Conflict resolution in TestSplineFilter.cpp

.azure-pipelines/azure-pipelines-windows.yml

@@ -2,9 +2,9 @@ jobs:
 
   # Configure, build, install, and test job
   - job: 'windows_build'
-    displayName: 'Windows VS2015'
+    displayName: 'Windows VS2017'
     pool:
-      vmImage: 'vs2015-win2012r2'
+      vmImage: 'vs2017-win2016'
     timeoutInMinutes: 360
     variables:
       llvm.version: '7.0.1'
@@ -54,7 +54,7 @@ jobs:
       # Download OpenCL Headers and build the ICD loader
       - script: |
           setlocal EnableDelayedExpansion
-          call "C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\vcvarsall.bat" x86_amd64
+          call "C:\Program Files (x86)\Microsoft Visual Studio\2017\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" x64
           mkdir opencl
           cd opencl
           wget https://www.khronos.org/registry/cl/specs/opencl-icd-1.2.11.0.tgz -O opencl-icd-1.2.11.0.tgz
@@ -67,7 +67,9 @@ jobs:
           move .\OpenCL-Headers-master\CL\*.h .\inc\CL\
           mkdir lib > $null
           cd lib
-          cmake -G Ninja ..
+          cmake -G Ninja .. ^
+                -DCMAKE_CXX_COMPILER=cl.exe ^
+                -DCMAKE_C_COMPILER=cl.exe
           cmake --build . ^
                 -- -j %NUMBER_OF_PROCESSORS%
         displayName: "Download and install OpenCL"
@@ -75,12 +77,14 @@ jobs:
       # Configure
       - script: |
           setlocal EnableDelayedExpansion
-          call "C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\vcvarsall.bat" x86_amd64
+          call "C:\Program Files (x86)\Microsoft Visual Studio\2017\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" x64
           mkdir build & cd build
           cmake -G Ninja ^
                 -DOPENCL_INCLUDE_DIR=$(Pipeline.Workspace)/opencl/inc ^
                 -DOPENCL_LIBRARY=$(Pipeline.Workspace)/opencl/lib/OpenCL.lib ^
                 -DCMAKE_BUILD_TYPE=$(cmake.build.type) ^
+                -DCMAKE_CXX_COMPILER=cl.exe ^
+                -DCMAKE_C_COMPILER=cl.exe ^
                 -DCMAKE_INSTALL_PREFIX=../install ^
                 -DOPENMM_BUILD_EXAMPLES=OFF ^
                 -DOPENMM_BUILD_OPENCL_TESTS=OFF ^
@@ -89,7 +93,7 @@ jobs:
         workingDirectory: $(Build.BinariesDirectory)
       # Build
       - script: |
-          call "C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\vcvarsall.bat" x86_amd64
+          call "C:\Program Files (x86)\Microsoft Visual Studio\2017\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" x64
           cmake --build . ^
                 --config $(cmake.build.type) ^
                 -- -j %NUMBER_OF_PROCESSORS%

.travis.yml

@@ -3,7 +3,6 @@ language: python
 addons:
   apt:
     packages:
-      - gromacs
       - doxygen
       - python-numpy
       - python-scipy
@@ -104,6 +103,24 @@ jobs:
            CXX=$CCACHE/g++
            CMAKE_FLAGS=""
 
