From f509f133afd9bdeed268de1582ca0adb4f4e0692 Mon Sep 17 00:00:00 2001
From: Steve Plimpton <sjplimp@sandia.gov>
Date: Tue, 13 Dec 2016 17:14:28 -0700
Subject: [PATCH] patch 13Dec16: neighbor refactor, Stan pppm/disp bug fix, M
Brown INTEL package updates
---
doc/src/Manual.txt | 4 +-
src/MAKE/MACHINES/Makefile.cori2 | 127 ++++++++++++++++++
src/MAKE/OPTIONS/Makefile.intel_cpu | 123 +++++++++++++++++
.../OPTIONS/Makefile.intel_knl_coprocessor | 123 +++++++++++++++++
src/USER-INTEL/fix_intel.cpp | 32 +++--
src/USER-INTEL/intel_intrinsics.h | 2 +-
src/USER-INTEL/intel_preprocess.h | 15 ++-
src/USER-INTEL/verlet_lrt_intel.cpp | 15 ++-
src/USER-INTEL/verlet_lrt_intel.h | 2 +-
src/force.cpp | 30 ++++-
src/memory.cpp | 15 ++-
src/modify.cpp | 15 ++-
src/neighbor.cpp | 11 +-
src/version.h | 2 +-
14 files changed, 484 insertions(+), 32 deletions(-)
create mode 100755 src/MAKE/MACHINES/Makefile.cori2
create mode 100755 src/MAKE/OPTIONS/Makefile.intel_cpu
create mode 100644 src/MAKE/OPTIONS/Makefile.intel_knl_coprocessor
diff --git a/doc/src/Manual.txt b/doc/src/Manual.txt
index 40288268a1..ab6aa91b7c 100644
--- a/doc/src/Manual.txt
+++ b/doc/src/Manual.txt
@@ -1,7 +1,7 @@
<!-- HTML_ONLY -->
<HEAD>
<TITLE>LAMMPS Users Manual</TITLE>
-<META NAME="docnumber" CONTENT="30 Nov 2016 version">
+<META NAME="docnumber" CONTENT="13 Dec 2016 version">
<META NAME="author" CONTENT="http://lammps.sandia.gov - Sandia National Laboratories">
<META NAME="copyright" CONTENT="Copyright (2003) Sandia Corporation. This software and manual is distributed under the GNU General Public License.">
</HEAD>
@@ -21,7 +21,7 @@
<H1></H1>
LAMMPS Documentation :c,h3
-30 Nov 2016 version :c,h4
+13 Dec 2016 version :c,h4
Version info: :h4
diff --git a/src/MAKE/MACHINES/Makefile.cori2 b/src/MAKE/MACHINES/Makefile.cori2
new file mode 100755
index 0000000000..a367d54080
--- /dev/null
+++ b/src/MAKE/MACHINES/Makefile.cori2
@@ -0,0 +1,127 @@
+# cori2 = NERSC Cori II KNL, static build, FFTW (single precision)
+
+# ---------------------------------------------------------------------
+# module swap craype-haswell craype-mic-knl
+# module load fftw
+# module load craype-hugepages2M
+# Recommend using #SBATCH -S 2 for core specialization
+# ---------------------------------------------------------------------
+
+SHELL = /bin/sh
+
+# ---------------------------------------------------------------------
+# compiler/linker settings
+# specify flags and libraries needed for your compiler
+
+CC = CC
+OPTFLAGS = -xMIC-AVX512 -O2 -fp-model fast=2 -no-prec-div -qoverride-limits
+CCFLAGS = -g -qopenmp -DLAMMPS_MEMALIGN=64 -qno-offload \
+ -fno-alias -ansi-alias -restrict $(OPTFLAGS) -DLMP_INTEL_NO_TBB
+SHFLAGS = -fPIC
+DEPFLAGS = -M
+
+LINK = CC
+LINKFLAGS = -g -qopenmp $(OPTFLAGS)
+LIB =
+SIZE = size
+
+ARCHIVE = ar
+ARFLAGS = -rc
