Commit 2c4c464a by Alan Mishchenko

Adding structural flop priority heuristics in 'pdr' (bug fix).

parent 45bf0369
...@@ -43,15 +43,134 @@ ABC_NAMESPACE_IMPL_START ...@@ -43,15 +43,134 @@ ABC_NAMESPACE_IMPL_START
SeeAlso [] SeeAlso []
***********************************************************************/ ***********************************************************************/
Vec_Int_t * Pdr_ManDeriveFlopPriorities2( Gia_Man_t * p, int fMuxCtrls, int * pnPrioShift ) Vec_Int_t * Pdr_ManDeriveFlopPriorities3( Gia_Man_t * p, int fMuxCtrls )
{ {
int fDiscount = 0;
Vec_Wec_t * vLevels;
Vec_Int_t * vRes, * vLevel, * vCosts;
Gia_Obj_t * pObj, * pCtrl, * pData0, * pData1;
int i, k, Entry, MaxEntry = 0;
Gia_ManCreateRefs(p);
// discount references
if ( fDiscount )
Gia_ManForEachAnd( p, pObj, i )
{
if ( !Gia_ObjIsMuxType(pObj) )
continue;
pCtrl = Gia_Regular(Gia_ObjRecognizeMux(pObj, &pData1, &pData0));
pData0 = Gia_Regular(pData0);
pData1 = Gia_Regular(pData1);
p->pRefs[Gia_ObjId(p, pCtrl)]--;
if ( pData0 == pData1 )
p->pRefs[Gia_ObjId(p, pData0)]--;
}
// create flop costs
vCosts = Vec_IntAlloc( Gia_ManRegNum(p) );
Gia_ManForEachRo( p, pObj, i )
{
Vec_IntPush( vCosts, Gia_ObjRefNum(p, pObj) );
MaxEntry = Abc_MaxInt( MaxEntry, Gia_ObjRefNum(p, pObj) );
//printf( "%d(%d) ", i, Gia_ObjRefNum(p, pObj) );
}
//printf( "\n" );
MaxEntry++;
// add costs due to MUX inputs
if ( fMuxCtrls )
{
int fVerbose = 0;
Vec_Bit_t * vCtrls = Vec_BitStart( Gia_ManObjNum(p) );
Vec_Bit_t * vDatas = Vec_BitStart( Gia_ManObjNum(p) );
Gia_Obj_t * pCtrl, * pData1, * pData0;
int nCtrls = 0, nDatas = 0, nBoth = 0;
Gia_ManForEachAnd( p, pObj, i )
{
if ( !Gia_ObjIsMuxType(pObj) )
continue;
pCtrl = Gia_ObjRecognizeMux( pObj, &pData1, &pData0 );
pCtrl = Gia_Regular(pCtrl);
pData1 = Gia_Regular(pData1);
pData0 = Gia_Regular(pData0);
Vec_BitWriteEntry( vCtrls, Gia_ObjId(p, pCtrl), 1 );
Vec_BitWriteEntry( vDatas, Gia_ObjId(p, pData1), 1 );
Vec_BitWriteEntry( vDatas, Gia_ObjId(p, pData0), 1 );
}
Gia_ManForEachRo( p, pObj, i )
if ( Vec_BitEntry(vCtrls, Gia_ObjId(p, pObj)) )
Vec_IntAddToEntry( vCosts, i, MaxEntry );
//else if ( Vec_BitEntry(vDatas, Gia_ObjId(p, pObj)) )
// Vec_IntAddToEntry( vCosts, i, MaxEntry );
MaxEntry = 2*MaxEntry + 1;
// print out
if ( fVerbose )
{
Gia_ManForEachRo( p, pObj, i )
{
if ( Vec_BitEntry(vCtrls, Gia_ObjId(p, pObj)) )
nCtrls++;
if ( Vec_BitEntry(vDatas, Gia_ObjId(p, pObj)) )
nDatas++;
if ( Vec_BitEntry(vCtrls, Gia_ObjId(p, pObj)) && Vec_BitEntry(vDatas, Gia_ObjId(p, pObj)) )
nBoth++;
}