+    - sudo: required
+      dist: bionic
+      python: "3.8"
+      name: "PPC"
+      arch: ppc64le
+      env: OPENCL=false
+           CUDA=false
+           CMAKE_FLAGS=""
+
+    - sudo: required
+      dist: bionic
+      python: "3.8"
+      name: "ARM"
+      arch: arm64
+      env: OPENCL=false
+           CUDA=false
+           CMAKE_FLAGS=""
+
 before_install:
   - START_TIME=$(date +%s)
   - wget http://anaconda.org/omnia/ccache/3.2.4/download/${TRAVIS_OS_NAME}-64/ccache-3.2.4-0.tar.bz2
@@ -114,10 +131,21 @@ before_install:
       brew install -y https://raw.githubusercontent.com/Homebrew/homebrew-core/5b680fb58fedfb00cd07a7f69f5a621bb9240f3b/Formula/doxygen.rb;
       sudo pip install -U pytest --ignore-installed six;
     fi
-  - if [[ "${TRAVIS_OS_NAME}" == "linux" ]]; then
-      CMAKE_URL="http://cmake.org/files/v3.7/cmake-3.7.2-Linux-x86_64.tar.gz";
-      mkdir $HOME/cmake && travis_retry wget --no-check-certificate --quiet -O - ${CMAKE_URL} | tar --strip-components=1 -xz -C $HOME/cmake;
-      export PATH=${HOME}/cmake/bin:${PATH};
+  # The cmake version installed by apt on ARM and PPC is very old,
+  # so download a newer version.
+  - if [[ "${TRAVIS_CPU_ARCH}" == "ppc64le" ]]; then
+      sudo apt-get install libuv1 rhash libstdc++6;
+      wget https://anaconda.org/conda-forge/cmake/3.17.0/download/linux-ppc64le/cmake-3.17.0-hfb1cb51_0.tar.bz2;
+      mkdir $HOME/cmake;
+      tar -xjvf cmake-3.17.0-hfb1cb51_0.tar.bz2 -C $HOME/cmake;
+      export PATH=$HOME/cmake/bin:$PATH;
+    fi
+  - if [[ "${TRAVIS_CPU_ARCH}" == "arm64" ]]; then
+      sudo apt-get install libuv1 rhash libstdc++6;
+      wget https://anaconda.org/conda-forge/cmake/3.17.0/download/linux-aarch64/cmake-3.17.0-h28c56e5_0.tar.bz2;
+      mkdir $HOME/cmake;
+      tar -xjvf cmake-3.17.0-h28c56e5_0.tar.bz2 -C $HOME/cmake;
+      export PATH=$HOME/cmake/bin:$PATH;
     fi
   - if [[ "$OPENCL" == "true" ]]; then
       wget http://s3.amazonaws.com/omnia-ci/AMD-APP-SDKInstaller-v3.0.130.135-GA-linux64.tar.bz2;
@@ -132,17 +160,20 @@ before_install:
       ${AMDAPPSDK}/bin/x86_64/clinfo;
       sudo apt-get install -y libgl1-mesa-dev;
     fi
-  # Install SWIG and Cython for Python wrappers. However, testing CUDA and
-  # OpenCL, we skip the Python wrapper for speed. We're not using anaconda
-  # python, but this is a fast way to get an apparently functional precompiled
-  # build of swig that's more modern than what's in apt.
-  - if [[ "$OPENCL" == "false" && "$CUDA" == "false" && "$TRAVIS_OS_NAME" == "linux" ]]; then
+  # Install packages needed for Python: SWIG, Cython, and Gromacs (used by some tests).
+  # We do this differently on different platforms.  Possibly some of this could be unified.
+  - if [[ "$OPENCL" == "false" && "$CUDA" == "false" && "$TRAVIS_OS_NAME" == "linux" && "${TRAVIS_CPU_ARCH}" != "ppc64le" && "${TRAVIS_CPU_ARCH}" != "arm64" ]]; then
       wget http://anaconda.org/omnia/swig/3.0.7/download/linux-64/swig-3.0.7-0.tar.bz2;
       mkdir $HOME/swig;
       tar -xjvf swig-3.0.7-0.tar.bz2 -C $HOME/swig;
       export PATH=$HOME/swig/bin:$PATH;
       export SWIG_LIB=$HOME/swig/share/swig/3.0.7;
       pip install cython;
+      sudo apt-get install gromacs;
+    fi
+  - if [[ "${TRAVIS_CPU_ARCH}" == "ppc64le" || "${TRAVIS_CPU_ARCH}" == "arm64" ]]; then
+      sudo apt-get install swig;
+      pip install cython;
     fi
   - if [[ "$OPENCL" == "false" && "$CUDA" == "false" && "$TRAVIS_OS_NAME" == "osx" ]]; then
       wget http://anaconda.org/omnia/swig/3.0.7/download/osx-64/swig-3.0.7-0.tar.bz2;

CMakeLists.txt

@@ -41,11 +41,15 @@ if ("${TARGET_ARCH}" MATCHES "x86_64|i386")
 endif()
 if ("${TARGET_ARCH}" MATCHES "arm")
     set(ARM ON)
-    add_compile_definitions(__ARM__=1)
+    if ("${TARGET_ARCH}" MATCHES "armv8")
+        add_definitions(-D__ARM64__=1)
+    else()
+        add_definitions(-D__ARM__=1)
+    endif()
 endif()
 if ("${TARGET_ARCH}" MATCHES "ppc")
     set(PPC ON)
-    add_compile_definitions(__PPC__=1)
+    add_definitions(-D__PPC__=1)
 endif()
 
 # Where to install

cmake_modules/TargetArch.cmake

@@ -24,19 +24,40 @@
 # OF THE POSSIBILITY OF SUCH DAMAGE.
 