+SHLIBFLAGS = -shared
+
+# ---------------------------------------------------------------------
+# LAMMPS-specific settings, all OPTIONAL
+# specify settings for LAMMPS features you will use
+# if you change any -D setting, do full re-compile after "make clean"
+
+# LAMMPS ifdef settings
+# see possible settings in Section 2.2 (step 4) of manual
+
+LMP_INC = #-DLAMMPS_GZIP -DLAMMPS_JPEG
+
+# MPI library
+# see discussion in Section 2.2 (step 5) of manual
+# MPI wrapper compiler/linker can provide this info
+# can point to dummy MPI library in src/STUBS as in Makefile.serial
+# use -D MPICH and OMPI settings in INC to avoid C++ lib conflicts
+# INC = path for mpi.h, MPI compiler settings
+# PATH = path for MPI library
+# LIB = name of MPI library
+
+MPI_INC = -DMPICH_SKIP_MPICXX -DOMPI_SKIP_MPICXX=1
+MPI_PATH =
+MPI_LIB =
+
+# FFT library
+# see discussion in Section 2.2 (step 6) of manaul
+# can be left blank to use provided KISS FFT library
+# INC = -DFFT setting, e.g. -DFFT_FFTW, FFT compiler settings
+# PATH = path for FFT library
+# LIB = name of FFT library
+
+FFT_INC = -DFFT_FFTW3 -DFFT_SINGLE
+FFT_PATH =
+FFT_LIB = -lfftw3f
+
+# JPEG and/or PNG library
+# see discussion in Section 2.2 (step 7) of manual
+# only needed if -DLAMMPS_JPEG or -DLAMMPS_PNG listed with LMP_INC
+# INC = path(s) for jpeglib.h and/or png.h
+# PATH = path(s) for JPEG library and/or PNG library
+# LIB = name(s) of JPEG library and/or PNG library
+
+JPG_INC =
+JPG_PATH =
+JPG_LIB = #-ljpeg
+
+# ---------------------------------------------------------------------
+# build rules and dependencies
+# do not edit this section
+
+include Makefile.package.settings
+include Makefile.package
+
+EXTRA_INC = $(LMP_INC) $(PKG_INC) $(MPI_INC) $(FFT_INC) $(JPG_INC) $(PKG_SYSINC)
+EXTRA_PATH = $(PKG_PATH) $(MPI_PATH) $(FFT_PATH) $(JPG_PATH) $(PKG_SYSPATH)
+EXTRA_LIB = $(PKG_LIB) $(MPI_LIB) $(FFT_LIB) $(JPG_LIB) $(PKG_SYSLIB)
+
+# Path to src files
+
+vpath %.cpp ..
+vpath %.h ..
+
+# Link target
+
+$(EXE): $(OBJ)
+ $(LINK) $(LINKFLAGS) $(EXTRA_PATH) $(OBJ) $(EXTRA_LIB) $(LIB) -o $(EXE)
+ $(SIZE) $(EXE)
+
+# Library targets
+
+lib: $(OBJ)
+ $(ARCHIVE) $(ARFLAGS) $(EXE) $(OBJ)
+
+shlib: $(OBJ)
+ $(CC) $(CCFLAGS) $(SHFLAGS) $(SHLIBFLAGS) $(EXTRA_PATH) -o $(EXE) \
+ $(OBJ) $(EXTRA_LIB) $(LIB)
+
+# Compilation rules
+
+%.o:%.cpp
+ $(CC) $(CCFLAGS) $(SHFLAGS) $(EXTRA_INC) -c $<
+
+%.d:%.cpp
+ $(CC) $(CCFLAGS) $(EXTRA_INC) $(DEPFLAGS) $< > $@
+
+%.o:%.cu
+ $(CC) $(CCFLAGS) $(SHFLAGS) $(EXTRA_INC) -c $<
+
+# Individual dependencies
+
+depend : fastdep.exe $(SRC)
+ @./fastdep.exe $(EXTRA_INC) -- $^ > .depend || exit 1
+
+fastdep.exe: ../DEPEND/fastdep.c
+ cc -O -o $@ $<
+
+sinclude .depend
diff --git a/src/MAKE/OPTIONS/Makefile.intel_cpu b/src/MAKE/OPTIONS/Makefile.intel_cpu
new file mode 100755
