Merge pull request #2 from peastman/ljpme

More updates to dispersion PME

docs-source/usersguide/application.rst

@@ -916,8 +916,9 @@ Value                      Meaning
 :code:`NoCutoff`           No cutoff is applied.
 :code:`CutoffNonPeriodic`  The reaction field method is used to eliminate all interactions beyond a cutoff distance.  Not valid for AMOEBA.
 :code:`CutoffPeriodic`     The reaction field method is used to eliminate all interactions beyond a cutoff distance.  Periodic boundary conditions are applied, so each atom interacts only with the nearest periodic copy of every other atom.  Not valid for AMOEBA.
-:code:`Ewald`              Periodic boundary conditions are applied.  Ewald summation is used to compute long range interactions.  (This option is rarely used, since PME is much faster for all but the smallest systems.)  Not valid for AMOEBA.
-:code:`PME`                Periodic boundary conditions are applied.  The Particle Mesh Ewald method is used to compute long range interactions.
+:code:`Ewald`              Periodic boundary conditions are applied.  Ewald summation is used to compute long range Coulomb interactions.  (This option is rarely used, since PME is much faster for all but the smallest systems.)  Not valid for AMOEBA.
+:code:`PME`                Periodic boundary conditions are applied.  The Particle Mesh Ewald method is used to compute long range Coulomb interactions.
+:code:`LJPME`              Periodic boundary conditions are applied.  The Particle Mesh Ewald method is used to compute long range interactions for both Coulomb and Lennard-Jones.
 =========================  ===========================================================================================================================================================================================================================================
 
 
@@ -926,7 +927,7 @@ cutoff distance.  Be sure to specify units, as shown in the examples above. For
 example, :code:`nonbondedCutoff=1.5*nanometers` or
 :code:`nonbondedCutoff=12*angstroms` are legal values.
 
-When using :code:`Ewald` or :code:`PME`\ , you can optionally specify an
+When using :code:`Ewald`, :code:`PME`, or :code:`LJPME`\ , you can optionally specify an
 error tolerance for the force computation.  For example:
 ::
 

docs-source/usersguide/references.bib

@@ -517,3 +517,15 @@
    year = {2014},
    type = {Journal Article}
 }
+
+@article{Wennberg2015
+   author = {Wennberg, Christian L. and Murtola, Teemu and Páll, Szilárd and Abraham, Mark J. and Hess, Berk and Lindahl, Erik},
+   title = {Direct-Space Corrections Enable Fast and Accurate {Lorentz–Berthelot} Combination Rule {Lennard-Jones} Lattice Summation},
+   journal = {Journal of Chemical Theory and Computation},
+   volume = {11},
+   number = {12},
+   pages = {5737-5746},
+   year = {2015},
+   type = {Journal Article}
+}
+

docs-source/usersguide/theory.rst

@@ -412,7 +412,7 @@ and the number of nodes in the mesh along each dimension as
 
 
 .. math::
-   n_\mathit{mesh}=\frac{2\alpha d}{{3d}^{1/5}}
+   n_\mathit{mesh}=\frac{2\alpha d}{{3\delta}^{1/5}}
 
 
 where *d* is the width of the periodic box along that dimension.  Alternatively,
@@ -432,6 +432,38 @@ to numerical round-off error than Ewald summation.  For Platforms that do
 calculations in single precision, making :math:`\delta` too small (typically below about
 5·10\ :sup:`-5`\ ) can actually cause the error to increase.
 
