diff --git a/.gitignore b/.gitignore
index aa3676fe292246d0ab53022dfd2a4a3ce305174e..db06575cf9f291d1fb9fa253fb5146c065523dd9 100644
--- a/.gitignore
+++ b/.gitignore
@@ -36,6 +36,7 @@ examples/*/*/*.xmf
examples/*/*/*.hdf5
examples/*/*/*.txt
examples/*/*/used_parameters.yml
+examples/*/*.png
tests/testPair
tests/brute_force_standard.dat
@@ -72,7 +73,11 @@ tests/testPair.sh
tests/testPairPerturbed.sh
tests/testParser.sh
tests/testReading.sh
-
+tests/testAdiabaticIndex
+tests/testRiemannExact
+tests/testRiemannTRRS
+tests/testRiemannHLLC
+tests/testMatrixInversion
theory/latex/swift.pdf
theory/kernel/kernels.pdf
diff --git a/configure.ac b/configure.ac
index a798372c2f219a8da71f2b79321c140aba23e790..82382447fde7c411f61dbd62f7db388a6a8d9cf9 100644
--- a/configure.ac
+++ b/configure.ac
@@ -16,7 +16,7 @@
# along with this program. If not, see <http://www.gnu.org/licenses/>.
# Init the project.
-AC_INIT([SWIFT],[0.3.0])
+AC_INIT([SWIFT],[0.4.0])
AC_CONFIG_SRCDIR([src/space.c])
AC_CONFIG_AUX_DIR([.])
AM_INIT_AUTOMAKE
@@ -466,7 +466,7 @@ if test "$enable_warn" != "no"; then
# We will do this by hand instead and only default to the macro for unknown compilers
case "$ax_cv_c_compiler_vendor" in
gnu | clang)
- CFLAGS="$CFLAGS -Wall"
+ CFLAGS="$CFLAGS -Wall -Wextra -Wno-unused-parameter"
;;
intel)
CFLAGS="$CFLAGS -w2 -Wunused-variable"
diff --git a/examples/Disk-Patch/GravityOnly/makeIC.py b/examples/Disk-Patch/GravityOnly/makeIC.py
index 135d1d4e34e0ec4c32d75a2c49d6e55f46db4e1c..702a50ff53b73d004ff36be8049823515675cccf 100644
--- a/examples/Disk-Patch/GravityOnly/makeIC.py
+++ b/examples/Disk-Patch/GravityOnly/makeIC.py
@@ -110,7 +110,7 @@ grp.attrs["Time"] = 0.0
grp.attrs["NumFilesPerSnapshot"] = 1
grp.attrs["MassTable"] = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
grp.attrs["Flag_Entropy_ICs"] = [0, 0, 0, 0, 0, 0]
-
+grp.attrs["Dimension"] = 3
#Runtime parameters
grp = file.create_group("/RuntimePars")
diff --git a/examples/Disk-Patch/HydroStatic/makeIC.py b/examples/Disk-Patch/HydroStatic/makeIC.py
index 7404f76aeaa485037b19052d84636c1656847a2b..40b1d1f1ff9e08dae0c4b0b1539ca773c93009b4 100644
--- a/examples/Disk-Patch/HydroStatic/makeIC.py
+++ b/examples/Disk-Patch/HydroStatic/makeIC.py
@@ -175,6 +175,7 @@ grp.attrs["Time"] = 0.0
grp.attrs["NumFilesPerSnapshot"] = 1
grp.attrs["MassTable"] = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
grp.attrs["Flag_Entropy_ICs"] = [entropy_flag]
+grp.attrs["Dimension"] = 3
#Runtime parameters
grp = file.create_group("/RuntimePars")
diff --git a/examples/Gradients/gradientsCartesian.yml b/examples/Gradients/gradientsCartesian.yml
new file mode 100644
index 0000000000000000000000000000000000000000..917a4803004c2ce89984beb857cb1691d9a1ec1b
--- /dev/null
+++ b/examples/Gradients/gradientsCartesian.yml
@@ -0,0 +1,36 @@
+# Define the system of units to use internally.
+InternalUnitSystem:
+ UnitMass_in_cgs: 1 # Grams
+ UnitLength_in_cgs: 1 # Centimeters
+ UnitVelocity_in_cgs: 1 # Centimeters per second
+ UnitCurrent_in_cgs: 1 # Amperes
+ UnitTemp_in_cgs: 1 # Kelvin
+
+# Parameters governing the time integration
+TimeIntegration:
+ time_begin: 0. # The starting time of the simulation (in internal units).
+ time_end: 1e-6 # The end time of the simulation (in internal units).
+ dt_min: 1e-6 # The minimal time-step size of the simulation (in internal units).
+ dt_max: 1e-6 # The maximal time-step size of the simulation (in internal units).
+
+# Parameters governing the snapshots
+Snapshots:
+ basename: gradients_cartesian # Common part of the name of output files
+ time_first: 0. # Time of the first output (in internal units)
+ delta_time: 5e-7 # Time difference between consecutive outputs (in internal units)
+
+# Parameters governing the conserved quantities statistics
+Statistics:
+ delta_time: 1e-6 # Time between statistics output
+
+# Parameters for the hydrodynamics scheme
+SPH:
+ resolution_eta: 1.2348 # Target smoothing length in units of the mean inter-particle separation (1.2348 == 48Ngbs with the cubic spline kernel).
+ delta_neighbours: 0.1 # The tolerance for the targetted number of neighbours.
+ max_smoothing_length: 0.01 # Maximal smoothing length allowed (in internal units).
+ CFL_condition: 0.1 # Courant-Friedrich-Levy condition for time integration.
+
+# Parameters related to the initial conditions
+InitialConditions:
+ file_name: ./Gradients_cartesian.hdf5 # The file to read
+
diff --git a/examples/Gradients/gradientsRandom.yml b/examples/Gradients/gradientsRandom.yml
new file mode 100644
index 0000000000000000000000000000000000000000..209f30060f031f7d50a15ffbf8ad0e7fe5b013b8
--- /dev/null
+++ b/examples/Gradients/gradientsRandom.yml
@@ -0,0 +1,36 @@
+# Define the system of units to use internally.
+InternalUnitSystem:
+ UnitMass_in_cgs: 1 # Grams
+ UnitLength_in_cgs: 1 # Centimeters
+ UnitVelocity_in_cgs: 1 # Centimeters per second
+ UnitCurrent_in_cgs: 1 # Amperes
+ UnitTemp_in_cgs: 1 # Kelvin
+
+# Parameters governing the time integration
+TimeIntegration:
+ time_begin: 0. # The starting time of the simulation (in internal units).
+ time_end: 1e-6 # The end time of the simulation (in internal units).
+ dt_min: 1e-6 # The minimal time-step size of the simulation (in internal units).
+ dt_max: 1e-6 # The maximal time-step size of the simulation (in internal units).
+
+# Parameters governing the snapshots
+Snapshots:
+ basename: gradients_random # Common part of the name of output files
+ time_first: 0. # Time of the first output (in internal units)
+ delta_time: 5e-7 # Time difference between consecutive outputs (in internal units)
+
+# Parameters governing the conserved quantities statistics
+Statistics:
+ delta_time: 1e-6 # Time between statistics output
+
+# Parameters for the hydrodynamics scheme
+SPH:
+ resolution_eta: 1.2348 # Target smoothing length in units of the mean inter-particle separation (1.2348 == 48Ngbs with the cubic spline kernel).
+ delta_neighbours: 0.1 # The tolerance for the targetted number of neighbours.
+ max_smoothing_length: 0.01 # Maximal smoothing length allowed (in internal units).
+ CFL_condition: 0.1 # Courant-Friedrich-Levy condition for time integration.
+
+# Parameters related to the initial conditions
+InitialConditions:
+ file_name: ./Gradients_random.hdf5 # The file to read
+
diff --git a/examples/Gradients/gradientsStretched.yml b/examples/Gradients/gradientsStretched.yml
new file mode 100644
index 0000000000000000000000000000000000000000..592a70762988fca764c3ec7dcbc9bfcc9a8f2751
--- /dev/null
+++ b/examples/Gradients/gradientsStretched.yml
@@ -0,0 +1,36 @@
+# Define the system of units to use internally.
+InternalUnitSystem:
+ UnitMass_in_cgs: 1 # Grams
+ UnitLength_in_cgs: 1 # Centimeters
+ UnitVelocity_in_cgs: 1 # Centimeters per second
+ UnitCurrent_in_cgs: 1 # Amperes
+ UnitTemp_in_cgs: 1 # Kelvin
+
+# Parameters governing the time integration
+TimeIntegration:
+ time_begin: 0. # The starting time of the simulation (in internal units).
+ time_end: 1e-6 # The end time of the simulation (in internal units).
+ dt_min: 1e-6 # The minimal time-step size of the simulation (in internal units).
+ dt_max: 1e-6 # The maximal time-step size of the simulation (in internal units).
+
+# Parameters governing the snapshots
+Snapshots:
+ basename: gradients_stretched # Common part of the name of output files
+ time_first: 0. # Time of the first output (in internal units)
+ delta_time: 5e-7 # Time difference between consecutive outputs (in internal units)
+
+# Parameters governing the conserved quantities statistics
+Statistics:
+ delta_time: 1e-6 # Time between statistics output
+
+# Parameters for the hydrodynamics scheme
+SPH:
+ resolution_eta: 1.2348 # Target smoothing length in units of the mean inter-particle separation (1.2348 == 48Ngbs with the cubic spline kernel).
+ delta_neighbours: 0.1 # The tolerance for the targetted number of neighbours.
+ max_smoothing_length: 0.01 # Maximal smoothing length allowed (in internal units).
+ CFL_condition: 0.1 # Courant-Friedrich-Levy condition for time integration.
+
+# Parameters related to the initial conditions
+InitialConditions:
+ file_name: ./Gradients_stretched.hdf5 # The file to read
+
diff --git a/examples/Gradients/makeICs.py b/examples/Gradients/makeICs.py
new file mode 100644
index 0000000000000000000000000000000000000000..38d035d2ad2dd3dd6daacfd6f58d824e9daf6742
--- /dev/null
+++ b/examples/Gradients/makeICs.py
@@ -0,0 +1,177 @@
+################################################################################
+# This file is part of SWIFT.