index 0000000000..b34ff47761
--- /dev/null
+++ b/src/MAKE/OPTIONS/Makefile.intel_cpu
@@ -0,0 +1,123 @@
+# intel_cpu_intelmpi = USER-INTEL package, Intel MPI, MKL FFT
+
+SHELL = /bin/sh
+
+# ---------------------------------------------------------------------
+# compiler/linker settings
+# specify flags and libraries needed for your compiler
+
+CC = mpiicpc
+OPTFLAGS = -xHost -O2 -fp-model fast=2 -no-prec-div -qoverride-limits
+CCFLAGS = -g -qopenmp -DLAMMPS_MEMALIGN=64 -no-offload \
+ -fno-alias -ansi-alias -restrict $(OPTFLAGS)
+SHFLAGS = -fPIC
+DEPFLAGS = -M
+
+LINK = mpiicpc
+LINKFLAGS = -g -qopenmp $(OPTFLAGS)
+LIB = -ltbbmalloc -ltbbmalloc_proxy
+SIZE = size
+
+ARCHIVE = ar
+ARFLAGS = -rc
+SHLIBFLAGS = -shared
+
+# ---------------------------------------------------------------------
+# LAMMPS-specific settings, all OPTIONAL
+# specify settings for LAMMPS features you will use
+# if you change any -D setting, do full re-compile after "make clean"
+
+# LAMMPS ifdef settings
+# see possible settings in Section 2.2 (step 4) of manual
+
+LMP_INC = -DLAMMPS_GZIP -DLAMMPS_JPEG
+
+# MPI library
+# see discussion in Section 2.2 (step 5) of manual
+# MPI wrapper compiler/linker can provide this info
+# can point to dummy MPI library in src/STUBS as in Makefile.serial
+# use -D MPICH and OMPI settings in INC to avoid C++ lib conflicts
+# INC = path for mpi.h, MPI compiler settings
+# PATH = path for MPI library
+# LIB = name of MPI library
+
+MPI_INC = -DMPICH_SKIP_MPICXX -DOMPI_SKIP_MPICXX=1
+MPI_PATH =
+MPI_LIB =
+
+# FFT library
+# see discussion in Section 2.2 (step 6) of manaul
+# can be left blank to use provided KISS FFT library
+# INC = -DFFT setting, e.g. -DFFT_FFTW, FFT compiler settings
+# PATH = path for FFT library
+# LIB = name of FFT library
+
+FFT_INC = -DFFT_MKL -DFFT_SINGLE
+FFT_PATH =
+FFT_LIB = -L$MKLROOT/lib/intel64/ -lmkl_intel_ilp64 \
+ -lmkl_sequential -lmkl_core
+
+# JPEG and/or PNG library
+# see discussion in Section 2.2 (step 7) of manual
+# only needed if -DLAMMPS_JPEG or -DLAMMPS_PNG listed with LMP_INC
+# INC = path(s) for jpeglib.h and/or png.h
+# PATH = path(s) for JPEG library and/or PNG library
+# LIB = name(s) of JPEG library and/or PNG library
+
+JPG_INC =
+JPG_PATH =
+JPG_LIB = -ljpeg
+
+# ---------------------------------------------------------------------
+# build rules and dependencies
+# do not edit this section
+
+include Makefile.package.settings
+include Makefile.package
+
+EXTRA_INC = $(LMP_INC) $(PKG_INC) $(MPI_INC) $(FFT_INC) $(JPG_INC) $(PKG_SYSINC)
+EXTRA_PATH = $(PKG_PATH) $(MPI_PATH) $(FFT_PATH) $(JPG_PATH) $(PKG_SYSPATH)
+EXTRA_LIB = $(PKG_LIB) $(MPI_LIB) $(FFT_LIB) $(JPG_LIB) $(PKG_SYSLIB)
+EXTRA_CPP_DEPENDS = $(PKG_CPP_DEPENDS)
+EXTRA_LINK_DEPENDS = $(PKG_LINK_DEPENDS)
+
+# Path to src files
+
+vpath %.cpp ..
+vpath %.h ..
+
+# Link target
+
+$(EXE): $(OBJ) $(EXTRA_LINK_DEPENDS)
+ $(LINK) $(LINKFLAGS) $(EXTRA_PATH) $(OBJ) $(EXTRA_LIB) $(LIB) -o $(EXE)
+ $(SIZE) $(EXE)
+
+# Library targets
+
+lib: $(OBJ) $(EXTRA_LINK_DEPENDS)
+ $(ARCHIVE) $(ARFLAGS) $(EXE) $(OBJ)
+
+shlib: $(OBJ) $(EXTRA_LINK_DEPENDS)
+ $(CC) $(CCFLAGS) $(SHFLAGS) $(SHLIBFLAGS) $(EXTRA_PATH) -o $(EXE) \
+ $(OBJ) $(EXTRA_LIB) $(LIB)
+
+# Compilation rules
+
+%.o:%.cpp $(EXTRA_CPP_DEPENDS)
+ $(CC) $(CCFLAGS) $(SHFLAGS) $(EXTRA_INC) -c $<
+
+%.d:%.cpp $(EXTRA_CPP_DEPENDS)
+ $(CC) $(CCFLAGS) $(EXTRA_INC) $(DEPFLAGS) $< > $@
+
+%.o:%.cu $(EXTRA_CPP_DEPENDS)
+ $(CC) $(CCFLAGS) $(SHFLAGS) $(EXTRA_INC) -c $<
+
+# Individual dependencies
+
+depend : fastdep.exe $(SRC)
+ @./fastdep.exe $(EXTRA_INC) -- $^ > .depend || exit 1
+
+fastdep.exe: ../DEPEND/fastdep.c
+ cc -O -o $@ $<
+
+sinclude .depend
diff --git a/src/MAKE/OPTIONS/Makefile.intel_knl_coprocessor b/src/MAKE/OPTIONS/Makefile.intel_knl_coprocessor
new file mode 100644
index 0000000000..b7f3cd6846
--- /dev/null
+++ b/src/MAKE/OPTIONS/Makefile.intel_knl_coprocessor
@@ -0,0 +1,123 @@
+# intel_phi = USER-INTEL with Phi x200 (KNL) offload support,Intel MPI,MKL FFT
+
+SHELL = /bin/sh
+
+# ---------------------------------------------------------------------
+# compiler/linker settings
+# specify flags and libraries needed for your compiler
+
+CC = mpiicpc
+MIC_OPT = -qoffload-arch=mic-avx512 -fp-model fast=2
+CCFLAGS = -g -O3 -qopenmp -DLMP_INTEL_OFFLOAD -DLAMMPS_MEMALIGN=64 \
+ -xHost -fno-alias -ansi-alias -restrict \
+ -qoverride-limits $(MIC_OPT)
+SHFLAGS = -fPIC
+DEPFLAGS = -M
+
+LINK = mpiicpc
+LINKFLAGS = -g -O3 -xHost -qopenmp -qoffload $(MIC_OPT)
+LIB = -ltbbmalloc
+SIZE = size
+
+ARCHIVE = ar
+ARFLAGS = -rc
+SHLIBFLAGS = -shared
+
+# ---------------------------------------------------------------------
+# LAMMPS-specific settings, all OPTIONAL
+# specify settings for LAMMPS features you will use
+# if you change any -D setting, do full re-compile after "make clean"
+
+# LAMMPS ifdef settings
+# see possible settings in Section 2.2 (step 4) of manual
+
+LMP_INC = -DLAMMPS_GZIP -DLAMMPS_JPEG
+
+# MPI library
+# see discussion in Section 2.2 (step 5) of manual
+# MPI wrapper compiler/linker can provide this info
+# can point to dummy MPI library in src/STUBS as in Makefile.serial
+# use -D MPICH and OMPI settings in INC to avoid C++ lib conflicts
+# INC = path for mpi.h, MPI compiler settings
+# PATH = path for MPI library
+# LIB = name of MPI library
+
+MPI_INC = -DMPICH_SKIP_MPICXX -DOMPI_SKIP_MPICXX=1
+MPI_PATH =
+MPI_LIB =
+
+# FFT library
+# see discussion in Section 2.2 (step 6) of manaul
+# can be left blank to use provided KISS FFT library
+# INC = -DFFT setting, e.g. -DFFT_FFTW, FFT compiler settings
+# PATH = path for FFT library
+# LIB = name of FFT library
+
+FFT_INC = -DFFT_MKL -DFFT_SINGLE
+FFT_PATH =
+FFT_LIB = -L$(MKLROOT)/lib/intel64/ -lmkl_intel_ilp64 -lmkl_sequential -lmkl_core
+
+# JPEG and/or PNG library
+# see discussion in Section 2.2 (step 7) of manual
+# only needed if -DLAMMPS_JPEG or -DLAMMPS_PNG listed with LMP_INC
+# INC = path(s) for jpeglib.h and/or png.h
+# PATH = path(s) for JPEG library and/or PNG library
+# LIB = name(s) of JPEG library and/or PNG library
+
+JPG_INC =
+JPG_PATH =
+JPG_LIB = -ljpeg
+
+# ---------------------------------------------------------------------
+# build rules and dependencies
+# do not edit this section
+
+include Makefile.package.settings
+include Makefile.package
+
+EXTRA_INC = $(LMP_INC) $(PKG_INC) $(MPI_INC) $(FFT_INC) $(JPG_INC) $(PKG_SYSINC)
+EXTRA_PATH = $(PKG_PATH) $(MPI_PATH) $(FFT_PATH) $(JPG_PATH) $(PKG_SYSPATH)
+EXTRA_LIB = $(PKG_LIB) $(MPI_LIB) $(FFT_LIB) $(JPG_LIB) $(PKG_SYSLIB)
+EXTRA_CPP_DEPENDS = $(PKG_CPP_DEPENDS)
+EXTRA_LINK_DEPENDS = $(PKG_LINK_DEPENDS)
+
+# Path to src files
+
+vpath %.cpp ..