+Lennard-Jones Interaction With Particle Mesh Ewald
+==================================================
+
+The PME algorithm can also be used for Lennard-Jones interactions.  Usually this
+is not necessary, since Lennard-Jones forces are short ranged, but there are
+situations (such as membrane simulations) where neglecting interactions beyond
+the cutoff can measurably affect results.
+
+For computational efficiency, certain approximations are made\ :cite:`Wennberg2015`.
+Interactions beyond the cutoff distance include only the attractive :math:`1/r^6`
+term, not the repulsive :math:`1/r^{12}` term.  Since the latter is much smaller
+than the former at long distances, this usually has negligible effect.  Also,
+the interaction between particles farther apart than the cutoff distance is
+computed using geometric combination rules:
+
+.. math::
+   \sigma=\sqrt{\sigma_1 \sigma_2}
+
+The effect of this approximation is also quite small, and it is still far more
+accurate than ignoring the interactions altogether (which is what would happen
+with PME).
+
+The formula used to compute the number of nodes along each dimension of the mesh
+is slightly different from the one used for Coulomb interactions:
+
+.. math::
+   n_\mathit{mesh}=\frac{\alpha d}{{3\delta}^{1/20}}
+
+As before, this is an empirical formula.  It will usually produce an average
+relative error in the forces less than or similar to :math:`\delta`\ , but that
+is not guaranteed.
+
 .. _gbsaobcforce:
 
 GBSAOBCForce

openmmapi/include/openmm/NonbondedForce.h

@@ -101,18 +101,18 @@ public:
          */
         CutoffPeriodic = 2,
         /**
-         * Periodic boundary conditions are used, and Ewald summation is used to compute the interaction of each particle
+         * Periodic boundary conditions are used, and Ewald summation is used to compute the Coulomb interaction of each particle
          * with all periodic copies of every other particle.
          */
         Ewald = 3,
         /**
-         * Periodic boundary conditions are used, and Particle-Mesh Ewald (PME) summation is used to compute the interaction of each particle
+         * Periodic boundary conditions are used, and Particle-Mesh Ewald (PME) summation is used to compute the Coulomb interaction of each particle
          * with all periodic copies of every other particle.
          */
         PME = 4,
         /**
          * Periodic boundary conditions are used, and Particle-Mesh Ewald (PME) summation is used to compute the interaction of each particle
-         * with all periodic copies of every other particle for both electrostatics and dispersion.  No switching is used for either interaction.
+         * with all periodic copies of every other particle for both Coulomb and Lennard-Jones.  No switching is used for either interaction.
          */
         LJPME = 5
     };
@@ -412,7 +412,8 @@ public:
     bool usesPeriodicBoundaryConditions() const {
         return nonbondedMethod == NonbondedForce::CutoffPeriodic ||
                nonbondedMethod == NonbondedForce::Ewald ||
-               nonbondedMethod == NonbondedForce::PME;
+               nonbondedMethod == NonbondedForce::PME ||
+               nonbondedMethod == NonbondedForce::LJPME;
     }
 protected:
     ForceImpl* createImpl() const;

openmmapi/src/NonbondedForce.cpp

@@ -48,7 +48,8 @@ using std::stringstream;
 using std::vector;
 
 NonbondedForce::NonbondedForce() : nonbondedMethod(NoCutoff), cutoffDistance(1.0), switchingDistance(-1.0), rfDielectric(78.3),
-        ewaldErrorTol(5e-4), alpha(0.0), useSwitchingFunction(false), useDispersionCorrection(true), recipForceGroup(-1), nx(0), ny(0), nz(0) {
+        ewaldErrorTol(5e-4), alpha(0.0), dalpha(0.0), useSwitchingFunction(false), useDispersionCorrection(true), recipForceGroup(-1),
+        nx(0), ny(0), nz(0), dnx(0), dny(0), dnz(0) {
 }
 