+# Copyright (c) 2015 Bert Vandenbroucke (bert.vandenbroucke@ugent.be)
+#
+# 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/>.
+#
+################################################################################
+
+import h5py
+import random
+import numpy as np
+import sys
+
+# Generates a swift IC file with some density gradients, to check the gradient
+# reconstruction
+
+# Parameters
+periodic= 1 # 1 For periodic box
+gamma = 5./3. # Gas adiabatic index
+gridtype = "cartesian"
+if len(sys.argv) > 1:
+ gridtype = sys.argv[1]
+
+# stretched cartesian box ######################################################
+if gridtype == "stretched":
+ fileName = "Gradients_stretched.hdf5"
+ factor = 8
+ boxSize = [ 1.0 , 1.0/factor , 1.0/factor ]
+ L = 20
+ nx1 = factor*L/2
+ ny1 = L
+ nz1 = L
+ numfac = 2.
+ nx2 = int(nx1/numfac)
+ ny2 = int(ny1/numfac)
+ nz2 = int(nz1/numfac)
+ npart = nx1*ny1*nz1 + nx2*ny2*nz2
+ vol = boxSize[0] * boxSize[1] * boxSize[2]
+ partVol1 = 0.5*vol/(nx1*ny1*nz1)
+ partVol2 = 0.5*vol/(nx2*ny2*nz2)
+
+ coords = np.zeros((npart,3))
+ h = np.zeros((npart,1))
+ ids = np.zeros((npart,1), dtype='L')
+ idx = 0
+ dcell = 0.5/nx1
+ for i in range(nx1):
+ for j in range(ny1):
+ for k in range(nz1):
+ coords[idx,0] = (i+0.5)*dcell
+ coords[idx,1] = (j+0.5)*dcell
+ coords[idx,2] = (k+0.5)*dcell
+ h[idx] = 0.56/nx1
+ ids[idx] = idx
+ idx += 1
+ dcell = 0.5/nx2
+ for i in range(nx2):
+ for j in range(ny2):
+ for k in range(nz2):
+ coords[idx,0] = 0.5+(i+0.5)*dcell
+ coords[idx,1] = (j+0.5)*dcell
+ coords[idx,2] = (k+0.5)*dcell
+ h[idx] = 0.56/nx2
+ ids[idx] = idx
+ idx += 1
+
+# cartesian box ################################################################
+if gridtype == "cartesian":
+ fileName = "Gradients_cartesian.hdf5"
+ boxSize = [ 1.0 , 1.0 , 1.0 ]
+ nx = 20
+ npart = nx**3
+ partVol = 1./npart
+ coords = np.zeros((npart,3))
+ h = np.zeros((npart,1))
+ ids = np.zeros((npart,1), dtype='L')
+ idx = 0
+ dcell = 1./nx
+ for i in range(nx):
+ for j in range(nx):
+ for k in range(nx):
+ coords[idx,0] = (i+0.5)*dcell
+ coords[idx,1] = (j+0.5)*dcell
+ coords[idx,2] = (k+0.5)*dcell
+ h[idx] = 1.12/nx
+ ids[idx] = idx
+ idx += 1
+
+# random box ###################################################################
+if gridtype == "random":
+ fileName = "Gradients_random.hdf5"
+ boxSize = [ 1.0 , 1.0 , 1.0 ]
+ glass = h5py.File("../Glass/glass_50000.hdf5", "r")
+ coords = np.array(glass["/PartType0/Coordinates"])
+ npart = len(coords)
+ partVol = 1./npart
+ h = np.zeros((npart,1))
+ ids = np.zeros((npart,1), dtype='L')
+ for i in range(npart):
+ h[i] = 0.019
+ ids[i] = i
+
+v = np.zeros((npart,3))
+m = np.zeros((npart,1))
+rho = np.zeros((npart,1))
+u = np.zeros((npart,1))
+
+for i in range(npart):
+ rhox = coords[i,0]
+ if coords[i,0] < 0.75:
+ rhox = 0.75
+ if coords[i,0] < 0.25:
+ rhox = 1.-coords[i,0]
+ rhoy = 1.+boxSize[1]-coords[i,1]
+ if coords[i,1] < 0.75*boxSize[1]:
+ rhoy = 1. + 0.25*boxSize[1]
+ if coords[i,1] < 0.25*boxSize[1]:
+ rhoy = 1.+coords[i,1]
+ rhoz = 1.
+ rho[i] = rhox + rhoy + rhoz
+ P = 1.
+ u[i] = P / ((gamma-1.)*rho[i])
+ if gridtype == "stretched":
+ if coords[i,0] < 0.5:
+ m[i] = rho[i] * partVol1
+ else:
+ m[i] = rho[i] * partVol2
+ else:
+ m[i] = rho[i] * partVol
+
+#File
+file = h5py.File(fileName, 'w')
+
+# Header
+grp = file.create_group("/Header")
+grp.attrs["BoxSize"] = boxSize
+grp.attrs["NumPart_Total"] = [npart, 0, 0, 0, 0, 0]
+grp.attrs["NumPart_Total_HighWord"] = [0, 0, 0, 0, 0, 0]
+grp.attrs["NumPart_ThisFile"] = [npart, 0, 0, 0, 0, 0]
+grp.attrs["Time"] = 0.0
+grp.attrs["NumFilesPerSnapshot"] = 1
+grp.attrs["MassTable"] = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
+grp.attrs["Flag_Entropy_ICs"] = [0, 0, 0, 0, 0, 0]
+
+#Runtime parameters
+grp = file.create_group("/RuntimePars")
+grp.attrs["PeriodicBoundariesOn"] = periodic
+
+#Particle group
+grp = file.create_group("/PartType0")
+ds = grp.create_dataset('Coordinates', (npart, 3), 'd')
+ds[()] = coords
+ds = grp.create_dataset('Velocities', (npart, 3), 'f')
+ds[()] = v
+ds = grp.create_dataset('Masses', (npart,1), 'f')
+ds[()] = m
+ds = grp.create_dataset('Density', (npart,1), 'd')
+ds[()] = rho
+ds = grp.create_dataset('SmoothingLength', (npart,1), 'f')
+ds[()] = h
+ds = grp.create_dataset('InternalEnergy', (npart,1), 'd')
+ds[()] = u
+ds = grp.create_dataset('ParticleIDs', (npart,1), 'L')
+ds[()] = ids[:]
+
+file.close()
diff --git a/examples/Gradients/plot.py b/examples/Gradients/plot.py
new file mode 100644
index 0000000000000000000000000000000000000000..d6750ffc581437ebbf402ec44bcb1d14cb82a698
--- /dev/null
+++ b/examples/Gradients/plot.py
@@ -0,0 +1,50 @@
+################################################################################
+# This file is part of SWIFT.
+# Copyright (c) 2015 Bert Vandenbroucke (bert.vandenbroucke@ugent.be)
+#
+# 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/>.
+#
+################################################################################
+
+import scipy as sp
+import pylab as pl
+import numpy as np
+import h5py
+import sys
+
+# this file plots the gradients of the density in the x and y direction for
+# the given input file and saves the result as gradiens_NAME.png
+
+inputfile = sys.argv[1]
+outputfile = "gradients_{0}.png".format(sys.argv[2])
+
+f = h5py.File(inputfile, "r")
+rho = np.array(f["/PartType0/Density"])
+gradrho = np.array(f["/PartType0/GradDensity"])
+coords = np.array(f["/PartType0/Coordinates"])
+
+fig, ax = pl.subplots(1,2, sharey=True)
+
+ax[0].plot(coords[:,0], rho, "r.", label="density")
+ax[0].plot(coords[:,0], gradrho[:,0], "b.", label="grad density x")
+ax[0].set_xlabel("x")
+ax[0].legend(loc="best")
+
+ax[1].plot(coords[:,1], rho, "r.", label="density")
+ax[1].plot(coords[:,1], gradrho[:,1], "b.", label="grad density y")
+ax[1].set_xlabel("y")
+ax[1].legend(loc="best")
+
+pl.tight_layout()
+pl.savefig(outputfile)
diff --git a/examples/Gradients/run.sh b/examples/Gradients/run.sh
new file mode 100755
index 0000000000000000000000000000000000000000..cc1adc676427b257445f64a011ed8ebee87285ab
--- /dev/null
+++ b/examples/Gradients/run.sh
@@ -0,0 +1,13 @@
+#! /bin/bash
+
+python makeICs.py stretched
+../swift -s -t 2 gradientsStretched.yml
+python plot.py gradients_stretched_001.hdf5 stretched
+
+python makeICs.py cartesian
+../swift -s -t 2 gradientsCartesian.yml
+python plot.py gradients_cartesian_001.hdf5 cartesian
+
+python makeICs.py random
+../swift -s -t 2 gradientsRandom.yml
+python plot.py gradients_random_001.hdf5 random
diff --git a/examples/main.c b/examples/main.c
index 561f20ae3fefa723c0fb2c0677f12b9eae724306..7dcfee85478235e48351f7b8f9cf973cf5f0fda4 100644
--- a/examples/main.c
+++ b/examples/main.c
@@ -273,18 +273,6 @@ int main(int argc, char *argv[]) {
message("sizeof(struct cell) is %4zi bytes.", sizeof(struct cell));
}
-/* Temporary abort to handle absence of vectorized functions */
-#ifdef WITH_VECTORIZATION
-
-#ifdef MINIMAL_SPH
- error(
- "Vectorized version of Minimal SPH routines not implemented yet. "
- "Reconfigure with --disable-vec and recompile or use DEFAULT_SPH.");
-#endif
-
-#endif
- /* End temporary fix */
-
/* How vocal are we ? */
const int talking = (verbose == 1 && myrank == 0) || (verbose == 2);
@@ -487,6 +475,8 @@ int main(int argc, char *argv[]) {
#endif
if (myrank == 0)
message("Time integration ready to start. End of dry-run.");
+ engine_clean(&e);
+ free(params);
return 0;
}
@@ -553,7 +543,7 @@ int main(int argc, char *argv[]) {
if (!e.sched.tasks[l].skip && !e.sched.tasks[l].implicit) {
fprintf(
file_thread, " %03i %i %i %i %i %lli %lli %i %i %i %i %i\n",
- myrank, e.sched.tasks[l].last_rid, e.sched.tasks[l].type,
+ myrank, e.sched.tasks[l].rid, e.sched.tasks[l].type,
e.sched.tasks[l].subtype, (e.sched.tasks[l].cj == NULL),
e.sched.tasks[l].tic, e.sched.tasks[l].toc,
(e.sched.tasks[l].ci != NULL) ? e.sched.tasks[l].ci->count
@@ -589,7 +579,7 @@ int main(int argc, char *argv[]) {
if (!e.sched.tasks[l].skip && !e.sched.tasks[l].implicit)
fprintf(
file_thread, " %i %i %i %i %lli %lli %i %i %i %i\n",
- e.sched.tasks[l].last_rid, e.sched.tasks[l].type,
+ e.sched.tasks[l].rid, e.sched.tasks[l].type,
e.sched.tasks[l].subtype, (e.sched.tasks[l].cj == NULL),
e.sched.tasks[l].tic, e.sched.tasks[l].toc,
(e.sched.tasks[l].ci == NULL) ? 0 : e.sched.tasks[l].ci->count,
diff --git a/examples/plot_tasks.py b/examples/plot_tasks.py
index 6e71f476a106937f43bd4bd5973af01f65218afe..fb8b2ce57a47b4d397284bba9960b098c1e3ce62 100755
--- a/examples/plot_tasks.py
+++ b/examples/plot_tasks.py
@@ -56,9 +56,8 @@ pl.rcParams.update(PLOT_PARAMS)