+vpath %.h ..
+
+# Link target
+
+$(EXE): $(OBJ) $(EXTRA_LINK_DEPENDS)
+ $(LINK) $(LINKFLAGS) $(EXTRA_PATH) $(OBJ) $(EXTRA_LIB) $(LIB) -o $(EXE)
+ $(SIZE) $(EXE)
+
+# Library targets
+
+lib: $(OBJ) $(EXTRA_LINK_DEPENDS)
+ $(ARCHIVE) $(ARFLAGS) $(EXE) $(OBJ)
+
+shlib: $(OBJ) $(EXTRA_LINK_DEPENDS)
+ $(CC) $(CCFLAGS) $(SHFLAGS) $(SHLIBFLAGS) $(EXTRA_PATH) -o $(EXE) \
+ $(OBJ) $(EXTRA_LIB) $(LIB)
+
+# Compilation rules
+
+%.o:%.cpp $(EXTRA_CPP_DEPENDS)
+ $(CC) $(CCFLAGS) $(SHFLAGS) $(EXTRA_INC) -c $<
+
+%.d:%.cpp $(EXTRA_CPP_DEPENDS)
+ $(CC) $(CCFLAGS) $(EXTRA_INC) $(DEPFLAGS) $< > $@
+
+%.o:%.cu $(EXTRA_CPP_DEPENDS)
+ $(CC) $(CCFLAGS) $(SHFLAGS) $(EXTRA_INC) -c $<
+
+# Individual dependencies
+
+depend : fastdep.exe $(SRC)
+ @./fastdep.exe $(EXTRA_INC) -- $^ > .depend || exit 1
+
+fastdep.exe: ../DEPEND/fastdep.c
+ cc -O -o $@ $<
+
+sinclude .depend
diff --git a/src/USER-INTEL/fix_intel.cpp b/src/USER-INTEL/fix_intel.cpp
index 06bd23abcc..84f4994901 100644
--- a/src/USER-INTEL/fix_intel.cpp
+++ b/src/USER-INTEL/fix_intel.cpp
@@ -63,7 +63,7 @@ FixIntel::FixIntel(LAMMPS *lmp, int narg, char **arg) : Fix(lmp, narg, arg)
_nbor_pack_width = 1;
_precision_mode = PREC_MODE_MIXED;
- _offload_balance = 1.0;
+ _offload_balance = -1.0;
_overflow_flag[LMP_OVERFLOW] = 0;
_off_overflow_flag[LMP_OVERFLOW] = 0;
@@ -189,10 +189,18 @@ FixIntel::FixIntel(LAMMPS *lmp, int narg, char **arg) : Fix(lmp, narg, arg)
offload_cores = omp_get_num_procs();
omp_set_num_threads(offload_cores);
max_offload_threads = omp_get_max_threads();
+ #ifdef __AVX512F__
+ if ( (offload_cores / 4) % 2 == 1) {
+ offload_cores += 4;
+ max_offload_threads += 4;
+ }
+ #endif
}
_max_offload_threads = max_offload_threads;
_offload_cores = offload_cores;
if (_offload_threads == 0) _offload_threads = offload_cores;
+ if (_offload_cores > 244 && _offload_tpc > 2)
+ _offload_tpc = 2;
}
#endif
@@ -317,6 +325,8 @@ void FixIntel::init()
error->all(FLERR,
"Currently, cannot use more than one intel style with hybrid.");
+ neighbor->fix_intel = (void *)this;
+
check_neighbor_intel();
if (_precision_mode == PREC_MODE_SINGLE)
_single_buffers->zero_ev();
@@ -1004,8 +1014,10 @@ void FixIntel::set_offload_affinity()
int offload_threads = _offload_threads;
int offload_tpc = _offload_tpc;
int offload_affinity_balanced = _offload_affinity_balanced;
+ int offload_cores = _offload_cores;
#pragma offload target(mic:_cop) mandatory \
- in(node_rank,offload_threads,offload_tpc,offload_affinity_balanced)
+ in(node_rank,offload_threads,offload_tpc,offload_affinity_balanced, \
+ offload_cores)
{
omp_set_num_threads(offload_threads);
#pragma omp parallel
@@ -1013,20 +1025,24 @@ void FixIntel::set_offload_affinity()
int tnum = omp_get_thread_num();
kmp_affinity_mask_t mask;
kmp_create_affinity_mask(&mask);
- int proc;
- if (offload_affinity_balanced) {
- proc = offload_threads * node_rank + tnum;
+ int proc = offload_threads * node_rank + tnum;
+ #ifdef __AVX512F__
+ proc = (proc / offload_tpc) + (proc % offload_tpc) *
+ ((offload_cores) / 4);
+ proc += 68;
+ #else
+ if (offload_affinity_balanced)
proc = proc * 4 - (proc / 60) * 240 + proc / 60 + 1;
- } else {
- proc = offload_threads * node_rank + tnum;
+ else
proc += (proc / 4) * (4 - offload_tpc) + 1;
- }
+ #endif
kmp_set_affinity_mask_proc(proc, &mask);
if (kmp_set_affinity(&mask) != 0)
printf("Could not set affinity on rank %d thread %d to %d\n",
node_rank, tnum, proc);
}
}
+
if (_precision_mode == PREC_MODE_SINGLE)
_single_buffers->set_off_params(offload_threads, _cop, _separate_buffers);
else if (_precision_mode == PREC_MODE_MIXED)
diff --git a/src/USER-INTEL/intel_intrinsics.h b/src/USER-INTEL/intel_intrinsics.h
index 754b390722..44a9605961 100644
--- a/src/USER-INTEL/intel_intrinsics.h
+++ b/src/USER-INTEL/intel_intrinsics.h
@@ -28,7 +28,7 @@
// implementations.
