diff --git a/configure.ac b/configure.ac
index 44d21dc30d408260a11a7b7b100df455b98295d5..6abf5a35bb102c6bcfd1716b2506a4ed0ec82905 100644
--- a/configure.ac
+++ b/configure.ac
@@ -236,6 +236,7 @@ AX_PTHREAD([LIBS="$PTHREAD_LIBS $LIBS" CFLAGS="$CFLAGS $PTHREAD_CFLAGS"
# Check for metis. Note AX_LIB_METIS exists, but cannot be configured
# to be default off (i.e. given no option it tries to locate METIS), so we
# don't use that.
+have_metis="no"
AC_ARG_WITH([metis],
[AS_HELP_STRING([--with-metis=PATH],
[root directory where metis is installed @<:@yes/no@:>@]
@@ -251,6 +252,7 @@ if test "x$with_metis" != "xno"; then
METIS_LIBS="-lmetis"
METIS_INCS=""
fi
+ have_metis="yes"
AC_CHECK_LIB([metis],[METIS_PartGraphKway],
AC_DEFINE([HAVE_METIS],1,[The metis library appears to be present.]),
AC_MSG_ERROR(something is wrong with the metis library!),$METIS_LIBS)
@@ -380,7 +382,7 @@ AC_MSG_RESULT([
MPI enabled : $enable_mpi
HDF5 enabled : $with_hdf5
- parallel : $have_parallel_hdf5
- Metis enabled : $with_metis
+ Metis enabled : $have_metis
libNUMA enabled : $have_numa
])
diff --git a/examples/Makefile.am b/examples/Makefile.am
index 44d1b22333767f0d4b4282c7c054a70aa7a114be..b2c09e54800c556ba03bdb1e2436af3144263dcb 100644
--- a/examples/Makefile.am
+++ b/examples/Makefile.am
@@ -20,7 +20,7 @@
MYFLAGS = -DTIMER
# Add the source directory and debug to CFLAGS
-AM_CFLAGS = -I../src -DCPU_TPS=2.67e9 $(HDF5_CPPFLAGS)
+AM_CFLAGS = -I../src $(HDF5_CPPFLAGS)
AM_LDFLAGS =
diff --git a/examples/main.c b/examples/main.c
index a28faa628b856b3f1917331aecc6b95835c628b3..77cb1fbd494c1746d9c7c8be496e51eb5546ce9a 100644
--- a/examples/main.c
+++ b/examples/main.c
@@ -46,11 +46,6 @@
/* Local headers. */
#include "swift.h"
-/* Ticks per second on this machine. */
-#ifndef CPU_TPS
-#define CPU_TPS 2.40e9
-#endif
-
/* Engine policy flags. */
#ifndef ENGINE_POLICY
#define ENGINE_POLICY engine_policy_none
@@ -75,13 +70,14 @@ int main(int argc, char *argv[]) {
struct space s;
struct engine e;
struct UnitSystem us;
+ struct clocks_time tic, toc;
char ICfileName[200] = "";
char dumpfile[30];
float dt_max = 0.0f, dt_min = 0.0f;
- ticks tic;
int nr_nodes = 1, myrank = 0;
FILE *file_thread;
int with_outputs = 1;
+ unsigned long long cpufreq = 0;
#ifdef WITH_MPI
struct partition initial_partition;
@@ -154,7 +150,7 @@ int main(int argc, char *argv[]) {
bzero(&s, sizeof(struct space));
/* Parse the options */
- while ((c = getopt(argc, argv, "a:c:d:e:f:m:oP:q:R:s:t:w:y:z:")) != -1)
+ while ((c = getopt(argc, argv, "a:c:d:e:f:h:m:oP:q:R:s:t:w:y:z:")) != -1)
switch (c) {
case 'a':
if (sscanf(optarg, "%lf", &scaling) != 1)
@@ -171,7 +167,7 @@ int main(int argc, char *argv[]) {
case 'd':
if (sscanf(optarg, "%f", &dt_min) != 1)
error("Error parsing minimal timestep.");
- if (myrank == 0) message("dt_min set to %e.", dt_max);
+ if (myrank == 0) message("dt_min set to %e.", dt_min);
fflush(stdout);
break;
case 'e':
@@ -183,6 +179,12 @@ int main(int argc, char *argv[]) {
case 'f':
if (!strcpy(ICfileName, optarg)) error("Error parsing IC file name.");
break;
+ case 'h':
+ if (sscanf(optarg, "%llu", &cpufreq) != 1)
+ error("Error parsing CPU frequency.");
+ if (myrank == 0) message("CPU frequency set to %llu.", cpufreq);
+ fflush(stdout);
+ break;
case 'm':
if (sscanf(optarg, "%lf", &h_max) != 1) error("Error parsing h_max.");
if (myrank == 0) message("maximum h set to %e.", h_max);
@@ -192,8 +194,8 @@ int main(int argc, char *argv[]) {
with_outputs = 0;
break;
case 'P':
- /* Partition type is one of "g", "m", "w", or "v"; "g" can be
- * followed by three numbers defining the grid. */
+/* Partition type is one of "g", "m", "w", or "v"; "g" can be
+ * followed by three numbers defining the grid. */
#ifdef WITH_MPI
switch (optarg[0]) {
case 'g':
@@ -224,7 +226,7 @@ int main(int argc, char *argv[]) {
error("Error parsing number of queues.");
break;
case 'R':
- /* Repartition type "n", "b", "v", "e" or "x".
+ /* Repartition type "n", "b", "v", "e" or "x".
* Note only none is available without METIS. */
#ifdef WITH_MPI
switch (optarg[0]) {
@@ -323,6 +325,12 @@ int main(int argc, char *argv[]) {
aFactor(&us, UNIT_CONV_ENTROPY), hFactor(&us, UNIT_CONV_ENTROPY));
}
+ /* Initialize CPU frequency. */
+ clocks_set_cpufreq(cpufreq);
+ cpufreq = clocks_get_cpufreq();
+ if (myrank == 0)
+ message("CPU frequency used for tick conversion: %llu Hz", cpufreq);
+
/* Check we have sensible time step bounds */
if (dt_min > dt_max)
error("Minimal time step size must be large than maximal time step size ");
@@ -332,7 +340,9 @@ int main(int argc, char *argv[]) {
error("An IC file name must be provided via the option -f");
/* Read particles and space information from (GADGET) IC */
- tic = getticks();
+
+ if (myrank == 0)
+ clocks_gettime(&tic);
#if defined(WITH_MPI)
#if defined(HAVE_PARALLEL_HDF5)
read_ic_parallel(ICfileName, dim, &parts, &N, &periodic, myrank, nr_nodes,
@@ -345,10 +355,12 @@ int main(int argc, char *argv[]) {
read_ic_single(ICfileName, dim, &parts, &N, &periodic);
#endif
- if (myrank == 0)
- message("reading particle properties took %.3f ms.",
- ((double)(getticks() - tic)) / CPU_TPS * 1000);
- fflush(stdout);
+ if (myrank == 0) {
+ clocks_gettime(&toc);
+ message("reading particle properties took %.3f %s.",
+ clocks_diff(&tic, &toc), clocks_getunit());
+ fflush(stdout);
+ }
#if defined(WITH_MPI)
long long N_long = N;
@@ -374,12 +386,15 @@ int main(int argc, char *argv[]) {
if (nr_queues < 0) nr_queues = nr_threads;
/* Initialize the space with this data. */
- tic = getticks();
- space_init(&s, dim, parts, N, periodic, h_max, myrank == 0);
if (myrank == 0)
- message("space_init took %.3f ms.",
- ((double)(getticks() - tic)) / CPU_TPS * 1000);
- fflush(stdout);
+ clocks_gettime(&tic);
+ space_init(&s, dim, parts, N, periodic, h_max, myrank == 0);
+ if (myrank == 0) {
+ clocks_gettime(&toc);
+ message("space_init took %.3f %s.", clocks_diff(&tic, &toc),
+ clocks_getunit());
+ fflush(stdout);
+ }
/* Say a few nice things about the space we just created. */
if (myrank == 0) {
@@ -408,15 +423,18 @@ int main(int argc, char *argv[]) {
}
/* Initialize the engine with this space. */
- tic = getticks();
+ if (myrank == 0)
+ clocks_gettime(&tic);
if (myrank == 0) message("nr_nodes is %i.", nr_nodes);
engine_init(&e, &s, dt_max, nr_threads, nr_queues, nr_nodes, myrank,
ENGINE_POLICY | engine_policy_steal | engine_policy_hydro, 0,
time_end, dt_min, dt_max);
- if (myrank == 0)
- message("engine_init took %.3f ms.",
- ((double)(getticks() - tic)) / CPU_TPS * 1000);
- fflush(stdout);
+ if (myrank == 0) {
+ clocks_gettime(&toc);
+ message("engine_init took %.3f %s.", clocks_diff(&tic, &toc),
+ clocks_getunit());
+ fflush(stdout);
+ }
#ifdef WITH_MPI
/* Split the space. */
@@ -427,7 +445,8 @@ int main(int argc, char *argv[]) {
if (with_outputs) {
/* Write the state of the system as it is before starting time integration.