# Tasks and subtypes. Indexed as in tasks.h.
TASKTYPES = ["none", "sort", "self", "pair", "sub_self", "sub_pair", "init", "ghost",
- "drift", "kick", "kick_fixdt", "send", "recv", "grav_gather_m", "grav_fft",
- "grav_mm", "grav_up", "grav_external", "part_sort", "gpart_sort",
- "split_cell", "rewait", "count"]
+ "kick", "kick_fixdt", "send", "recv", "grav_gather_m", "grav_fft",
+ "grav_mm", "grav_up", "grav_external", "count"]
TASKCOLOURS = {"none": "black",
"sort": "lightblue",
@@ -68,7 +67,6 @@ TASKCOLOURS = {"none": "black",
"sub_pair": "navy",
"init": "indigo",
"ghost": "cyan",
- "drift": "maroon",
"kick": "green",
"kick_fixdt": "green",
"send": "yellow",
@@ -78,20 +76,17 @@ TASKCOLOURS = {"none": "black",
"grav_mm": "mediumturquoise",
"grav_up": "mediumvioletred",
"grav_external": "darkred",
- "part_sort": "steelblue",
- "gpart_sort": "teal" ,
- "split_cell": "seagreen",
- "rewait": "olive",
"count": "powerblue"}
-SUBTYPES = ["none", "density", "force", "grav", "tend", "count"]
+SUBTYPES = ["none", "density", "gradient", "force", "grav", "tend", "count"]
SUBCOLOURS = {"none": "black",
"density": "red",
+ "gradient": "powerblue",
"force": "blue",
"grav": "indigo",
- "tend": "grey"
- "count": "purple"}
+ "tend": "grey",
+ "count": "black"}
# Show docs if help is requested.
if len( sys.argv ) == 2 and ( sys.argv[1][0:2] == "-h" or sys.argv[1][0:3] == "--h" ):
diff --git a/examples/plot_tasks_MPI.py b/examples/plot_tasks_MPI.py
index 7550899da2d4a34a5f73b192cbd7c348426786b7..398324cdc773d1dc4b7f26c58866c9df6469cc0b 100755
--- a/examples/plot_tasks_MPI.py
+++ b/examples/plot_tasks_MPI.py
@@ -62,9 +62,8 @@ pl.rcParams.update(PLOT_PARAMS)
# Tasks and subtypes. Indexed as in tasks.h.
TASKTYPES = ["none", "sort", "self", "pair", "sub_self", "sub_pair", "init", "ghost",
- "drift", "kick", "kick_fixdt", "send", "recv", "grav_gather_m", "grav_fft",
- "grav_mm", "grav_up", "grav_external", "part_sort", "gpart_sort",
- "split_cell", "rewait", "count"]
+ "kick", "kick_fixdt", "send", "recv", "grav_gather_m", "grav_fft",
+ "grav_mm", "grav_up", "grav_external", "count"]
TASKCOLOURS = {"none": "black",
"sort": "lightblue",
@@ -74,7 +73,6 @@ TASKCOLOURS = {"none": "black",
"sub_pair": "navy",
"init": "indigo",
"ghost": "cyan",
- "drift": "maroon",
"kick": "green",
"kick_fixdt": "green",
"send": "yellow",
@@ -84,20 +82,17 @@ TASKCOLOURS = {"none": "black",
"grav_mm": "mediumturquoise",
"grav_up": "mediumvioletred",
"grav_external": "darkred",
- "part_sort": "steelblue",
- "gpart_sort": "teal" ,
- "split_cell": "seagreen",
- "rewait": "olive",
"count": "powerblue"}
-SUBTYPES = ["none", "density", "force", "grav", "tend", "count"]
+SUBTYPES = ["none", "density", "gradient", "force", "grav", "tend", "count"]
SUBCOLOURS = {"none": "black",
"density": "red",
+ "gradient": "powerblue",
"force": "blue",
"grav": "indigo",
- "tend": "grey"
- "count": "purple"}
+ "tend": "grey",
+ "count": "black"}
# Show docs if help is requested.
if len( sys.argv ) == 2 and ( sys.argv[1][0:2] == "-h" or sys.argv[1][0:3] == "--h" ):
diff --git a/src/adiabatic_index.h b/src/adiabatic_index.h
index 78d1cc9d2deaf60c5d933b74a089c77446b44414..a0c9ce09e3e004af07e8b208ef9f1af5f46c9e81 100644
--- a/src/adiabatic_index.h
+++ b/src/adiabatic_index.h
@@ -1,6 +1,7 @@
/*******************************************************************************
* This file is part of SWIFT.
* Copyright (c) 2016 Matthieu Schaller (matthieu.schaller@durham.ac.uk).
+ * Bert Vandenbroucke (bert.vandenbroucke@gmail.com).
*
* 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
@@ -34,6 +35,7 @@
/* Local headers. */
#include "const.h"
#include "debug.h"
+#include "error.h"
#include "inline.h"
/* First define some constants */
@@ -42,18 +44,56 @@
#define hydro_gamma 1.66666666666666667f
#define hydro_gamma_minus_one 0.66666666666666667f
#define hydro_one_over_gamma_minus_one 1.5f
+#define hydro_gamma_plus_one_over_two_gamma 0.8f
+#define hydro_gamma_minus_one_over_two_gamma 0.2f
+#define hydro_gamma_minus_one_over_gamma_plus_one 0.25f
+#define hydro_two_over_gamma_plus_one 0.75f
+#define hydro_two_over_gamma_minus_one 3.f
+#define hydro_gamma_minus_one_over_two 0.33333333333333333f
+#define hydro_two_gamma_over_gamma_minus_one 5.f
+#define hydro_one_over_gamma 0.6f
+
+#elif defined(HYDRO_GAMMA_7_5)
+
+#define hydro_gamma 1.4f
+#define hydro_gamma_minus_one 0.4f
+#define hydro_one_over_gamma_minus_one 2.5f
+#define hydro_gamma_plus_one_over_two_gamma 0.857142857f
+#define hydro_gamma_minus_one_over_two_gamma 0.142857143f
+#define hydro_gamma_minus_one_over_gamma_plus_one 0.166666667f
+#define hydro_two_over_gamma_plus_one 0.833333333
+#define hydro_two_over_gamma_minus_one 5.f
+#define hydro_gamma_minus_one_over_two 0.2f
+#define hydro_two_gamma_over_gamma_minus_one 7.f
+#define hydro_one_over_gamma 0.714285714f
#elif defined(HYDRO_GAMMA_4_3)
#define hydro_gamma 1.33333333333333333f
#define hydro_gamma_minus_one 0.33333333333333333f
#define hydro_one_over_gamma_minus_one 3.f
+#define hydro_gamma_plus_one_over_two_gamma 0.875f
+#define hydro_gamma_minus_one_over_two_gamma 0.125f
+#define hydro_gamma_minus_one_over_gamma_plus_one 0.142857143f
+#define hydro_two_over_gamma_plus_one 0.857142857f
+#define hydro_two_over_gamma_minus_one 6.f
+#define hydro_gamma_minus_one_over_two 0.166666666666666666f
+#define hydro_two_gamma_over_gamma_minus_one 8.f
+#define hydro_one_over_gamma 0.75f
#elif defined(HYDRO_GAMMA_2_1)
#define hydro_gamma 2.f
#define hydro_gamma_minus_one 1.f
#define hydro_one_over_gamma_minus_one 1.f
+#define hydro_gamma_plus_one_over_two_gamma 0.75f
+#define hydro_gamma_minus_one_over_two_gamma 0.25f
+#define hydro_gamma_minus_one_over_gamma_plus_one 0.33333333333333333f
+#define hydro_two_over_gamma_plus_one 0.66666666666666666f
+#define hydro_two_over_gamma_minus_one 2.f
+#define hydro_gamma_minus_one_over_two 0.5f
+#define hydro_two_gamma_over_gamma_minus_one 4.f
+#define hydro_one_over_gamma 0.5f
#else
@@ -73,6 +113,10 @@ __attribute__((always_inline)) INLINE static float pow_gamma(float x) {
const float cbrt = cbrtf(x); /* x^(1/3) */
return cbrt * cbrt * x; /* x^(5/3) */
+#elif defined(HYDRO_GAMMA_7_5)
+
+ return powf(x, 1.4f); /* x^(7/5) */
+
#elif defined(HYDRO_GAMMA_4_3)
return cbrtf(x) * x; /* x^(4/3) */
@@ -103,6 +147,10 @@ __attribute__((always_inline)) INLINE static float pow_gamma_minus_one(
const float cbrt = cbrtf(x); /* x^(1/3) */
return cbrt * cbrt; /* x^(2/3) */
+#elif defined(HYDRO_GAMMA_7_5)
+
+ return powf(x, 0.4f); /* x^(2/5) */
+
#elif defined(HYDRO_GAMMA_4_3)
return cbrtf(x); /* x^(1/3) */
@@ -133,6 +181,10 @@ __attribute__((always_inline)) INLINE static float pow_minus_gamma_minus_one(
const float cbrt_inv = 1.f / cbrtf(x); /* x^(-1/3) */
return cbrt_inv * cbrt_inv; /* x^(-2/3) */
+#elif defined(HYDRO_GAMMA_7_5)
+
+ return powf(x, -0.4f); /* x^(-2/5) */
+
#elif defined(HYDRO_GAMMA_4_3)
return 1.f / cbrtf(x); /* x^(-1/3) */
@@ -149,4 +201,235 @@ __attribute__((always_inline)) INLINE static float pow_minus_gamma_minus_one(
#endif
}
+/**
+ * @brief Returns one over the argument to the power given by the adiabatic
+ * index
+ *
+ * Computes \f$x^{-\gamma}\f$.