// Vector classes provided with the intel compiler
-#ifdef __MIC__
+#if defined(__MIC__) && !defined(__AVX512F__)
#include <mic/micvec.h>
#else
#include <dvec.h> // icc-mmic hates generating movq
diff --git a/src/USER-INTEL/intel_preprocess.h b/src/USER-INTEL/intel_preprocess.h
index 48378d3840..ad07dfd7c2 100644
--- a/src/USER-INTEL/intel_preprocess.h
+++ b/src/USER-INTEL/intel_preprocess.h
@@ -22,6 +22,11 @@
#ifdef __INTEL_OFFLOAD
#ifdef LMP_INTEL_OFFLOAD
#define _LMP_INTEL_OFFLOAD
+#ifdef __TARGET_ARCH_MIC
+#ifndef __MIC__
+#define __MIC__ 1
+#endif
+#endif
#endif
#endif
@@ -62,6 +67,7 @@ enum {TIME_PACK, TIME_HOST_NEIGHBOR, TIME_HOST_PAIR, TIME_OFFLOAD_NEIGHBOR,
#define INTEL_MAX_STENCIL_CHECK 4096
#define INTEL_P3M_MAXORDER 5
+#ifdef __INTEL_COMPILER
#ifdef __AVX__
#undef INTEL_VECTOR_WIDTH
#define INTEL_VECTOR_WIDTH 8
@@ -90,6 +96,13 @@ enum {TIME_PACK, TIME_HOST_NEIGHBOR, TIME_HOST_PAIR, TIME_OFFLOAD_NEIGHBOR,
#endif
#endif
+#else
+
+#undef INTEL_VECTOR_WIDTH
+#define INTEL_VECTOR_WIDTH 1
+
+#endif
+
#define INTEL_DATA_ALIGN 64
#define INTEL_ONEATOM_FACTOR 2
#define INTEL_MIC_NBOR_PAD INTEL_MIC_VECTOR_WIDTH
@@ -97,7 +110,7 @@ enum {TIME_PACK, TIME_HOST_NEIGHBOR, TIME_HOST_PAIR, TIME_OFFLOAD_NEIGHBOR,
#define INTEL_LB_MEAN_WEIGHT 0.1
#define INTEL_BIGP 1e15
#define INTEL_MAX_HOST_CORE_COUNT 512
-#define INTEL_MAX_COI_CORES 2
+#define INTEL_MAX_COI_CORES 36
#define IP_PRE_get_stride(stride, n, datasize, torque) \
{ \
diff --git a/src/USER-INTEL/verlet_lrt_intel.cpp b/src/USER-INTEL/verlet_lrt_intel.cpp
index fcd9dff08a..afb7852f98 100644
--- a/src/USER-INTEL/verlet_lrt_intel.cpp
+++ b/src/USER-INTEL/verlet_lrt_intel.cpp
@@ -43,7 +43,20 @@ using namespace LAMMPS_NS;
/* ---------------------------------------------------------------------- */
VerletLRTIntel::VerletLRTIntel(LAMMPS *lmp, int narg, char **arg) :
- Verlet(lmp, narg, arg) {}
+ Verlet(lmp, narg, arg) {
+ #if defined(_LMP_INTEL_LRT_PTHREAD)
+ pthread_mutex_init(&_kmutex,NULL);
+ #endif
+}
+
+/* ---------------------------------------------------------------------- */
+
+VerletLRTIntel::~VerletLRTIntel()
+{
+ #if defined(_LMP_INTEL_LRT_PTHREAD)
+ pthread_mutex_destroy(&_kmutex);
+ #endif
+}
/* ----------------------------------------------------------------------
initialization before run
diff --git a/src/USER-INTEL/verlet_lrt_intel.h b/src/USER-INTEL/verlet_lrt_intel.h
index 112b299cd1..a699c20796 100644
--- a/src/USER-INTEL/verlet_lrt_intel.h
+++ b/src/USER-INTEL/verlet_lrt_intel.h
@@ -40,7 +40,7 @@ namespace LAMMPS_NS {
class VerletLRTIntel : public Verlet {
public:
VerletLRTIntel(class LAMMPS *, int, char **);
- virtual ~VerletLRTIntel() {}
+ virtual ~VerletLRTIntel();
virtual void init();
virtual void setup();
virtual void run(int);
diff --git a/src/force.cpp b/src/force.cpp
index 3154139d54..95a6ff4f6d 100644
--- a/src/force.cpp
+++ b/src/force.cpp