*/
- tic = getticks();
+ if (myrank == 0)
+ clocks_gettime(&tic);
#if defined(WITH_MPI)
#if defined(HAVE_PARALLEL_HDF5)
write_output_parallel(&e, &us, myrank, nr_nodes, MPI_COMM_WORLD,
@@ -439,10 +458,12 @@ int main(int argc, char *argv[]) {
#else
write_output_single(&e, &us);
#endif
- if (myrank == 0)
- message("writing particle properties took %.3f ms.",
- ((double)(getticks() - tic)) / CPU_TPS * 1000);
- fflush(stdout);
+ if (myrank == 0) {
+ clocks_gettime(&toc);
+ message("writing particle properties took %.3f %s.",
+ clocks_diff(&tic, &toc), clocks_getunit());
+ fflush(stdout);
+ }
}
/* Init the runner history. */
@@ -465,10 +486,10 @@ int main(int argc, char *argv[]) {
if (myrank == 0)
printf(
"# Step Time time-step Number of updates CPU Wall-clock time "
- "[ms]\n");
+ "[%s]\n", clocks_getunit());
/* Let loose a runner on the space. */
- for (j = 0; e.time < time_end; j++) {
+ for (j = 0; !engine_is_done(&e); j++) {
/* Repartition the space amongst the nodes? */
#ifdef WITH_MPI
@@ -571,14 +592,14 @@ int main(int argc, char *argv[]) {
* e.count_step, */
/* e.dt_min, e.dt_max); */
/* for (k = 0; k < timer_count; k++) */
- /* printf(" %.3f", ((double)timers[k]) / CPU_TPS * 1000); */
+ /* printf(" %.3f", clocks_from_ticks(timers[k]); */
/* printf("\n"); */
/* fflush(stdout); */
/* } */
/* if (myrank == 0) { */
/* printf("%i %e", j, e.time); */
- /* printf(" %.3f", ((double)timers[timer_count - 1]) / CPU_TPS * 1000); */
+ /* printf(" %.3f", clocks_from_ticks(timers[timer_count - 1]); */
/* printf("\n"); */
/* fflush(stdout); */
/* } */
diff --git a/src/Makefile.am b/src/Makefile.am
index f4b293925bfcfdf989889db768d60e3de778c985..8a4e7cdc4b849e8f99bdf9a3dbda9b30fa986332 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -17,7 +17,7 @@
# along with this program. If not, see <http://www.gnu.org/licenses/>.
# Add the debug flag to the whole thing
-AM_CFLAGS = -DTIMER -DCOUNTER -DCPU_TPS=2.30e9 $(HDF5_CPPFLAGS)
+AM_CFLAGS = -DTIMER -DCOUNTER $(HDF5_CPPFLAGS)
# Assign a "safe" version number
AM_LDFLAGS = $(LAPACK_LIBS) $(BLAS_LIBS) $(HDF5_LDFLAGS) -version-info 0:0:0 # -fsanitize=address
@@ -35,13 +35,13 @@ endif
# List required headers
include_HEADERS = space.h runner.h queue.h task.h lock.h cell.h part.h const.h \
engine.h swift.h serial_io.h timers.h debug.h scheduler.h proxy.h parallel_io.h \
- common_io.h single_io.h multipole.h map.h tools.h partition.h
+ common_io.h single_io.h multipole.h map.h tools.h partition.h clocks.h
# Common source files
AM_SOURCES = space.c runner.c queue.c task.c cell.c engine.c \
serial_io.c timers.c debug.c scheduler.c proxy.c parallel_io.c \
units.c common_io.c single_io.c multipole.c version.c map.c \
- kernel.c tools.c part.c partition.c
+ kernel.c tools.c part.c partition.c clocks.c
# Include files for distribution, not installation.
nobase_noinst_HEADERS = approx_math.h atomic.h cycle.h error.h inline.h kernel.h vector.h \
diff --git a/src/clocks.c b/src/clocks.c
new file mode 100644
index 0000000000000000000000000000000000000000..2d2cac90ddb5e7dca1fb5feddb1c8086c57458d8
--- /dev/null
+++ b/src/clocks.c
@@ -0,0 +1,224 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (c) 2016 Peter W. Draper (p.w.draper@durham.ac.uk)
+ *
+ * This program 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 3 of the License, or
+ * (at your option) any later version.
+ *
+ * This program 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ *
+ ******************************************************************************/
+
+/**
+ * @file clocks.c
+ * @brief support for measuring intervals in milli seconds, when that
+ * is possible, otherwise ticks.
+ *
+ * Use cycle.h or timers.h for relative times.
+ */
+
+/* Config parameters. */
+#include "../config.h"
+
+/* Standard headers. */
+#include <stdio.h>
+#include <unistd.h>
+
+/* Local headers. */
+#include "clocks.h"
+
+/* 0.25 of a second in nanoseconds. */
+#define SLEEPTIME 250000000
+
+/* The CPU frequency used to convert ticks to seconds. */
+static unsigned long long clocks_cpufreq = 0;
+
+/* The units of any returned times. */
+static char *clocks_units[] = {"ms", "ticks"};
+static int clocks_units_index = 0;
+static double clocks_units_scale = 1000.0;
+
+/* Local prototypes. */
+static void clocks_estimate_cpufreq();
+
+/**
+ * @brief Get the current time.
+ *
+ * @param time the current time.
+ */
+void clocks_gettime(struct clocks_time *time) {
+
+#ifdef HAVE_CLOCK_GETTIME
+ clock_gettime(CLOCK_REALTIME, &time->time);
+#else
+ time->time = getticks();
+#endif
+}
+
+/**
+ * @brief Get difference in between two times.
+ *
+ * @param start the start time.
+ * @param end the end time.
+ *
+ * @return the difference.
+ */
+double clocks_diff(struct clocks_time *start, struct clocks_time *end) {
+#ifdef HAVE_CLOCK_GETTIME
+ struct timespec temp;
+ if ((end->time.tv_nsec - start->time.tv_nsec) < 0) {
+ temp.tv_sec = end->time.tv_sec - start->time.tv_sec - 1;
+ temp.tv_nsec = 1000000000 + end->time.tv_nsec - start->time.tv_nsec;
+ } else {
+ temp.tv_sec = end->time.tv_sec - start->time.tv_sec;
+ temp.tv_nsec = end->time.tv_nsec - start->time.tv_nsec;
+ }
+ return (double)temp.tv_sec * 1000.0 + (double)temp.tv_nsec * 1.0E-6;
+#else
+ return elapsed(end->time, start->time) / clocks_get_cpufreq() * clocks_units_scale;
+#endif
+}
+
+/**
+ * @brief Set the CPU frequency.
+ *
+ * This function should be called at least once to set the CPU frequency.
+ * To use the builtin estimation techniques give a value of 0.
+ *
+ * @param freq the CPU frequency in Hz or 0 to estimate one.
+ */
+void clocks_set_cpufreq(unsigned long long freq) {
+ if (freq > 0) {
+ clocks_cpufreq = freq;
+ } else {
+ clocks_estimate_cpufreq();
+ }
+}
+
+/**
+ * @brief Get the CPU frequency in Hz.
+ *
+ * @result the CPU frequency.
+ */
+unsigned long long clocks_get_cpufreq() {
+
+ if (clocks_cpufreq > 0) return clocks_cpufreq;
+
+ /* It not already set estimate it. */
+ clocks_estimate_cpufreq();
+ return clocks_cpufreq;
+}
+
+/**
+ * @brief Estimate the CPU frequency in Hz.
+ *
+ * If already set return the CPU frequency, then estimate the CPU frequency.
+ *
+ * The technique is either use a clock timed nanosleep (this was the best
+ * method on i7), to read the value from the cpuinfo_max_freq
+ * file (probably a overestimate) or finally just use a value of 1 with
+ * time units of ticks.
+ */
+static void clocks_estimate_cpufreq() {
+
+#ifdef HAVE_CLOCK_GETTIME
+ /* Try to time a nanosleep() in ticks. */
+ struct clocks_time time1;
+ struct clocks_time time2;
+
+ struct timespec sleep;
+ sleep.tv_sec = 0;
+ sleep.tv_nsec = SLEEPTIME;
+
+ clocks_gettime(&time1);
+ ticks tic = getticks();
+
+ /* Could do some calculation, but constant_tsc should protect us. */
+ nanosleep(&sleep, NULL);
+
+ clocks_gettime(&time2);
+ ticks toc = getticks();
+ double realsleep = clocks_diff(&time1, &time2);
+
+ clocks_cpufreq =
+ (signed long long)(double)(toc - tic) * 1.0 / realsleep * 1000.0;
+ clocks_units_index = 0;
+ clocks_units_scale = 1000.0;
+#endif
+
+/* Look for the system value, if available. Tends to be too large. */
+#ifdef __linux__
+ if (clocks_cpufreq == 0) {
+ FILE *file =
+ fopen("/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq", "r");
+ if (file != NULL) {
+ unsigned long long maxfreq;
+ if (fscanf(file, "%llu", &maxfreq) == 1) {
+ clocks_cpufreq = maxfreq * 1000;
+ clocks_units_index = 0;
+ clocks_units_scale = 1000.0;
+ }
+ fclose(file);
+ }
+ }
+#endif
+
+ /* If all fails just report ticks in any times. */
+ if (clocks_cpufreq == 0) {
+ clocks_cpufreq = 1;
+ clocks_units_index = 1;
+ clocks_units_scale = 1.0;
+ }
+}
+
+/**
+ * @brief Return the difference between two ticks.
+ *
+ * Only an approximation as based on how well we have estimated the
+ * rtc frequency. Should be good for machines that support constant_rtc
+ * and clock_gettime().
+ *
+ * @param tic a number of ticks returned by the cycle.h getticks() function.
+ * @param toc a number of ticks returned by the cycle.h getticks() function.
+ *
+ * @result the difference.
+ */
+double clocks_diff_ticks(ticks tic, ticks toc) {
+ return clocks_from_ticks(tic - toc);
+}
+
+/**
+ * @brief Convert a number of ticks into milli seconds, if possible.
+ *
+ * Only an approximation as based on how well we have estimated the
+ * rtc frequency. Should be good for machines that support constant_rtc
+ * and clock_gettime(), and reasonable for most Linux machines, otherwise
+ * ticks will just be returned. See clocks_getunit() for the actual units.
+ *
+ * @param tics a number of ticks returned by the cycle.h getticks() function.
+ *
+ * @result the milli seconds, if possible.
+ */
+double clocks_from_ticks(ticks tics) {
+ return ((double)tics / (double)clocks_get_cpufreq() * clocks_units_scale);
+}
+
+/**
+ * @brief return the time units.
+ *
+ * Normally "ms" for milliseconds, but can be "ticks" when no conversion
+ * factor is available.
+ *
+ * @result the current time units.
+ */
+const char *clocks_getunit() {
+ return clocks_units[clocks_units_index];
+}
diff --git a/src/clocks.h b/src/clocks.h
new file mode 100644
index 0000000000000000000000000000000000000000..59b9f52b0303ac74dd237c7fd19cf689cac99cc3
--- /dev/null
+++ b/src/clocks.h
@@ -0,0 +1,43 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (c) 2016 Peter W. Draper (p.w.draper@durham.ac.uk)
+ *
+ * This program 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 3 of the License, or
+ * (at your option) any later version.
+ *
+ * This program 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ *
+ ******************************************************************************/
+#ifndef SWIFT_CLOCKS_H
+#define SWIFT_CLOCKS_H
+
+#include <time.h>
+#include "cycle.h"
+
+/* Struct to record a time for the clocks functions. */
+struct clocks_time {
+#ifdef HAVE_CLOCK_GETTIME
+ struct timespec time;
+#else
+ ticks time;
+#endif
+};
+
+void clocks_gettime(struct clocks_time *time);
+double clocks_diff(struct clocks_time *start, struct clocks_time *end);
+const char *clocks_getunit();
+
+void clocks_set_cpufreq(unsigned long long freq);
+unsigned long long clocks_get_cpufreq();
+double clocks_from_ticks(ticks tics);
+double clocks_diff_ticks(ticks tic, ticks toc);
+
+#endif /* SWIFT_CLOCKS_H */
diff --git a/src/engine.c b/src/engine.c
index 272d5c0c9ff07b98befed3c68810d4c909e78c57..3c8c1f60a1529735d8dfe9992a6538197d90ba9e 100644
--- a/src/engine.c
+++ b/src/engine.c
@@ -46,6 +46,7 @@
/* Local headers. */
#include "atomic.h"
#include "cell.h"
+#include "clocks.h"
#include "cycle.h"
#include "debug.h"
#include "error.h"
@@ -285,7 +286,6 @@ void engine_redistribute(struct engine *e) {
#endif
}
-
/**
* @brief Repartition the cells amongst the nodes.
*
@@ -1116,7 +1116,7 @@ int engine_marktasks(struct engine *e) {
}
}
- // message( "took %.3f ms." , (double)(getticks() - tic)/CPU_TPS*1000 );
+ // message( "took %.3f %s." , clocks_from_ticks(getticks() - tic), clocks_getunit());
/* All is well... */
return 0;
@@ -1167,29 +1167,29 @@ void engine_rebuild(struct engine *e) {
/* Re-build the space. */
// tic = getticks();
space_rebuild(e->s, 0.0, e->nodeID == 0);
-// message( "space_rebuild took %.3f ms." , (double)(getticks() -
-// tic)/CPU_TPS*1000 );
+ // message( "space_rebuild took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
/* If in parallel, exchange the cell structure. */
#ifdef WITH_MPI
// tic = getticks();
engine_exchange_cells(e);
-// message( "engine_exchange_cells took %.3f ms." , (double)(getticks() -
-// tic)/CPU_TPS*1000 );
+ // message( "engine_exchange_cells took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
#endif
/* Re-build the tasks. */
// tic = getticks();
engine_maketasks(e);
- // message( "engine_maketasks took %.3f ms." , (double)(getticks() -
- // tic)/CPU_TPS*1000 );
+ // message( "engine_maketasks took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
/* Run through the tasks and mark as skip or not. */
// tic = getticks();
if (engine_marktasks(e))
error("engine_marktasks failed after space_rebuild.");
- // message( "engine_marktasks took %.3f ms." , (double)(getticks() -
- // tic)/CPU_TPS*1000 );
+ // message( "engine_marktasks took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
/* Print the status of the system */
engine_print(e);
@@ -1210,8 +1210,8 @@ void engine_prepare(struct engine *e) {
/* Run through the tasks and mark as skip or not. */
// tic = getticks();
rebuild = (e->forcerebuild || engine_marktasks(e));
-// message( "space_marktasks took %.3f ms." , (double)(getticks() -
-// tic)/CPU_TPS*1000 );
+ // message( "space_marktasks took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
/* Collect the values of rebuild from all nodes. */
#ifdef WITH_MPI
@@ -1221,8 +1221,8 @@ void engine_prepare(struct engine *e) {
MPI_SUCCESS)
error("Failed to aggregate the rebuild flag across nodes.");
rebuild = buff;
-// message( "rebuild allreduce took %.3f ms." , (double)(getticks() -
-// tic)/CPU_TPS*1000 );
+ // message( "rebuild allreduce took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
#endif
e->tic_step = getticks();
@@ -1230,16 +1230,16 @@ void engine_prepare(struct engine *e) {
if (rebuild) {
// tic = getticks();
engine_rebuild(e);
- // message( "engine_rebuild took %.3f ms." , (double)(getticks() -
- // tic)/CPU_TPS*1000 );
+ // message( "engine_rebuild took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
}
/* Re-rank the tasks every now and then. */
if (e->tasks_age % engine_tasksreweight == 1) {
// tic = getticks();
scheduler_reweight(&e->sched);
- // message( "scheduler_reweight took %.3f ms." , (double)(getticks() -
- // tic)/CPU_TPS*1000 );
+ // message( "scheduler_reweight took %.3f %s." ,
+ //clocks_from_ticks(getticks() -tic), clocks_getunit());
}
e->tasks_age += 1;
@@ -1393,7 +1393,7 @@ void engine_init_particles(struct engine *e) {
struct space *s = e->s;
- if(e->nodeID == 0) message("Initialising particles");
+ if (e->nodeID == 0) message("Initialising particles");
/* Make sure all particles are ready to go */
/* i.e. clean-up any stupid state in the ICs */
@@ -1482,6 +1482,9 @@ void engine_step(struct engine *e) {
TIMER_TIC2;
+ struct clocks_time time1, time2;
+ clocks_gettime(&time1);
+
/* Collect the cell data. */
for (k = 0; k < s->nr_cells; k++)
if (s->cells[k].nodeID == e->nodeID) {
@@ -1620,11 +1623,20 @@ void engine_step(struct engine *e) {
TIMER_TOC2(timer_step);
- e->wallclock_time = ((double)timers[timer_count - 1]) / CPU_TPS * 1000;
+ clocks_gettime(&time2);
+
+ e->wallclock_time = (float) clocks_diff(&time1, &time2);
// printParticle(e->s->parts, e->s->xparts,1000, e->s->nr_parts);
// printParticle(e->s->parts, e->s->xparts,515050, e->s->nr_parts);
}
+/**
+ * @brief Returns 1 if the simulation has reached its end point, 0 otherwise
+ */
+int engine_is_done(struct engine *e) {
+ return !(e->ti_current < max_nr_timesteps);
+}
+
/**
* @brief Create and fill the proxies.
*
@@ -1831,7 +1843,7 @@ void engine_init(struct engine *e, struct space *s, float dt, int nr_threads,
int home = numa_node_of_cpu(sched_getcpu()), half = nr_cores / 2;
bool done = false, swap_hyperthreads = hyperthreads_present();
if (swap_hyperthreads && nodeID == 0)
- message("prefer physical cores to hyperthreads");
+ message("prefer physical cores to hyperthreads");
while (!done) {
done = true;
@@ -1928,15 +1940,25 @@ void engine_init(struct engine *e, struct space *s, float dt, int nr_threads,
engine_print_policy(e);
/* Print information about the hydro scheme */
- if (e->nodeID == 0)
- message("Hydrodynamic scheme: %s", SPH_IMPLEMENTATION);
+ if (e->nodeID == 0) message("Hydrodynamic scheme: %s", SPH_IMPLEMENTATION);
+
+ /* Check we have sensible time bounds */
+ if (timeBegin >= timeEnd)
+ error(
+ "Final simulation time (t_end = %e) must be larger than the start time "
+ "(t_beg = %e)",
+ timeEnd, timeBegin);
+
+ /* Check we have sensible time step bounds */
+ if (e->dt_min > e->dt_max)
+ error(
+ "Minimal time step size must be smaller than maximal time step size ");
/* Deal with timestep */
e->timeBase = (timeEnd - timeBegin) / max_nr_timesteps;
e->ti_current = 0;
- if (e->nodeID == 0)
- message("Absolute minimal timestep size: %e", e->timeBase);
+ /* Fixed time-step case */
if ((e->policy & engine_policy_fixdt) == engine_policy_fixdt) {
e->dt_min = e->dt_max;
@@ -1947,11 +1969,32 @@ void engine_init(struct engine *e, struct space *s, float dt, int nr_threads,
e->dt_min = e->dt_max = dti_timeline * e->timeBase;
if (e->nodeID == 0) message("Timestep set to %e", e->dt_max);
+ } else {
+
+ if (e->nodeID == 0) {
+ message("Absolute minimal timestep size: %e", e->timeBase);
+
+ float dt_min = timeEnd - timeBegin;
+ while (dt_min > e->dt_min) dt_min /= 2.f;
+
+ message("Minimal timestep size (on time-line): %e", dt_min);
+
+ float dt_max = timeEnd - timeBegin;
+ while (dt_max > e->dt_max) dt_max /= 2.f;
+
+ message("Maximal timestep size (on time-line): %e", dt_max);
+ }
}
if (e->dt_min < e->timeBase && e->nodeID == 0)
- error("Minimal timestep smaller than the absolue possible minimum dt=%e",
- e->timeBase);
+ error(
+ "Minimal time-step size smaller than the absolute possible minimum "
+ "dt=%e",
+ e->timeBase);
+
+ if (e->dt_max > (e->timeEnd - e->timeBegin) && e->nodeID == 0)
+ error("Maximal time-step size larger than the simulation run time t=%e",
+ e->timeEnd - e->timeBegin);
/* Construct types for MPI communications */
#ifdef WITH_MPI
diff --git a/src/engine.h b/src/engine.h
index 13c7ec40612713d3771543150763da6924144c1a..cd189f6ac64b0809eec7f9db8cba898a6377cf27 100644
--- a/src/engine.h
+++ b/src/engine.h
@@ -63,7 +63,6 @@ extern const char *engine_policy_names[];
#define engine_maxproxies 64
#define engine_tasksreweight 10
-
/* The rank of the engine as a global variable (for messages). */
extern int engine_rank;
@@ -186,5 +185,6 @@ void engine_makeproxies(struct engine *e);
void engine_redistribute(struct engine *e);
struct link *engine_addlink(struct engine *e, struct link *l, struct task *t);
void engine_print_policy(struct engine *e);
+int engine_is_done(struct engine *e);
#endif /* SWIFT_ENGINE_H */
diff --git a/src/error.h b/src/error.h
index 14c6518efa99b4c5b2a5862619426f30d7f69b7b..2e9a6804303a9b60123403f7a5bbacfc40f9366c 100644
--- a/src/error.h
+++ b/src/error.h
@@ -56,25 +56,25 @@ extern int engine_rank;
* followed by the MPI error string and aborts.