+ *
+ * @param x Argument
+ * @return One over the argument to the power given by the adiabatic index
+ */
+__attribute__((always_inline)) INLINE static float pow_minus_gamma(float x) {
+
+#if defined(HYDRO_GAMMA_5_3)
+
+ const float cbrt_inv = 1.f / cbrtf(x); /* x^(-1/3) */
+ const float cbrt_inv2 = cbrt_inv * cbrt_inv; /* x^(-2/3) */
+ return cbrt_inv * cbrt_inv2 * cbrt_inv2; /* x^(-5/3) */
+
+#elif defined(HYDRO_GAMMA_7_5)
+
+ return powf(x, -1.4f); /* x^(-7/5) */
+
+#elif defined(HYDRO_GAMMA_4_3)
+
+ const float cbrt_inv = 1.f / cbrtf(x); /* x^(-1/3) */
+ const float cbrt_inv2 = cbrt_inv * cbrt_inv; /* x^(-2/3) */
+ return cbrt_inv2 * cbrt_inv2; /* x^(-4/3) */
+
+#elif defined(HYDRO_GAMMA_2_1)
+
+ const float inv = 1.f / x;
+ return inv * inv;
+
+#else
+
+ error("The adiabatic index is not defined !");
+ return 0.f;
+
+#endif
+}
+
+/**
+ * @brief Return the argument to the power given by two divided by the adiabatic
+ * index minus one
+ *
+ * Computes \f$x^{\frac{2}{\gamma - 1}}\f$.
+ *
+ * @param x Argument
+ * @return Argument to the power two divided by the adiabatic index minus one
+ */
+__attribute__((always_inline)) INLINE static float pow_two_over_gamma_minus_one(
+ float x) {
+
+#if defined(HYDRO_GAMMA_5_3)
+
+ return x * x * x; /* x^3 */
+
+#elif defined(HYDRO_GAMMA_7_5)
+
+ const float x2 = x * x;
+ const float x3 = x2 * x;
+ return x2 * x3;
+
+#elif defined(HYDRO_GAMMA_4_3)
+
+ const float x3 = x * x * x; /* x^3 */
+ return x3 * x3; /* x^6 */
+
+#elif defined(HYDRO_GAMMA_2_1)
+
+ return x * x; /* x^2 */
+
+#else
+
+ error("The adiabatic index is not defined !");
+ return 0.f;
+
+#endif
+}
+
+/**
+ * @brief Return the argument to the power given by two times the adiabatic
+ * index divided by the adiabatic index minus one
+ *
+ * Computes \f$x^{\frac{2\gamma}{\gamma - 1}}\f$.
+ *
+ * @param x Argument
+ * @return Argument to the power two times the adiabatic index divided by the
+ * adiabatic index minus one
+ */
+__attribute__((always_inline)) INLINE static float
+pow_two_gamma_over_gamma_minus_one(float x) {
+
+#if defined(HYDRO_GAMMA_5_3)
+
+ const float x2 = x * x;
+ const float x3 = x2 * x;
+ return x2 * x3;
+
+#elif defined(HYDRO_GAMMA_7_5)
+
+ const float x2 = x * x;
+ const float x4 = x2 * x2;
+ return x4 * x2 * x;
+
+#elif defined(HYDRO_GAMMA_4_3)
+
+ const float x2 = x * x;
+ const float x4 = x2 * x2;
+ return x4 * x4; /* x^8 */
+
+#elif defined(HYDRO_GAMMA_2_1)
+
+ const float x2 = x * x;
+ return x2 * x2; /* x^4 */
+
+#else
+
+ error("The adiabatic index is not defined !");
+ return 0.f;
+
+#endif
+}
+
+/**
+ * @brief Return the argument to the power given by the adiabatic index minus
+ * one divided by two times the adiabatic index
+ *
+ * Computes \f$x^{\frac{\gamma - 1}{2\gamma}}\f$.
+ *
+ * @param x Argument
+ * @return Argument to the power the adiabatic index minus one divided by two
+ * times the adiabatic index
+ */
+__attribute__((always_inline)) INLINE static float
+pow_gamma_minus_one_over_two_gamma(float x) {
+
+#if defined(HYDRO_GAMMA_5_3)
+
+ return powf(x, 0.2f); /* x^0.2 */
+
+#elif defined(HYDRO_GAMMA_7_5)
+
+ return powf(x, hydro_gamma_minus_one_over_two_gamma);
+
+#elif defined(HYDRO_GAMMA_4_3)
+
+ return powf(x, 0.125f); /* x^0.125 */
+
+#elif defined(HYDRO_GAMMA_2_1)
+
+ return powf(x, 0.25f); /* x^0.25 */
+
+#else
+
+ error("The adiabatic index is not defined !");
+ return 0.f;
+
+#endif
+}
+
+/**
+ * @brief Return the inverse argument to the power given by the adiabatic index
+ * plus one divided by two times the adiabatic index
+ *
+ * Computes \f$x^{-\frac{\gamma + 1}{2\gamma}}\f$.
+ *
+ * @param x Argument
+ * @return Inverse argument to the power the adiabatic index plus one divided by
+ * two times the adiabatic index
+ */
+__attribute__((always_inline)) INLINE static float
+pow_minus_gamma_plus_one_over_two_gamma(float x) {
+
+#if defined(HYDRO_GAMMA_5_3)
+
+ return powf(x, -0.8f); /* x^-0.8 */
+
+#elif defined(HYDRO_GAMMA_7_5)
+
+ return powf(x, -hydro_gamma_plus_one_over_two_gamma);
+
+#elif defined(HYDRO_GAMMA_4_3)
+
+ return powf(x, -0.875f); /* x^-0.875 */
+
+#elif defined(HYDRO_GAMMA_2_1)
+
+ return powf(x, -0.75f); /* x^-0.75 */
+
+#else
+
+ error("The adiabatic index is not defined !");
+ return 0.f;
+
+#endif
+}
+
+/**
+ * @brief Return the argument to the power one over the adiabatic index
+ *
+ * Computes \f$x^{\frac{1}{\gamma}}\f$.
+ *
+ * @param x Argument
+ * @return Argument to the power one over the adiabatic index
+ */
+__attribute__((always_inline)) INLINE static float pow_one_over_gamma(float x) {
+
+#if defined(HYDRO_GAMMA_5_3)
+
+ return powf(x, 0.6f); /* x^(3/5) */
+
+#elif defined(HYDRO_GAMMA_7_5)
+
+ return powf(x, hydro_one_over_gamma);
+
+#elif defined(HYDRO_GAMMA_4_3)
+
+ return powf(x, 0.75f); /* x^(3/4) */
+
+#elif defined(HYDRO_GAMMA_2_1)
+
+ return sqrtf(x); /* x^(1/2) */
+
+#else
+
+ error("The adiabatic index is not defined !");
+ return 0.f;
+
+#endif
+}
+
#endif /* SWIFT_ADIABATIC_INDEX_H */
diff --git a/src/cell.c b/src/cell.c
index 9cf64b89005c497d70feccb6bc4067bdbfb8ee02..094645ba50e35cd06e3a3f0c333741cb34d4858d 100644
--- a/src/cell.c
+++ b/src/cell.c
@@ -704,6 +704,9 @@ void cell_clean_links(struct cell *c, void *data) {
c->density = NULL;
c->nr_density = 0;
+ c->gradient = NULL;
+ c->nr_gradient = 0;
+
c->force = NULL;
c->nr_force = 0;
}
diff --git a/src/cell.h b/src/cell.h
index 605d0c85444c1ac3b4147ef2fb23df52b2694ff5..19e347225d59d6bdb86091efaa4a06bbf981422a 100644
--- a/src/cell.h
+++ b/src/cell.h
@@ -134,18 +134,20 @@ struct cell {
int sortsize, gsortsize;
/* The tasks computing this cell's density. */
- struct link *density, *force, *grav;
- int nr_density, nr_force, nr_grav;
+ struct link *density, *gradient, *force, *grav;
+ int nr_density, nr_gradient, nr_force, nr_grav;
/* The hierarchical tasks. */
- struct task *ghost, *init, *kick;
+ struct task *extra_ghost, *ghost, *init, *kick;
#ifdef WITH_MPI
- /* Task receiving data. */
- struct task *recv_xv, *recv_rho, *recv_ti;
+
+/* Task receiving data. */
+ struct task *recv_xv, *recv_rho, *recv_gradient, *recv_ti;
/* Task send data. */
- struct link *send_xv, *send_rho, *send_ti;
+ struct link *send_xv, *send_rho, *send_gradient, *send_ti;
+
#endif
/* Tasks for gravity tree. */
@@ -215,6 +217,8 @@ 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);
+void cell_init_parts(struct cell *c, void *data);
+void cell_init_gparts(struct cell *c, void *data);
void cell_convert_hydro(struct cell *c, void *data);
void cell_clean_links(struct cell *c, void *data);
int cell_are_neighbours(const struct cell *restrict ci,
diff --git a/src/common_io.c b/src/common_io.c
index 3c001d9da106a46ef5033c8cdec9346d68c54ecd..37e2837fbaeee87916ddea9264439c824149479c 100644
--- a/src/common_io.c
+++ b/src/common_io.c
@@ -42,6 +42,7 @@
/* Local includes. */
#include "const.h"
#include "error.h"
+#include "hydro.h"
#include "kernel_hydro.h"
#include "part.h"
#include "units.h"
@@ -515,13 +516,13 @@ void writeXMFgroupheader(FILE* xmfFile, char* hdfFileName, size_t N,
fprintf(xmfFile, "\n<Grid Name=\"%s\" GridType=\"Uniform\">\n",
particle_type_names[ptype]);
fprintf(xmfFile,
- "<Topology TopologyType=\"Polyvertex\" Dimensions=\"%zi\"/>\n", N);
+ "<Topology TopologyType=\"Polyvertex\" Dimensions=\"%zu\"/>\n", N);
fprintf(xmfFile, "<Geometry GeometryType=\"XYZ\">\n");
fprintf(xmfFile,
- "<DataItem Dimensions=\"%zi 3\" NumberType=\"Double\" "
+ "<DataItem Dimensions=\"%zu 3\" NumberType=\"Double\" "
"Precision=\"8\" "
"Format=\"HDF\">%s:/PartType%d/Coordinates</DataItem>\n",
- N, hdfFileName, ptype);
+ N, hdfFileName, (int)ptype);
fprintf(xmfFile,
"</Geometry>\n <!-- Done geometry for %s, start of particle fields "
"list -->\n",
@@ -555,12 +556,12 @@ void writeXMFline(FILE* xmfFile, const char* fileName,
name, dim == 1 ? "Scalar" : "Vector");
if (dim == 1)
fprintf(xmfFile,
- "<DataItem Dimensions=\"%zi\" NumberType=\"Double\" "
+ "<DataItem Dimensions=\"%zu\" NumberType=\"Double\" "
"Precision=\"%d\" Format=\"HDF\">%s:%s/%s</DataItem>\n",
N, type == FLOAT ? 4 : 8, fileName, partTypeGroupName, name);
else
fprintf(xmfFile,
- "<DataItem Dimensions=\"%zi %d\" NumberType=\"Double\" "
+ "<DataItem Dimensions=\"%zu %d\" NumberType=\"Double\" "
"Precision=\"%d\" Format=\"HDF\">%s:%s/%s</DataItem>\n",
N, dim, type == FLOAT ? 4 : 8, fileName, partTypeGroupName, name);
fprintf(xmfFile, "</Attribute>\n");
@@ -582,7 +583,7 @@ void prepare_dm_gparts(struct gpart* const gparts, size_t Ndm) {
for (size_t i = 0; i < Ndm; ++i) {
/* 0 or negative ids are not allowed */
if (gparts[i].id_or_neg_offset <= 0)
- error("0 or negative ID for DM particle %zd: ID=%lld", i,
+ error("0 or negative ID for DM particle %zu: ID=%lld", i,
gparts[i].id_or_neg_offset);
}
}
@@ -614,7 +615,7 @@ void duplicate_hydro_gparts(struct part* const parts,
gparts[i + Ndm].v_full[1] = parts[i].v[1];
gparts[i + Ndm].v_full[2] = parts[i].v[2];
- gparts[i + Ndm].mass = parts[i].mass;
+ gparts[i + Ndm].mass = hydro_get_mass(&parts[i]);
/* Link the particles */
gparts[i + Ndm].id_or_neg_offset = -i;
@@ -650,7 +651,7 @@ void collect_dm_gparts(const struct gpart* const gparts, size_t Ntot,
/* Check that everything is fine */
if (count != Ndm)
- error("Collected the wrong number of dm particles (%zd vs. %zd expected)",
+ error("Collected the wrong number of dm particles (%zu vs. %zu expected)",
count, Ndm);
}
diff --git a/src/const.h b/src/const.h
index cbaa0054fe3bc9f829c59b19b71197264fd0c5af..5f7bf0d26431887c8e37d96b20b0eba460de55a6 100644
--- a/src/const.h
+++ b/src/const.h
@@ -46,6 +46,7 @@
/* Hydrodynamical adiabatic index. */
#define HYDRO_GAMMA_5_3
+//#define HYDRO_GAMMA_7_5
//#define HYDRO_GAMMA_4_3
//#define HYDRO_GAMMA_2_1
@@ -65,6 +66,22 @@
//#define MINIMAL_SPH
#define GADGET2_SPH
//#define DEFAULT_SPH
+//#define GIZMO_SPH
+
+/* Riemann solver to use (GIZMO_SPH only) */
+#define RIEMANN_SOLVER_EXACT
+//#define RIEMANN_SOLVER_TRRS
+//#define RIEMANN_SOLVER_HLLC
+
+/* Type of gradients to use (GIZMO_SPH only) */
+/* If no option is chosen, no gradients are used (first order scheme) */
+//#define GRADIENTS_SPH
+#define GRADIENTS_GIZMO
+
+/* Types of slope limiter to use (GIZMO_SPH only) */
+/* Different slope limiters can be combined */
+#define SLOPE_LIMITER_PER_FACE
+#define SLOPE_LIMITER_CELL_WIDE
/* Self gravity stuff. */
#define const_gravity_multipole_order 2
diff --git a/src/debug.c b/src/debug.c
index 4601f9bacc2d51d988ee0212b4ccf778bcf15f5c..15354b7d419544a8456543b79c38235eb3b68b2c 100644
--- a/src/debug.c
+++ b/src/debug.c
@@ -46,6 +46,8 @@
#include "./hydro/Gadget2/hydro_debug.h"
#elif defined(DEFAULT_SPH)
#include "./hydro/Default/hydro_debug.h"
+#elif defined(GIZMO_SPH)
+#include "./hydro/Gizmo/hydro_debug.h"
#else
#error "Invalid choice of SPH variant"
#endif
@@ -71,7 +73,7 @@ void printParticle(const struct part *parts, struct xpart *xparts,
/* Look for the particle. */
for (size_t i = 0; i < N; i++)
if (parts[i].id == id) {
- printf("## Particle[%zd]:\n id=%lld ", i, parts[i].id);
+ printf("## Particle[%zu]:\n id=%lld ", i, parts[i].id);
hydro_debug_particle(&parts[i], &xparts[i]);
found = 1;
break;
@@ -100,13 +102,13 @@ void printgParticle(const struct gpart *gparts, const struct part *parts,
/* Look for the particle. */
for (size_t i = 0; i < N; i++)
if (gparts[i].id_or_neg_offset == id) {
- printf("## gParticle[%zd] (DM) :\n id=%lld", i, id);
+ printf("## gParticle[%zu] (DM) :\n id=%lld", i, id);
gravity_debug_particle(&gparts[i]);
found = 1;
break;
} 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);
+ printf("## gParticle[%zu] (hydro) :\n id=%lld", i, id);
gravity_debug_particle(&gparts[i]);
found = 1;
break;
@@ -159,7 +161,7 @@ int checkSpacehmax(struct space *s) {
/* Now all particles. */
float part_h_max = 0.0f;
- for (int k = 0; k < s->nr_parts; k++) {
+ for (size_t k = 0; k < s->nr_parts; k++) {
if (s->parts[k].h > part_h_max) {
part_h_max = s->parts[k].h;
}
@@ -178,7 +180,7 @@ int checkSpacehmax(struct space *s) {
}
}
- for (int k = 0; k < s->nr_parts; k++) {
+ for (size_t k = 0; k < s->nr_parts; k++) {
if (s->parts[k].h > cell_h_max) {
message("part %lld is inconsistent (%f > %f)", s->parts[k].id,
s->parts[k].h, cell_h_max);
diff --git a/src/dimension.h b/src/dimension.h
index 6395d4d04e50d40b733e7a74dafb7d0ab277d204..0fae2c5602b87622ff67f6f5feb325efc6422472 100644
--- a/src/dimension.h
+++ b/src/dimension.h
@@ -33,6 +33,8 @@
#include "inline.h"
#include "vector.h"
+#include <math.h>
+
/* First define some constants */
#if defined(HYDRO_DIMENSION_3D)
@@ -114,6 +116,92 @@ __attribute__((always_inline)) INLINE static float pow_dimension_plus_one(
#endif
}
+/**
+ * @brief Inverts the given dimension by dimension matrix (in place)
+ *
+ * @param A A 3x3 matrix of which we want to invert the top left dxd part