printf( "%10s : Flops = %5d. Ctrls = %5d. Datas = %5d. Both = %5d.\n", Gia_ManName(p), Gia_ManRegNum(p), nCtrls, nDatas, nBoth );
}
Vec_BitFree( vCtrls );
Vec_BitFree( vDatas );
}
// create levelized structure
vLevels = Vec_WecStart( MaxEntry );
Vec_IntForEachEntry( vCosts, Entry, i )
Vec_WecPush( vLevels, Entry, i );
// collect in this order
MaxEntry = 0;
vRes = Vec_IntStart( Gia_ManRegNum(p) );
Vec_WecForEachLevel( vLevels, vLevel, i )
Vec_IntForEachEntry( vLevel, Entry, k )
Vec_IntWriteEntry( vRes, Entry, MaxEntry++ );
//printf( "%d ", Gia_ObjRefNum(p, Gia_ManCi(p, Gia_ManPiNum(p)+Entry)) );
//printf( "\n" );
assert( MaxEntry == Gia_ManRegNum(p) );
Vec_WecFree( vLevels );
Vec_IntFree( vCosts );
ABC_FREE( p->pRefs );
//Vec_IntPrint( vRes );
return vRes;
}
/**Function*************************************************************
Synopsis [Structural analysis.]
Description []
SideEffects []
SeeAlso []
***********************************************************************/
Vec_Int_t * Pdr_ManDeriveFlopPriorities2( Gia_Man_t * p, int fMuxCtrls )
{
int fDiscount = 0;
Vec_Int_t * vRes = NULL; Vec_Int_t * vRes = NULL;
Gia_Obj_t * pObj; Gia_Obj_t * pObj, * pCtrl, * pData0, * pData1;
int MaxEntry = 0; int MaxEntry = 0;
int i, * pPerm; int i, * pPerm;
// create flop costs // create flop costs
Vec_Int_t * vCosts = Vec_IntStart( Gia_ManRegNum(p) ); Vec_Int_t * vCosts = Vec_IntStart( Gia_ManRegNum(p) );
Gia_ManCreateRefs(p); Gia_ManCreateRefs(p);
// discount references
if ( fDiscount )
Gia_ManForEachAnd( p, pObj, i )
{
if ( !Gia_ObjIsMuxType(pObj) )
continue;
pCtrl = Gia_Regular(Gia_ObjRecognizeMux(pObj, &pData1, &pData0));
pData0 = Gia_Regular(pData0);
pData1 = Gia_Regular(pData1);
p->pRefs[Gia_ObjId(p, pCtrl)]--;
if ( pData0 == pData1 )
p->pRefs[Gia_ObjId(p, pData0)]--;
}
Gia_ManForEachRo( p, pObj, i ) Gia_ManForEachRo( p, pObj, i )
{ {
Vec_IntWriteEntry( vCosts, i, Gia_ObjRefNum(p, pObj) ); Vec_IntWriteEntry( vCosts, i, Gia_ObjRefNum(p, pObj) );
...@@ -81,10 +200,9 @@ Vec_Int_t * Pdr_ManDeriveFlopPriorities2( Gia_Man_t * p, int fMuxCtrls, int * pn ...@@ -81,10 +200,9 @@ Vec_Int_t * Pdr_ManDeriveFlopPriorities2( Gia_Man_t * p, int fMuxCtrls, int * pn
} }
Gia_ManForEachRo( p, pObj, i ) Gia_ManForEachRo( p, pObj, i )
if ( Vec_BitEntry(vCtrls, Gia_ObjId(p, pObj)) ) if ( Vec_BitEntry(vCtrls, Gia_ObjId(p, pObj)) )
Vec_IntAddToEntry( vCosts, i, 2*MaxEntry ); Vec_IntAddToEntry( vCosts, i, MaxEntry );
//else if ( Vec_BitEntry(vDatas, Gia_ObjId(p, pObj)) ) //else if ( Vec_BitEntry(vDatas, Gia_ObjId(p, pObj)) )
// Vec_IntAddToEntry( vCosts, i, MaxEntry ); // Vec_IntAddToEntry( vCosts, i, MaxEntry );