 NonbondedForce::NonbondedMethod NonbondedForce::getNonbondedMethod() const {

openmmapi/src/NonbondedForceImpl.cpp

@@ -90,9 +90,7 @@ void NonbondedForceImpl::initialize(ContextImpl& context) {
         exceptions[particle1].insert(particle2);
         exceptions[particle2].insert(particle1);
     }
-    if (owner.getNonbondedMethod() == NonbondedForce::CutoffPeriodic ||
-            owner.getNonbondedMethod() == NonbondedForce::Ewald ||
-            owner.getNonbondedMethod() == NonbondedForce::PME) {
+    if (owner.getNonbondedMethod() != NonbondedForce::NoCutoff && owner.getNonbondedMethod() != NonbondedForce::CutoffNonPeriodic) {
         Vec3 boxVectors[3];
         system.getDefaultPeriodicBoxVectors(boxVectors[0], boxVectors[1], boxVectors[2]);
         double cutoff = owner.getCutoffDistance();
@@ -161,12 +159,19 @@ void NonbondedForceImpl::calcPMEParameters(const System& system, const Nonbonded
         system.getDefaultPeriodicBoxVectors(boxVectors[0], boxVectors[1], boxVectors[2]);
         double tol = force.getEwaldErrorTolerance();
         alpha = (1.0/force.getCutoffDistance())*std::sqrt(-log(2.0*tol));
+        if (lj) {
+            xsize = (int) ceil(alpha*boxVectors[0][0]/(3*pow(tol, 0.05)));
+            ysize = (int) ceil(alpha*boxVectors[1][1]/(3*pow(tol, 0.05)));
+            zsize = (int) ceil(alpha*boxVectors[2][2]/(3*pow(tol, 0.05)));
+        }
+        else {
             xsize = (int) ceil(2*alpha*boxVectors[0][0]/(3*pow(tol, 0.2)));
             ysize = (int) ceil(2*alpha*boxVectors[1][1]/(3*pow(tol, 0.2)));
             zsize = (int) ceil(2*alpha*boxVectors[2][2]/(3*pow(tol, 0.2)));
-        xsize = max(xsize, 5);
-        ysize = max(ysize, 5);
-        zsize = max(zsize, 5);
+        }
+        xsize = max(xsize, 6);
+        ysize = max(ysize, 6);
+        zsize = max(zsize, 6);
     }
 }
 

platforms/cpu/tests/TestCpuDispersionPME.cpp

+/* -------------------------------------------------------------------------- *
+ *                                   OpenMM                                   *
+ * -------------------------------------------------------------------------- *
+ * This is part of the OpenMM molecular simulation toolkit originating from   *
+ * Simbios, the NIH National Center for Physics-Based Simulation of           *
+ * Biological Structures at Stanford, funded under the NIH Roadmap for        *
+ * Medical Research, grant U54 GM072970. See https://simtk.org.               *
+ *                                                                            *
+ * Portions copyright (c) 2017 Stanford University and the Authors.           *
+ * Authors: Peter Eastman                                                     *
+ * Contributors:                                                              *
+ *                                                                            *
+ * Permission is hereby granted, free of charge, to any person obtaining a    *
+ * copy of this software and associated documentation files (the "Software"), *
+ * to deal in the Software without restriction, including without limitation  *
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,   *
+ * and/or sell copies of the Software, and to permit persons to whom the      *
+ * Software is furnished to do so, subject to the following conditions:       *
+ *                                                                            *
+ * The above copyright notice and this permission notice shall be included in *
+ * all copies or substantial portions of the Software.                        *
+ *                                                                            *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR *
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,   *
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL    *
+ * THE AUTHORS, CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,    *
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR      *
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE  *
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.                                     *
+ * -------------------------------------------------------------------------- */
+
+#include "CpuTests.h"
+#include "TestDispersionPME.h"
+
+void runPlatformTests() {
+}
+

platforms/cuda/include/CudaKernels.h

@@ -599,8 +599,8 @@ class CudaCalcNonbondedForceKernel : public CalcNonbondedForceKernel {
 public:
     CudaCalcNonbondedForceKernel(std::string name, const Platform& platform, CudaContext& cu, const System& system) : CalcNonbondedForceKernel(name, platform),
             cu(cu), hasInitializedFFT(false), sigmaEpsilon(NULL), exceptionParams(NULL), cosSinSums(NULL), directPmeGrid(NULL), reciprocalPmeGrid(NULL),
-            pmeBsplineModuliX(NULL), pmeBsplineModuliY(NULL), pmeBsplineModuliZ(NULL),  pmeAtomRange(NULL), pmeAtomGridIndex(NULL),
-            pmeEnergyBuffer(NULL), sort(NULL), dispersionFft(NULL), fft(NULL), pmeio(NULL) {
+            pmeBsplineModuliX(NULL), pmeBsplineModuliY(NULL), pmeBsplineModuliZ(NULL), pmeDispersionBsplineModuliX(NULL), pmeDispersionBsplineModuliY(NULL),
+            pmeDispersionBsplineModuliZ(NULL), pmeAtomRange(NULL), pmeAtomGridIndex(NULL), pmeEnergyBuffer(NULL), sort(NULL), dispersionFft(NULL), fft(NULL), pmeio(NULL) {
     }
     ~CudaCalcNonbondedForceKernel();
     /**
@@ -672,6 +672,9 @@ private:
     CudaArray* pmeBsplineModuliX;
     CudaArray* pmeBsplineModuliY;
     CudaArray* pmeBsplineModuliZ;
+    CudaArray* pmeDispersionBsplineModuliX;
+    CudaArray* pmeDispersionBsplineModuliY;
+    CudaArray* pmeDispersionBsplineModuliZ;
     CudaArray* pmeAtomRange;
     CudaArray* pmeAtomGridIndex;
     CudaArray* pmeEnergyBuffer;