*
*/
-#define mpi_error(res,s, ...) \
+#define mpi_error(res, s, ...) \
{ \
fprintf(stderr, "[%03i] %s:%s():%i: " s "\n", engine_rank, __FILE__, \
__FUNCTION__, __LINE__, ##__VA_ARGS__); \
int len = 1024; \
char buf[len]; \
- MPI_Error_string( res, buf, &len ); \
- fprintf(stderr, "%s\n\n", buf ); \
+ MPI_Error_string(res, buf, &len); \
+ fprintf(stderr, "%s\n\n", buf); \
MPI_Abort(MPI_COMM_WORLD, -1); \
}
-#define mpi_error_string(res,s, ...) \
+#define mpi_error_string(res, s, ...) \
{ \
fprintf(stderr, "[%03i] %s:%s():%i: " s "\n", engine_rank, __FILE__, \
__FUNCTION__, __LINE__, ##__VA_ARGS__); \
int len = 1024; \
char buf[len]; \
- MPI_Error_string( res, buf, &len ); \
- fprintf(stderr, "%s\n\n", buf ); \
+ MPI_Error_string(res, buf, &len); \
+ fprintf(stderr, "%s\n\n", buf); \
}
#endif
diff --git a/src/partition.c b/src/partition.c
index f1fe2b2a8436994de13501dc4ce6714ec1d2e92d..0f8eb3ebe334d71228510307dd9ccc4e56e234b3 100644
--- a/src/partition.c
+++ b/src/partition.c
@@ -58,11 +58,8 @@
/* Simple descriptions of initial partition types for reports. */
const char *initial_partition_name[] = {
- "gridded cells",
- "vectorized point associated cells",
- "METIS particle weighted cells",
- "METIS unweighted cells"
-};
+ "gridded cells", "vectorized point associated cells",
+ "METIS particle weighted cells", "METIS unweighted cells"};
/* Simple descriptions of repartition types for reports. */
const char *repartition_name[] = {
@@ -70,8 +67,7 @@ const char *repartition_name[] = {
"METIS edge and vertex time weighted cells",
"METIS particle count vertex weighted cells",
"METIS time edge weighted cells",
- "METIS particle count vertex and time edge cells"
-};
+ "METIS particle count vertex and time edge cells"};
/* Local functions, if needed. */
static int check_complete(struct space *s, int verbose, int nregions);
@@ -229,7 +225,6 @@ static void graph_init_metis(struct space *s, idx_t *adjncy, idx_t *xadj) {
if (xadj != NULL) {
xadj[0] = 0;
for (int k = 0; k < s->nr_cells; k++) xadj[k + 1] = xadj[k] + 26;
-
}
}
#endif
@@ -283,8 +278,7 @@ static void accumulate_counts(struct space *s, int *counts) {
*/
static void split_metis(struct space *s, int nregions, int *celllist) {
- for (int i = 0; i < s->nr_cells; i++)
- s->cells[i].nodeID = celllist[i];
+ for (int i = 0; i < s->nr_cells; i++) s->cells[i].nodeID = celllist[i];
}
#endif
@@ -404,7 +398,6 @@ static void pick_metis(struct space *s, int nregions, int *vertexw, int *edgew,
free(xadj);
free(adjncy);
free(regionid);
-
}
#endif
@@ -423,10 +416,9 @@ static void pick_metis(struct space *s, int nregions, int *vertexw, int *edgew,
* @param tasks the completed tasks from the last engine step for our node.
* @param nr_tasks the number of tasks.
*/
-static void repart_edge_metis(int partweights, int bothweights,
- int nodeID, int nr_nodes, struct space *s,
- struct task *tasks, int nr_tasks) {
-
+static void repart_edge_metis(int partweights, int bothweights, int nodeID,
+ int nr_nodes, struct space *s, struct task *tasks,
+ int nr_tasks) {
/* Create weight arrays using task ticks for vertices and edges (edges
* assume the same graph structure as used in the part_ calls). */
@@ -453,7 +445,7 @@ static void repart_edge_metis(int partweights, int bothweights,
bzero(weights_v, sizeof(int) * nr_cells);
}
if ((weights_e = (int *)malloc(sizeof(int) * 26 * nr_cells)) == NULL)
- error("Failed to allocate edge weights arrays.");
+ error("Failed to allocate edge weights arrays.");
bzero(weights_e, sizeof(int) * 26 * nr_cells);
/* Generate task weights for vertices. */
@@ -503,18 +495,15 @@ static void repart_edge_metis(int partweights, int bothweights,
if (t->type == task_type_ghost || t->type == task_type_drift ||
t->type == task_type_kick) {
/* Particle updates add only to vertex weight. */
- if (taskvweights)
- weights_v[cid] += w;
+ if (taskvweights) weights_v[cid] += w;
}
/* Self interaction? */
else if ((t->type == task_type_self && ci->nodeID == nodeID) ||
- (t->type == task_type_sub && cj == NULL &&
- ci->nodeID == nodeID)) {
+ (t->type == task_type_sub && cj == NULL && ci->nodeID == nodeID)) {
/* Self interactions add only to vertex weight. */
- if (taskvweights)
- weights_v[cid] += w;
+ if (taskvweights) weights_v[cid] += w;
}
@@ -524,8 +513,7 @@ static void repart_edge_metis(int partweights, int bothweights,
/* In-cell pair? */
if (ci == cj) {
/* Add weight to vertex for ci. */
- if (taskvweights)
- weights_v[cid] += w;
+ if (taskvweights) weights_v[cid] += w;
}
@@ -585,8 +573,8 @@ static void repart_edge_metis(int partweights, int bothweights,
}
if ((res = MPI_Reduce((nodeID == 0) ? MPI_IN_PLACE : weights_e, weights_e,
- 26 * nr_cells, MPI_INT, MPI_SUM, 0,
- MPI_COMM_WORLD)) != MPI_SUCCESS)
+ 26 * nr_cells, MPI_INT, MPI_SUM, 0, MPI_COMM_WORLD)) !=
+ MPI_SUCCESS)
mpi_error(res, "Failed to allreduce edge weights.");
/* Allocate cell list for the partition. */
@@ -655,8 +643,9 @@ static void repart_edge_metis(int partweights, int bothweights,
/* If partition failed continue with the current one, but make this
* clear. */
if (failed) {
- message("WARNING: METIS repartition has failed, continuing with "
- "the current partition, load balance will not be optimal");
+ message(
+ "WARNING: METIS repartition has failed, continuing with "
+ "the current partition, load balance will not be optimal");
for (int k = 0; k < nr_cells; k++) celllist[k] = cells[k].nodeID;
}
}
@@ -697,16 +686,15 @@ static void repart_vertex_metis(struct space *s, int nodeID, int nr_nodes) {
/* Get all the counts from all the nodes. */
int res;
- if ((res = MPI_Allreduce(MPI_IN_PLACE, weights, s->nr_cells, MPI_INT,
- MPI_SUM, MPI_COMM_WORLD)) != MPI_SUCCESS)
+ if ((res = MPI_Allreduce(MPI_IN_PLACE, weights, s->nr_cells, MPI_INT, MPI_SUM,
+ MPI_COMM_WORLD)) != MPI_SUCCESS)
mpi_error(res, "Failed to allreduce particle cell weights.");
/* Main node does the partition calculation. */
int *celllist = (int *)malloc(sizeof(int) * s->nr_cells);
if (celllist == NULL) error("Failed to allocate celllist");
- if (nodeID == 0)
- pick_metis(s, nr_nodes, weights, NULL, celllist);
+ if (nodeID == 0) pick_metis(s, nr_nodes, weights, NULL, celllist);
/* Distribute the celllist partition and apply. */
if ((res = MPI_Bcast(celllist, s->nr_cells, MPI_INT, 0, MPI_COMM_WORLD)) !=
@@ -721,14 +709,14 @@ static void repart_vertex_metis(struct space *s, int nodeID, int nr_nodes) {
}
#endif
-
/**
* @brief Repartition the space using the given repartition type.
*
* Note that at the end of this process all the cells will be re-distributed
* across the nodes, but the particles themselves will not be.
*
- * @param reparttype the type of repartition to attempt, see the repart_type enum.
+ * @param reparttype the type of repartition to attempt, see the repart_type
+ *enum.
* @param nodeID our nodeID.
* @param nr_nodes the number of nodes.
* @param s the space of cells holding our local particles.
@@ -797,8 +785,8 @@ void partition_initial_partition(struct partition *initial_partition,
struct cell *c;
/* If we've got the wrong number of nodes, fail. */
- if (nr_nodes != initial_partition->grid[0] * initial_partition->grid[1]
- * initial_partition->grid[2])
+ if (nr_nodes != initial_partition->grid[0] * initial_partition->grid[1] *
+ initial_partition->grid[2])
error("Grid size does not match number of nodes.");
/* Run through the cells and set their nodeID. */
@@ -806,9 +794,9 @@ void partition_initial_partition(struct partition *initial_partition,
for (k = 0; k < s->nr_cells; k++) {
c = &s->cells[k];
for (j = 0; j < 3; j++)
- ind[j] = c->loc[j] / s->dim[j] * initial_partition->grid[j];
+ ind[j] = c->loc[j] / s->dim[j] * initial_partition->grid[j];
c->nodeID = ind[0] + initial_partition->grid[0] *
- (ind[1] + initial_partition->grid[1] * ind[2]);
+ (ind[1] + initial_partition->grid[1] * ind[2]);
// message("cell at [%e,%e,%e]: ind = [%i,%i,%i], nodeID = %i", c->loc[0],
// c->loc[1], c->loc[2], ind[0], ind[1], ind[2], c->nodeID);
}
@@ -833,7 +821,7 @@ void partition_initial_partition(struct partition *initial_partition,
* not. */
int *weights = NULL;
if (initial_partition->type == INITPART_METIS_WEIGHT) {
- if ((weights = (int *)malloc(sizeof(int) * s->nr_cells)) == NULL)
+ if ((weights = (int *)malloc(sizeof(int) * s->nr_cells)) == NULL)
error("Failed to allocate weights buffer.");
bzero(weights, sizeof(int) * s->nr_cells);
@@ -864,14 +852,12 @@ void partition_initial_partition(struct partition *initial_partition,
if (MPI_Allreduce(MPI_IN_PLACE, weights, s->nr_cells, MPI_INT, MPI_SUM,
MPI_COMM_WORLD) != MPI_SUCCESS)
error("Failed to allreduce particle cell weights.");
-
}
/* Main node does the partition calculation. */
int *celllist = (int *)malloc(sizeof(int) * s->nr_cells);
if (celllist == NULL) error("Failed to allocate celllist");
- if (nodeID == 0)
- pick_metis(s, nr_nodes, weights, NULL, celllist);
+ if (nodeID == 0) pick_metis(s, nr_nodes, weights, NULL, celllist);
/* Distribute the celllist partition and apply. */
int res = MPI_Bcast(celllist, s->nr_cells, MPI_INT, 0, MPI_COMM_WORLD);
@@ -908,8 +894,7 @@ void partition_initial_partition(struct partition *initial_partition,
}
/* Share the samplecells around all the nodes. */
- int res =
- MPI_Bcast(samplecells, nr_nodes * 3, MPI_INT, 0, MPI_COMM_WORLD);
+ int res = MPI_Bcast(samplecells, nr_nodes * 3, MPI_INT, 0, MPI_COMM_WORLD);
if (res != MPI_SUCCESS)
mpi_error(res, "Failed to bcast the partition sample cells.");
diff --git a/src/partition.h b/src/partition.h
index 492af04ac9ef017ceddabb410ee53496e64bb018..3ab5f5a817bf5b77d45c7fb3313158b83d98e251 100644
--- a/src/partition.h
+++ b/src/partition.h
@@ -51,8 +51,8 @@ enum repartition_type {
extern const char *repartition_name[];
void partition_repartition(enum repartition_type reparttype, int nodeID,
- int nr_nodes, struct space *s,
- struct task *tasks, int nr_tasks);
+ int nr_nodes, struct space *s, struct task *tasks,
+ int nr_tasks);
void partition_initial_partition(struct partition *initial_partition,
int nodeID, int nr_nodes, struct space *s);
diff --git a/src/runner.c b/src/runner.c
index 1cd52e1dd89f08172c3a3b679f15a096a0007b1b..b2f5eae0ffa39588ff7028c07b81a326e8061dde 100644
--- a/src/runner.c
+++ b/src/runner.c
@@ -311,20 +311,7 @@ void runner_dosort(struct runner *r, struct cell *c, int flags, int clock) {
}
} */
-#ifdef TIMER_VERBOSE
- message(
- "runner %02i: %i parts at depth %i (flags = %i%i%i%i%i%i%i%i%i%i%i%i%i) "
- "took %.3f ms.",
- r->id, count, c->depth, (flags & 0x1000) >> 12, (flags & 0x800) >> 11,
- (flags & 0x400) >> 10, (flags & 0x200) >> 9, (flags & 0x100) >> 8,
- (flags & 0x80) >> 7, (flags & 0x40) >> 6, (flags & 0x20) >> 5,
- (flags & 0x10) >> 4, (flags & 0x8) >> 3, (flags & 0x4) >> 2,
- (flags & 0x2) >> 1, (flags & 0x1) >> 0,
- ((double)TIMER_TOC(timer_dosort)) / CPU_TPS * 1000);
- fflush(stdout);
-#else
if (clock) TIMER_TOC(timer_dosort);
-#endif
}
void runner_dogsort(struct runner *r, struct cell *c, int flags, int clock) {
@@ -469,20 +456,7 @@ void runner_dogsort(struct runner *r, struct cell *c, int flags, int clock) {
}
} */
-#ifdef TIMER_VERBOSE
- message(
- "runner %02i: %i parts at depth %i (flags = %i%i%i%i%i%i%i%i%i%i%i%i%i) "
- "took %.3f ms.",
- r->id, count, c->depth, (flags & 0x1000) >> 12, (flags & 0x800) >> 11,
- (flags & 0x400) >> 10, (flags & 0x200) >> 9, (flags & 0x100) >> 8,
- (flags & 0x80) >> 7, (flags & 0x40) >> 6, (flags & 0x20) >> 5,
- (flags & 0x10) >> 4, (flags & 0x8) >> 3, (flags & 0x4) >> 2,
- (flags & 0x2) >> 1, (flags & 0x1) >> 0,
- ((double)TIMER_TOC(timer_dosort)) / CPU_TPS * 1000);
- fflush(stdout);
-#else
if (clock) TIMER_TOC(timer_dosort);
-#endif
}
/**
@@ -522,15 +496,7 @@ void runner_doinit(struct runner *r, struct cell *c, int timer) {
}
}
- if (timer) {
-#ifdef TIMER_VERBOSE
- message("runner %02i: %i parts at depth %i took %.3f ms.", r->id, c->count,
- c->depth, ((double)TIMER_TOC(timer_init)) / CPU_TPS * 1000);
- fflush(stdout);
-#else
- TIMER_TOC(timer_init);
-#endif
- }
+ if (timer) TIMER_TOC(timer_init);
}
/**
@@ -679,13 +645,7 @@ void runner_doghost(struct runner *r, struct cell *c) {
if (count)
message("Smoothing length failed to converge on %i particles.", count);
-#ifdef TIMER_VERBOSE
- message("runner %02i: %i parts at depth %i took %.3f ms.", r->id, c->count,
- c->depth, ((double)TIMER_TOC(timer_doghost)) / CPU_TPS * 1000);
- fflush(stdout);
-#else
TIMER_TOC(timer_doghost);
-#endif
}
/**
@@ -779,15 +739,7 @@ void runner_dodrift(struct runner *r, struct cell *c, int timer) {
c->h_max = h_max;
c->dx_max = dx_max;
- if (timer) {
-#ifdef TIMER_VERBOSE
- message("runner %02i: %i parts at depth %i took %.3f ms.", r->id, c->count,
- c->depth, ((double)TIMER_TOC(timer_drift)) / CPU_TPS * 1000);
- fflush(stdout);
-#else
- TIMER_TOC(timer_drift);
-#endif
- }
+ if (timer) TIMER_TOC(timer_drift);
}
/**
@@ -991,15 +943,7 @@ void runner_dokick(struct runner *r, struct cell *c, int timer) {
c->ti_end_min = ti_end_min;
c->ti_end_max = ti_end_max;
- if (timer) {
-#ifdef TIMER_VERBOSE
- message("runner %02i: %i parts at depth %i took %.3f ms.", r->id, c->count,
- c->depth, ((double)TIMER_TOC(timer_kick)) / CPU_TPS * 1000);
- fflush(stdout);
-#else
- TIMER_TOC(timer_kick);
-#endif
- }
+ if (timer) TIMER_TOC(timer_kick);
}
/**
diff --git a/src/runner_doiact.h b/src/runner_doiact.h
index 90bd2bb8a7a618f8539d98b91ffdf347f17eca17..cf5d56e94169b44e6cd2974a3422a0bc5e4610ac 100644
--- a/src/runner_doiact.h
+++ b/src/runner_doiact.h
@@ -202,15 +202,7 @@ void DOPAIR_NAIVE(struct runner *r, struct cell *restrict ci,
IACT(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf(
- "runner_dopair_naive[%02i]: %i/%i parts at depth %i (r_max=%.3f/%.3f) "
- "took %.3f ms.\n",
- r->id, count_i, count_j, ci->depth, ci->h_max, cj->h_max,
- ((double)TIMER_TOC(TIMER_DOPAIR)) / CPU_TPS * 1000);
-#else
TIMER_TOC(TIMER_DOPAIR);
-#endif
}
void DOSELF_NAIVE(struct runner *r, struct cell *restrict c) {
@@ -305,12 +297,7 @@ void DOSELF_NAIVE(struct runner *r, struct cell *restrict c) {
IACT(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf("runner_doself[%02i]: %i parts at depth %i took %.3f ms.\n", r->id,
- count, c->depth, ((double)TIMER_TOC(TIMER_DOSELF)) / CPU_TPS * 1000);
-#else
TIMER_TOC(TIMER_DOSELF);
-#endif
}
/**
@@ -510,15 +497,7 @@ void DOPAIR_SUBSET(struct runner *r, struct cell *restrict ci,
IACT_NONSYM(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf(
- "runner_dopair_subset[%02i]: %i/%i parts at depth %i (r_max=%.3f/%.3f) "
- "took %.3f ms.\n",
- r->id, count, count_j, ci->depth, ci->h_max, cj->h_max,
- ((double)TIMER_TOC(TIMER_DOPAIR)) / CPU_TPS * 1000);
-#else
TIMER_TOC(timer_dopair_subset);
-#endif
}
/**
@@ -630,15 +609,7 @@ void DOPAIR_SUBSET_NAIVE(struct runner *r, struct cell *restrict ci,
IACT_NONSYM(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf(
- "runner_dopair_subset[%02i]: %i/%i parts at depth %i (r_max=%.3f/%.3f) "
- "took %.3f ms.\n",
- r->id, count, count_j, ci->depth, ci->h_max, cj->h_max,
- ((double)TIMER_TOC(TIMER_DOPAIR)) / CPU_TPS * 1000);
-#else
TIMER_TOC(timer_dopair_subset);
-#endif
}
/**
@@ -741,13 +712,7 @@ void DOSELF_SUBSET(struct runner *r, struct cell *restrict ci,
IACT_NONSYM(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf("runner_doself_subset[%02i]: %i/%i parts at depth %i took %.3f ms.\n",
- r->id, count, ci->count, ci->depth,
- ((double)TIMER_TOC(TIMER_DOSELF)) / CPU_TPS * 1000);
-#else
TIMER_TOC(timer_dopair_subset);
-#endif
}
/**
@@ -870,8 +835,8 @@ void DOPAIR1(struct runner *r, struct cell *ci, struct cell *cj) {
} /* loop over the parts in ci. */
- /* printf( "runner_dopair: first half took %.3f ms...\n" ,
- ((double)(getticks() - tic)) / CPU_TPS * 1000 );
+ /* printf( "runner_dopair: first half took %.3f %s...\n" ,
+ clocks_from_ticks(getticks() - tic), clocks_getunit());
tic = getticks(); */
/* Loop over the parts in cj. */
@@ -941,16 +906,7 @@ void DOPAIR1(struct runner *r, struct cell *ci, struct cell *cj) {
IACT_NONSYM(r2q[k], &dxq[3 * k], hiq[k], hjq[k], piq[k], pjq[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf(
- "runner_dopair[%02i]: %i/%i parts at depth %i (r_max=%.3f/%.3f, h=%.3f) "
- "took %.3f ms.\n",
- r->id, count_i, count_j, ci->depth, ci->h_max, cj->h_max,
- fmax(ci->h[0], fmax(ci->h[1], ci->h[2])),
- ((double)(TIMER_TOC(TIMER_DOPAIR))) / CPU_TPS * 1000);
-#else
TIMER_TOC(TIMER_DOPAIR);
-#endif
}
void DOPAIR2(struct runner *r, struct cell *ci, struct cell *cj) {
@@ -1182,8 +1138,8 @@ void DOPAIR2(struct runner *r, struct cell *ci, struct cell *cj) {
} /* loop over the parts in ci. */
- /* printf( "runner_dopair: first half took %.3f ms...\n" ,
- ((double)(getticks() - tic)) / CPU_TPS * 1000 );
+ /* printf( "runner_dopair: first half took %.3f %s...\n" ,
+ clocks_from_ticks(getticks() - tic), clocks_getunit());
tic = getticks(); */
/* Loop over the parts in cj. */
@@ -1336,16 +1292,7 @@ void DOPAIR2(struct runner *r, struct cell *ci, struct cell *cj) {
IACT(r2q2[k], &dxq2[3 * k], hiq2[k], hjq2[k], piq2[k], pjq2[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf(
- "runner_dopair[%02i]: %i/%i parts at depth %i (r_max=%.3f/%.3f, h=%.3f) "
- "took %.3f ms.\n",
- r->id, count_i, count_j, ci->depth, ci->h_max, cj->h_max,
- fmax(ci->h[0], fmax(ci->h[1], ci->h[2])),
- ((double)(TIMER_TOC(TIMER_DOPAIR))) / CPU_TPS * 1000);
-#else
TIMER_TOC(TIMER_DOPAIR);
-#endif
}
/**
@@ -1570,12 +1517,7 @@ void DOSELF1(struct runner *r, struct cell *restrict c) {
IACT(r2q2[k], &dxq2[3 * k], hiq2[k], hjq2[k], piq2[k], pjq2[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf("runner_doself1[%02i]: %i parts at depth %i took %.3f ms.\n", r->id,
- count, c->depth, ((double)TIMER_TOC(TIMER_DOSELF)) / CPU_TPS * 1000);
-#else
TIMER_TOC(TIMER_DOSELF);
-#endif
}
void DOSELF2(struct runner *r, struct cell *restrict c) {
@@ -1767,12 +1709,7 @@ void DOSELF2(struct runner *r, struct cell *restrict c) {
IACT(r2q2[k], &dxq2[3 * k], hiq2[k], hjq2[k], piq2[k], pjq2[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf("runner_doself2[%02i]: %i parts at depth %i took %.3f ms.\n", r->id,
- count, c->depth, ((double)TIMER_TOC(TIMER_DOSELF)) / CPU_TPS * 1000);
-#else
TIMER_TOC(TIMER_DOSELF);
-#endif
}
/**
@@ -2054,13 +1991,7 @@ void DOSUB1(struct runner *r, struct cell *ci, struct cell *cj, int sid,
} /* otherwise, pair interaction. */
- if (gettimer)
-#ifdef TIMER_VERBOSE
- printf("runner_dosub1[%02i]: flags=%i at depth %i took %.3f ms.\n", r->id,
- sid, ci->depth, ((double)TIMER_TOC(TIMER_DOSUB)) / CPU_TPS * 1000);
-#else
- TIMER_TOC(TIMER_DOSUB);
-#endif
+ if (gettimer) TIMER_TOC(TIMER_DOSUB);
}
void DOSUB2(struct runner *r, struct cell *ci, struct cell *cj, int sid,
@@ -2329,13 +2260,7 @@ void DOSUB2(struct runner *r, struct cell *ci, struct cell *cj, int sid,
} /* otherwise, pair interaction. */
- if (gettimer)
-#ifdef TIMER_VERBOSE
- printf("runner_dosub2[%02i]: flags=%i at depth %i took %.3f ms.\n", r->id,
- sid, ci->depth, ((double)TIMER_TOC(TIMER_DOSUB)) / CPU_TPS * 1000);
-#else
- TIMER_TOC(TIMER_DOSUB);
-#endif
+ if (gettimer) TIMER_TOC(TIMER_DOSUB);
}
void DOSUB_SUBSET(struct runner *r, struct cell *ci, struct part *parts,
@@ -2933,11 +2858,5 @@ void DOSUB_SUBSET(struct runner *r, struct cell *ci, struct part *parts,
} /* otherwise, pair interaction. */
- if (gettimer)
-#ifdef TIMER_VERBOSE
- printf("runner_dosub[%02i]: flags=%i at depth %i took %.3f ms.\n", r->id,
- sid, ci->depth, ((double)TIMER_TOC(TIMER_DOSUB)) / CPU_TPS * 1000);
-#else
- TIMER_TOC(TIMER_DOSUB);
-#endif
+ if (gettimer) TIMER_TOC(TIMER_DOSUB);
}
diff --git a/src/runner_doiact_grav.h b/src/runner_doiact_grav.h
index 4f4e8557892b554a7d3d4c534f4abe5032de32f5..f374339da75e31b39a5295fcd8bbc23c34d8d67d 100644
--- a/src/runner_doiact_grav.h
+++ b/src/runner_doiact_grav.h
@@ -21,6 +21,7 @@
/* Includes. */
#include "cell.h"
+#include "clocks.h"
#include "part.h"
/**
@@ -192,16 +193,7 @@ void runner_dopair_grav_new(struct runner *r, struct cell *ci,
} /* loop over the parts in cj and interact with the multipole. */
-#ifdef TIMER_VERBOSE
- printf(
- "runner_dopair[%02i]: %i/%i parts at depth %i (r_max=%.3f/%.3f, h=%.3f) "
- "took %.3f ms.\n",
- r->id, count_i, count_j, ci->depth, ci->h_max, cj->h_max,
- fmax(ci->h[0], fmax(ci->h[1], ci->h[2])),
- ((double)(TIMER_TOC(TIMER_DOPAIR))) / CPU_TPS * 1000);
-#else
TIMER_TOC(TIMER_DOPAIR);
-#endif
}
/**
@@ -429,15 +421,7 @@ void runner_dopair_grav(struct runner *r, struct cell *restrict ci,
runner_iact_grav(r2q[k], &dxq[3 * k], piq[k], pjq[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf(
- "runner_dopair_naive_grav[%02i]: %i/%i parts at depth %i "
- "(r_max=%.3f/%.3f) took %.3f ms.\n",