MaxEntry *= 3;
// print out // print out
if ( fVerbose ) if ( fVerbose )
{ {
...@@ -102,7 +220,6 @@ Vec_Int_t * Pdr_ManDeriveFlopPriorities2( Gia_Man_t * p, int fMuxCtrls, int * pn ...@@ -102,7 +220,6 @@ Vec_Int_t * Pdr_ManDeriveFlopPriorities2( Gia_Man_t * p, int fMuxCtrls, int * pn
Vec_BitFree( vCtrls ); Vec_BitFree( vCtrls );
Vec_BitFree( vDatas ); Vec_BitFree( vDatas );
} }
*pnPrioShift = 1 + Abc_Base2Log( MaxEntry );
// create ordering based on costs // create ordering based on costs
pPerm = Abc_MergeSortCost( Vec_IntArray(vCosts), Vec_IntSize(vCosts) ); pPerm = Abc_MergeSortCost( Vec_IntArray(vCosts), Vec_IntSize(vCosts) );
vRes = Vec_IntAllocArray( pPerm, Vec_IntSize(vCosts) ); vRes = Vec_IntAllocArray( pPerm, Vec_IntSize(vCosts) );
...@@ -112,10 +229,10 @@ Vec_Int_t * Pdr_ManDeriveFlopPriorities2( Gia_Man_t * p, int fMuxCtrls, int * pn ...@@ -112,10 +229,10 @@ Vec_Int_t * Pdr_ManDeriveFlopPriorities2( Gia_Man_t * p, int fMuxCtrls, int * pn
//Vec_IntPrint( vCosts ); //Vec_IntPrint( vCosts );
return vCosts; return vCosts;
} }
Vec_Int_t * Pdr_ManDeriveFlopPriorities( Aig_Man_t * pAig, int fMuxCtrls, int * pnPrioShift ) Vec_Int_t * Pdr_ManDeriveFlopPriorities( Aig_Man_t * pAig, int fMuxCtrls )
{ {
Gia_Man_t * pGia = Gia_ManFromAigSimple(pAig); Gia_Man_t * pGia = Gia_ManFromAigSimple(pAig);
Vec_Int_t * vRes = Pdr_ManDeriveFlopPriorities2(pGia, fMuxCtrls, pnPrioShift); Vec_Int_t * vRes = Pdr_ManDeriveFlopPriorities2(pGia, fMuxCtrls);
Gia_ManStop( pGia ); Gia_ManStop( pGia );
return vRes; return vRes;
} }
...@@ -145,11 +262,11 @@ Pdr_Man_t * Pdr_ManStart( Aig_Man_t * pAig, Pdr_Par_t * pPars, Vec_Int_t * vPrio ...@@ -145,11 +262,11 @@ Pdr_Man_t * Pdr_ManStart( Aig_Man_t * pAig, Pdr_Par_t * pPars, Vec_Int_t * vPrio
if ( !p->pPars->fMonoCnf ) if ( !p->pPars->fMonoCnf )
p->vVLits = Vec_WecStart( 1+Abc_MaxInt(1, Aig_ManLevels(pAig)) ); p->vVLits = Vec_WecStart( 1+Abc_MaxInt(1, Aig_ManLevels(pAig)) );
// internal use // internal use
p->nPrioShift = 0; p->nPrioShift = Abc_Base2Log(Aig_ManRegNum(pAig));
if ( vPrioInit ) if ( vPrioInit )
p->vPrio = vPrioInit; p->vPrio = vPrioInit;
else if ( pPars->fFlopPrio ) else if ( pPars->fFlopPrio )
p->vPrio = Pdr_ManDeriveFlopPriorities(pAig, 0, &p->nPrioShift); p->vPrio = Pdr_ManDeriveFlopPriorities(pAig, 1);
else else
p->vPrio = Vec_IntStart( Aig_ManRegNum(pAig) ); p->vPrio = Vec_IntStart( Aig_ManRegNum(pAig) );
p->vLits = Vec_IntAlloc( 100 ); // array of literals p->vLits = Vec_IntAlloc( 100 ); // array of literals
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment