Skip to content

GitLab

Commit bf6d95c2 authored by John Chodera (MSKCC)'s avatar John Chodera (MSKCC)
Browse files

Merge remote-tracking branch 'upstream/master'

parents 896413aa 227c86bf
Expand all Show whitespace changes
Inline Side-by-side
Showing with 1495 additions and 1527 deletions
CMakeLists.txt
View file @ bf6d95c2
......@@ -81,20 +81,6 @@ ENDIF(${CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT})
SET(OPENMM_SOURCE_SUBDIRS . openmmapi olla libraries/jama libraries/quern libraries/lepton libraries/sfmt libraries/lbfgs libraries/hilbert libraries/csha1 platforms/reference serialization libraries/validate libraries/irrxml)
IF(WIN32)
SET(OPENMM_SOURCE_SUBDIRS ${OPENMM_SOURCE_SUBDIRS} libraries/pthreads)
ADD_CUSTOM_TARGET(PthreadsLibraries ALL)
FILE(GLOB PTHREADS_LIBRARIES libraries/pthreads/lib/*.dll libraries/pthreads/lib/*.lib)
ADD_CUSTOM_COMMAND(TARGET PthreadsLibraries COMMAND ${CMAKE_COMMAND} ARGS -E make_directory "${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_CFG_INTDIR}/")
FOREACH(lib ${PTHREADS_LIBRARIES})
ADD_CUSTOM_COMMAND(TARGET PthreadsLibraries COMMAND ${CMAKE_COMMAND} ARGS -E copy ${lib} "${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_CFG_INTDIR}/")
INSTALL(FILES ${lib} DESTINATION "lib/")
ENDFOREACH(lib)
LINK_DIRECTORIES("${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_CFG_INTDIR}")
IF(CMAKE_SIZEOF_VOID_P EQUAL 8)
SET(PTHREADS_LIB pthreadVC2_x64)
ELSE(CMAKE_SIZEOF_VOID_P EQUAL 8)
SET(PTHREADS_LIB pthreadVC2)
SET(PTHREADS_LIB_STATIC pthreadVC2_static_mt)
ENDIF(CMAKE_SIZEOF_VOID_P EQUAL 8)
ELSE(WIN32)
IF (NOT ANDROID)
SET(PTHREADS_LIB pthread)
......@@ -303,17 +289,14 @@ ENDIF (CMAKE_SYSTEM_NAME MATCHES "Linux")
IF(OPENMM_BUILD_SHARED_LIB)
ADD_LIBRARY(${SHARED_TARGET} SHARED ${SOURCE_FILES} ${SOURCE_INCLUDE_FILES} ${API_ABS_INCLUDE_FILES})
SET_TARGET_PROPERTIES(${SHARED_TARGET} PROPERTIES LINK_FLAGS "${EXTRA_LINK_FLAGS}" COMPILE_FLAGS "${EXTRA_COMPILE_FLAGS} -DOPENMM_BUILDING_SHARED_LIBRARY -DLEPTON_BUILDING_SHARED_LIBRARY -DOPENMM_VALIDATE_BUILDING_SHARED_LIBRARY")
IF(WIN32)
ADD_DEPENDENCIES(${SHARED_TARGET} PthreadsLibraries)
ENDIF(WIN32)
SET_TARGET_PROPERTIES(${SHARED_TARGET} PROPERTIES LINK_FLAGS "${EXTRA_LINK_FLAGS}" COMPILE_FLAGS "${EXTRA_COMPILE_FLAGS} -DOPENMM_BUILDING_SHARED_LIBRARY -DLEPTON_BUILDING_SHARED_LIBRARY -DOPENMM_VALIDATE_BUILDING_SHARED_LIBRARY -DPTHREAD_BUILDING_SHARED_LIBRARY")
ENDIF(OPENMM_BUILD_SHARED_LIB)
SET(OPENMM_BUILD_STATIC_LIB OFF CACHE BOOL "Whether to build static OpenMM libraries")
IF(OPENMM_BUILD_STATIC_LIB)
ADD_LIBRARY(${STATIC_TARGET} STATIC ${SOURCE_FILES} ${SOURCE_INCLUDE_FILES} ${API_ABS_INCLUDE_FILES})
SET(EXTRA_COMPILE_FLAGS "${EXTRA_COMPILE_FLAGS} -DOPENMM_USE_STATIC_LIBRARIES -DLEPTON_USE_STATIC_LIBRARIES -DPTW32_STATIC_LIB")
SET_TARGET_PROPERTIES(${STATIC_TARGET} PROPERTIES LINK_FLAGS "${EXTRA_LINK_FLAGS}" COMPILE_FLAGS "${EXTRA_COMPILE_FLAGS} -DOPENMM_BUILDING_STATIC_LIBRARY -DLEPTON_BUILDING_STATIC_LIBRARY -DOPENMMM_VALIDATE_BUILDING_STATIC_LIBRARY -DOPENMM_VALIDATE_BUILDING_STATIC_LIBRARY")
SET_TARGET_PROPERTIES(${STATIC_TARGET} PROPERTIES LINK_FLAGS "${EXTRA_LINK_FLAGS}" COMPILE_FLAGS "${EXTRA_COMPILE_FLAGS} -DOPENMM_BUILDING_STATIC_LIBRARY -DLEPTON_BUILDING_STATIC_LIBRARY -DOPENMMM_VALIDATE_BUILDING_STATIC_LIBRARY -DOPENMM_VALIDATE_BUILDING_STATIC_LIBRARY -DPTHREAD_BUILDING_STATIC_LIBRARY")
ENDIF(OPENMM_BUILD_STATIC_LIB)
IF(OPENMM_BUILD_C_AND_FORTRAN_WRAPPERS)
......
cmake_modules/FindOpenCL.cmake
View file @ bf6d95c2
......@@ -33,6 +33,7 @@ find_path(OPENCL_INCLUDE_DIR
"/usr/local/cuda"
"/usr/local/streamsdk"
"/usr"
"${CUDA_TOOLKIT_ROOT_DIR}"
PATH_SUFFIXES "include"
)
......
docs-source/usersguide/application.rst
View file @ bf6d95c2
......@@ -1121,6 +1121,25 @@ integrator is:
The parameter is the integration error tolerance (0.001), whose meaning is the
same as for the Langevin integrator.
Multiple Time Step Integrator
-----------------------------
The :class:`MTSIntegrator` class implements the rRESPA multiple time step
algorithm\ :cite:`Tuckerman1992`. This allows some forces in the system to be evaluated more
frequently than others. For details on how to use it, consult the API
documentation.
aMD Integrator
--------------
There are three different integrator types that implement variations of the
aMD\ :cite:`Hamelberg2007` accelerated sampling algorithm: :class:`AMDIntegrator`,
:class:`AMDForceGroupIntegrator`, and :class:`DualAMDIntegrator`. They
perform integration on a modified potential energy surface to allow much faster
sampling of conformations. For details on how to use them, consult the API
documentation.
Temperature Coupling
====================
......
docs-source/usersguide/references.bib
View file @ bf6d95c2
......@@ -90,6 +90,17 @@
type = {Journal Article}
}
@article{Hamelberg2007,
author={Hamelberg, Donald and de Oliveira, Cesar Augusto F. and McCammon, J. Andrew},
title={Sampling of slow diffusive conformational transitions with accelerated molecular dynamics},
journal={Journal of Chemical Physics},
volume={127},
number={15},
pages={155102},
year={2007},
type = {Journal Article}
}
@article{Hawkins1995
author = {Hawkins, Gregory D. and Cramer, Christopher J. and Truhlar, Donald G.},
title = {Pairwise solute descreening of solute charges from a dielectric medium},
......@@ -440,6 +451,17 @@
type = {Journal Article}
}
@article{Tuckerman1992,
author={Tuckerman, M. and Berne, Bruce J. and Martyna, Glenn J.},
title={Reversible multiple time scale molecular dynamics},
journal = {Journal of Chemical Physics},
volume={97},
number={3},
pages={1990-2001},
year={1992},
type = {Journal Article}
}
@article{Uberuaga2004,
author = {Blas P. Uberuaga and Marian Anghel and Arthur
F. Voter},
......
libraries/pthreads/include/pthread.h
View file @ bf6d95c2
This diff is collapsed.
libraries/pthreads/include/sched.h deleted 100644 → 0
View file @ 896413aa
/*
* Module: sched.h
*
* Purpose:
* Provides an implementation of POSIX realtime extensions
* as defined in
*
* POSIX 1003.1b-1993 (POSIX.1b)
*
* --------------------------------------------------------------------------
*
* Pthreads-win32 - POSIX Threads Library for Win32
* Copyright(C) 1998 John E. Bossom
* Copyright(C) 1999,2005 Pthreads-win32 contributors
*
* Contact Email: rpj@callisto.canberra.edu.au
*
* The current list of contributors is contained
* in the file CONTRIBUTORS included with the source
* code distribution. The list can also be seen at the
* following World Wide Web location:
* http://sources.redhat.com/pthreads-win32/contributors.html
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library in the file COPYING.LIB;
* if not, write to the Free Software Foundation, Inc.,
* 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
*/
#ifndef _SCHED_H
#define _SCHED_H
#undef PTW32_LEVEL
#if defined(_POSIX_SOURCE)
#define PTW32_LEVEL 0
/* Early POSIX */
#endif
#if defined(_POSIX_C_SOURCE) && _POSIX_C_SOURCE >= 199309
#undef PTW32_LEVEL
#define PTW32_LEVEL 1
/* Include 1b, 1c and 1d */
#endif
#if defined(INCLUDE_NP)
#undef PTW32_LEVEL
#define PTW32_LEVEL 2
/* Include Non-Portable extensions */
#endif
#define PTW32_LEVEL_MAX 3
#if !defined(PTW32_LEVEL)
#define PTW32_LEVEL PTW32_LEVEL_MAX
/* Include everything */
#endif
#if __GNUC__ && ! defined (__declspec)
# error Please upgrade your GNU compiler to one that supports __declspec.
#endif
/*
* When building the DLL code, you should define PTW32_BUILD so that
* the variables/functions are exported correctly. When using the DLL,
* do NOT define PTW32_BUILD, and then the variables/functions will
* be imported correctly.
*/
#ifndef PTW32_STATIC_LIB
# ifdef PTW32_BUILD
# define PTW32_DLLPORT __declspec (dllexport)
# else
# define PTW32_DLLPORT __declspec (dllimport)
# endif
#else
# define PTW32_DLLPORT
#endif
/*
* This is a duplicate of what is in the autoconf config.h,
* which is only used when building the pthread-win32 libraries.
*/
#ifndef PTW32_CONFIG_H
# if defined(WINCE)
# define NEED_ERRNO
# define NEED_SEM
# endif
# if defined(_UWIN) || defined(__MINGW32__)
# define HAVE_MODE_T
# endif
#endif
/*
*
*/
#if PTW32_LEVEL >= PTW32_LEVEL_MAX
#ifdef NEED_ERRNO
#include "need_errno.h"
#else
#include <errno.h>
#endif
#endif /* PTW32_LEVEL >= PTW32_LEVEL_MAX */
#if defined(__MINGW32__) || defined(_UWIN)
#if PTW32_LEVEL >= PTW32_LEVEL_MAX
/* For pid_t */
# include <sys/types.h>
/* Required by Unix 98 */
# include <time.h>
#endif /* PTW32_LEVEL >= PTW32_LEVEL_MAX */
#else
typedef int pid_t;
#endif
/* Thread scheduling policies */
enum {
SCHED_OTHER = 0,
SCHED_FIFO,
SCHED_RR,
SCHED_MIN = SCHED_OTHER,
SCHED_MAX = SCHED_RR
};
struct sched_param {
int sched_priority;
};
#ifdef __cplusplus
extern "C"
{
#endif /* __cplusplus */
PTW32_DLLPORT int __cdecl sched_yield (void);
PTW32_DLLPORT int __cdecl sched_get_priority_min (int policy);
PTW32_DLLPORT int __cdecl sched_get_priority_max (int policy);
PTW32_DLLPORT int __cdecl sched_setscheduler (pid_t pid, int policy);
PTW32_DLLPORT int __cdecl sched_getscheduler (pid_t pid);
/*
* Note that this macro returns ENOTSUP rather than
* ENOSYS as might be expected. However, returning ENOSYS
* should mean that sched_get_priority_{min,max} are
* not implemented as well as sched_rr_get_interval.
