Merge pull request #752 from ye-luo/adjust-J1

Adjust one body Jastrow SoA algorithm

src/Particle/ParticleSet.cpp

@@ -456,7 +456,6 @@ void ParticleSet::update(bool skipSK)
   if (!skipSK && SK)
     SK->UpdateAllPart(*this);
 
-  Ready4Measure=true;
   activePtcl=-1;
 }
 
@@ -471,7 +470,6 @@ void ParticleSet::update(const ParticlePos_t& pos)
   if (SK && !SK->DoUpdate)
     SK->UpdateAllPart(*this);
 
-  Ready4Measure=true;
   activePtcl=-1;
 }
 
@@ -590,7 +588,6 @@ bool ParticleSet::makeMove(const Walker_t& awalker
 bool ParticleSet::makeMove(const Walker_t& awalker
                            , const ParticlePos_t& deltaR, const std::vector<RealType>& dt)
 {
-  Ready4Measure=false;
   activePtcl=-1;
   if (UseBoundBox)
   {
@@ -632,7 +629,6 @@ bool ParticleSet::makeMoveWithDrift(const Walker_t& awalker
                                     , const ParticlePos_t& drift , const ParticlePos_t& deltaR
                                     , RealType dt)
 {
-  Ready4Measure=false;
   activePtcl=-1;
   if (UseBoundBox)
   {
@@ -669,7 +665,6 @@ bool ParticleSet::makeMoveWithDrift(const Walker_t& awalker
                                     , const ParticlePos_t& drift , const ParticlePos_t& deltaR
                                     , const std::vector<RealType>& dt)
 {
-  Ready4Measure=false;
   activePtcl=-1;
   if (UseBoundBox)
   {
@@ -765,7 +760,6 @@ void ParticleSet::donePbyP(bool skipSK)
     DistTables[i]->donePbyP();
   if (!skipSK && SK && !SK->DoUpdate)
     SK->UpdateAllPart(*this);
-  Ready4Measure=true;
   activePtcl=-1;
   myTimers[2]->stop();
 }
@@ -799,7 +793,6 @@ void ParticleSet::loadWalker(Walker_t& awalker, bool pbyp)
       SK->UpdateAllPart(*this);
   }
 
-  Ready4Measure=false;
   activePtcl=-1;
 }
 

src/Particle/ParticleSet.h

@@ -131,8 +131,6 @@ class ParticleSet
   bool IsGrouped;
   ///true if the particles have the same mass
   bool SameMass;
-  /// true if all the internal state is ready for estimators
-  bool Ready4Measure;
   ///threa id
   Index_t ThreadID;
   ///the index of the active particle for particle-by-particle moves

src/QMCWaveFunctions/Jastrow/J1OrbitalSoA.h

@@ -54,7 +54,7 @@ struct  J1OrbitalSoA : public OrbitalBase
   posT curGrad;
 
   ///\f$Vat[i] = sum_(j) u_{i,j}\f$
-  aligned_vector<RealType> Vat;
+  Vector<valT> Vat;
   aligned_vector<valT> U, dU, d2U;
   aligned_vector<valT> DistCompressed;
   aligned_vector<int> DistIndice;
@@ -136,14 +136,6 @@ struct  J1OrbitalSoA : public OrbitalBase
   {
     UpdateMode=ORB_PBYP_RATIO;
     curAt = computeU(P.DistTables[myTableID]->Temp_r.data());
-
-    if(!P.Ready4Measure)
-    {//need to compute per atom
-      computeU3(P,iat,P.DistTables[myTableID]->Distances[iat]);
-      Lap[iat]=accumulateGL(dU.data(),d2U.data(),P.DistTables[myTableID]->Displacements[iat],Grad[iat]);
-      Vat[iat]=simd::accumulate_n(U.data(),Nions,valT());
-    }
-
     return std::exp(Vat[iat]-curAt);
   }
 