- r->id, count_i, count_j, ci->depth, ci->h_max, cj->h_max,
- ((double)TIMER_TOC(TIMER_DOPAIR)) / CPU_TPS * 1000);
-#else
TIMER_TOC(timer_dopair_grav);
-#endif
}
/**
@@ -525,15 +509,7 @@ void runner_doself_grav(struct runner *r, struct cell *restrict c) {
runner_iact_grav(r2q[k], &dxq[3 * k], piq[k], pjq[k]);
#endif
-#ifdef TIMER_VERBOSE
- printf(
- "runner_doself_grav[%02i]: %i/%i parts at depth %i (r_max=%.3f/%.3f) "
- "took %.3f ms.\n",
- r->id, count_i, count_j, ci->depth, ci->h_max, cj->h_max,
- ((double)TIMER_TOC(TIMER_DOPAIR)) / CPU_TPS * 1000);
-#else
TIMER_TOC(timer_doself_grav);
-#endif
}
/**
@@ -616,13 +592,7 @@ void runner_dosub_grav(struct runner *r, struct cell *ci, struct cell *cj,
runner_dograv_mm(r, ci, cj);
}
- if (gettimer)
-#ifdef TIMER_VERBOSE
- printf("runner_dosub_grav[%02i]: at depth %i took %.3f ms.\n", r->id,
- ci->depth, ((double)TIMER_TOC(TIMER_DOSUB)) / CPU_TPS * 1000);
-#else
- TIMER_TOC(timer_dosub_grav);
-#endif
+ if (gettimer) TIMER_TOC(timer_dosub_grav);
}
#endif /* SWIFT_RUNNER_DOIACT_GRAV_H */
diff --git a/src/scheduler.c b/src/scheduler.c
index 047a402abbdc6877958ea8e2730472cf0a995ebe..a9a4827e7fa368061ae0a5b1bdfec51940e5d716 100644
--- a/src/scheduler.c
+++ b/src/scheduler.c
@@ -936,8 +936,8 @@ void scheduler_reweight(struct scheduler *s) {
if (t->type == task_type_send) t->weight = INT_MAX / 8;
if (t->type == task_type_recv) t->weight *= 1.41;
}
- // message( "weighting tasks took %.3f ms." , (double)( getticks() - tic ) /
- // CPU_TPS * 1000 );
+ // message( "weighting tasks took %.3f %s." ,
+ //clocks_from_ticks( getticks() - tic ), clocks_getunit());
/* int min = tasks[0].weight, max = tasks[0].weight;
for ( k = 1 ; k < nr_tasks ; k++ )
@@ -973,8 +973,8 @@ void scheduler_start(struct scheduler *s, unsigned int mask,
s->tasks[k].wait = 1;
s->tasks[k].rid = -1;
}
- // message( "waiting tasks took %.3f ms." , (double)( getticks() - tic ) /
- // CPU_TPS * 1000 );
+ // message( "waiting tasks took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit() );
/* Enqueue a set of extraenous tasks to set the task waits. */
struct task *rewait_tasks = &s->tasks[s->nr_tasks];
@@ -1016,8 +1016,8 @@ void scheduler_start(struct scheduler *s, unsigned int mask,
pthread_cond_wait(&s->sleep_cond, &s->sleep_mutex);
}
pthread_mutex_unlock(&s->sleep_mutex);
- /* message("waiting tasks took %.3f ms.",
- (double)(getticks() - tic) / CPU_TPS * 1000); */
+ /* message("waiting tasks took %.3f %s.",
+ clocks_from_ticks(getticks() - tic), clocks_getunit());*/
s->mask = mask;
s->submask = submask | (1 << task_subtype_none);
@@ -1033,8 +1033,8 @@ void scheduler_start(struct scheduler *s, unsigned int mask,
}
}
- // message( "enqueueing tasks took %.3f ms." , (double)( getticks() - tic ) /
- // CPU_TPS * 1000 );
+ // message( "enqueueing tasks took %.3f %s." ,
+ //clocks_from_ticks( getticks() - tic ), clocks_getunit());
}
/**
diff --git a/src/space.c b/src/space.c
index f5674b969fb0ec427320b647b8cae80f3b019b44..0785740335f6e782a875b7ab658a9a0564288aab 100644
--- a/src/space.c
+++ b/src/space.c
@@ -294,8 +294,8 @@ void space_regrid(struct space *s, double cell_max, int verbose) {
}
#endif
} /* re-build upper-level cells? */
- // message( "rebuilding upper-level cells took %.3f ms." , (double)(getticks()
- // - tic) / CPU_TPS * 1000 );
+ // message( "rebuilding upper-level cells took %.3f %s." ,
+ // clocks_from_ticks(double)(getticks() - tic), clocks_getunit());
/* Otherwise, just clean up the cells. */
else {
@@ -377,8 +377,8 @@ void space_rebuild(struct space *s, double cell_max, int verbose) {
cell_getid(cdim, p->x[0] * ih[0], p->x[1] * ih[1], p->x[2] * ih[2]);
cells[ind[k]].count++;
}
-// message( "getting particle indices took %.3f ms." , (double)(getticks() -
-// tic) / CPU_TPS * 1000 );
+ // message( "getting particle indices took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit()):
#ifdef WITH_MPI
/* Move non-local parts to the end of the list. */
@@ -430,8 +430,8 @@ void space_rebuild(struct space *s, double cell_max, int verbose) {
/* Sort the parts according to their cells. */
// tic = getticks();
space_parts_sort(s, ind, nr_parts, 0, s->nr_cells - 1);
- // message( "parts_sort took %.3f ms." , (double)(getticks() - tic) / CPU_TPS
- // * 1000 );
+ // message( "parts_sort took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
/* Re-link the gparts. */
for (k = 0; k < nr_parts; k++)
@@ -465,16 +465,16 @@ void space_rebuild(struct space *s, double cell_max, int verbose) {
cell_getid(cdim, gp->x[0] * ih[0], gp->x[1] * ih[1], gp->x[2] * ih[2]);
cells[ind[k]].gcount++;
}
- // message( "getting particle indices took %.3f ms." , (double)(getticks() -
- // tic) / CPU_TPS * 1000 );
+ // message( "getting particle indices took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
/* TODO: Here we should exchange the gparts as well! */
/* Sort the parts according to their cells. */
// tic = getticks();
gparts_sort(s->gparts, ind, nr_gparts, 0, s->nr_cells - 1);
- // message( "gparts_sort took %.3f ms." , (double)(getticks() - tic) / CPU_TPS
- // * 1000 );
+ // message( "gparts_sort took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
/* Re-link the parts. */
for (k = 0; k < nr_gparts; k++)
@@ -497,8 +497,8 @@ void space_rebuild(struct space *s, double cell_max, int verbose) {
xfinger = &xfinger[c->count];
gfinger = &gfinger[c->gcount];
}
- // message( "hooking up cells took %.3f ms." , (double)(getticks() - tic) /
- // CPU_TPS * 1000 );
+ // message( "hooking up cells took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
/* At this point, we have the upper-level cells, old or new. Now make
sure that the parts in each cell are ok. */
@@ -509,8 +509,8 @@ void space_rebuild(struct space *s, double cell_max, int verbose) {
0, &cells[k], NULL, 0);
engine_launch(s->e, s->e->nr_threads, 1 << task_type_split_cell, 0);
- // message( "space_split took %.3f ms." , (double)(getticks() - tic) / CPU_TPS
- // * 1000 );
+ // message( "space_split took %.3f %s." ,
+ //clocks_from_ticks(getticks() - tic), clocks_getunit());
}
/**
diff --git a/src/swift.h b/src/swift.h
index 03960e83bf75488b3456327157288160e995951b..9ab090dccd195ff4927d3e614e446b36d273f824 100644
--- a/src/swift.h
+++ b/src/swift.h
@@ -25,6 +25,7 @@
/* Local headers. */
#include "atomic.h"
#include "cell.h"
+#include "clocks.h"
#include "const.h"
#include "const.h"
#include "cycle.h"
diff --git a/src/timers.h b/src/timers.h
index 95db02c5b8ea055e4b8264a7055c678460c29fdf..de2d9edb9ed54717472e5ae1222dfb33235c3e95 100644
--- a/src/timers.h
+++ b/src/timers.h
@@ -57,11 +57,6 @@ extern ticks timers[timer_count];
#define timers_mask_all ((1 << timer_count) - 1)
/* Define the timer macros. */
-#ifdef TIMER_VERBOSE
-#ifndef TIMER
-#define TIMER
-#endif
-#endif
#ifdef TIMER
#define TIMER_TIC_ND tic = getticks();
#define TIMER_TIC2_ND ticks tic2 = getticks();
diff --git a/tests/Makefile.am b/tests/Makefile.am
index 7dcabd8f3e35963e5fd6331994791dccc2a94c1b..f0bfbefd3c7f4591134d1707c4ac9bf63278e855 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -15,7 +15,7 @@