* This is not the case, since we just don't support
* round-robin scheduling. Therefore I have chosen to
* return the same value as sched_setscheduler when
* SCHED_RR is passed to it.
*/
#define sched_rr_get_interval(_pid, _interval) \
( errno = ENOTSUP, (int) -1 )
#ifdef __cplusplus
} /* End of extern "C" */
#endif /* __cplusplus */
#undef PTW32_LEVEL
#undef PTW32_LEVEL_MAX
#endif /* !_SCHED_H */
libraries/pthreads/include/semaphore.h deleted 100644 → 0
View file @ 896413aa
/*
* Module: semaphore.h
*
* Purpose:
* Semaphores aren't actually part of the PThreads standard.
* They are defined by the POSIX Standard:
*
* POSIX 1003.1b-1993 (POSIX.1b)
*
* --------------------------------------------------------------------------
*
* Pthreads-win32 - POSIX Threads Library for Win32
* Copyright(C) 1998 John E. Bossom
* Copyright(C) 1999,2005 Pthreads-win32 contributors
*
* Contact Email: rpj@callisto.canberra.edu.au
*
* The current list of contributors is contained
* in the file CONTRIBUTORS included with the source
* code distribution. The list can also be seen at the
* following World Wide Web location:
* http://sources.redhat.com/pthreads-win32/contributors.html
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library in the file COPYING.LIB;
* if not, write to the Free Software Foundation, Inc.,
* 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
*/
#if !defined( SEMAPHORE_H )
#define SEMAPHORE_H
#undef PTW32_LEVEL
#if defined(_POSIX_SOURCE)
#define PTW32_LEVEL 0
/* Early POSIX */
#endif
#if defined(_POSIX_C_SOURCE) && _POSIX_C_SOURCE >= 199309
#undef PTW32_LEVEL
#define PTW32_LEVEL 1
/* Include 1b, 1c and 1d */
#endif
#if defined(INCLUDE_NP)
#undef PTW32_LEVEL
#define PTW32_LEVEL 2
/* Include Non-Portable extensions */
#endif
#define PTW32_LEVEL_MAX 3
#if !defined(PTW32_LEVEL)
#define PTW32_LEVEL PTW32_LEVEL_MAX
/* Include everything */
#endif
#if __GNUC__ && ! defined (__declspec)
# error Please upgrade your GNU compiler to one that supports __declspec.
#endif
/*
* When building the DLL code, you should define PTW32_BUILD so that
* the variables/functions are exported correctly. When using the DLL,
* do NOT define PTW32_BUILD, and then the variables/functions will
* be imported correctly.
*/
#ifndef PTW32_STATIC_LIB
# ifdef PTW32_BUILD
# define PTW32_DLLPORT __declspec (dllexport)
# else
# define PTW32_DLLPORT __declspec (dllimport)
# endif
#else
# define PTW32_DLLPORT
#endif
/*
* This is a duplicate of what is in the autoconf config.h,
* which is only used when building the pthread-win32 libraries.