platforms/cuda/src/CudaKernels.cpp

@@ -1607,6 +1607,12 @@ CudaCalcNonbondedForceKernel::~CudaCalcNonbondedForceKernel() {
         delete pmeBsplineModuliY;
     if (pmeBsplineModuliZ != NULL)
         delete pmeBsplineModuliZ;
+    if (pmeDispersionBsplineModuliX != NULL)
+        delete pmeDispersionBsplineModuliX;
+    if (pmeDispersionBsplineModuliY != NULL)
+        delete pmeDispersionBsplineModuliY;
+    if (pmeDispersionBsplineModuliZ != NULL)
+        delete pmeDispersionBsplineModuliZ;
     if (pmeAtomRange != NULL)
         delete pmeAtomRange;
     if (pmeAtomGridIndex != NULL)
@@ -1861,6 +1867,11 @@ void CudaCalcNonbondedForceKernel::initialize(const System& system, const Nonbon
                 pmeBsplineModuliX = new CudaArray(cu, gridSizeX, elementSize, "pmeBsplineModuliX");
                 pmeBsplineModuliY = new CudaArray(cu, gridSizeY, elementSize, "pmeBsplineModuliY");
                 pmeBsplineModuliZ = new CudaArray(cu, gridSizeZ, elementSize, "pmeBsplineModuliZ");
+                if (doLJPME) {
+                    pmeDispersionBsplineModuliX = new CudaArray(cu, dispersionGridSizeX, elementSize, "pmeDispersionBsplineModuliX");
+                    pmeDispersionBsplineModuliY = new CudaArray(cu, dispersionGridSizeY, elementSize, "pmeDispersionBsplineModuliY");
+                    pmeDispersionBsplineModuliZ = new CudaArray(cu, dispersionGridSizeZ, elementSize, "pmeDispersionBsplineModuliZ");
+                }
                 pmeAtomRange = CudaArray::create<int>(cu, gridSizeX*gridSizeY*gridSizeZ+1, "pmeAtomRange");
                 pmeAtomGridIndex = CudaArray::create<int2>(cu, numParticles, "pmeAtomGridIndex");
                 int energyElementSize = (cu.getUseDoublePrecision() || cu.getUseMixedPrecision() ? sizeof(double) : sizeof(float));
@@ -1911,7 +1922,28 @@ void CudaCalcNonbondedForceKernel::initialize(const System& system, const Nonbon
                 hasInitializedFFT = true;
                 // Initialize the b-spline moduli.
 
-                int maxSize = max(max(gridSizeX, gridSizeY), gridSizeZ);
+                for (int grid = 0; grid < 2; grid++) {
+                    int xsize, ysize, zsize;
+                    CudaArray *xmoduli, *ymoduli, *zmoduli;
+                    if (grid == 0) {
+                        xsize = gridSizeX;
+                        ysize = gridSizeY;
+                        zsize = gridSizeZ;
+                        xmoduli = pmeBsplineModuliX;
+                        ymoduli = pmeBsplineModuliY;
+                        zmoduli = pmeBsplineModuliZ;
+                    }
+                    else {
+                        if (!doLJPME)
+                            continue;
+                        xsize = dispersionGridSizeX;
+                        ysize = dispersionGridSizeY;
+                        zsize = dispersionGridSizeZ;
+                        xmoduli = pmeDispersionBsplineModuliX;
+                        ymoduli = pmeDispersionBsplineModuliY;
+                        zmoduli = pmeDispersionBsplineModuliZ;
+                    }
+                    int maxSize = max(max(xsize, ysize), zsize);
                     vector<double> data(PmeOrder);
                     vector<double> ddata(PmeOrder);
                     vector<double> bsplines_data(maxSize);
@@ -1944,7 +1976,7 @@ void CudaCalcNonbondedForceKernel::initialize(const System& system, const Nonbon
                     // Evaluate the actual bspline moduli for X/Y/Z.
 