# along with this program. If not, see <http://www.gnu.org/licenses/>.
# Add the source directory and debug to CFLAGS
-AM_CFLAGS = -I../src -DCPU_TPS=2.67e9 $(HDF5_CPPFLAGS) -DTIMER
+AM_CFLAGS = -I../src $(HDF5_CPPFLAGS) -DTIMER
AM_LDFLAGS = ../src/.libs/libswiftsim.a $(HDF5_LDFLAGS) $(HDF5_LIBS)
diff --git a/tests/testReading.c b/tests/testReading.c
index c8123a26acd12692ccd6cd3d893f94e0f95fb856..17b33d88f6fd702e0bad6280f4adbe7ebc50ef2a 100644
--- a/tests/testReading.c
+++ b/tests/testReading.c
@@ -31,7 +31,7 @@ int main() {
const double boxSize = 1.;
const int L = 4;
const double rho = 2.;
-
+
/* Read data */
read_ic_single("input.hdf5", dim, &parts, &N, &periodic);
diff --git a/tests/testSPHStep.c b/tests/testSPHStep.c
index 32e7d62815533fb0ad50e2e85ebd10f452ec8d30..984b8ea867250d0bda1bc14d2600279a27321b2c 100644
--- a/tests/testSPHStep.c
+++ b/tests/testSPHStep.c
@@ -22,43 +22,45 @@
#include <stdlib.h>
#include <string.h>
-/**
+/**
* @brief Constructs a cell with N SPH particles
*/
struct cell *make_cell(size_t N, float cellSize, int offset[3], int id_offset) {
- size_t count = N*N*N;
- struct cell *cell = malloc(sizeof( struct cell) );
+ size_t count = N * N * N;
+ struct cell *cell = malloc(sizeof(struct cell));
struct part *part;
struct xpart *xpart;
float h;
size_t x, y, z, size;
- size = count*sizeof(struct part);
- if (posix_memalign((void**)&cell->parts, 32, size) != 0) {
+ size = count * sizeof(struct part);
+ if (posix_memalign((void **)&cell->parts, 32, size) != 0) {
error("couldn't allocate particles");
}
- size = count*sizeof(struct xpart);
- if (posix_memalign((void**)&cell->xparts, 32, size) != 0) {
+ size = count * sizeof(struct xpart);
+ if (posix_memalign((void **)&cell->xparts, 32, size) != 0) {
error("couldn't allocate extended particles");
}
-
h = 1.127 * cellSize / N;
-
+
part = cell->parts;
xpart = cell->xparts;
- memset(part, 0, count*sizeof(struct part));
- memset(xpart, 0, count*sizeof(struct xpart));
+ memset(part, 0, count * sizeof(struct part));
+ memset(xpart, 0, count * sizeof(struct xpart));
for (x = 0; x < N; ++x) {
for (y = 0; y < N; ++y) {
for (z = 0; z < N; ++z) {
- part->x[0] = offset[0] * cellSize + x * cellSize / N + cellSize / (2*N);
- part->x[1] = offset[1] * cellSize + y * cellSize / N + cellSize / (2*N);
- part->x[2] = offset[2] * cellSize + z * cellSize / N + cellSize / (2*N);
- part->h = h;
- part->id = x*N*N + y*N + z + id_offset;
- ++part;
+ part->x[0] =
+ offset[0] * cellSize + x * cellSize / N + cellSize / (2 * N);
+ part->x[1] =
+ offset[1] * cellSize + y * cellSize / N + cellSize / (2 * N);
+ part->x[2] =
+ offset[2] * cellSize + z * cellSize / N + cellSize / (2 * N);
+ part->h = h;
+ part->id = x * N * N + y * N + z + id_offset;
+ ++part;
}
}
}
@@ -69,7 +71,7 @@ struct cell *make_cell(size_t N, float cellSize, int offset[3], int id_offset) {
cell->h[0] = cellSize;
cell->h[1] = cellSize;
cell->h[2] = cellSize;
-
+
return cell;
}
@@ -78,37 +80,38 @@ struct cell *make_cell(size_t N, float cellSize, int offset[3], int id_offset) {
/* Run a full time step integration for one cell */
int main() {
- int i,j,k, offset[3];
+ int i, j, k, offset[3];
struct part *p;
int N = 10;
float dim = 1.;
float rho = 2.;
float P = 1.;
-
+
/* Create cells */
struct cell *cells[27];
- for (i=0; i<3; i++)
- for (j=0; j<3; j++)
- for (k=0; k<3; k++) {
- offset[0] = i;
- offset[1] = j;
- offset[2] = k;
- cells[i*9 + j*3 + k] = make_cell(N, dim, offset, (i*9+j*3+k) * N*N*N);
+ for (i = 0; i < 3; i++)
+ for (j = 0; j < 3; j++)
+ for (k = 0; k < 3; k++) {
+ offset[0] = i;
+ offset[1] = j;
+ offset[2] = k;
+ cells[i * 9 + j * 3 + k] =
+ make_cell(N, dim, offset, (i * 9 + j * 3 + k) * N * N * N);
}
/* Set particle properties */
- for (j=0 ; j < 27; ++j)
- for (i=0 ; i< cells[j]->count; ++i) {
- cells[j]->parts[i].mass = dim * dim * dim * rho / (N * N * N);
- cells[j]->parts[i].u = P / ((const_hydro_gamma - 1.)*rho);
+ for (j = 0; j < 27; ++j)
+ for (i = 0; i < cells[j]->count; ++i) {
+ cells[j]->parts[i].mass = dim * dim * dim * rho / (N * N * N);
+ cells[j]->parts[i].u = P / ((const_hydro_gamma - 1.) * rho);
}
message("m=%f", dim * dim * dim * rho / (N * N * N));
/* Pick the central cell */
struct cell *ci = cells[13];
-
+
/* Create the infrastructure */
struct engine e;
struct runner r;
@@ -123,40 +126,35 @@ int main() {
e.dt_max = 1e10;
/* The tracked particle */
- p = &(ci->parts[N*N*N / 2 + N*N / 2 + N / 2]);
+ p = &(ci->parts[N * N * N / 2 + N * N / 2 + N / 2]);
message("Studying particle p->id=%lld", p->id);
-
/* Initialise the particles */
- for (j=0 ; j<27; ++j)
- {
- runner_doinit(&r, cells[j]);
- }
-
+ for (j = 0; j < 27; ++j) {
+ runner_doinit(&r, cells[j]);
+ }
+
/* Compute density */
runner_doself1_density(&r, ci);
runner_doghost(&r, ci);
-
+
message("h=%f rho=%f N_ngb=%f", p->h, p->rho, p->density.wcount);
message("c=%f", p->force.c);
runner_doself2_force(&r, ci);
runner_dokick(&r, ci, 1);
-
+
message("t_end=%f", p->t_end);
-
+
free(ci->parts);
free(ci->xparts);
-
+
return 0;
}
#else
-int main() {
-
- return 0;
-}
+int main() { return 0; }
#endif
diff --git a/tests/testSingle.c b/tests/testSingle.c
index 1bab13959b5d04f23170b761b953de0fa43561b9..c85b77ff1c5b2285c33fa7787bbd53deab463039 100644
--- a/tests/testSingle.c
+++ b/tests/testSingle.c
@@ -40,11 +40,6 @@
/* Local headers. */
#include "swift.h"
-/* Ticks per second on this machine. */
-#ifndef CPU_TPS
-#define CPU_TPS 2.67e9
-#endif
-
/* Engine policy flags. */
#ifndef ENGINE_POLICY
#define ENGINE_POLICY engine_policy_none
diff --git a/tests/testTimeIntegration.c b/tests/testTimeIntegration.c
index 4fcd253e877ed167b9534117832675f8f6ac5ef4..f3802888bccc40a424d659cde2605d12c9268e47 100644
--- a/tests/testTimeIntegration.c
+++ b/tests/testTimeIntegration.c
@@ -41,14 +41,13 @@ int main() {
float v_max = 30287.; /* [m/s] */
// float v_min = 29291.; /* [m/s] */
-
/* Derived quantities */
float e = (r_max - r_min) / (r_max + r_min); /* Eccentricity */
float b = sqrtf(r_max * r_min); /* Semi-minor axis */
float p = b * sqrtf(1 - e * e); /* Semi-lactus rectum */
float a = p / (1 - e * e); /* Semi-major axis */
float T = sqrtf(4 * M_PI * M_PI * a * a * a /
- (G * (M_sun + M_earth))); /* Period [s] */
+ (G * (M_sun + M_earth))); /* Period [s] */
/* Print some info */
message("Semi-major axis: a=%e [m]", a);
@@ -98,8 +97,8 @@ int main() {
eng.time = 0.;
eng.timeBegin = 0.;
eng.timeEnd = N_orbits * T;
- eng.dt_min = dt; /* This forces the time-step to be dt */
- eng.dt_max = dt; /* irrespective of the state of the particle */
+ eng.dt_min = dt; /* This forces the time-step to be dt */
+ eng.dt_max = dt; /* irrespective of the state of the particle */
/* Simulate ! */
for (i = 0; i < N; i++) {
diff --git a/tests/testVectorize.c b/tests/testVectorize.c
index 1c77e9d54effe34fd2c24e97678d905890b291ba..a18b6e8af5ac3f7b94bd7be3bdf8fd21e49681ff 100644
--- a/tests/testVectorize.c
+++ b/tests/testVectorize.c
@@ -34,8 +34,8 @@ struct cell *make_cell(size_t n, double *offset, double h,
part->h = h;
part->id = ++(*partId);
part->mass = 1.0f;
- part->t_begin = 0.f;
- part->t_end = 0.1f;
+ part->ti_begin = 0;
+ part->ti_end = 1;
++part;
}
}
@@ -177,13 +177,13 @@ int main(int argc, char *argv[]) {
toc = getticks();
time += toc - tic;
-
+
/* Dump if necessary */
if (i % 50 == 0) dump_particle_fields("swift_dopair.dat", ci, cj);
}
/* Output timing */
- message("SWIFT calculation took %lli ticks." , time / runs);
+ message("SWIFT calculation took %lli ticks.", time / runs);
/* Now perform a brute-force version for accuracy tests */
@@ -202,7 +202,7 @@ int main(int argc, char *argv[]) {
dump_particle_fields("brute_force.dat", ci, cj);
/* Output timing */
- message("Brute force calculation took %lli ticks." , toc - tic);
+ message("Brute force calculation took %lli ticks.", toc - tic);
/* Clean things to make the sanitizer happy ... */
clean_up(ci);