@@ -276,9 +268,6 @@ struct  J1OrbitalSoA : public OrbitalBase
    */
   GradType evalGrad(ParticleSet& P, int iat)
   {
-    computeU3(P,iat,P.DistTables[myTableID]->Distances[iat]);
-    Lap[iat]=accumulateGL(dU.data(),d2U.data(),P.DistTables[myTableID]->Displacements[iat],Grad[iat]);
-    Vat[iat]=simd::accumulate_n(U.data(),Nions,valT());
     return GradType(Grad[iat]);
   }
 
@@ -319,15 +308,39 @@ struct  J1OrbitalSoA : public OrbitalBase
   }
 
 
-  inline void registerData(ParticleSet& P, WFBufferType& buf) { }
+  inline void registerData(ParticleSet& P, WFBufferType& buf)
+  {
+    if ( Bytes_in_WFBuffer == 0 )
+    {
+      Bytes_in_WFBuffer = buf.current();
+      buf.add(Vat.begin(), Vat.end());
+      buf.add(Grad.begin(), Grad.end());
+      buf.add(Lap.begin(), Lap.end());
+      Bytes_in_WFBuffer = buf.current()-Bytes_in_WFBuffer;
+      // free local space
+      Vat.free();
+      Grad.free();
+      Lap.free();
+    }
+    else
+    {
+      buf.forward(Bytes_in_WFBuffer);
+    }
+  }
 
   inline RealType updateBuffer(ParticleSet& P, WFBufferType& buf, bool fromscratch=false)
   {
     evaluateGL(P, P.G, P.L, false);
+    buf.forward(Bytes_in_WFBuffer);
     return LogValue;
   }
 
-  inline void copyFromBuffer(ParticleSet& P, WFBufferType& buf) { }
+  inline void copyFromBuffer(ParticleSet& P, WFBufferType& buf)
+  {
+    Vat.attachReference(buf.lendReference<valT>(Nelec), Nelec);
+    Grad.attachReference(buf.lendReference<posT>(Nelec), Nelec);
+    Lap.attachReference(buf.lendReference<valT>(Nelec), Nelec);
+  }
 
   OrbitalBasePtr makeClone(ParticleSet& tqp) const
   {

src/Sandbox/diff_j1.cpp

@@ -195,8 +195,6 @@ int main(int argc, char** argv)
       double r_ratio=0.0;
       double g_ratio=0.0;
 
-      els.Ready4Measure=false;
-
       int naccepted=0;
       for(int iel=0; iel<nels; ++iel)
       {

src/Sandbox/diff_wfs.cpp

@@ -90,6 +90,8 @@ int main(int argc, char** argv)
 
   {
     ParticleSet ions, els;
+    ions.setName("ion");
+    els.setName("e");
     OHMMS_PRECISION scale=1.0;
 
     int np=omp_get_num_threads();
@@ -177,8 +179,6 @@ int main(int argc, char** argv)
       double r_ratio=0.0;
       double g_ratio=0.0;
 
-      els.Ready4Measure=false;
-
       int naccepted=0;
       for(int iel=0; iel<nels; ++iel)
       {

src/Sandbox/miniqmc.cpp

@@ -165,6 +165,8 @@ int main(int argc, char** argv)
     clock.restart();
 
     ParticleSet ions, els;
+    ions.setName("ion");
+    els.setName("e");
     const OHMMS_PRECISION scale=1.0;
 
     const int np=omp_get_num_threads();

src/simd/algorithm.hpp

@@ -44,13 +44,13 @@ namespace qmcplusplus {
         return res;
       }
 
-  ///inner product
-  template<typename T1, typename T2, typename T3>
-    inline T3 inner_product_n(const T1* restrict a, const T2* restrict b, int n, T3 res)
-    {
-      for(int i=0; i<n; ++i) res += a[i]*b[i];
-      return res;
-    }
+    ///inner product
+    template<typename T1, typename T2, typename T3>
+      inline T3 inner_product_n(const T1* restrict a, const T2* restrict b, int n, T3 res)
+      {
+        for(int i=0; i<n; ++i) res += a[i]*b[i];
+        return res;
+      }
 
   } //simd namepsace
 }