Skip to content
GitLab
Commit e5c20fd3 authored by Matthieu Schaller's avatar Matthieu Schaller
Browse files

Merge branch 'master' into new_cosmo_ICs

parents e16665f5 e85ecae0
Hide whitespace changes
Inline Side-by-side
README
View file @ e5c20fd3
......@@ -26,6 +26,8 @@ Valid options are:
-f {int} Overwrite the CPU frequency (Hz) to be used for time measurements
-g Run with an external gravitational potential
-G Run with self-gravity
-n {int} Execute a fixed number of time steps. When unset use the time_end
parameter to stop.
-s Run with SPH
-t {int} The number of threads to use on each MPI rank. Defaults to 1 if not specified.
-v [12] Increase the level of verbosity 1: MPI-rank 0 writes
......
configure.ac
View file @ e5c20fd3
......@@ -72,9 +72,6 @@ if test "$enable_ipo" = "yes"; then
fi
fi
# Add libtool support.
LT_INIT
# Check for MPI. Need to do this before characterising the compiler (C99 mode),
# as this changes the compiler.
# We should consider using AX_PROG_CC_MPI to replace AC_PROG_CC when compiling
......@@ -151,6 +148,9 @@ AM_CONDITIONAL([HAVEMPI],[test $enable_mpi = "yes"])
# Indicate that MPIRUN can be modified by an environement variable
AC_ARG_VAR(MPIRUN, Path to the mpirun command if non-standard)
# Add libtool support (now that CC is defined).
LT_INIT
# Need C99 and inline support.
AC_PROG_CC_C99
AC_C_INLINE
......@@ -179,7 +179,7 @@ if test "$enable_opt" = "yes" ; then
ac_test_CFLAGS="yes"
CFLAGS="$old_CFLAGS $CFLAGS"
# Check SSE & AVX support (some overlap with AX_CC_MAXOPT).
# Check SSE & AVX support (some overlap with AX_CC_MAXOPT).
# Don't use the SIMD_FLAGS result with Intel compilers. The -x<code>
# value from AX_CC_MAXOPT should be sufficient.
AX_EXT
......@@ -287,12 +287,75 @@ AC_SUBST([METIS_LIBS])
AC_SUBST([METIS_INCS])
AM_CONDITIONAL([HAVEMETIS],[test -n "$METIS_LIBS"])
# # Check for zlib.
# AC_CHECK_LIB([z],[gzopen],[
# AC_DEFINE([HAVE_LIBZ],[1],[Set to 1 if zlib is installed.])
# LDFLAGS="$LDFLAGS -lz"
# ],[])
# Check for tcmalloc a fast malloc that is part of the gperftools.
have_tcmalloc="no"
AC_ARG_WITH([tcmalloc],
[AS_HELP_STRING([--with-tcmalloc],
[use tcmalloc library or specify the directory with lib @<:@yes/no@:>@]
)],
[with_tcmalloc="$withval"],
[with_tcmalloc="no"]
)
if test "x$with_tcmalloc" != "xno"; then
if test "x$with_tcmalloc" != "xyes" && test "x$with_tcmalloc" != "x"; then
tclibs="-L$with_tcmalloc -ltcmalloc"
else
tclibs="-ltcmalloc"
fi
AC_CHECK_LIB([tcmalloc],[tc_cfree],[have_tcmalloc="yes"],[have_tcmalloc="no"],
$tclibs)
# Could just have the minimal version.
if test "$have_tcmalloc" = "no"; then
if test "x$with_tcmalloc" != "xyes" && test "x$with_tcmalloc" != "x"; then
tclibs="-L$with_tcmalloc -ltcmalloc_minimal"
else
tclibs="-ltcmalloc_minimal"
fi
AC_CHECK_LIB([tcmalloc],[tc_cfree],[have_tcmalloc="yes"],[have_tcmalloc="no"],
$tclibs)
fi
if test "$have_tcmalloc" = "yes"; then
TCMALLOC_LIBS="$tclibs"
# These are recommended for GCC.
if test "$ax_cv_c_compiler_vendor" = "gnu"; then
CFLAGS="$CFLAGS -fno-builtin-malloc -fno-builtin-calloc -fno-builtin-realloc -fno-builtin-free"
fi
else
TCMALLOC_LIBS=""
fi
fi
AC_SUBST([TCMALLOC_LIBS])
AM_CONDITIONAL([HAVETCMALLOC],[test -n "$TCMALLOC_LIBS"])
# Check for -lprofiler usually part of the gpreftools along with tcmalloc.
have_profiler="no"
AC_ARG_WITH([profiler],
[AS_HELP_STRING([--with-profiler],
[use cpu profiler library or specify the directory with lib @<:@yes/no@:>@]
)],
[with_profiler="$withval"],
[with_profiler="yes"]
)
if test "x$with_profiler" != "xno"; then
if test "x$with_profiler" != "xyes" && test "x$with_profiler" != "x"; then
proflibs="-L$with_profiler -lprofiler"
else
proflibs="-lprofiler"
fi
AC_CHECK_LIB([profiler],[ProfilerFlush],[have_profiler="yes"],[have_profiler="no"],
$proflibs)
if test "$have_profiler" = "yes"; then
PROFILER_LIBS="$proflibs"
else
PROFILER_LIBS=""
fi
fi
AC_SUBST([PROFILER_LIBS])
AM_CONDITIONAL([HAVEPROFILER],[test -n "$PROFILER_LIBS"])
# Check for HDF5. This is required.
AX_LIB_HDF5
......@@ -410,6 +473,8 @@ AC_MSG_RESULT([
- parallel : $have_parallel_hdf5
Metis enabled : $have_metis
libNUMA enabled : $have_numa
Using tcmalloc : $have_tcmalloc
CPU profiler : $have_profiler
])
# Generate output.
......
examples/Makefile.am
View file @ e5c20fd3
......@@ -21,13 +21,17 @@ MYFLAGS = -DTIMER
# Add the source directory and debug to CFLAGS
AM_CFLAGS = -I../src $(HDF5_CPPFLAGS)
AM_LDFLAGS =
AM_LDFLAGS = $(HDF5_LDFLAGS)
MPI_THREAD_LIBS = @MPI_THREAD_LIBS@
# Extra libraries.
EXTRA_LIBS = $(HDF5_LIBS) $(PROFILER_LIBS) $(TCMALLOC_LIBS)
# MPI libraries.
MPI_LIBS = $(METIS_LIBS) $(MPI_THREAD_LIBS)
MPI_FLAGS = -DWITH_MPI $(METIS_INCS)
# Set-up the library
# Programs.
bin_PROGRAMS = swift swift_fixdt
# Build MPI versions as well?
......@@ -45,20 +49,20 @@ endif
# Sources for swift
swift_SOURCES = main.c
swift_CFLAGS = $(MYFLAGS) $(AM_CFLAGS) -DENGINE_POLICY="engine_policy_keep $(ENGINE_POLICY_SETAFFINITY)"
swift_LDADD = ../src/.libs/libswiftsim.a $(HDF5_LDFLAGS) $(HDF5_LIBS)
swift_LDADD = ../src/.libs/libswiftsim.a $(EXTRA_LIBS)
swift_fixdt_SOURCES = main.c
swift_fixdt_CFLAGS = $(MYFLAGS) $(AM_CFLAGS) -DENGINE_POLICY="engine_policy_fixdt | engine_policy_keep $(ENGINE_POLICY_SETAFFINITY)"
swift_fixdt_LDADD = ../src/.libs/libswiftsim.a $(HDF5_LDFLAGS) $(HDF5_LIBS)
swift_fixdt_LDADD = ../src/.libs/libswiftsim.a $(EXTRA_LIBS)
# Sources for swift_mpi, do we need an affinity policy for MPI?
swift_mpi_SOURCES = main.c
swift_mpi_CFLAGS = $(MYFLAGS) $(AM_CFLAGS) $(MPI_FLAGS) -DENGINE_POLICY="engine_policy_keep $(ENGINE_POLICY_SETAFFINITY)"
swift_mpi_LDADD = ../src/.libs/libswiftsim_mpi.a $(HDF5_LDFLAGS) $(HDF5_LIBS) $(MPI_LIBS)
swift_mpi_LDADD = ../src/.libs/libswiftsim_mpi.a $(MPI_LIBS) $(EXTRA_LIBS)
swift_fixdt_mpi_SOURCES = main.c
swift_fixdt_mpi_CFLAGS = $(MYFLAGS) $(AM_CFLAGS) $(MPI_FLAGS) -DENGINE_POLICY="engine_policy_fixdt | engine_policy_keep $(ENGINE_POLICY_SETAFFINITY)"
swift_fixdt_mpi_LDADD = ../src/.libs/libswiftsim_mpi.a $(HDF5_LDFLAGS) $(HDF5_LIBS) $(MPI_LIBS)
swift_fixdt_mpi_LDADD = ../src/.libs/libswiftsim_mpi.a $(MPI_LIBS) $(EXTRA_LIBS)
# Scripts to generate ICs
EXTRA_DIST = UniformBox/makeIC.py UniformBox/run.sh UniformBox/uniformBox.yml \
......
examples/main.c
View file @ e5c20fd3
......@@ -74,6 +74,8 @@ void print_help_message() {
printf(" %2s %8s %s\n", "-g", "",
"Run with an external gravitational potential");
printf(" %2s %8s %s\n", "-G", "", "Run with self-gravity");
printf(" %2s %8s %s\n", "-n", "{int}",
"Execute a fixed number of time steps");
printf(" %2s %8s %s\n", "-s", "", "Run with SPH");
printf(" %2s %8s %s\n", "-t", "{int}",
"The number of threads to use on each MPI rank. Defaults to 1 if not "
......@@ -138,6 +140,7 @@ int main(int argc, char *argv[]) {
int with_aff = 0;
int dry_run = 0;
int dump_tasks = 0;
int nsteps = -2;
int with_cosmology = 0;
int with_external_gravity = 0;
int with_self_gravity = 0;
......@@ -150,7 +153,7 @@ int main(int argc, char *argv[]) {
/* Parse the parameters */
int c;
while ((c = getopt(argc, argv, "acdef:gGhst:v:y:")) != -1) switch (c) {
while ((c = getopt(argc, argv, "acdef:gGhn:st:v:y:")) != -1) switch (c) {
case 'a':
with_aff = 1;
break;
......@@ -179,6 +182,13 @@ int main(int argc, char *argv[]) {
case 'h':
if (myrank == 0) print_help_message();
return 0;
case 'n':
if (sscanf(optarg, "%d", &nsteps) != 1) {
if (myrank == 0) printf("Error parsing fixed number of steps.\n");
if (myrank == 0) print_help_message();
return 1;
}
break;
case 's':
with_hydro = 1;
break;
......@@ -322,18 +332,21 @@ int main(int argc, char *argv[]) {
size_t Ngas = 0, Ngpart = 0;
double dim[3] = {0., 0., 0.};
int periodic = 0;
int flag_entropy_ICs = 0;
if (myrank == 0) clocks_gettime(&tic);
#if defined(WITH_MPI)
#if defined(HAVE_PARALLEL_HDF5)
read_ic_parallel(ICfileName, dim, &parts, &gparts, &Ngas, &Ngpart, &periodic,
myrank, nr_nodes, MPI_COMM_WORLD, MPI_INFO_NULL, dry_run);
&flag_entropy_ICs, myrank, nr_nodes, MPI_COMM_WORLD,
MPI_INFO_NULL, dry_run);
#else
read_ic_serial(ICfileName, dim, &parts, &gparts, &Ngas, &Ngpart, &periodic,
myrank, nr_nodes, MPI_COMM_WORLD, MPI_INFO_NULL, dry_run);
&flag_entropy_ICs, myrank, nr_nodes, MPI_COMM_WORLD,
MPI_INFO_NULL, dry_run);
#endif
#else
read_ic_single(ICfileName, dim, &parts, &gparts, &Ngas, &Ngpart, &periodic,
dry_run);
&flag_entropy_ICs, dry_run);
#endif
if (myrank == 0) {
clocks_gettime(&toc);
......@@ -354,7 +367,7 @@ int main(int argc, char *argv[]) {
free(parts);
parts = NULL;
for (size_t k = 0; k < Ngpart; ++k)
if (gparts[k].id > 0) error("Linking problem");
if (gparts[k].id_or_neg_offset < 0) error("Linking problem");
Ngas = 0;
}
......@@ -466,7 +479,7 @@ int main(int argc, char *argv[]) {
#endif
/* Initialise the particles */
engine_init_particles(&e);
engine_init_particles(&e, flag_entropy_ICs);
/* Legend */
if (myrank == 0)
......@@ -474,7 +487,7 @@ int main(int argc, char *argv[]) {
"Updates", "g-Updates", "Wall-clock time", clocks_getunit());
/* Main simulation loop */
for (int j = 0; !engine_is_done(&e); j++) {
for (int j = 0; !engine_is_done(&e) && e.step != nsteps; j++) {
/* Repartition the space amongst the nodes? */
#ifdef WITH_MPI
......
src/Makefile.am
View file @ e5c20fd3
......@@ -24,6 +24,13 @@ AM_LDFLAGS = $(HDF5_LDFLAGS) -version-info 0:0:0
# The git command, if available.
GIT_CMD = @GIT_CMD@
# Additional dependencies for shared libraries.
EXTRA_LIBS = $(HDF5_LIBS) $(PROFILER_LIBS) $(TCMALLOC_LIBS)
# MPI libraries.
MPI_LIBS = $(METIS_LIBS) $(MPI_THREAD_LIBS)
MPI_FLAGS = -DWITH_MPI $(METIS_INCS)
# Build the libswiftsim library
lib_LTLIBRARIES = libswiftsim.la
# Build a MPI-enabled version too?
......@@ -65,11 +72,13 @@ nobase_noinst_HEADERS = approx_math.h atomic.h cycle.h error.h inline.h kernel_h
# Sources and flags for regular library
libswiftsim_la_SOURCES = $(AM_SOURCES)
libswiftsim_la_CFLAGS = $(AM_CFLAGS)
libswiftsim_la_LDFLAGS = $(AM_LDFLAGS) $(EXTRA_LIBS)
# Sources and flags for MPI library
libswiftsim_mpi_la_SOURCES = $(AM_SOURCES)
libswiftsim_mpi_la_CFLAGS = $(AM_CFLAGS) -DWITH_MPI $(METIS_INCS)
libswiftsim_mpi_la_LDFLAGS = $(AM_LDFLAGS) -DWITH_MPI $(METIS_LIBS)
libswiftsim_mpi_la_CFLAGS = $(AM_CFLAGS) $(MPI_FLAGS)
libswiftsim_mpi_la_LDFLAGS = $(AM_LDFLAGS) $(MPI_LIBS) $(EXTRA_LIBS)
libswiftsim_mpi_la_SHORTNAME = mpi
......
src/cell.c
View file @ e5c20fd3
......@@ -128,6 +128,7 @@ int cell_unpack(struct pcell *pc, struct cell *c, struct space *s) {
}
/* Return the total number of unpacked cells. */
c->pcell_size = count;
return count;
}
......@@ -213,6 +214,39 @@ int cell_pack(struct cell *c, struct pcell *pc) {
pc->progeny[k] = -1;
/* Return the number of packed cells used. */
c->pcell_size = count;
return count;
}
int cell_pack_ti_ends(struct cell *c, int *ti_ends) {
/* Pack this cell's data. */
ti_ends[0] = c->ti_end_min;
/* Fill in the progeny, depth-first recursion. */
int count = 1;
for (int k = 0; k < 8; k++)
if (c->progeny[k] != NULL) {
count += cell_pack_ti_ends(c->progeny[k], &ti_ends[count]);
}
/* Return the number of packed values. */
return count;
}
int cell_unpack_ti_ends(struct cell *c, int *ti_ends) {
/* Unpack this cell's data. */
c->ti_end_min = ti_ends[0];
/* Fill in the progeny, depth-first recursion. */
int count = 1;
for (int k = 0; k < 8; k++)
if (c->progeny[k] != NULL) {
count += cell_unpack_ti_ends(c->progeny[k], &ti_ends[count]);
}
/* Return the number of packed values. */
return count;
}
......@@ -376,9 +410,11 @@ void cell_gunlocktree(struct cell *c) {
* @brief Sort the parts into eight bins along the given pivots.
*
* @param c The #cell array to be sorted.
* @param parts_offset Offset of the cell parts array relative to the
* space's parts array, i.e. c->parts - s->parts.
*/
void cell_split(struct cell *c) {
void cell_split(struct cell *c, ptrdiff_t parts_offset) {
int i, j;
const int count = c->count, gcount = c->gcount;
......@@ -496,8 +532,7 @@ void cell_split(struct cell *c) {
}
/* Re-link the gparts. */
for (int k = 0; k < count; k++)
if (parts[k].gpart != NULL) parts[k].gpart->part = &parts[k];
part_relink_gparts(parts, count, parts_offset);
#ifdef SWIFT_DEBUG_CHECKS
/* Verify that _all_ the parts have been assigned to a cell. */
......@@ -592,8 +627,7 @@ void cell_split(struct cell *c) {
}
/* Re-link the parts. */
for (int k = 0; k < gcount; k++)
if (gparts[k].id > 0) gparts[k].part->gpart = &gparts[k];
part_relink_parts(gparts, gcount, parts - parts_offset);
}
/**
......
src/cell.h
View file @ e5c20fd3
......@@ -24,6 +24,9 @@
#define SWIFT_CELL_H
/* Includes. */
#include <stddef.h>
/* Local includes. */
#include "lock.h"
#include "multipole.h"
#include "part.h"
......@@ -122,7 +125,10 @@ struct cell {
struct task *ghost, *init, *drift, *kick;
/* Task receiving data. */
struct task *recv_xv, *recv_rho;
struct task *recv_xv, *recv_rho, *recv_ti;
/* Task send data. */
struct link *send_xv, *send_rho, *send_ti;
/* Tasks for gravity tree. */
struct task *grav_up, *grav_down;
......@@ -175,13 +181,15 @@ struct cell {
((int)(k) + (cdim)[2] * ((int)(j) + (cdim)[1] * (int)(i)))
/* Function prototypes. */
void cell_split(struct cell *c);
void cell_split(struct cell *c, ptrdiff_t parts_offset);
int cell_locktree(struct cell *c);
void cell_unlocktree(struct cell *c);
int cell_glocktree(struct cell *c);
void cell_gunlocktree(struct cell *c);
int cell_pack(struct cell *c, struct pcell *pc);
int cell_unpack(struct pcell *pc, struct cell *c, struct space *s);
int cell_pack_ti_ends(struct cell *c, int *ti_ends);
int cell_unpack_ti_ends(struct cell *c, int *ti_ends);
int cell_getsize(struct cell *c);
int cell_link_parts(struct cell *c, struct part *parts);
int cell_link_gparts(struct cell *c, struct gpart *gparts);
......
src/common_io.c
View file @ e5c20fd3
......@@ -516,12 +516,10 @@ void prepare_dm_gparts(struct gpart* const gparts, size_t Ndm) {
/* Let's give all these gparts a negative id */
for (size_t i = 0; i < Ndm; ++i) {
/* 0 or negative ids are not allowed */
if (gparts[i].id <= 0)
error("0 or negative ID for DM particle %zd: ID=%lld", i, gparts[i].id);
gparts[i].id = -gparts[i].id;
if (gparts[i].id_or_neg_offset <= 0)
error("0 or negative ID for DM particle %zd: ID=%lld", i,
gparts[i].id_or_neg_offset);
}
}
......@@ -555,7 +553,7 @@ void duplicate_hydro_gparts(struct part* const parts,
gparts[i + Ndm].mass = parts[i].mass;
/* Link the particles */
gparts[i + Ndm].part = &parts[i];
gparts[i + Ndm].id_or_neg_offset = -i;
parts[i].gpart = &gparts[i + Ndm];
}
}
......@@ -580,9 +578,8 @@ void collect_dm_gparts(const struct gpart* const gparts, size_t Ntot,
* gparts[i].part); */
/* And collect the DM ones */
if (gparts[i].id < 0LL) {
memcpy(&dmparts[count], &gparts[i], sizeof(struct gpart));
dmparts[count].id = -dmparts[count].id;
if (gparts[i].id_or_neg_offset > 0) {
dmparts[count] = gparts[i];
count++;
}
}
......
src/debug.c
View file @ e5c20fd3
......@@ -59,8 +59,8 @@
*
* (Should be used for debugging only as it runs in O(N).)
*/
void printParticle(struct part *parts, struct xpart *xparts, long long int id,
size_t N) {
void printParticle(const struct part *parts, struct xpart *xparts,
long long int id, size_t N) {
int found = 0;
......@@ -82,24 +82,27 @@ void printParticle(struct part *parts, struct xpart *xparts, long long int id,
* the standard output.
*
* @param gparts The array of g-particles.
* @param parts The array of particles.
* @param id The id too look for.
* @param N The size of the array of g-particles.
*
* (Should be used for debugging only as it runs in O(N).)
*/
void printgParticle(struct gpart *gparts, long long int id, size_t N) {
void printgParticle(const struct gpart *gparts, const struct part *parts,
long long int id, size_t N) {
int found = 0;
/* Look for the particle. */
for (size_t i = 0; i < N; i++)
if (gparts[i].id == -id) {
printf("## gParticle[%zd] (DM) :\n id=%lld ", i, -gparts[i].id);
if (gparts[i].id_or_neg_offset == id) {
printf("## gParticle[%zd] (DM) :\n id=%lld", i, id);
gravity_debug_particle(&gparts[i]);
found = 1;
break;
} else if (gparts[i].id > 0 && gparts[i].part->id == id) {
printf("## gParticle[%zd] (hydro) :\n id=%lld ", i, gparts[i].id);
} else if (gparts[i].id_or_neg_offset < 0 &&
parts[-gparts[i].id_or_neg_offset].id == id) {
printf("## gParticle[%zd] (hydro) :\n id=%lld", i, id);
gravity_debug_particle(&gparts[i]);
found = 1;
break;
......@@ -114,9 +117,9 @@ void printgParticle(struct gpart *gparts, long long int id, size_t N) {
* @param p The particle to print
* @param xp The extended data ot the particle to print
*/
void printParticle_single(struct part *p, struct xpart *xp) {
void printParticle_single(const struct part *p, const struct xpart *xp) {
printf("## Particle: id=%lld ", p->id);
printf("## Particle: id=%lld", p->id);
hydro_debug_particle(p, xp);
printf("\n");
}
......@@ -128,7 +131,7 @@ void printParticle_single(struct part *p, struct xpart *xp) {
*/
void printgParticle_single(struct gpart *gp) {
printf("## g-Particle: id=%lld ", gp->id);
printf("## g-Particle: id=%lld ", gp->id_or_neg_offset);
gravity_debug_particle(gp);
printf("\n");
}
......
src/debug.h
View file @ e5c20fd3
......@@ -19,14 +19,15 @@
#ifndef SWIFT_DEBUG_H
#define SWIFT_DEBUG_H
struct part;
struct gpart;
struct xpart;
/* Includes. */
#include "cell.h"
#include "part.h"
void printParticle(struct part *parts, struct xpart *xparts, long long int id,
size_t N);
void printgParticle(struct gpart *parts, long long int id, size_t N);
void printParticle_single(struct part *p, struct xpart *xp);
void printParticle(const struct part *parts, struct xpart *xparts,
long long int id, size_t N);
void printgParticle(const struct gpart *gparts, const struct part *parts,
long long int id, size_t N);
void printParticle_single(const struct part *p, const struct xpart *xp);
void printgParticle_single(struct gpart *gp);
#ifdef HAVE_METIS
......