@@ -237,7 +237,10 @@ Pair *Force::new_pair(const char *style, int trysuffix, int &sflag)
return pair_creator(lmp);
}
- error->all(FLERR,"Unknown pair style");
+ char str[128];
+ sprintf(str,"Unknown pair style %s",style);
+ error->all(FLERR,str);
+
return NULL;
}
@@ -362,7 +365,10 @@ Bond *Force::new_bond(const char *style, int trysuffix, int &sflag)
return bond_creator(lmp);
}
- error->all(FLERR,"Unknown bond style");
+ char str[128];
+ sprintf(str,"Unknown bond style %s",style);
+ error->all(FLERR,str);
+
return NULL;
}
@@ -440,7 +446,10 @@ Angle *Force::new_angle(const char *style, int trysuffix, int &sflag)
return angle_creator(lmp);
}
- error->all(FLERR,"Unknown angle style");
+ char str[128];
+ sprintf(str,"Unknown angle style %s",style);
+ error->all(FLERR,str);
+
return NULL;
}
@@ -519,7 +528,10 @@ Dihedral *Force::new_dihedral(const char *style, int trysuffix, int &sflag)
return dihedral_creator(lmp);
}
- error->all(FLERR,"Unknown dihedral style");
+ char str[128];
+ sprintf(str,"Unknown dihedral style %s",style);
+ error->all(FLERR,str);
+
return NULL;
}
@@ -597,7 +609,10 @@ Improper *Force::new_improper(const char *style, int trysuffix, int &sflag)
return improper_creator(lmp);
}
- error->all(FLERR,"Unknown improper style");
+ char str[128];
+ sprintf(str,"Unknown improper style %s",style);
+ error->all(FLERR,str);
+
return NULL;
}
@@ -679,7 +694,10 @@ KSpace *Force::new_kspace(int narg, char **arg, int trysuffix, int &sflag)
return kspace_creator(lmp, narg-1, &arg[1]);
}
- error->all(FLERR,"Unknown kspace style");
+ char str[128];
+ sprintf(str,"Unknown kspace style %s",arg[0]);
+ error->all(FLERR,str);
+
return NULL;
}
diff --git a/src/memory.cpp b/src/memory.cpp
index 1ab5074079..ed0993202b 100644
--- a/src/memory.cpp
+++ b/src/memory.cpp
@@ -18,12 +18,12 @@
#include "error.h"
#if defined(LMP_USER_INTEL) && defined(__INTEL_COMPILER)
+#ifndef LMP_INTEL_NO_TBB
#define LMP_USE_TBB_ALLOCATOR
#include "tbb/scalable_allocator.h"
+#else
+#include <malloc.h>
#endif
-
-#if defined(LMP_USER_INTEL) && !defined(LAMMPS_MEMALIGN)
-#define LAMMPS_MEMALIGN 64
#endif
using namespace LAMMPS_NS;
@@ -75,6 +75,15 @@ void *Memory::srealloc(void *ptr, bigint nbytes, const char *name)
#if defined(LMP_USE_TBB_ALLOCATOR)
ptr = scalable_aligned_realloc(ptr, nbytes, LAMMPS_MEMALIGN);
+#elif defined(LMP_INTEL_NO_TBB) && defined(LAMMPS_MEMALIGN)
+ ptr = realloc(ptr, nbytes);
+ uintptr_t offset = ((uintptr_t)(const void *)(ptr)) % LAMMPS_MEMALIGN;
+ if (offset) {
+ void *optr = ptr;
+ ptr = smalloc(nbytes, name);
+ memcpy(ptr, optr, MIN(nbytes,malloc_usable_size(optr)));
+ free(optr);
+ }
#else
ptr = realloc(ptr,nbytes);
#endif
diff --git a/src/modify.cpp b/src/modify.cpp
index 6370998a23..2e5bd78504 100644
--- a/src/modify.cpp
+++ b/src/modify.cpp
@@ -767,7 +767,8 @@ void Modify::add_fix(int narg, char **arg, int trysuffix)