*/
#ifndef PTW32_CONFIG_H
# if defined(WINCE)
# define NEED_ERRNO
# define NEED_SEM
# endif
# if defined(_UWIN) || defined(__MINGW32__)
# define HAVE_MODE_T
# endif
#endif
/*
*
*/
#if PTW32_LEVEL >= PTW32_LEVEL_MAX
#ifdef NEED_ERRNO
#include "need_errno.h"
#else
#include <errno.h>
#endif
#endif /* PTW32_LEVEL >= PTW32_LEVEL_MAX */
#define _POSIX_SEMAPHORES
#ifdef __cplusplus
extern "C"
{
#endif /* __cplusplus */
#ifndef HAVE_MODE_T
typedef unsigned int mode_t;
#endif
typedef struct sem_t_ * sem_t;
PTW32_DLLPORT int __cdecl sem_init (sem_t * sem,
int pshared,
unsigned int value);
PTW32_DLLPORT int __cdecl sem_destroy (sem_t * sem);
PTW32_DLLPORT int __cdecl sem_trywait (sem_t * sem);
PTW32_DLLPORT int __cdecl sem_wait (sem_t * sem);
PTW32_DLLPORT int __cdecl sem_timedwait (sem_t * sem,
const struct timespec * abstime);
PTW32_DLLPORT int __cdecl sem_post (sem_t * sem);
PTW32_DLLPORT int __cdecl sem_post_multiple (sem_t * sem,
int count);
PTW32_DLLPORT int __cdecl sem_open (const char * name,
int oflag,
mode_t mode,
unsigned int value);
PTW32_DLLPORT int __cdecl sem_close (sem_t * sem);
PTW32_DLLPORT int __cdecl sem_unlink (const char * name);
PTW32_DLLPORT int __cdecl sem_getvalue (sem_t * sem,
int * sval);
#ifdef __cplusplus
} /* End of extern "C" */
#endif /* __cplusplus */
#undef PTW32_LEVEL
#undef PTW32_LEVEL_MAX
#endif /* !SEMAPHORE_H */
libraries/pthreads/src/pthread.cpp 0 → 100644
View file @ bf6d95c2
#include "pthread.h"
PTHREAD_EXPORT volatile long _pthread_cancelling;
PTHREAD_EXPORT int _pthread_concur;
/* Will default to zero as needed */
PTHREAD_EXPORT pthread_once_t _pthread_tls_once;
PTHREAD_EXPORT DWORD _pthread_tls;
/* Note initializer is zero, so this works */
PTHREAD_EXPORT pthread_rwlock_t _pthread_key_lock;
PTHREAD_EXPORT long _pthread_key_max;
PTHREAD_EXPORT long _pthread_key_sch;
PTHREAD_EXPORT void (**_pthread_key_dest)(void *);
platforms/cpu/include/CpuNonbondedForce.h
View file @ bf6d95c2
/* Portions copyright (c) 2006-2013 Stanford University and Simbios.
/* Portions copyright (c) 2006-2015 Stanford University and Simbios.
* Contributors: Pande Group
*
* Permission is hereby granted, free of charge, to any person obtaining
......@@ -174,8 +174,8 @@ protected:
float alphaEwald;
int numRx, numRy, numRz;
int meshDim[3];
std::vector<float> ewaldScaleTable;
float ewaldDX, ewaldDXInv;
std::vector<float> erfcTable, ewaldScaleTable;
float ewaldDX, ewaldDXInv, erfcDXInv;
std::vector<double> threadEnergy;
// The following variables are used to make information accessible to the individual threads.
int numberOfAtoms;
......@@ -241,7 +241,7 @@ protected:
/**
* Compute a fast approximation to erfc(x).
*/
static float erfcApprox(float x);
float erfcApprox(float x);
};
} // namespace OpenMM
......
platforms/cpu/include/CpuNonbondedForceVec4.h
View file @ bf6d95c2
/* Portions copyright (c) 2006-2014 Stanford University and Simbios.