src/engine.c
View file @ e5c20fd3
......@@ -128,10 +128,12 @@ void engine_make_gravity_hierarchical_tasks(struct engine *e, struct cell *c,
engine_policy_external_gravity;
const int is_fixdt = (e->policy & engine_policy_fixdt) == engine_policy_fixdt;
/* Is this the super-cell? */
if (super == NULL && (c->grav != NULL || (c->gcount > 0 && !c->split))) {
/* Am I the super-cell? */
/* TODO(pedro): Add a condition for gravity tasks as well. */
if (super == NULL &&
(c->density != NULL || (!c->split && (c->count > 0 || c->gcount > 0)))) {
/* This is the super cell, i.e. the first with gravity tasks attached. */
/* This is the super cell, i.e. the first with density tasks attached. */
super = c;
/* Local tasks only... */
......@@ -331,11 +333,11 @@ void engine_redistribute(struct engine *e) {
}
#ifdef SWIFT_DEBUG_CHECKS
if (s->parts[k].gpart->id < 0)
if (s->parts[k].gpart->id_or_neg_offset >= 0)
error("Trying to link a partnerless gpart !");
#endif
s->parts[k].gpart->id = count_this_dest;
s->parts[k].gpart->id_or_neg_offset = -count_this_dest;
count_this_dest++;
}
}
......@@ -519,13 +521,14 @@ void engine_redistribute(struct engine *e) {
for (size_t k = offset_gparts; k < offset_gparts + count_gparts; ++k) {
/* Does this gpart have a partner ? */
if (gparts_new[k].id >= 0) {
if (gparts_new[k].id_or_neg_offset <= 0) {
const size_t partner_index = offset_parts + gparts_new[k].id;
const ptrdiff_t partner_index =
offset_parts - gparts_new[k].id_or_neg_offset;
/* Re-link */
gparts_new[k].part = &parts_new[partner_index];
gparts_new[k].part->gpart = &gparts_new[k];
gparts_new[k].id_or_neg_offset = -partner_index;
parts_new[partner_index].gpart = &gparts_new[k];
}
}
......@@ -546,22 +549,22 @@ void engine_redistribute(struct engine *e) {
/* Verify that the links are correct */
for (size_t k = 0; k < nr_gparts; ++k) {
if (gparts_new[k].id > 0) {
if (gparts_new[k].id_or_neg_offset <= 0) {
struct part *part = &parts_new[-gparts_new[k].id_or_neg_offset];
if (gparts_new[k].part->gpart != &gparts_new[k])
error("Linking problem !");
if (part->gpart != &gparts_new[k]) error("Linking problem !");
if (gparts_new[k].x[0] != gparts_new[k].part->x[0] ||
gparts_new[k].x[1] != gparts_new[k].part->x[1] ||
gparts_new[k].x[2] != gparts_new[k].part->x[2])
if (gparts_new[k].x[0] != part->x[0] ||
gparts_new[k].x[1] != part->x[1] || gparts_new[k].x[2] != part->x[2])
error("Linked particles are not at the same position !");
}
}
for (size_t k = 0; k < nr_parts; ++k) {
if (parts_new[k].gpart != NULL) {
if (parts_new[k].gpart->part != &parts_new[k]) error("Linking problem !");
if (parts_new[k].gpart != NULL &&
parts_new[k].gpart->id_or_neg_offset != -k) {
error("Linking problem !");
}
}
#endif
......@@ -673,44 +676,63 @@ void engine_addtasks_grav(struct engine *e, struct cell *c, struct task *up,
*
* @param e The #engine.
* @param ci The sending #cell.
* @param cj The receiving #cell
* @param cj Dummy cell containing the nodeID of the receiving node.
* @param t_xv The send_xv #task, if it has already been created.
* @param t_rho The send_rho #task, if it has already been created.
* @param t_ti The send_ti #task, if required and has already been created.
*/
void engine_addtasks_send(struct engine *e, struct cell *ci, struct cell *cj) {
void engine_addtasks_send(struct engine *e, struct cell *ci, struct cell *cj,
struct task *t_xv, struct task *t_rho,