                     for(int dim = 0; dim < 3; dim++) {
-                    int ndata = (dim == 0 ? gridSizeX : dim == 1 ? gridSizeY : gridSizeZ);
+                        int ndata = (dim == 0 ? xsize : dim == 1 ? ysize : zsize);
                         vector<double> moduli(ndata);
                         for (int i = 0; i < ndata; i++) {
                             double sc = 0.0;
@@ -1961,22 +1993,23 @@ void CudaCalcNonbondedForceKernel::initialize(const System& system, const Nonbon
                                 moduli[i] = (moduli[i-1]+moduli[i+1])*0.5;
                         if (cu.getUseDoublePrecision()) {
                             if (dim == 0)
-                            pmeBsplineModuliX->upload(moduli);
+                                xmoduli->upload(moduli);
                             else if (dim == 1)
-                            pmeBsplineModuliY->upload(moduli);
+                                ymoduli->upload(moduli);
                             else
-                            pmeBsplineModuliZ->upload(moduli);
+                                zmoduli->upload(moduli);
                         }
                         else {
                             vector<float> modulif(ndata);
                             for (int i = 0; i < ndata; i++)
                                 modulif[i] = (float) moduli[i];
                             if (dim == 0)
-                            pmeBsplineModuliX->upload(modulif);
+                                xmoduli->upload(modulif);
                             else if (dim == 1)
-                            pmeBsplineModuliY->upload(modulif);
+                                ymoduli->upload(modulif);
                             else
-                            pmeBsplineModuliZ->upload(modulif);
+                                zmoduli->upload(modulif);
+                        }
                     }
                 }
             }

platforms/cuda/src/kernels/coulombLennardJones.cu

@@ -30,7 +30,7 @@
             const real dar6 = dar4*dar2;
             const real invR2 = invR*invR;
             const real expDar2 = EXP(-dar2);
-            const real2 sigExpProd = sigmaEpsilon1*sigmaEpsilon2;
+            const float2 sigExpProd = sigmaEpsilon1*sigmaEpsilon2;
             const real c6 = 64*sigExpProd.x*sigExpProd.x*sigExpProd.x*sigExpProd.y;
             const real coef = invR2*invR2*invR2*c6;
             const real eprefac = 1.0f + dar2 + 0.5f*dar4;

platforms/cuda/src/kernels/pme.cu

@@ -23,7 +23,7 @@ extern "C" __global__ void gridSpreadCharge(const real4* __restrict__ posq, real
         real4 periodicBoxSize, real4 invPeriodicBoxSize, real4 periodicBoxVecX, real4 periodicBoxVecY, real4 periodicBoxVecZ,
         real3 recipBoxVecX, real3 recipBoxVecY, real3 recipBoxVecZ, const int2* __restrict__ pmeAtomGridIndex
 #ifdef USE_LJPME
-        , const real2* __restrict__ sigmaEpsilon
+        , const float2* __restrict__ sigmaEpsilon
 #endif
         ) {
     real3 data[PME_ORDER];
@@ -68,7 +68,7 @@ extern "C" __global__ void gridSpreadCharge(const real4* __restrict__ posq, real
         // Spread the charge from this atom onto each grid point.
         