+ */
+__attribute__((always_inline)) INLINE static void
+invert_dimension_by_dimension_matrix(float A[3][3]) {
+
+#if defined(HYDRO_DIMENSION_3D)
+
+ float detA, Ainv[3][3];
+
+ detA = A[0][0] * A[1][1] * A[2][2] + A[0][1] * A[1][2] * A[2][0] +
+ A[0][2] * A[1][0] * A[2][1] - A[0][2] * A[1][1] * A[2][0] -
+ A[0][1] * A[1][0] * A[2][2] - A[0][0] * A[1][2] * A[2][1];
+
+ if (detA && !isnan(detA)) {
+ Ainv[0][0] = (A[1][1] * A[2][2] - A[1][2] * A[2][1]) / detA;
+ Ainv[0][1] = (A[0][2] * A[2][1] - A[0][1] * A[2][2]) / detA;
+ Ainv[0][2] = (A[0][1] * A[1][2] - A[0][2] * A[1][1]) / detA;
+ Ainv[1][0] = (A[1][2] * A[2][0] - A[1][0] * A[2][2]) / detA;
+ Ainv[1][1] = (A[0][0] * A[2][2] - A[0][2] * A[2][0]) / detA;
+ Ainv[1][2] = (A[0][2] * A[1][0] - A[0][0] * A[1][2]) / detA;
+ Ainv[2][0] = (A[1][0] * A[2][1] - A[1][1] * A[2][0]) / detA;
+ Ainv[2][1] = (A[0][1] * A[2][0] - A[0][0] * A[2][1]) / detA;
+ Ainv[2][2] = (A[0][0] * A[1][1] - A[0][1] * A[1][0]) / detA;
+ } else {
+ Ainv[0][0] = 0.0f;
+ Ainv[0][1] = 0.0f;
+ Ainv[0][2] = 0.0f;
+ Ainv[1][0] = 0.0f;
+ Ainv[1][1] = 0.0f;
+ Ainv[1][2] = 0.0f;
+ Ainv[2][0] = 0.0f;
+ Ainv[2][1] = 0.0f;
+ Ainv[2][2] = 0.0f;
+ }
+
+ A[0][0] = Ainv[0][0];
+ A[0][1] = Ainv[0][1];
+ A[0][2] = Ainv[0][2];
+ A[1][0] = Ainv[1][0];
+ A[1][1] = Ainv[1][1];
+ A[1][2] = Ainv[1][2];
+ A[2][0] = Ainv[2][0];
+ A[2][1] = Ainv[2][1];
+ A[2][2] = Ainv[2][2];
+
+#elif defined(HYDRO_DIMENSION_2D)
+
+ float detA, Ainv[2][2];
+
+ detA = A[0][0] * A[1][1] - A[0][1] * A[1][0];
+
+ if (detA && !isnan(detA)) {
+ Ainv[0][0] = A[1][1] / detA;
+ Ainv[0][1] = -A[0][1] / detA;
+ Ainv[1][0] = -A[1][0] / detA;
+ Ainv[1][1] = A[0][0] / detA;
+ } else {
+ Ainv[0][0] = 0.0f;
+ Ainv[0][1] = 0.0f;
+ Ainv[1][0] = 0.0f;
+ Ainv[1][1] = 0.0f;
+ }
+
+ A[0][0] = Ainv[0][0];
+ A[0][1] = Ainv[0][1];
+ A[1][0] = Ainv[1][0];
+ A[1][1] = Ainv[1][1];
+
+#elif defined(HYDRO_DIMENSION_1D)
+
+ if (A[0][0] && !isnan(A[0][0])) {
+ A[0][0] = 1.0f / A[0][0];
+ } else {
+ A[0][0] = 0.0f;
+ }
+
+#else
+
+ error("The dimension is not defined !");
+
+#endif
+}
+
/* ------------------------------------------------------------------------- */
#ifdef WITH_VECTORIZATION
diff --git a/src/drift.h b/src/drift.h
index 880595dc59e3e5174ee5e888e595a9204ad383e2..bd1b35926740d49a67291ede4676f3387cd66748 100644
--- a/src/drift.h
+++ b/src/drift.h
@@ -65,8 +65,6 @@ __attribute__((always_inline)) INLINE static void drift_gpart(
__attribute__((always_inline)) INLINE static void drift_part(
struct part *restrict p, struct xpart *restrict xp, float dt,
double timeBase, int ti_old, int ti_current) {
- /* Useful quantity */
- const float h_inv = 1.0f / p->h;
/* Drift... */
p->x[0] += xp->v_full[0] * dt;
@@ -78,22 +76,8 @@ __attribute__((always_inline)) INLINE static void drift_part(
p->v[1] += p->a_hydro[1] * dt;
p->v[2] += p->a_hydro[2] * dt;
- /* Predict smoothing length */
- const float w1 = p->force.h_dt * h_inv * dt;
- if (fabsf(w1) < 0.2f)
- p->h *= approx_expf(w1); /* 4th order expansion of exp(w) */
- else
- p->h *= expf(w1);
-
- /* Predict density */
- const float w2 = -hydro_dimension * w1;
- if (fabsf(w2) < 0.2f)
- p->rho *= approx_expf(w2); /* 4th order expansion of exp(w) */
- else
- p->rho *= expf(w2);
-
/* Predict the values of the extra fields */
- hydro_predict_extra(p, xp, ti_old, ti_current, timeBase);
+ hydro_predict_extra(p, xp, dt, ti_old, ti_current, timeBase);
/* Compute offset since last cell construction */
xp->x_diff[0] -= xp->v_full[0] * dt;
diff --git a/src/engine.c b/src/engine.c
index 2d4b8460c0a1373cfa67832c4437472c64394d7f..0b09e2766e52e7ac2c354f038aab8d13c0a52553 100644
--- a/src/engine.c
+++ b/src/engine.c
@@ -219,6 +219,12 @@ void engine_make_hydro_hierarchical_tasks(struct engine *e, struct cell *c,
/* Generate the ghost task. */
c->ghost = scheduler_addtask(s, task_type_ghost, task_subtype_none, 0, 0,
c, NULL, 0);
+
+#ifdef EXTRA_HYDRO_LOOP
+ /* Generate the extra ghost task. */
+ c->extra_ghost = scheduler_addtask(s, task_type_extra_ghost,
+ task_subtype_none, 0, 0, c, NULL, 0);
+#endif
}
}
@@ -297,7 +303,7 @@ void engine_redistribute(struct engine *e) {
parts[k].x[2] * iwidth[2]);
#ifdef SWIFT_DEBUG_CHECKS
if (cid < 0 || cid >= s->nr_cells)
- error("Bad cell id %i for part %zi at [%.3e,%.3e,%.3e].", cid, k,
+ error("Bad cell id %i for part %zu at [%.3e,%.3e,%.3e].", cid, k,
parts[k].x[0], parts[k].x[1], parts[k].x[2]);
#endif
@@ -353,7 +359,7 @@ void engine_redistribute(struct engine *e) {
gparts[k].x[2] * iwidth[2]);
#ifdef SWIFT_DEBUG_CHECKS
if (cid < 0 || cid >= s->nr_cells)
- error("Bad cell id %i for part %zi at [%.3e,%.3e,%.3e].", cid, k,
+ error("Bad cell id %i for part %zu at [%.3e,%.3e,%.3e].", cid, k,
gparts[k].x[0], gparts[k].x[1], gparts[k].x[2]);
#endif
@@ -588,7 +594,7 @@ void engine_redistribute(struct engine *e) {
int my_cells = 0;
for (int k = 0; k < nr_cells; k++)
if (cells[k].nodeID == nodeID) my_cells += 1;
- message("node %i now has %zi parts and %zi gparts in %i cells.", nodeID,
+ message("node %i now has %zu parts and %zu gparts in %i cells.", nodeID,
nr_parts, nr_gparts, my_cells);
}
@@ -676,11 +682,12 @@ void engine_addtasks_grav(struct engine *e, struct cell *c, struct task *up,
* @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_gradient The send_gradient #task, if already 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,
struct task *t_xv, struct task *t_rho,
- struct task *t_ti) {
+ struct task *t_gradient, struct task *t_ti) {
#ifdef WITH_MPI
struct link *l = NULL;
@@ -699,21 +706,42 @@ void engine_addtasks_send(struct engine *e, struct cell *ci, struct cell *cj,
/* Create the tasks and their dependencies? */
if (t_xv == NULL) {
t_xv = scheduler_addtask(s, task_type_send, task_subtype_none,
- 3 * ci->tag, 0, ci, cj, 0);
+ 4 * ci->tag, 0, ci, cj, 0);
t_rho = scheduler_addtask(s, task_type_send, task_subtype_none,
- 3 * ci->tag + 1, 0, ci, cj, 0);
+ 4 * ci->tag + 1, 0, ci, cj, 0);
if (!(e->policy & engine_policy_fixdt))
t_ti = scheduler_addtask(s, task_type_send, task_subtype_tend,
- 3 * ci->tag + 2, 0, ci, cj, 0);
+ 4 * ci->tag + 2, 0, ci, cj, 0);
+#ifdef EXTRA_HYDRO_LOOP
+ t_gradient = scheduler_addtask(s, task_type_send, task_subtype_none,
+ 4 * ci->tag + 3, 0, ci, cj, 0);
+#endif
+
+#ifdef EXTRA_HYDRO_LOOP
+
+ scheduler_addunlock(s, t_gradient, ci->super->kick);
+
+ scheduler_addunlock(s, ci->super->extra_ghost, t_gradient);
+
+ /* The send_rho task should unlock the super-cell's extra_ghost task. */
+ scheduler_addunlock(s, t_rho, ci->super->extra_ghost);
/* The send_rho task depends on the cell's ghost task. */
scheduler_addunlock(s, ci->super->ghost, t_rho);
+ /* The send_xv task should unlock the super-cell's ghost task. */
+ scheduler_addunlock(s, t_xv, ci->super->ghost);
+
+#else
/* The send_rho task should unlock the super-cell's kick task. */
scheduler_addunlock(s, t_rho, ci->super->kick);
+ /* The send_rho task depends on the cell's ghost task. */
+ scheduler_addunlock(s, ci->super->ghost, t_rho);
+
/* The send_xv task should unlock the super-cell's ghost task. */
scheduler_addunlock(s, t_xv, ci->super->ghost);
+#endif
/* The super-cell's kick task should unlock the send_ti task. */
if (t_ti != NULL) scheduler_addunlock(s, ci->super->kick, t_ti);
@@ -722,6 +750,9 @@ void engine_addtasks_send(struct engine *e, struct cell *ci, struct cell *cj,
/* Add them to the local cell. */
engine_addlink(e, &ci->send_xv, t_xv);
engine_addlink(e, &ci->send_rho, t_rho);
+#ifdef EXTRA_HYDRO_LOOP
+ engine_addlink(e, &ci->send_gradient, t_gradient);
+#endif
if (t_ti != NULL) engine_addlink(e, &ci->send_ti, t_ti);
}
@@ -729,7 +760,8 @@ void engine_addtasks_send(struct engine *e, struct cell *ci, struct cell *cj,
if (ci->split)
for (int k = 0; k < 8; k++)
if (ci->progeny[k] != NULL)
- engine_addtasks_send(e, ci->progeny[k], cj, t_xv, t_rho, t_ti);
+ engine_addtasks_send(e, ci->progeny[k], cj, t_xv, t_rho, t_gradient,
+ t_ti);
#else
error("SWIFT was not compiled with MPI support.");
@@ -743,11 +775,13 @@ void engine_addtasks_send(struct engine *e, struct cell *ci, struct cell *cj,
* @param c The foreign #cell.
* @param t_xv The recv_xv #task, if it has already been created.
* @param t_rho The recv_rho #task, if it has already been created.
+ * @param t_gradient The recv_gradient #task, if it has already been created.
* @param t_ti The recv_ti #task, if required and has already been created.