/* Portions copyright (c) 2006-2015 Stanford University and Simbios.
* Contributors: Pande Group
*
* Permission is hereby granted, free of charge, to any person obtaining
......@@ -87,7 +87,7 @@ protected:
/**
* Compute a fast approximation to erfc(x).
*/
static fvec4 erfcApprox(const fvec4& x);
fvec4 erfcApprox(const fvec4& x);
/**
* Evaluate the scale factor used with Ewald and PME: erfc(alpha*r) + 2*alpha*r*exp(-alpha*alpha*r*r)/sqrt(PI)
......
platforms/cpu/include/CpuNonbondedForceVec8.h
View file @ bf6d95c2
/* Portions copyright (c) 2006-2014 Stanford University and Simbios.
/* Portions copyright (c) 2006-2015 Stanford University and Simbios.
* Contributors: Pande Group
*
* Permission is hereby granted, free of charge, to any person obtaining
......@@ -86,7 +86,7 @@ protected:
/**
* Compute a fast approximation to erfc(x).
*/
static fvec8 erfcApprox(const fvec8& x);
fvec8 erfcApprox(const fvec8& x);
/**
* Evaluate the scale factor used with Ewald and PME: erfc(alpha*r) + 2*alpha*r*exp(-alpha*alpha*r*r)/sqrt(PI)
......
platforms/cpu/src/CpuNonbondedForce.cpp
View file @ bf6d95c2
......@@ -171,17 +171,20 @@ void CpuNonbondedForce::setUseSwitchingFunction(float distance) {
}
void CpuNonbondedForce::tabulateEwaldScaleFactor() {
void CpuNonbondedForce::tabulateEwaldScaleFactor() {
if (tableIsValid)
return;
tableIsValid = true;
ewaldDX = cutoffDistance/NUM_TABLE_POINTS;
ewaldDXInv = 1.0f/ewaldDX;
erfcDXInv = 1.0f/(ewaldDX*alphaEwald);
erfcTable.resize(NUM_TABLE_POINTS+4);
ewaldScaleTable.resize(NUM_TABLE_POINTS+4);
for (int i = 0; i < NUM_TABLE_POINTS+4; i++) {
double r = i*ewaldDX;
double alphaR = alphaEwald*r;
ewaldScaleTable[i] = erfc(alphaR) + TWO_OVER_SQRT_PI*alphaR*exp(-alphaR*alphaR);
erfcTable[i] = erfc(alphaR);
ewaldScaleTable[i] = erfcTable[i] + TWO_OVER_SQRT_PI*alphaR*exp(-alphaR*alphaR);
}
}
......@@ -473,14 +476,10 @@ void CpuNonbondedForce::getDeltaR(const fvec4& posI, const fvec4& posJ, fvec4& d
}
float CpuNonbondedForce::erfcApprox(float x) {
// This approximation for erfc is from Abramowitz and Stegun (1964) p. 299. They cite the following as
// the original source: C. Hastings, Jr., Approximations for Digital Computers (1955). It has a maximum
// error of 3e-7.
float t = 1.0f+(0.0705230784f+(0.0422820123f+(0.0092705272f+(0.0001520143f+(0.0002765672f+0.0000430638f*x)*x)*x)*x)*x)*x;
t *= t;
t *= t;
t *= t;
return 1.0f/(t*t);
float x1 = x*erfcDXInv;
int index = min((int) floor(x1), NUM_TABLE_POINTS);
float coeff2 = x1-index;
float coeff1 = 1.0f-coeff2;
return coeff1*erfcTable[index] + coeff2*erfcTable[index+1];
}
platforms/cpu/src/CpuNonbondedForceVec4.cpp
View file @ bf6d95c2
/* Portions copyright (c) 2006-2014 Stanford University and Simbios.
/* Portions copyright (c) 2006-2015 Stanford University and Simbios.