 #ifdef USE_LJPME
-        const real2 sigEps = sigmaEpsilon[atom];
+        const float2 sigEps = sigmaEpsilon[atom];
         const real charge = 8*sigEps.x*sigEps.x*sigEps.x*sigEps.y;
 #else
         const real charge = pos.w;
@@ -250,7 +250,7 @@ void gridInterpolateForce(const real4* __restrict__ posq, unsigned long long* __
         real4 periodicBoxSize, real4 invPeriodicBoxSize, real4 periodicBoxVecX, real4 periodicBoxVecY, real4 periodicBoxVecZ,
         real3 recipBoxVecX, real3 recipBoxVecY, real3 recipBoxVecZ, const int2* __restrict__ pmeAtomGridIndex
 #ifdef USE_LJPME
-        , const real2* __restrict__ sigmaEpsilon
+        , const float2* __restrict__ sigmaEpsilon
 #endif
         ) {
     real3 data[PME_ORDER];
@@ -325,7 +325,7 @@ void gridInterpolateForce(const real4* __restrict__ posq, unsigned long long* __
             }
         }
 #ifdef USE_LJPME
-        const real2 sigEps = sigmaEpsilon[atom];
+        const float2 sigEps = sigmaEpsilon[atom];
         real q = 8*sigEps.x*sigEps.x*sigEps.x*sigEps.y;
 #else
         real q = pos.w*EPSILON_FACTOR;

platforms/cuda/tests/TestCudaDispersionPME.cpp

+/* -------------------------------------------------------------------------- *
+ *                                   OpenMM                                   *
+ * -------------------------------------------------------------------------- *
+ * This is part of the OpenMM molecular simulation toolkit originating from   *
+ * Simbios, the NIH National Center for Physics-Based Simulation of           *
+ * Biological Structures at Stanford, funded under the NIH Roadmap for        *
+ * Medical Research, grant U54 GM072970. See https://simtk.org.               *
+ *                                                                            *
+ * Portions copyright (c) 2017 Stanford University and the Authors.           *
+ * Authors: Peter Eastman                                                     *
+ * Contributors:                                                              *
+ *                                                                            *
+ * Permission is hereby granted, free of charge, to any person obtaining a    *
+ * copy of this software and associated documentation files (the "Software"), *
+ * to deal in the Software without restriction, including without limitation  *
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,   *
+ * and/or sell copies of the Software, and to permit persons to whom the      *
+ * Software is furnished to do so, subject to the following conditions:       *
+ *                                                                            *
+ * The above copyright notice and this permission notice shall be included in *
+ * all copies or substantial portions of the Software.                        *
+ *                                                                            *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR *
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,   *
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL    *
+ * THE AUTHORS, CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,    *
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR      *
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE  *
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.                                     *
+ * -------------------------------------------------------------------------- */
+
+#include "CudaTests.h"
+#include "TestDispersionPME.h"
+
+void runPlatformTests() {
+}

platforms/opencl/tests/TestOpenCLDispersionPME.cpp

+/* -------------------------------------------------------------------------- *
+ *                                   OpenMM                                   *
+ * -------------------------------------------------------------------------- *
+ * This is part of the OpenMM molecular simulation toolkit originating from   *
+ * Simbios, the NIH National Center for Physics-Based Simulation of           *
+ * Biological Structures at Stanford, funded under the NIH Roadmap for        *
+ * Medical Research, grant U54 GM072970. See https://simtk.org.               *
+ *                                                                            *
+ * Portions copyright (c) 2017 Stanford University and the Authors.           *
+ * Authors: Peter Eastman                                                     *
+ * Contributors:                                                              *
+ *                                                                            *
+ * Permission is hereby granted, free of charge, to any person obtaining a    *
+ * copy of this software and associated documentation files (the "Software"), *
+ * to deal in the Software without restriction, including without limitation  *
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,   *
+ * and/or sell copies of the Software, and to permit persons to whom the      *
+ * Software is furnished to do so, subject to the following conditions:       *
+ *                                                                            *
+ * The above copyright notice and this permission notice shall be included in *
+ * all copies or substantial portions of the Software.                        *
+ *                                                                            *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR *
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,   *
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL    *
+ * THE AUTHORS, CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,    *
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR      *
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE  *
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.                                     *
+ * -------------------------------------------------------------------------- */
+
+#include "OpenCLTests.h"
+#include "TestDispersionPME.h"
+
+void runPlatformTests() {
+}