 # Based on the Qt 5 processor detection code, so should be very accurate
-# https://qt.gitorious.org/qt/qtbase/blobs/master/src/corelib/global/qprocessordetection.h
-# Currently handles arm (v5, v6, v7), x86 (32/64), ia64, and ppc (32/64)
+# https://github.com/qt/qtbase/blob/9a6a847/src/corelib/global/qprocessordetection.h
+# Currently handles arm / aarch64 (v5, v6, v7, v8), x86 (32/64), ia64, and ppc (32/64)
 
 # Regarding POWER/PowerPC, just as is noted in the Qt source,
 # "There are many more known variants/revisions that we do not handle/detect."
 
 set(archdetect_c_code "
-#if defined(__arm__) || defined(__TARGET_ARCH_ARM)
-    #if defined(__ARM_ARCH_7__) \\
+#define _STR(x) #x
+#define STR(x) _STR(x)
+
+#if defined(__arm__) || defined(__TARGET_ARCH_ARM) || defined(_M_ARM) || \\
+    defined(_M_ARM64) || defined(__aarch64__) || defined(__ARM64__)
+
+    #if defined(__ARM_ARCH) && __ARM_ARCH > 1
+        #pragma message \"cmake_ARCH armv\" STR(__ARM_ARCH)
+        #error
+    #elif defined(__TARGET_ARCH_ARM) && __TARGET_ARCH_ARM > 1
+        #pragma message \"cmake_ARCH armv\" STR(__TARGET_ARCH_ARM)
+        #error
+    #elif defined(_M_ARM) && _M_ARM > 1
+        #error cmake_ARCH arm ## __M_ARM
+    #elif defined(__ARM64_ARCH_8__) \\
+        || defined(__aarch64__) \\
+        || defined(__ARMv8__) \\
+        || defined(__ARMv8_A__) \\
+        || defined(_M_ARM64)
+        #error cmake_ARCH armv8
+    #elif defined(__ARM_ARCH_7__) \\
         || defined(__ARM_ARCH_7A__) \\
         || defined(__ARM_ARCH_7R__) \\
         || defined(__ARM_ARCH_7M__) \\
-        || (defined(__TARGET_ARCH_ARM) && __TARGET_ARCH_ARM-0 >= 7)
+        || defined(__ARM_ARCH_7S__) \\
+        || defined(_ARM_ARCH_7) \\
+        || defined(__CORE_CORTEXA__)
         #error cmake_ARCH armv7
     #elif defined(__ARM_ARCH_6__) \\
         || defined(__ARM_ARCH_6J__) \\
@@ -44,11 +65,10 @@ set(archdetect_c_code "
         || defined(__ARM_ARCH_6Z__) \\
         || defined(__ARM_ARCH_6K__) \\
         || defined(__ARM_ARCH_6ZK__) \\
-        || defined(__ARM_ARCH_6M__) \\
-        || (defined(__TARGET_ARCH_ARM) && __TARGET_ARCH_ARM-0 >= 6)
+        || defined(__ARM_ARCH_6M__)
         #error cmake_ARCH armv6
     #elif defined(__ARM_ARCH_5TEJ__) \\
-        || (defined(__TARGET_ARCH_ARM) && __TARGET_ARCH_ARM-0 >= 5)
+        || defined(__ARM_ARCH_5TE__)
         #error cmake_ARCH armv5
     #else
         #error cmake_ARCH arm

devtools/forcefield-scripts/processAmberForceField.py

@@ -272,7 +272,7 @@ removeType = [False]*len(types)
 for res in residueAtoms:
     if res not in residueBonds:
         continue
-    atomBonds = [[] for atom in residueAtoms[res]]
+    atomBonds = [[] for _ in residueAtoms[res]]
     for bond in residueBonds[res]:
         atomBonds[bond[0]].append(bond[1])
         atomBonds[bond[1]].append(bond[0])
@@ -361,7 +361,7 @@ for tor in reversed(torsions):
     tag = "  <Proper class1=\"%s\" class2=\"%s\" class3=\"%s\" class4=\"%s\"" % signature
     i = 4
     while i < len(tor):
-        index = i/3
+        index = i//3
         periodicity = int(float(tor[i+2]))
         phase = float(tor[i+1])*math.pi/180.0
         k = tor[i]*4.184
@@ -380,7 +380,7 @@ for tor in reversed(impropers):
     tag = "  <Improper class1=\"%s\" class2=\"%s\" class3=\"%s\" class4=\"%s\"" % signature
     i = 4
     while i < len(tor):
-        index = i/3
+        index = i//3
         periodicity = int(float(tor[i+2]))
         phase = float(tor[i+1])*math.pi/180.0
         k = float(tor[i])*4.184

devtools/forcefield-scripts/processCharmmForceField.py

@@ -23,7 +23,7 @@ nbfixes = {}
 
 def getFieldPairs(fields):
     pairs = []
-    for i in range(len(fields)/2):
+    for i in range(len(fields)//2):
         pairs.append((fields[2*i], fields[2*i+1]))
     return pairs
 
@@ -361,7 +361,7 @@ print(' <CMAPTorsionForce>')
 for values in sorted(uniqueCmaps, key=lambda x: uniqueCmaps[x]):
     print('  <Map>')
     size = int(math.sqrt(len(values)))
-    shift = size/2
+    shift = size//2
     scale = kilocalories_per_mole.conversion_factor_to(kilojoules_per_mole)
     # Convert the ordering from the one used by CHARMM to the one used by OpenMM.
     reordered = [0]*len(values)

docs-source/usersguide/application.rst

@@ -1123,6 +1123,13 @@ algorithm\ :cite:`Tuckerman1992`.  This allows some forces in the system to be e
 frequently than others.  For details on how to use it, consult the API
 documentation.
 
