Commit b98345ce by Alan Mishchenko

Improvements to buffering and sizing.

parent 95684b04
......@@ -102,7 +102,7 @@ int Abc_SclCheckImprovement( SC_Man * p, Abc_Obj_t * pObj, Vec_Int_t * vNodes, V
{
Abc_Obj_t * pTemp;
SC_Cell * pCellOld, * pCellNew;
float dGain, dGainBest, gGainCur;
float dGain, dGainBest;
int i, k, gateBest;
abctime clk;
clk = Abc_Clock();
......@@ -132,20 +132,6 @@ clk = Abc_Clock();
Abc_SclObjSetCell( pObj, pCellOld );
Abc_SclLoadRestore( p, pObj );
// evaluate gain
/*
dGain = 0.0;
Abc_NtkForEachObjVec( vEvals, p->pNtk, pTemp, k )
if ( Abc_SclObjLegal(p, pTemp, p->MaxDelay0) )
{
gGainCur = Abc_SclObjGain( p, pTemp );
dGain += (gGainCur > 0) ? gGainCur : 1.0 * gGainCur;
}
else
break;
if ( k < Vec_IntSize(vEvals) )
continue;
dGain /= Vec_IntSize(vEvals);
*/
dGain = Abc_SclEvalPerformLegal( p, vEvals, p->MaxDelay0 );
if ( dGain == -1 )
continue;
......
......@@ -47,27 +47,24 @@ struct SC_Man_
Abc_Ntk_t * pNtk; // network
int nObjs; // allocated size
// get assignment
// Vec_Int_t * vGates; // mapping of objId into gateId
Vec_Int_t * vGatesBest; // best gate sizes found so far
Vec_Int_t * vUpdates; // sizing updates in this round
Vec_Int_t * vUpdates2; // sizing updates in this round
// timing information
SC_WireLoad * pWLoadUsed; // name of the used WireLoad model
SC_Pair * pLoads; // loads for each gate
SC_Pair * pDepts; // departures for each gate
SC_Pair * pTimes; // arrivals for each gate
SC_Pair * pSlews; // slews for each gate
SC_Pair * pTimes2; // arrivals for each gate
SC_Pair * pSlews2; // slews for each gate
Vec_Flt_t * vTimes2; // backup storage for times
Vec_Flt_t * vTimes3; // backup storage for slews
Vec_Flt_t * vLoads2; // backup storage for loads
Vec_Flt_t * vLoads3; // backup storage for loads
float * pSlack; // slacks for each gatt
float * pInDrive; // maximum input drive strength
Vec_Int_t * vBestFans; // best fanouts
Vec_Flt_t * vTimesOut; // output arrival times
Vec_Que_t * vQue; // outputs by their time
SC_WireLoad * pWLoadUsed; // name of the used WireLoad model
Vec_Flt_t * vTimesOut; // output arrival times
// backup information
Vec_Flt_t * vLoads2; // backup storage for loads
Vec_Flt_t * vLoads3; // backup storage for loads
Vec_Flt_t * vTimes2; // backup storage for times
Vec_Flt_t * vTimes3; // backup storage for slews
// buffer trees
float EstLoadMax; // max ratio of Cout/Cin when this kicks in
float EstLoadAve; // average load of the gate
......@@ -78,6 +75,7 @@ struct SC_Man_
Vec_Flt_t * vNode2Gain; // mapping node into its gain
Vec_Int_t * vNode2Gate; // mapping node into its best gate
Vec_Int_t * vNodeIter; // the last iteration the node was upsized
Vec_Int_t * vBestFans; // best fanouts
// optimization parameters
float SumArea; // total area
float MaxDelay; // max delay
......@@ -109,8 +107,6 @@ static inline SC_Pair * Abc_SclObjLoad( SC_Man * p, Abc_Obj_t * pObj )
static inline SC_Pair * Abc_SclObjDept( SC_Man * p, Abc_Obj_t * pObj ) { return p->pDepts + Abc_ObjId(pObj); }
static inline SC_Pair * Abc_SclObjTime( SC_Man * p, Abc_Obj_t * pObj ) { return p->pTimes + Abc_ObjId(pObj); }
static inline SC_Pair * Abc_SclObjSlew( SC_Man * p, Abc_Obj_t * pObj ) { return p->pSlews + Abc_ObjId(pObj); }
static inline SC_Pair * Abc_SclObjTime2( SC_Man * p, Abc_Obj_t * pObj ) { return p->pTimes2 + Abc_ObjId(pObj); }