*/
void engine_addtasks_recv(struct engine *e, struct cell *c, struct task *t_xv,
- struct task *t_rho, struct task *t_ti) {
+ struct task *t_rho, struct task *t_gradient,
+ struct task *t_ti) {
#ifdef WITH_MPI
struct scheduler *s = &e->sched;
@@ -758,19 +792,39 @@ void engine_addtasks_recv(struct engine *e, struct cell *c, struct task *t_xv,
if (t_xv == NULL && c->density != NULL) {
/* Create the tasks. */
- t_xv = scheduler_addtask(s, task_type_recv, task_subtype_none, 3 * c->tag,
+ t_xv = scheduler_addtask(s, task_type_recv, task_subtype_none, 4 * c->tag,
0, c, NULL, 0);
t_rho = scheduler_addtask(s, task_type_recv, task_subtype_none,
- 3 * c->tag + 1, 0, c, NULL, 0);
+ 4 * c->tag + 1, 0, c, NULL, 0);
if (!(e->policy & engine_policy_fixdt))
t_ti = scheduler_addtask(s, task_type_recv, task_subtype_tend,
- 3 * c->tag + 2, 0, c, NULL, 0);
+ 4 * c->tag + 2, 0, c, NULL, 0);
+#ifdef EXTRA_HYDRO_LOOP
+ t_gradient = scheduler_addtask(s, task_type_recv, task_subtype_none,
+ 4 * c->tag + 3, 0, c, NULL, 0);
+#endif
}
c->recv_xv = t_xv;
c->recv_rho = t_rho;
+ c->recv_gradient = t_gradient;
c->recv_ti = t_ti;
- /* Add dependencies. */
+/* Add dependencies. */
+#ifdef EXTRA_HYDRO_LOOP
+ for (struct link *l = c->density; l != NULL; l = l->next) {
+ scheduler_addunlock(s, t_xv, l->t);
+ scheduler_addunlock(s, l->t, t_rho);
+ }
+ for (struct link *l = c->gradient; l != NULL; l = l->next) {
+ scheduler_addunlock(s, t_rho, l->t);
+ scheduler_addunlock(s, l->t, t_gradient);
+ }
+ for (struct link *l = c->force; l != NULL; l = l->next) {
+ scheduler_addunlock(s, t_gradient, l->t);
+ if (t_ti != NULL) scheduler_addunlock(s, l->t, t_ti);
+ }
+ if (c->sorts != NULL) scheduler_addunlock(s, t_xv, c->sorts);
+#else
for (struct link *l = c->density; l != NULL; l = l->next) {
scheduler_addunlock(s, t_xv, l->t);
scheduler_addunlock(s, l->t, t_rho);
@@ -780,12 +834,13 @@ void engine_addtasks_recv(struct engine *e, struct cell *c, struct task *t_xv,
if (t_ti != NULL) scheduler_addunlock(s, l->t, t_ti);
}
if (c->sorts != NULL) scheduler_addunlock(s, t_xv, c->sorts);
+#endif
/* Recurse? */
if (c->split)
for (int k = 0; k < 8; k++)
if (c->progeny[k] != NULL)
- engine_addtasks_recv(e, c->progeny[k], t_xv, t_rho, t_ti);
+ engine_addtasks_recv(e, c->progeny[k], t_xv, t_rho, t_gradient, t_ti);
#else
error("SWIFT was not compiled with MPI support.");
@@ -879,7 +934,7 @@ void engine_exchange_cells(struct engine *e) {
/* Count the number of particles we need to import and re-allocate
the buffer if needed. */
- int count_parts_in = 0, count_gparts_in = 0;
+ size_t count_parts_in = 0, count_gparts_in = 0;
for (int k = 0; k < nr_proxies; k++)
for (int j = 0; j < e->proxies[k].nr_cells_in; j++) {
count_parts_in += e->proxies[k].cells_in[j]->count;
@@ -969,7 +1024,7 @@ void engine_exchange_strays(struct engine *e, size_t offset_parts,
if (pid < 0) {
error(
"Do not have a proxy for the requested nodeID %i for part with "
- "id=%llu, x=[%e,%e,%e].",
+ "id=%lld, x=[%e,%e,%e].",
node_id, s->parts[offset_parts + k].id,
s->parts[offset_parts + k].x[0], s->parts[offset_parts + k].x[1],
s->parts[offset_parts + k].x[2]);
@@ -1033,7 +1088,7 @@ void engine_exchange_strays(struct engine *e, size_t offset_parts,
count_gparts_in += e->proxies[k].nr_gparts_in;
}
if (e->verbose) {
- message("sent out %zi/%zi parts/gparts, got %i/%i back.", *Npart, *Ngpart,
+ message("sent out %zu/%zu parts/gparts, got %i/%i back.", *Npart, *Ngpart,
count_parts_in, count_gparts_in);
}
if (offset_parts + count_parts_in > s->size_parts) {
@@ -1127,13 +1182,13 @@ void engine_exchange_strays(struct engine *e, size_t offset_parts,
reqs_in[pid + 1] == MPI_REQUEST_NULL &&
reqs_in[pid + 2] == MPI_REQUEST_NULL) {
/* Copy the particle data to the part/xpart/gpart arrays. */
- struct proxy *p = &e->proxies[pid / 3];
- memcpy(&s->parts[offset_parts + count_parts], p->parts_in,
- sizeof(struct part) * p->nr_parts_in);
- memcpy(&s->xparts[offset_parts + count_parts], p->xparts_in,
- sizeof(struct xpart) * p->nr_parts_in);
- memcpy(&s->gparts[offset_gparts + count_gparts], p->gparts_in,
- sizeof(struct gpart) * p->nr_gparts_in);
+ struct proxy *prox = &e->proxies[pid / 3];
+ memcpy(&s->parts[offset_parts + count_parts], prox->parts_in,
+ sizeof(struct part) * prox->nr_parts_in);
+ memcpy(&s->xparts[offset_parts + count_parts], prox->xparts_in,
+ sizeof(struct xpart) * prox->nr_parts_in);
+ memcpy(&s->gparts[offset_gparts + count_gparts], prox->gparts_in,
+ sizeof(struct gpart) * prox->nr_gparts_in);
/* for (int k = offset; k < offset + count; k++)
message(
"received particle %lli, x=[%.3e %.3e %.3e], h=%.3e, from node %i.",
@@ -1141,8 +1196,8 @@ void engine_exchange_strays(struct engine *e, size_t offset_parts,
s->parts[k].x[2], s->parts[k].h, p->nodeID); */
/* Re-link the gparts. */
- for (int k = 0; k < p->nr_gparts_in; k++) {
- struct gpart *gp = &s->gparts[offset_gparts + count_gparts + k];
+ for (int kk = 0; kk < prox->nr_gparts_in; kk++) {
+ struct gpart *gp = &s->gparts[offset_gparts + count_gparts + kk];
if (gp->id_or_neg_offset <= 0) {
struct part *p =
&s->parts[offset_gparts + count_parts - gp->id_or_neg_offset];
@@ -1152,8 +1207,8 @@ void engine_exchange_strays(struct engine *e, size_t offset_parts,
}
/* Advance the counters. */
- count_parts += p->nr_parts_in;
- count_gparts += p->nr_gparts_in;
+ count_parts += prox->nr_parts_in;
+ count_gparts += prox->nr_gparts_in;
}
}
@@ -1471,19 +1526,46 @@ void engine_link_gravity_tasks(struct engine *e) {
}
}
+#ifdef EXTRA_HYDRO_LOOP
+
/**
* @brief Creates the dependency network for the hydro tasks of a given cell.
*
* @param sched The #scheduler.
* @param density The density task to link.
+ * @param gradient The gradient task to link.
* @param force The force task to link.
* @param c The cell.
*/
static inline void engine_make_hydro_loops_dependencies(struct scheduler *sched,
struct task *density,
+ struct task *gradient,
struct task *force,
struct cell *c) {
+ /* init --> density loop --> ghost --> gradient loop --> extra_ghost */
+ /* extra_ghost --> force loop --> kick */
+ scheduler_addunlock(sched, c->super->init, density);
+ scheduler_addunlock(sched, density, c->super->ghost);
+ scheduler_addunlock(sched, c->super->ghost, gradient);
+ scheduler_addunlock(sched, gradient, c->super->extra_ghost);
+ scheduler_addunlock(sched, c->super->extra_ghost, force);
+ scheduler_addunlock(sched, force, c->super->kick);
+}
+
+#else
+/**
+ * @brief Creates the dependency network for the hydro tasks of a given cell.
+ *
+ * @param sched The #scheduler.
+ * @param density The density task to link.
+ * @param force The force task to link.
+ * @param c The cell.