+Multiple Time Step Langevin Integrator
+--------------------------------------
+
+:class:`MTSLangevinIntegrator` is similar to :class:`MTSIntegrator`, but it uses
+the Langevin method to perform constant temperature dynamics.  For details on
+how to use it, consult the API documentation.
+
 Compound Integrator
 -------------------
 

libraries/irrxml/include/irrTypes.h

@@ -2,6 +2,8 @@
 // This file is part of the "Irrlicht Engine".
 // For conditions of distribution and use, see copyright notice in irrlicht.h
 
+// NOTE: This file has been changed from the orginal source distribution.
+
 #ifndef __IRR_TYPES_H_INCLUDED__
 #define __IRR_TYPES_H_INCLUDED__
 
@@ -80,7 +82,8 @@ typedef unsigned short wchar_t;
 
 //! define a break macro for debugging only in Win32 mode.
 #if defined(WIN32) && defined(_MSC_VER) && defined(_DEBUG)
-#define _IRR_DEBUG_BREAK_IF( _CONDITION_ ) if (_CONDITION_) {_asm int 3}
+// NOTE: This line has been changed and originally used '_asm int 3' to break into the debugger.
+#define _IRR_DEBUG_BREAK_IF( _CONDITION_ ) if (_CONDITION_) { __debugbreak(); }
 #else 
 #define _IRR_DEBUG_BREAK_IF( _CONDITION_ )
 #endif

libraries/jama/include/jama_cholesky.h

@@ -191,6 +191,9 @@ Array2D<Real> Cholesky<Real>::solve(const Array2D<Real> &B)
 	if (B.dim1() != n)
 		return Array2D<Real>();
 
+	if (!isspd)
+	    return Arran2D<Real>();
+
 
 	Array2D<Real> X = B.copy();
 	int nx = B.dim2();
@@ -224,9 +227,9 @@ Array2D<Real> Cholesky<Real>::solve(const Array2D<Real> &B)
 
 
       // Solve L*y = b;
-  	  for (int j=0; j< nx; j++)
+      for (int k = 0; k < n; k++)
 	  {
-      	for (int k = 0; k < n; k++) 
+  	    for (int j=0; j< nx; j++)
 		{
 			for (int i = 0; i < k; i++) 
                X[k][j] -= X[i][j]*L_[k][i];
@@ -235,9 +238,9 @@ Array2D<Real> Cholesky<Real>::solve(const Array2D<Real> &B)
       }
 
       // Solve L'*X = Y;
-     for (int j=0; j<nx; j++)
+     for (int k = n-1; k >= 0; k--)
 	 {
-      	for (int k = n-1; k >= 0; k--) 
+        for (int j=0; j<nx; j++)
 	  	{
          	for (int i = k+1; i < n; i++) 
                X[k][j] -= X[i][j]*L_[i][k];

libraries/jama/include/jama_eig.h

@@ -663,21 +663,20 @@ class Eigenvalue
             // Double QR step involving rows l:n and columns m:n
    
             for (int k = m; k <= n-1; k++) {
-               int notlast = (k != n-1);
+               bool notlast = (k != n-1);
                if (k != m) {
                   p = H[k][k-1];
                   q = H[k+1][k-1];
                   r = (notlast ? H[k+2][k-1] : 0.0);
                   x = abs(p) + abs(q) + abs(r);
-                  if (x != 0.0) {
-                     p = p / x;
-                     q = q / x;
-                     r = r / x;
+                  if (x == 0.0) {
+                     continue;
                   }
+                  p = p / x;
+                  q = q / x;
+                  r = r / x;
                }
-               if (x == 0.0) {
-                  break;
-               }
+
                s = sqrt(p * p + q * q + r * r);
                if (p < 0) {
                   s = -s;

libraries/jama/include/jama_svd.h

@@ -12,6 +12,8 @@
 // for min(), max() below
 #include <cmath>
 // for abs() below
+#include <limits>
+// for numeric_limits
 