if (strcmp(estyle,fix[ifix]->style) == 0) match = 1;
}
}
- if (!match) error->all(FLERR,"Replacing a fix, but new style != old style");
+ if (!match) error->all(FLERR,
+ "Replacing a fix, but new style != old style");
if (fix[ifix]->igroup != igroup && comm->me == 0)
error->warning(FLERR,"Replacing a fix, but new group != old group");
@@ -812,7 +813,11 @@ void Modify::add_fix(int narg, char **arg, int trysuffix)
fix[ifix] = fix_creator(lmp,narg,arg);
}
- if (fix[ifix] == NULL) error->all(FLERR,"Unknown fix style");
+ if (fix[ifix] == NULL) {
+ char str[128];
+ sprintf(str,"Unknown fix style %s",arg[2]);
+ error->all(FLERR,str);
+ }
// check if Fix is in restart_global list
// if yes, pass state info to the Fix so it can reset itself
@@ -994,7 +999,11 @@ void Modify::add_compute(int narg, char **arg, int trysuffix)
compute[ncompute] = compute_creator(lmp,narg,arg);
}
- if (compute[ncompute] == NULL) error->all(FLERR,"Unknown compute style");
+ if (compute[ncompute] == NULL) {
+ char str[128];
+ sprintf(str,"Unknown compute style %s",arg[2]);
+ error->all(FLERR,str);
+ }
ncompute++;
}
diff --git a/src/neighbor.cpp b/src/neighbor.cpp
index d014f1c8d0..f968be1dd0 100644
--- a/src/neighbor.cpp
+++ b/src/neighbor.cpp
@@ -890,7 +890,8 @@ void Neighbor::init_pair()
for (i = 0; i < nstencil; i++) {
flag = 0;
for (j = 0; j < npair_perpetual; j++)
- if (lists[plist[j]]->stencil_method == neigh_stencil[i]->istyle) flag = 1;
+ if (lists[plist[j]]->stencil_method == neigh_stencil[i]->istyle)
+ flag = 1;
if (flag) slist[nstencil_perpetual++] = i;
}
@@ -1139,8 +1140,8 @@ void Neighbor::print_pairwise_info()
} else {
sprintf(str," (%d) command %s",i+1,requests[i]->command_style);
}
- fprintf(out,"%s\n",str);
-
+ fprintf(out,"%s",str);
+
if (requests[i]->half) kind = "half";
else if (requests[i]->full) kind = "full";
else if (requests[i]->gran) kind = "size";
@@ -1328,7 +1329,6 @@ int Neighbor::choose_stencil(NeighRequest *rq)
if (style == BIN && !(mask & NS_BIN)) continue;
if (style == MULTI && !(mask & NS_MULTI)) continue;
-
if (dimension == 2 && !(mask & NS_2D)) continue;
if (dimension == 3 && !(mask & NS_3D)) continue;
@@ -1389,7 +1389,8 @@ int Neighbor::choose_pair(NeighRequest *rq)
// triclinic = orthgonal/triclinic box
int copyflag,skipflag,halfflag,fullflag,halffullflag,sizeflag,respaflag,
- ghostflag,off2onflag,onesideflag,ssaflag,ompflag,intelflag,kokkos_device_flag,kokkos_host_flag;
+ ghostflag,off2onflag,onesideflag,ssaflag,ompflag,intelflag,
+ kokkos_device_flag,kokkos_host_flag;
copyflag = skipflag = halfflag = fullflag = halffullflag = sizeflag =
ghostflag = respaflag = off2onflag = onesideflag = ssaflag =
diff --git a/src/version.h b/src/version.h
index 1562ca3d95..c7c630c699 100644
--- a/src/version.h
+++ b/src/version.h
@@ -1 +1 @@
-#define LAMMPS_VERSION "30 Nov 2016"
+#define LAMMPS_VERSION "13 Dec 2016"
--
GitLab