static inline SC_Pair * Abc_SclObjSlew2( SC_Man * p, Abc_Obj_t * pObj ) { return p->pSlews2 + Abc_ObjId(pObj); }
static inline float Abc_SclObjTimeMax( SC_Man * p, Abc_Obj_t * pObj ) { return Abc_MaxFloat(Abc_SclObjTime(p, pObj)->rise, Abc_SclObjTime(p, pObj)->fall); }
static inline float Abc_SclObjDepthMax( SC_Man * p, Abc_Obj_t * pObj ) { return Abc_MaxFloat(Abc_SclObjDept(p, pObj)->rise, Abc_SclObjDept(p, pObj)->fall); }
......@@ -119,10 +115,7 @@ static inline float Abc_SclObjGetSlackR( SC_Man * p, Abc_Obj_t * pObj, float
static inline float Abc_SclObjGetSlackF( SC_Man * p, Abc_Obj_t * pObj, float D ){ return D - (Abc_SclObjTime(p, pObj)->fall + Abc_SclObjDept(p, pObj)->fall); }
static inline float Abc_SclObjSlack( SC_Man * p, Abc_Obj_t * pObj ) { return p->pSlack[Abc_ObjId(pObj)]; }
static inline float Abc_SclObjLoadAve( SC_Man * p, Abc_Obj_t * pObj ) { return 0.5 * Abc_SclObjLoad(p, pObj)->rise + 0.5 * Abc_SclObjLoad(p, pObj)->fall; }
static inline void Abc_SclObjDupFanin( SC_Man * p, Abc_Obj_t * pObj ) { assert( Abc_ObjIsCo(pObj) ); *Abc_SclObjTime(p, pObj) = *Abc_SclObjTime(p, Abc_ObjFanin0(pObj)); }
static inline float Abc_SclObjGain( SC_Man * p, Abc_Obj_t * pObj ) { return 0.5*((Abc_SclObjTime2(p, pObj)->rise - Abc_SclObjTime(p, pObj)->rise) + (Abc_SclObjTime2(p, pObj)->fall - Abc_SclObjTime(p, pObj)->fall)); }
//static inline int Abc_SclObjLegal( SC_Man * p, Abc_Obj_t * pObj, float D ) { return Abc_SclObjTime(p, pObj)->rise <= Abc_SclObjTime2(p, pObj)->rise + Abc_SclObjGetSlackR(p, pObj, D) && Abc_SclObjTime(p, pObj)->fall <= Abc_SclObjTime2(p, pObj)->fall + Abc_SclObjGetSlackF(p, pObj, D); }
static inline double Abc_SclObjLoadFf( SC_Man * p, Abc_Obj_t * pObj, int fRise ) { return SC_LibCapFf( p->pLib, fRise ? Abc_SclObjLoad(p, pObj)->rise : Abc_SclObjLoad(p, pObj)->fall); }
static inline double Abc_SclObjTimePs( SC_Man * p, Abc_Obj_t * pObj, int fRise ) { return SC_LibTimePs(p->pLib, fRise ? Abc_SclObjTime(p, pObj)->rise : Abc_SclObjTime(p, pObj)->fall); }
......@@ -157,8 +150,6 @@ static inline SC_Man * Abc_SclManAlloc( SC_Lib * pLib, Abc_Ntk_t * pNtk )
p->pDepts = ABC_CALLOC( SC_Pair, p->nObjs );
p->pTimes = ABC_CALLOC( SC_Pair, p->nObjs );
p->pSlews = ABC_CALLOC( SC_Pair, p->nObjs );
p->pTimes2 = ABC_CALLOC( SC_Pair, p->nObjs );
p->pSlews2 = ABC_CALLOC( SC_Pair, p->nObjs );
p->pSlack = ABC_FALLOC( float, p->nObjs );
p->vBestFans = Vec_IntStart( p->nObjs );
p->vTimesOut = Vec_FltStart( Abc_NtkCoNum(pNtk) );
......@@ -189,10 +180,10 @@ static inline void Abc_SclManFree( SC_Man * p )
Vec_FltFreeP( &p->vNode2Gain );
Vec_IntFreeP( &p->vNode2Gate );
// intermediate data
Vec_FltFreeP( &p->vTimes2 );
Vec_FltFreeP( &p->vTimes3 );
Vec_FltFreeP( &p->vLoads2 );
Vec_FltFreeP( &p->vLoads3 );
Vec_FltFreeP( &p->vTimes2 );
Vec_FltFreeP( &p->vTimes3 );
Vec_IntFreeP( &p->vUpdates );
Vec_IntFreeP( &p->vUpdates2 );
Vec_IntFreeP( &p->vGatesBest );
......@@ -200,14 +191,11 @@ static inline void Abc_SclManFree( SC_Man * p )
Vec_QueCheck( p->vQue );
Vec_QueFreeP( &p->vQue );
Vec_FltFreeP( &p->vTimesOut );
// Vec_IntFreeP( &p->vGates );
Vec_IntFreeP( &p->vBestFans );
ABC_FREE( p->pLoads );
ABC_FREE( p->pDepts );
ABC_FREE( p->pTimes );
ABC_FREE( p->pSlews );
ABC_FREE( p->pTimes2 );
ABC_FREE( p->pSlews2 );