 using namespace TNT;
 using namespace std;
@@ -247,6 +249,7 @@ class SVD
       int pp = p-1;
       int iter = 0;
       Real eps(pow(2.0,-52.0));
+      Real tiny = numeric_limits<Real>::min();
       while (p > 0) {
          int k=0;
 		 int kase=0;
@@ -267,7 +270,7 @@ class SVD
             if (k == -1) {
                break;
             }
-            if (abs(e[k]) <= eps*(abs(s[k]) + abs(s[k+1]))) {
+            if (abs(e[k]) <= tiny + eps*(abs(s[k]) + abs(s[k+1]))) {
                e[k] = 0.0;
                break;
             }
@@ -282,7 +285,7 @@ class SVD
                }
                Real t( (ks != p ? abs(e[ks]) : 0.) + 
                           (ks != k+1 ? abs(e[ks-1]) : 0.));
-               if (abs(s[ks]) <= eps*t)  {
+               if (abs(s[ks]) <= tiny + eps*t)  {
                   s[ks] = 0.0;
                   break;
                }

libraries/lepton/src/MSVC_erfc.h

@@ -8,9 +8,13 @@
  * (VC11 has _MSC_VER=1700).
  */
 
-#if defined(_MSC_VER) 
+#if defined(_MSC_VER) || defined(__MINGW32__)
+#if !defined(M_PI)
 #define M_PI 3.14159265358979323846264338327950288
+#endif
+#endif
 
+#if defined(_MSC_VER)
 #if _MSC_VER <= 1700 // 1700 is VC11, 1800 is VC12 
 /***************************
 *   erf.cpp

libraries/vecmath/src/vecmath.cpp

-#if defined(__ANDROID__)
+#if defined(__ARM__) || defined(__ARM64__)
     #include "neon_mathfun.h"
 #else
     #if !defined(__PNACL__)

olla/include/openmm/kernels.h

@@ -1063,16 +1063,16 @@ public:
 /**
  * This kernel is invoked by NoseHooverIntegrator to take one time step.
  */
-class IntegrateVelocityVerletStepKernel : public KernelImpl {
+class IntegrateNoseHooverStepKernel : public KernelImpl {
 public:
     static std::string Name() {
-        return "IntegrateVelocityVerletStep";
+        return "IntegrateNoseHooverStep";
     }
-    IntegrateVelocityVerletStepKernel(std::string name, const Platform& platform) : KernelImpl(name, platform) {
+    IntegrateNoseHooverStepKernel(std::string name, const Platform& platform) : KernelImpl(name, platform) {
     }
     /**
      * Initialize the kernel.
-     * 
+     *
      * @param system     the System this kernel will be applied to
      * @param integrator the NoseHooverIntegrator this kernel will be used for
      */
@@ -1093,6 +1093,49 @@ public:
      * @param integrator the NoseHooverIntegrator this kernel is being used for
      */
     virtual double computeKineticEnergy(ContextImpl& context, const NoseHooverIntegrator& integrator) = 0;
+    /**
+     * Execute the kernel that propagates the Nose Hoover chain and determines the velocity scale factor.
+     *
+     * @param context  the context in which to execute this kernel
+     * @param noseHooverChain the object describing the chain to be propagated.
+     * @param kineticEnergy the {center of mass, relative} kineticEnergies of the particles being thermostated by this chain.
+     * @param timeStep the time step used by the integrator.
+     * @return the velocity scale factor to apply to the particles associated with this heat bath.
+     */
+    virtual std::pair<double, double> propagateChain(ContextImpl& context, const NoseHooverChain &noseHooverChain, std::pair<double, double> kineticEnergy, double timeStep) = 0;
+    /**
+     * Execute the kernal that computes the total (kinetic + potential) heat bath energy.
+     *
+     * @param context the context in which to execute this kernel
+     * @param noseHooverChain the chain whose energy is to be determined.
+     * @return the total heat bath energy.
+     */
+    virtual double computeHeatBathEnergy(ContextImpl& context, const NoseHooverChain &noseHooverChain) = 0;
+    /**
+     * Execute the kernel that computes the kinetic energy for a subset of atoms,
+     * or the relative kinetic energy of Drude particles with respect to their parent atoms
+     *
+     * @param context the context in which to execute this kernel
+     * @param noseHooverChain the chain whose energy is to be determined.
+     * @param downloadValue whether the computed value should be downloaded and returned.
+     */
+    virtual std::pair<double, double> computeMaskedKineticEnergy(ContextImpl& context, const NoseHooverChain &noseHooverChain, bool downloadValue) = 0;
+    /**
+     * Execute the kernel that scales the velocities of particles associated with a nose hoover chain
+     *
+     * @param context the context in which to execute this kernel
+     * @param noseHooverChain the chain whose energy is to be determined.
+     * @param scaleFactor the multiplicative factor by which {absolute, relative} velocities are scaled.
+     */
+    virtual void scaleVelocities(ContextImpl& context, const NoseHooverChain &noseHooverChain, std::pair<double, double> scaleFactor) = 0;
+    /**
+     * Write the chain states to a checkpoint.
+     */
+    virtual void createCheckpoint(ContextImpl& context, std::ostream& stream) const = 0;
+    /**
+     * Load the chain states from a checkpoint.
+     */
+    virtual void loadCheckpoint(ContextImpl& context, std::istream& stream) = 0;
 };
 
