtk) - 1 )
            fprintf( pFile, "," );
    }
    fprintf( pFile, "." );
    fprintf( pFile, "\n" );
}


void Io_WriteCellNet( Abc_Ntk_t * pNtk, char * pFileName )
{
    FILE * pFile;
    Abc_Obj_t * pObj, * pFanout;
    int i, k;

    assert( Abc_NtkIsLogic(pNtk)  );

    // start the output stream
    pFile = fopen( pFileName, "w" );
    if ( pFile == NULL )
    {
        fprintf( stdout, "Io_WriteCellNet(): Cannot open the output file \"%s\".\n", pFileName );
        return;
    }

    fprintf( pFile, "# CellNet file for network \"%s\" written by ABC on %s\n", pNtk->pName, Extra_TimeStamp() );

    // the only tricky part with writing is handling latches:
    // each latch comes with (a) single-input latch-input node, (b) latch proper, (c) single-input latch-output node
    // we arbitrarily decide to use the interger ID of the latch-input node to represent the latch in the file
    // (this ID is used for both the cell and the net driven by that cell)

    // write the PIs
    Abc_NtkForEachPi( pNtk, pObj, i )
        fprintf( pFile, "cell %d is 0\n", pObj->Id );
    // write the POs
    Abc_NtkForEachPo( pNtk, pObj, i )
        fprintf( pFile, "cell %d is 1\n", pObj->Id );
    // write the latches (use the ID of latch input)
    Abc_NtkForEachLatch( pNtk, pObj, i )
        fprintf( pFile, "cell %d is 2\n", Abc_ObjFanin0(pObj)->Id );
    // write the logic nodes
    Abc_NtkForEachNode( pNtk, pObj, i )
        fprintf( pFile, "cell %d is %d\n", pObj->Id, 3+Abc_ObjFaninNum(pObj) );

    // write the nets driven by PIs
    Abc_NtkForEachPi( pNtk, pObj, i )
    {
        fprintf( pFile, "net %d  %d 0", pObj->Id, pObj->Id );
        Abc_ObjForEachFanout( pObj, pFanout, k )
            fprintf( pFile, "  %d %d", pFanout->Id, 1 + Abc_ObjFanoutFaninNum(pFanout, pObj) );
            fprintf( pFile, "\n" );
    }
    // write the nets driven by latches
    Abc_NtkForEachLatch( pNtk, pObj, i )
    {
        fprintf( pFile, "net %d  %d 0", Abc_ObjFanin0(pObj)->Id, Abc_ObjFanin0(pObj)->Id );
        pObj = Abc_ObjFanout0(pObj);
        Abc_ObjForEachFanout( pObj, pFanout, k )
            fprintf( pFile, "  %d %d", pFanout->Id, 1 + Abc_ObjFanoutFaninNum(pFanout, pObj) );
            fprintf( pFile, "\n" );
    }
    // write the nets driven by nodes
    Abc_NtkForEachNode( pNtk, pObj, i )
    {
        fprintf( pFile, "net %d  %d 0", pObj->Id, pObj->Id );
        Abc_ObjForEachFanout( pObj, pFanout, k )
            fprintf( pFile, "  %d %d", pFanout->Id, 1 + Abc_ObjFanoutFaninNum(pFanout, pObj) );
            fprintf( pFile, "\n" );
    }

        fprintf( pFile, "\n" );
        fclose( pFile );
}

---