platforms/reference/tests/TestReferenceDispersionPME.cpp

+/* -------------------------------------------------------------------------- *
+ *                                   OpenMM                                   *
+ * -------------------------------------------------------------------------- *
+ * This is part of the OpenMM molecular simulation toolkit originating from   *
+ * Simbios, the NIH National Center for Physics-Based Simulation of           *
+ * Biological Structures at Stanford, funded under the NIH Roadmap for        *
+ * Medical Research, grant U54 GM072970. See https://simtk.org.               *
+ *                                                                            *
+ * Portions copyright (c) 2017 Stanford University and the Authors.           *
+ * Authors: Peter Eastman                                                     *
+ * Contributors:                                                              *
+ *                                                                            *
+ * Permission is hereby granted, free of charge, to any person obtaining a    *
+ * copy of this software and associated documentation files (the "Software"), *
+ * to deal in the Software without restriction, including without limitation  *
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,   *
+ * and/or sell copies of the Software, and to permit persons to whom the      *
+ * Software is furnished to do so, subject to the following conditions:       *
+ *                                                                            *
+ * The above copyright notice and this permission notice shall be included in *
+ * all copies or substantial portions of the Software.                        *
+ *                                                                            *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR *
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,   *
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL    *
+ * THE AUTHORS, CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,    *
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR      *
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE  *
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.                                     *
+ * -------------------------------------------------------------------------- */
+
+#include "ReferenceTests.h"
+#include "TestDispersionPME.h"
+
+void runPlatformTests() {
+}

serialization/src/NonbondedForceProxy.cpp

@@ -42,7 +42,7 @@ NonbondedForceProxy::NonbondedForceProxy() : SerializationProxy("NonbondedForce"
 }
 
 void NonbondedForceProxy::serialize(const void* object, SerializationNode& node) const {
-    node.setIntProperty("version", 1);
+    node.setIntProperty("version", 2);
     const NonbondedForce& force = *reinterpret_cast<const NonbondedForce*>(object);
     node.setIntProperty("forceGroup", force.getForceGroup());
     node.setIntProperty("method", (int) force.getNonbondedMethod());
@@ -59,6 +59,11 @@ void NonbondedForceProxy::serialize(const void* object, SerializationNode& node)
     node.setIntProperty("nx", nx);
     node.setIntProperty("ny", ny);
     node.setIntProperty("nz", nz);
+    force.getLJPMEParameters(alpha, nx, ny, nz);
+    node.setDoubleProperty("ljAlpha", alpha);
+    node.setIntProperty("ljnx", nx);
+    node.setIntProperty("ljny", ny);
+    node.setIntProperty("ljnz", nz);
     node.setIntProperty("recipForceGroup", force.getReciprocalSpaceForceGroup());
     SerializationNode& particles = node.createChildNode("Particles");
     for (int i = 0; i < force.getNumParticles(); i++) {
@@ -76,7 +81,8 @@ void NonbondedForceProxy::serialize(const void* object, SerializationNode& node)
 }
 
 void* NonbondedForceProxy::deserialize(const SerializationNode& node) const {
-    if (node.getIntProperty("version") != 1)
+    int version = node.getIntProperty("version");
+    if (version < 1 || version > 2)
         throw OpenMMException("Unsupported version number");
     NonbondedForce* force = new NonbondedForce();
     try {
@@ -93,6 +99,13 @@ void* NonbondedForceProxy::deserialize(const SerializationNode& node) const {
         int ny = node.getIntProperty("ny", 0);
         int nz = node.getIntProperty("nz", 0);
         force->setPMEParameters(alpha, nx, ny, nz);
+        if (version >= 2) {
+            alpha = node.getDoubleProperty("ljAlpha", 0.0);
+            nx = node.getIntProperty("ljnx", 0);
+            ny = node.getIntProperty("ljny", 0);
+            nz = node.getIntProperty("ljnz", 0);
+            force->setLJPMEParameters(alpha, nx, ny, nz);
+        }
         force->setReciprocalSpaceForceGroup(node.getIntProperty("recipForceGroup", -1));
         const SerializationNode& particles = node.getChildNode("Particles");
         for (int i = 0; i < (int) particles.getChildren().size(); i++) {