 /**
@@ -1360,58 +1403,6 @@ public:
     virtual void execute(ContextImpl& context) = 0;
 };
 
-/**
- * This kernel is invoked by NoseHooverChainThermostat at the beginning and end of each time step
- * to update the Nose-Hoover chain.
- */
-class NoseHooverChainKernel : public KernelImpl {
-public:
-    static std::string Name() {
-        return "NoseHooverChain";
-    }
-    NoseHooverChainKernel(std::string name, const Platform& platform) : KernelImpl(name, platform) {
-    }
-    /**
-     * Initialize the kernel.
-     */
-    virtual void initialize() = 0;
-    /**
-     * Execute the kernel that propagates the Nose Hoover chain and determines the velocity scale factor.
-     * 
-     * @param context  the context in which to execute this kernel
-     * @param noseHooverChain the object describing the chain to be propagated.
-     * @param kineticEnergy the {center of mass, relative} kineticEnergies of the particles being thermostated by this chain.
-     * @param timeStep the time step used by the integrator.
-     * @return the velocity scale factor to apply to the particles associated with this heat bath.
-     */
-    virtual std::pair<double, double> propagateChain(ContextImpl& context, const NoseHooverChain &noseHooverChain, std::pair<double, double> kineticEnergy, double timeStep) = 0;
-    /**
-     * Execute the kernal that computes the total (kinetic + potential) heat bath energy.
-     *
-     * @param context the context in which to execute this kernel
-     * @param noseHooverChain the chain whose energy is to be determined.
-     * @return the total heat bath energy.
-     */
-    virtual double computeHeatBathEnergy(ContextImpl& context, const NoseHooverChain &noseHooverChain) = 0;
-    /**
-     * Execute the kernel that computes the kinetic energy for a subset of atoms,
-     * or the relative kinetic energy of Drude particles with respect to their parent atoms
-     *
-     * @param context the context in which to execute this kernel
-     * @param noseHooverChain the chain whose energy is to be determined.
-     * @param downloadValue whether the computed value should be downloaded and returned.
-     */
-    virtual std::pair<double, double> computeMaskedKineticEnergy(ContextImpl& context, const NoseHooverChain &noseHooverChain, bool downloadValue) = 0;
-    /**
-     * Execute the kernel that scales the velocities of particles associated with a nose hoover chain
-     *
-     * @param context the context in which to execute this kernel
-     * @param noseHooverChain the chain whose energy is to be determined.
-     * @param scaleFactor the multiplicative factor by which {absolute, relative} velocities are scaled.
-     */
-    virtual void scaleVelocities(ContextImpl& context, const NoseHooverChain &noseHooverChain, std::pair<double, double> scaleFactor) = 0;
-};
-
 /**
  * This kernel is invoked by MonteCarloBarostat to adjust the periodic box volume
  */

openmmapi/include/openmm/CompoundIntegrator.h

@@ -9,7 +9,7 @@
  * Biological Structures at Stanford, funded under the NIH Roadmap for        *
  * Medical Research, grant U54 GM072970. See https://simtk.org.               *
  *                                                                            *
- * Portions copyright (c) 2015-2016 Stanford University and the Authors.      *
+ * Portions copyright (c) 2015-2020 Stanford University and the Authors.      *
  * Authors: Peter Eastman                                                     *
  * Contributors:                                                              *
  *                                                                            *
@@ -194,6 +194,16 @@ protected:
     double getVelocityTimeOffset() const {
         return getIntegrator(0).getVelocityTimeOffset();
     }
+    /**
+     * This is called while writing checkpoints.  It gives the integrator a chance to write
+     * its own data.
+     */
+    void createCheckpoint(std::ostream& stream) const;
+    /**
+     * This is called while loading a checkpoint.  The integrator should read in whatever
+     * data it wrote in createCheckpoint() and update its internal state accordingly.
+     */
+    void loadCheckpoint(std::istream& stream);
 private:
     int currentIntegrator;
     std::vector<Integrator*> integrators;

openmmapi/include/openmm/CustomIntegrator.h

@@ -668,6 +668,16 @@ protected:
      * Get whether computeKineticEnergy() expects forces to have been computed.
      */
     bool kineticEnergyRequiresForce() const;
+    /**
+     * This is called while writing checkpoints.  It gives the integrator a chance to write
+     * its own data.
+     */
+    void createCheckpoint(std::ostream& stream) const;
+    /**
+     * This is called while loading a checkpoint.  The integrator should read in whatever
+     * data it wrote in createCheckpoint() and update its internal state accordingly.
+     */
+    void loadCheckpoint(std::istream& stream);
 private:
     class ComputationInfo;
     class FunctionInfo;

