Very minor optimizations (#3602)

platforms/cpu/include/CpuNonbondedForce.h

@@ -210,7 +210,7 @@ protected:
         bool includeEnergy;
         float inverseRcut6;
         float inverseRcut6Expterm;
-        std::atomic<int> atomicCounter;
+        std::atomic<int> atomicCounter, atomicCounter2;
 
         static const float TWO_OVER_SQRT_PI;
         static const int NUM_TABLE_POINTS;

platforms/cpu/src/CpuNeighborList.cpp

@@ -92,11 +92,8 @@ public:
                 voxelSizeZ = (maxz-minz)/nz;
         }
         bins.resize(ny);
-        for (int i = 0; i < ny; i++) {
+        for (int i = 0; i < ny; i++)
             bins[i].resize(nz);
-            for (int j = 0; j < nz; j++)
-                bins[i][j].resize(0);
-        }
     }
 
     /**

platforms/cpu/src/CpuNonbondedForce.cpp

@@ -391,20 +391,13 @@ void CpuNonbondedForce::calculateDirectIxn(int numberOfAtoms, float* posq, const
     includeEnergy = (totalEnergy != NULL);
     threadEnergy.resize(threads.getNumThreads());
     atomicCounter = 0;
+    atomicCounter2 = 0;
     
     // Signal the threads to start running and wait for them to finish.
     
     threads.execute([&] (ThreadPool& threads, int threadIndex) { threadComputeDirect(threads, threadIndex); });
     threads.waitForThreads();
     
-    // Signal the threads to subtract the exclusions.
-    
-    if (ewald || pme) {
-        atomicCounter = 0;
-        threads.resumeThreads();
-        threads.waitForThreads();
-    }
-    
     // Combine the energies from all the threads.
     
     if (totalEnergy != NULL) {
@@ -436,10 +429,9 @@ void CpuNonbondedForce::threadComputeDirect(ThreadPool& threads, int threadIndex
 
         // Now subtract off the exclusions, since they were implicitly included in the reciprocal space sum.
 
-        threads.syncThreads();
         const int groupSize = max(1, numberOfAtoms/(10*numThreads));
         while (true) {
-            int start = atomicCounter.fetch_add(groupSize);
+            int start = atomicCounter2.fetch_add(groupSize);
             if (start >= numberOfAtoms)
                 break;
             int end = min(start+groupSize, numberOfAtoms);

platforms/cuda/src/CudaKernels.cpp

@@ -1024,7 +1024,7 @@ void CudaCalcNonbondedForceKernel::initialize(const System& system, const Nonbon
         replacements["CHARGE1"] = prefix+"charge1";
         replacements["CHARGE2"] = prefix+"charge2";
     }
-    if (hasCoulomb)
+    if (hasCoulomb && !usePosqCharges)
         cu.getNonbondedUtilities().addParameter(CudaNonbondedUtilities::ParameterInfo(prefix+"charge", "real", 1, charges.getElementSize(), charges.getDevicePointer()));
     sigmaEpsilon.initialize<float2>(cu, cu.getPaddedNumAtoms(), "sigmaEpsilon");
     if (hasLJ) {

platforms/opencl/src/OpenCLKernels.cpp

@@ -968,7 +968,7 @@ void OpenCLCalcNonbondedForceKernel::initialize(const System& system, const Nonb
         replacements["CHARGE1"] = prefix+"charge1";
         replacements["CHARGE2"] = prefix+"charge2";
     }
-    if (hasCoulomb)
+    if (hasCoulomb && !usePosqCharges)
         cl.getNonbondedUtilities().addParameter(OpenCLNonbondedUtilities::ParameterInfo(prefix+"charge", "real", 1, charges.getElementSize(), charges.getDeviceBuffer()));
     sigmaEpsilon.initialize<mm_float2>(cl, cl.getPaddedNumAtoms(), "sigmaEpsilon");
     if (hasLJ) {