* Contributors: Pande Group
*
* Permission is hereby granted, free of charge, to any person obtaining
......@@ -393,15 +393,16 @@ void CpuNonbondedForceVec4::getDeltaR(const fvec4& posI, const fvec4& x, const f
}
fvec4 CpuNonbondedForceVec4::erfcApprox(const fvec4& x) {
// This approximation for erfc is from Abramowitz and Stegun (1964) p. 299. They cite the following as
// the original source: C. Hastings, Jr., Approximations for Digital Computers (1955). It has a maximum
// error of 3e-7.
fvec4 t = 1.0f+(0.0705230784f+(0.0422820123f+(0.0092705272f+(0.0001520143f+(0.0002765672f+0.0000430638f*x)*x)*x)*x)*x)*x;
t *= t;
t *= t;
t *= t;
return 1.0f/(t*t);
fvec4 x1 = x*erfcDXInv;
ivec4 index = min(floor(x1), NUM_TABLE_POINTS);
fvec4 coeff2 = x1-index;
fvec4 coeff1 = 1.0f-coeff2;
fvec4 t1(&erfcTable[index[0]]);
fvec4 t2(&erfcTable[index[1]]);
fvec4 t3(&erfcTable[index[2]]);
fvec4 t4(&erfcTable[index[3]]);
transpose(t1, t2, t3, t4);
return coeff1*t1 + coeff2*t2;
}
fvec4 CpuNonbondedForceVec4::ewaldScaleFunction(const fvec4& x) {
......
platforms/cpu/src/CpuNonbondedForceVec8.cpp
View file @ bf6d95c2
/* Portions copyright (c) 2006-2014 Stanford University and Simbios.
/* Portions copyright (c) 2006-2015 Stanford University and Simbios.
* Contributors: Pande Group
*
* Permission is hereby granted, free of charge, to any person obtaining
......@@ -423,15 +423,23 @@ void CpuNonbondedForceVec8::getDeltaR(const fvec4& posI, const fvec8& x, const f
}
fvec8 CpuNonbondedForceVec8::erfcApprox(const fvec8& x) {
// This approximation for erfc is from Abramowitz and Stegun (1964) p. 299. They cite the following as
// the original source: C. Hastings, Jr., Approximations for Digital Computers (1955). It has a maximum
// error of 3e-7.
fvec8 t = 1.0f+(0.0705230784f+(0.0422820123f+(0.0092705272f+(0.0001520143f+(0.0002765672f+0.0000430638f*x)*x)*x)*x)*x)*x;
t *= t;
t *= t;
t *= t;
return 1.0f/(t*t);
fvec8 x1 = x*erfcDXInv;
ivec8 index = min(floor(x1), NUM_TABLE_POINTS);
fvec8 coeff2 = x1-index;
fvec8 coeff1 = 1.0f-coeff2;
ivec4 indexLower = index.lowerVec();
ivec4 indexUpper = index.upperVec();
fvec4 t1(&erfcTable[indexLower[0]]);
fvec4 t2(&erfcTable[indexLower[1]]);
fvec4 t3(&erfcTable[indexLower[2]]);
fvec4 t4(&erfcTable[indexLower[3]]);
fvec4 t5(&erfcTable[indexUpper[0]]);
fvec4 t6(&erfcTable[indexUpper[1]]);
fvec4 t7(&erfcTable[indexUpper[2]]);
fvec4 t8(&erfcTable[indexUpper[3]]);
fvec8 s1, s2, s3, s4;
transpose(t1, t2, t3, t4, t5, t6, t7, t8, s1, s2, s3, s4);
return coeff1*s1 + coeff2*s2;
}
fvec8 CpuNonbondedForceVec8::ewaldScaleFunction(const fvec8& x) {
......
platforms/cpu/src/gmx_atomic.h
View file @ bf6d95c2
......@@ -64,6 +64,7 @@
#include <stdio.h>
#define NOMINMAX
#include <pthread.h>
#ifdef __cplusplus
......
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