openmmapi/include/openmm/Integrator.h

@@ -34,6 +34,7 @@
 
 #include "State.h"
 #include "Vec3.h"
+#include <iosfwd>
 #include <map>
 #include <vector>
 #include "internal/windowsExport.h"
@@ -151,6 +152,18 @@ protected:
     virtual double getVelocityTimeOffset() const {
         return 0.0;
     }
+    /**
+     * This is called while writing checkpoints.  It gives the integrator a chance to write
+     * its own data.  The default implementation does nothing.
+     */
+    virtual void createCheckpoint(std::ostream& stream) const {
+    }
+    /**
+     * This is called while loading a checkpoint.  The integrator should read in whatever
+     * data it wrote in createCheckpoint() and update its internal state accordingly.
+     */
+    virtual void loadCheckpoint(std::istream& stream) {
+    }
 private:
     double stepSize, constraintTol;
 };

openmmapi/include/openmm/NoseHooverChain.h

@@ -72,7 +72,7 @@ public:
      *                                   interact with this chain
      * @param chainLength                the length of (number of particles in) this heat bath
      * @param numMTS                     the number of multi time steps used to propagate this chain
-     * @param numYoshidaSuzuki           the number of Yoshida Suzuki steps used to propagate this chain (1, 3, or 5).
+     * @param numYoshidaSuzuki           the number of Yoshida Suzuki steps used to propagate this chain (1, 3, 5, or 7).
      * @param chainID                    the chain id used to distinguish this Nose-Hoover chain from others that may
      *                                          be used to control a different set of particles, e.g. for Drude oscillators
      * @param thermostatedAtoms                 the list of atoms to be handled by this thermostat

openmmapi/include/openmm/NoseHooverIntegrator.h

@@ -45,13 +45,14 @@ namespace OpenMM {
 class System;
 /**
  * This is an Integrator which simulates a System using one or more Nose Hoover chain
- * thermostats, using the velocity Verlet propagation algorithm.
+ * thermostats, using the "middle" leapfrog propagation algorithm described in
+ * J. Phys. Chem. A 2019, 123, 6056-6079.
  */
 
 class OPENMM_EXPORT NoseHooverIntegrator : public Integrator {
 public:
     /**
-     * Create a NoseHooverIntegrator.  This version creates a bare velocity Verlet integrator
+     * Create a NoseHooverIntegrator.  This version creates a bare leapfrog integrator
      * with no thermostats; any thermostats should be added by calling addThermostat.
      * 
      * @param stepSize the step size with which to integrate the system (in picoseconds)
@@ -66,12 +67,12 @@ public:
      * @param chainLength the number of beads in the Nose-Hoover chain.
      * @param numMTS the number of step in the  multiple time step chain propagation algorithm.
      * @param numYoshidaSuzuki the number of terms in the Yoshida-Suzuki multi time step decomposition
-     *        used in the chain propagation algorithm (must be 1, 3, or 5).
+     *        used in the chain propagation algorithm (must be 1, 3, 5, or 7).
      */
     explicit NoseHooverIntegrator(double temperature, double collisionFrequency, double stepSize,
-                                  int chainLength = 3, int numMTS = 3, int numYoshidaSuzuki = 3);
+                                  int chainLength = 3, int numMTS = 3, int numYoshidaSuzuki = 7);
 