serialization/tests/TestSerializeNonbondedForce.cpp

@@ -53,6 +53,9 @@ void testSerialization() {
     double alpha = 0.5;
     int nx = 3, ny = 5, nz = 7;
     force.setPMEParameters(alpha, nx, ny, nz);
+    double dalpha = 0.8;
+    int dnx = 4, dny = 6, dnz = 7;
+    force.setLJPMEParameters(dalpha, dnx, dny, dnz);
     force.addParticle(1, 0.1, 0.01);
     force.addParticle(0.5, 0.2, 0.02);
     force.addParticle(-0.5, 0.3, 0.03);
@@ -84,6 +87,13 @@ void testSerialization() {
     ASSERT_EQUAL(nx, nx2);
     ASSERT_EQUAL(ny, ny2);
     ASSERT_EQUAL(nz, nz2);    
+    double dalpha2;
+    int dnx2, dny2, dnz2;
+    force2.getLJPMEParameters(dalpha2, dnx2, dny2, dnz2);
+    ASSERT_EQUAL(dalpha, dalpha2);
+    ASSERT_EQUAL(dnx, dnx2);
+    ASSERT_EQUAL(dny, dny2);
+    ASSERT_EQUAL(dnz, dnz2);    
     for (int i = 0; i < force.getNumParticles(); i++) {
         double charge1, sigma1, epsilon1;
         double charge2, sigma2, epsilon2;

wrappers/python/simtk/openmm/app/__init__.py

@@ -40,6 +40,7 @@ CutoffNonPeriodic = forcefield.CutoffNonPeriodic
 CutoffPeriodic = forcefield.CutoffPeriodic
 Ewald = forcefield.Ewald
 PME = forcefield.PME
+LJPME = forcefield.LJPME
 
 HBonds = forcefield.HBonds
 AllBonds = forcefield.AllBonds

wrappers/python/simtk/openmm/app/amberprmtopfile.py

@@ -169,7 +169,7 @@ class AmberPrmtopFile(object):
         ----------
         nonbondedMethod : object=NoCutoff
             The method to use for nonbonded interactions.  Allowed values are
-            NoCutoff, CutoffNonPeriodic, CutoffPeriodic, Ewald, or PME.
+            NoCutoff, CutoffNonPeriodic, CutoffPeriodic, Ewald, PME, or LJPME.
         nonbondedCutoff : distance=1*nanometer
             The cutoff distance to use for nonbonded interactions
         constraints : object=None
@@ -202,7 +202,7 @@ class AmberPrmtopFile(object):
             added to a hydrogen is subtracted from the heavy atom to keep their
             total mass the same.
         ewaldErrorTolerance : float=0.0005
-            The error tolerance to use if nonbondedMethod is Ewald or PME.
+            The error tolerance to use if nonbondedMethod is Ewald, PME, or LJPME.
         switchDistance : float=0*nanometers
             The distance at which the potential energy switching function is
             turned on for Lennard-Jones interactions. If the switchDistance is 0
@@ -222,10 +222,11 @@ class AmberPrmtopFile(object):
                      ff.CutoffNonPeriodic:'CutoffNonPeriodic',
                      ff.CutoffPeriodic:'CutoffPeriodic',
                      ff.Ewald:'Ewald',
-                     ff.PME:'PME'}
+                     ff.PME:'PME',
+                     ff.LJPME:'LJPME'}
         if nonbondedMethod not in methodMap:
             raise ValueError('Illegal value for nonbonded method')
-        if not self._prmtop.getIfBox() and nonbondedMethod in (ff.CutoffPeriodic, ff.Ewald, ff.PME):
+        if not self._prmtop.getIfBox() and nonbondedMethod in (ff.CutoffPeriodic, ff.Ewald, ff.PME, ff.LJPME):
             raise ValueError('Illegal nonbonded method for a non-periodic system')
         constraintMap = {None:None,
                          ff.HBonds:'h-bonds',