ABC_FREE( p->pSlack );
ABC_FREE( p->pInDrive );
ABC_FREE( p );
......@@ -297,17 +285,6 @@ static inline void Abc_SclLoadRestore3( SC_Man * p, Abc_Obj_t * pObj )
}
assert( Vec_FltSize(p->vLoads3) == k );
}
static inline void Abc_SclConeClear( SC_Man * p, Vec_Int_t * vCone )
{
SC_Pair Zero = { 0.0, 0.0 };
Abc_Obj_t * pObj;
int i;
Abc_NtkForEachObjVec( vCone, p->pNtk, pObj, i )
{
*Abc_SclObjTime(p, pObj) = Zero;
*Abc_SclObjSlew(p, pObj) = Zero;
}
}
static inline void Abc_SclConeStore( SC_Man * p, Vec_Int_t * vCone )
{
Abc_Obj_t * pObj;
......@@ -319,8 +296,6 @@ static inline void Abc_SclConeStore( SC_Man * p, Vec_Int_t * vCone )
Vec_FltPush( p->vTimes2, Abc_SclObjTime(p, pObj)->fall );
Vec_FltPush( p->vTimes2, Abc_SclObjSlew(p, pObj)->rise );
Vec_FltPush( p->vTimes2, Abc_SclObjSlew(p, pObj)->fall );
*Abc_SclObjTime2(p, pObj) = *Abc_SclObjTime(p, pObj);
*Abc_SclObjSlew2(p, pObj) = *Abc_SclObjSlew(p, pObj);
}
}
static inline void Abc_SclConeRestore( SC_Man * p, Vec_Int_t * vCone )
......@@ -333,8 +308,6 @@ static inline void Abc_SclConeRestore( SC_Man * p, Vec_Int_t * vCone )
Abc_SclObjTime(p, pObj)->fall = Vec_FltEntry(p->vTimes2, k++);
Abc_SclObjSlew(p, pObj)->rise = Vec_FltEntry(p->vTimes2, k++);
Abc_SclObjSlew(p, pObj)->fall = Vec_FltEntry(p->vTimes2, k++);
*Abc_SclObjTime(p, pObj) = *Abc_SclObjTime2(p, pObj);
*Abc_SclObjSlew(p, pObj) = *Abc_SclObjSlew2(p, pObj);
}
assert( Vec_FltSize(p->vTimes2) == k );
}
......@@ -374,11 +347,22 @@ static inline float Abc_SclEvalPerformLegal( SC_Man * p, Vec_Int_t * vCone, floa
Fall = Vec_FltEntry(p->vTimes3, k++) - Abc_SclObjTime(p, pObj)->fall;
if ( Rise + Abc_SclObjGetSlackR(p, pObj, D) < 0 || Fall + Abc_SclObjGetSlackF(p, pObj, D) < 0 )
return -1;
Eval = 0.5 * Rise + 0.5 * Fall;
Eval += 0.5 * Rise + 0.5 * Fall;
}
assert( Vec_FltSize(p->vTimes3) == k );
return Eval / Vec_IntSize(vCone);
}
static inline void Abc_SclConeClear( SC_Man * p, Vec_Int_t * vCone )
{
SC_Pair Zero = { 0.0, 0.0 };
Abc_Obj_t * pObj;
int i;
Abc_NtkForEachObjVec( vCone, p->pNtk, pObj, i )
{
*Abc_SclObjTime(p, pObj) = Zero;
*Abc_SclObjSlew(p, pObj) = Zero;
}
}
/**Function*************************************************************
......
......@@ -271,9 +271,8 @@ void Abc_SclFindNodesToUpdate( Abc_Obj_t * pPivot, Vec_Int_t ** pvNodes, Vec_Int
int Abc_SclFindBestCell( SC_Man * p, Abc_Obj_t * pObj, Vec_Int_t * vRecalcs, Vec_Int_t * vEvals, int Notches, int DelayGap, float * pGainBest )
{
SC_Cell * pCellOld, * pCellNew;
Abc_Obj_t * pTemp;
float dGain, dGainBest, gGainCur;
int k, n, gateBest;
float dGain, dGainBest;
int k, gateBest;
// save old gate, timing, fanin load
pCellOld = Abc_SclObjCell( pObj );
Abc_SclConeStore( p, vRecalcs );
......@@ -298,18 +297,8 @@ int Abc_SclFindBestCell( SC_Man * p, Abc_Obj_t * pObj, Vec_Int_t * vRecalcs, Vec
// set old cell
Abc_SclObjSetCell( pObj, pCellOld );
Abc_SclLoadRestore( p, pObj );
// evaluate gain
dGain = 0.0;
Abc_NtkForEachObjVec( vEvals, p->pNtk, pTemp, n )
{
gGainCur = Abc_SclObjGain( p, pTemp );
dGain += (gGainCur > 0) ? gGainCur : 2.0 * gGainCur;
}
dGain /= Vec_IntSize(vEvals);
// dGain = Abc_SclEvalPerform( p, vEvals );
// save best gain
dGain = Abc_SclEvalPerform( p, vEvals );
if ( dGainBest < dGain )
{
dGainBest = dGain;
......
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