-    virtual ~NoseHooverIntegrator();
+    ~NoseHooverIntegrator();
    /**
      * Advance a simulation through time by taking a series of time steps.
      * 
@@ -86,7 +87,7 @@ public:
      * @param chainLength the number of beads in the Nose-Hoover chain
      * @param numMTS the number of step in the  multiple time step chain propagation algorithm.
      * @param numYoshidaSuzuki the number of terms in the Yoshida-Suzuki multi time step decomposition
-     *        used in the chain propagation algorithm (must be 1, 3, or 5).
+     *        used in the chain propagation algorithm (must be 1, 3, 5, or 7).
      */
      int addThermostat(double temperature, double collisionFrequency,
                        int chainLength, int numMTS, int numYoshidaSuzuki);
@@ -110,13 +111,13 @@ public:
      * @param chainLength the number of beads in the Nose-Hoover chain.
      * @param numMTS the number of step in the  multiple time step chain propagation algorithm.
      * @param numYoshidaSuzuki the number of terms in the Yoshida-Suzuki multi time step decomposition
-     *        used in the chain propagation algorithm (must be 1, 3, or 5).
+     *        used in the chain propagation algorithm (must be 1, 3, 5, or 7).
      */
      int addSubsystemThermostat(const std::vector<int>& thermostatedParticles,
                                 const std::vector< std::pair< int, int> >& thermostatedPairs,
                                 double temperature, double collisionFrequency, double relativeTemperature,
                                 double relativeCollisionFrequency,
-                                int chainLength = 3, int numMTS = 3, int numYoshidaSuzuki = 3);
+                                int chainLength = 3, int numMTS = 3, int numYoshidaSuzuki = 7);
     /**
      * Get the temperature of the i-th chain for controling absolute particle motion (in Kelvin).
      * 
@@ -194,18 +195,6 @@ public:
      * @param chainID the index of the Nose-Hoover chain thermostat (default=0).
      */
     const NoseHooverChain& getThermostat(int chainID=0) const ;
-    /**
-     * This will be called by the Context when the user modifies aspects of the context state, such
-     * as positions, velocities, or parameters.  This gives the Integrator a chance to discard cached
-     * information.  This is <i>only</i> called when the user modifies information using methods of the Context
-     * object.  It is <i>not</i> called when a ForceImpl object modifies state information in its updateContextState()
-     * method (unless the ForceImpl calls a Context method to perform the modification).
-     * 
-     * @param changed     this specifies what aspect of the Context was changed
-     */
-    virtual void stateChanged(State::DataType changed) {
-       if (State::Positions == changed) forcesAreValid = false;
-    }
     /**
      * Return false, if this integrator was set up with the 'default constructor' that thermostats the whole system,
      * true otherwise. Required for serialization.
@@ -232,15 +221,6 @@ public:
      */
     const std::vector<std::tuple<int, int, double> > & getAllThermostatedPairs() const { return allPairs; }
 protected:
-   /**
-     * Advance any Nose-Hoover chains associated with this integrator and determine
-     * scale factor for the velocities.
-     * 
-     * @param kineticEnergy  the {absolute, relative} kinetic energies of the system that the chain is thermostating
-     * @param chainID        id of the Nose-Hoover-Chain
-     * @return the scale factor to be applied to the velocities of the particles thermostated by the chain.
-     */
-    std::pair<double, double> propagateChain(std::pair<double, double> kineticEnergy, int chainID=0);
     /**
      * This will be called by the Context when it is created.  It informs the Integrator
      * of what context it will be integrating, and gives it a chance to do any necessary initialization.
@@ -256,6 +236,18 @@ protected:
      * cleanup.  It will also get called again if the application calls reinitialize() on the Context.
      */
     void cleanup();
+    /**
+     * This will be called by the Context when the user modifies aspects of the context state, such
+     * as positions, velocities, or parameters.  This gives the Integrator a chance to discard cached
+     * information.  This is <i>only</i> called when the user modifies information using methods of the Context
+     * object.  It is <i>not</i> called when a ForceImpl object modifies state information in its updateContextState()
+     * method (unless the ForceImpl calls a Context method to perform the modification).
+     * 
+     * @param changed     this specifies what aspect of the Context was changed
+     */
+    void stateChanged(State::DataType changed) {
+       if (State::Positions == changed) forcesAreValid = false;
+    }
     /**
      * Get the names of all Kernels used by this Integrator.
      */
@@ -263,17 +255,27 @@ protected:
     /**
      * Compute the kinetic energy of the system at the current time.
      */
-    virtual double computeKineticEnergy();
+    double computeKineticEnergy();
     /**
      * Computing kinetic energy for this integrator does not require forces.
      */
-    bool kineticEnergyRequiresForce() const override;
+    bool kineticEnergyRequiresForce() const;
+    /**
+     * This is called while writing checkpoints.  It gives the integrator a chance to write
+     * its own data.
+     */
+    void createCheckpoint(std::ostream& stream) const;
+    /**
+     * This is called while loading a checkpoint.  The integrator should read in whatever
+     * data it wrote in createCheckpoint() and update its internal state accordingly.
+     */
+    void loadCheckpoint(std::istream& stream);
 
     std::vector<NoseHooverChain> noseHooverChains;
     std::vector<int> allAtoms;
     std::vector<std::tuple<int, int, double> > allPairs;
     bool forcesAreValid;
-    Kernel vvKernel, nhcKernel;
+    Kernel kernel;
     bool hasSubsystemThermostats_;
     double maxPairDistance_;
 };

openmmapi/include/openmm/internal/hardware.h

@@ -93,7 +93,8 @@ static int getNumProcessors() {
 #ifdef WIN32
 #define cpuid __cpuid
 #else
-#if !defined(__ANDROID__) && !defined(__PNACL__) && !defined(__PPC__)
+#if !defined(__ANDROID__) && !defined(__PNACL__) && !defined(__PPC__) \
+    && !defined(__ARM__) && !defined(__ARM64__)
     static void cpuid(int cpuInfo[4], int infoType){
     #ifdef __LP64__
         __asm__ __volatile__ (