+ */
+static inline void engine_make_hydro_loops_dependencies(struct scheduler *sched,
+ struct task *density,
+ struct task *force,
+ struct cell *c) {
/* init --> density loop --> ghost --> force loop --> kick */
scheduler_addunlock(sched, c->super->init, density);
scheduler_addunlock(sched, density, c->super->ghost);
@@ -1491,6 +1573,7 @@ static inline void engine_make_hydro_loops_dependencies(struct scheduler *sched,
scheduler_addunlock(sched, force, c->super->kick);
}
+#endif
/**
* @brief Duplicates the first hydro loop and construct all the
* dependencies for the hydro part
@@ -1518,6 +1601,24 @@ void engine_make_extra_hydroloop_tasks(struct engine *e) {
/* Self-interaction? */
if (t->type == task_type_self && t->subtype == task_subtype_density) {
+#ifdef EXTRA_HYDRO_LOOP
+ /* Start by constructing the task for the second and third hydro loop */
+ struct task *t2 = scheduler_addtask(
+ sched, task_type_self, task_subtype_gradient, 0, 0, t->ci, NULL, 0);
+ struct task *t3 = scheduler_addtask(
+ sched, task_type_self, task_subtype_force, 0, 0, t->ci, NULL, 0);
+
+ /* Add the link between the new loops and the cell */
+ engine_addlink(e, &t->ci->gradient, t2);
+ atomic_inc(&t->ci->nr_gradient);
+ engine_addlink(e, &t->ci->force, t3);
+ atomic_inc(&t->ci->nr_force);
+
+ /* Now, build all the dependencies for the hydro */
+ engine_make_hydro_loops_dependencies(sched, t, t2, t3, t->ci);
+
+#else
+
/* Start by constructing the task for the second hydro loop */
struct task *t2 = scheduler_addtask(
sched, task_type_self, task_subtype_force, 0, 0, t->ci, NULL, 0);
@@ -1528,11 +1629,40 @@ void engine_make_extra_hydroloop_tasks(struct engine *e) {
/* Now, build all the dependencies for the hydro */
engine_make_hydro_loops_dependencies(sched, t, t2, t->ci);
+#endif
}
/* Otherwise, pair interaction? */
else if (t->type == task_type_pair && t->subtype == task_subtype_density) {
+#ifdef EXTRA_HYDRO_LOOP
+ /* Start by constructing the task for the second and third hydro loop */
+ struct task *t2 = scheduler_addtask(
+ sched, task_type_pair, task_subtype_gradient, 0, 0, t->ci, t->cj, 0);
+ struct task *t3 = scheduler_addtask(
+ sched, task_type_pair, task_subtype_force, 0, 0, t->ci, t->cj, 0);
+
+ /* Add the link between the new loop and both cells */
+ engine_addlink(e, &t->ci->gradient, t2);
+ atomic_inc(&t->ci->nr_gradient);
+ engine_addlink(e, &t->cj->gradient, t2);
+ atomic_inc(&t->cj->nr_gradient);
+ engine_addlink(e, &t->ci->force, t3);
+ atomic_inc(&t->ci->nr_force);
+ engine_addlink(e, &t->cj->force, t3);
+ atomic_inc(&t->cj->nr_force);
+
+ /* Now, build all the dependencies for the hydro for the cells */
+ /* that are local and are not descendant of the same super-cells */
+ if (t->ci->nodeID == nodeID) {
+ engine_make_hydro_loops_dependencies(sched, t, t2, t3, t->ci);
+ }
+ if (t->cj->nodeID == nodeID && t->ci->super != t->cj->super) {
+ engine_make_hydro_loops_dependencies(sched, t, t2, t3, t->cj);
+ }
+
+#else
+
/* Start by constructing the task for the second hydro loop */
struct task *t2 = scheduler_addtask(
sched, task_type_pair, task_subtype_force, 0, 0, t->ci, t->cj, 0);
@@ -1551,12 +1681,38 @@ void engine_make_extra_hydroloop_tasks(struct engine *e) {
if (t->cj->nodeID == nodeID && t->ci->super != t->cj->super) {
engine_make_hydro_loops_dependencies(sched, t, t2, t->cj);
}
+
+#endif
+
}
/* Otherwise, sub-self interaction? */
else if (t->type == task_type_sub_self &&
t->subtype == task_subtype_density) {
+#ifdef EXTRA_HYDRO_LOOP
+
+ /* Start by constructing the task for the second and third hydro loop */
+ struct task *t2 =
+ scheduler_addtask(sched, task_type_sub_self, task_subtype_gradient,
+ t->flags, 0, t->ci, t->cj, 0);
+ struct task *t3 =
+ scheduler_addtask(sched, task_type_sub_self, task_subtype_force,
+ t->flags, 0, t->ci, t->cj, 0);
+
+ /* Add the link between the new loop and the cell */
+ engine_addlink(e, &t->ci->gradient, t2);
+ atomic_inc(&t->ci->nr_gradient);
+ engine_addlink(e, &t->ci->force, t3);
+ atomic_inc(&t->ci->nr_force);
+
+ /* Now, build all the dependencies for the hydro for the cells */
+ /* that are local and are not descendant of the same super-cells */
+ if (t->ci->nodeID == nodeID) {
+ engine_make_hydro_loops_dependencies(sched, t, t2, t3, t->ci);
+ }
+
+#else
/* Start by constructing the task for the second hydro loop */
struct task *t2 =
scheduler_addtask(sched, task_type_sub_self, task_subtype_force,
@@ -1571,12 +1727,43 @@ void engine_make_extra_hydroloop_tasks(struct engine *e) {
if (t->ci->nodeID == nodeID) {
engine_make_hydro_loops_dependencies(sched, t, t2, t->ci);
}
+#endif
}
/* Otherwise, sub-pair interaction? */
else if (t->type == task_type_sub_pair &&
t->subtype == task_subtype_density) {
+#ifdef EXTRA_HYDRO_LOOP
+
+ /* Start by constructing the task for the second and third hydro loop */
+ struct task *t2 =
+ scheduler_addtask(sched, task_type_sub_pair, task_subtype_gradient,
+ t->flags, 0, t->ci, t->cj, 0);
+ struct task *t3 =
+ scheduler_addtask(sched, task_type_sub_pair, task_subtype_force,
+ t->flags, 0, t->ci, t->cj, 0);
+
+ /* Add the link between the new loop and both cells */
+ engine_addlink(e, &t->ci->gradient, t2);
+ atomic_inc(&t->ci->nr_gradient);
+ engine_addlink(e, &t->cj->gradient, t2);
+ atomic_inc(&t->cj->nr_gradient);
+ engine_addlink(e, &t->ci->force, t3);
+ atomic_inc(&t->ci->nr_force);
+ engine_addlink(e, &t->cj->force, t3);
+ atomic_inc(&t->cj->nr_force);
+
+ /* Now, build all the dependencies for the hydro for the cells */
+ /* that are local and are not descendant of the same super-cells */
+ if (t->ci->nodeID == nodeID) {
+ engine_make_hydro_loops_dependencies(sched, t, t2, t3, t->ci);
+ }
+ if (t->cj->nodeID == nodeID && t->ci->super != t->cj->super) {
+ engine_make_hydro_loops_dependencies(sched, t, t2, t3, t->cj);
+ }
+
+#else
/* Start by constructing the task for the second hydro loop */
struct task *t2 =
scheduler_addtask(sched, task_type_sub_pair, task_subtype_force,
@@ -1596,8 +1783,8 @@ void engine_make_extra_hydroloop_tasks(struct engine *e) {
if (t->cj->nodeID == nodeID && t->ci->super != t->cj->super) {
engine_make_hydro_loops_dependencies(sched, t, t2, t->cj);
}
+#endif
}
-
/* External gravity tasks should depend on init and unlock the kick */
else if (t->type == task_type_grav_external) {
scheduler_addunlock(sched, t->ci->init, t);
@@ -1673,7 +1860,11 @@ void engine_maketasks(struct engine *e) {
is the number of cells (s->tot_cells) times the number of neighbours (27)
times the number of interaction types (2, density and force). */
if (e->links != NULL) free(e->links);
+#ifdef EXTRA_HYDRO_LOOP
+ e->size_links = s->tot_cells * 27 * 3;
+#else
e->size_links = s->tot_cells * 27 * 2;
+#endif
if ((e->links = malloc(sizeof(struct link) * e->size_links)) == NULL)
error("Failed to allocate cell-task links.");
e->nr_links = 0;
@@ -1719,13 +1910,13 @@ void engine_maketasks(struct engine *e) {
/* Loop through the proxy's incoming cells and add the
recv tasks. */
for (int k = 0; k < p->nr_cells_in; k++)
- engine_addtasks_recv(e, p->cells_in[k], NULL, NULL, NULL);
+ engine_addtasks_recv(e, p->cells_in[k], NULL, NULL, NULL, NULL);
/* Loop through the proxy's outgoing cells and add the
send tasks. */
for (int k = 0; k < p->nr_cells_out; k++)
engine_addtasks_send(e, p->cells_out[k], p->cells_in[0], NULL, NULL,
- NULL);
+ NULL, NULL);
}
}
#endif
@@ -2026,8 +2217,8 @@ void engine_print_task_counts(struct engine *e) {
printf(" %s=%i", taskID_names[k], counts[k]);
printf(" skipped=%i ]\n", counts[task_type_count]);
fflush(stdout);
- message("nr_parts = %zi.", e->s->nr_parts);
- message("nr_gparts = %zi.", e->s->nr_gparts);
+ message("nr_parts = %zu.", e->s->nr_parts);
+ message("nr_gparts = %zu.", e->s->nr_gparts);
}
/**
@@ -2508,11 +2699,11 @@ void engine_step(struct engine *e) {
if (e->nodeID == 0) {
/* Print some information to the screen */
- printf(" %6d %14e %14e %10zd %10zd %21.3f\n", e->step, e->time,
+ printf(" %6d %14e %14e %10zu %10zu %21.3f\n", e->step, e->time,
e->timeStep, e->updates, e->g_updates, e->wallclock_time);
fflush(stdout);
- fprintf(e->file_timesteps, " %6d %14e %14e %10zd %10zd %21.3f\n", e->step,
+ fprintf(e->file_timesteps, " %6d %14e %14e %10zu %10zu %21.3f\n", e->step,
e->time, e->timeStep, e->updates, e->g_updates, e->wallclock_time);
fflush(e->file_timesteps);
}
@@ -2554,6 +2745,11 @@ void engine_step(struct engine *e) {
submask |= 1 << task_subtype_density;
submask |= 1 << task_subtype_force;
+
+#ifdef EXTRA_HYDRO_LOOP
+ mask |= 1 << task_type_extra_ghost;
+ submask |= 1 << task_subtype_gradient;
+#endif
}
/* Add the tasks corresponding to self-gravity to the masks */
@@ -2727,7 +2923,7 @@ void engine_split(struct engine *e, struct partition *initial_partition) {
/* Re-allocate the local parts. */
if (e->verbose)
- message("Re-allocating parts array from %zi to %zi.", s->size_parts,
+ message("Re-allocating parts array from %zu to %zu.", s->size_parts,
(size_t)(s->nr_parts * 1.2));
s->size_parts = s->nr_parts * 1.2;
struct part *parts_new = NULL;
@@ -2749,7 +2945,7 @@ void engine_split(struct engine *e, struct partition *initial_partition) {
/* Re-allocate the local gparts. */
if (e->verbose)
- message("Re-allocating gparts array from %zi to %zi.", s->size_gparts,
+ message("Re-allocating gparts array from %zu to %zu.", s->size_gparts,
(size_t)(s->nr_gparts * 1.2));
s->size_gparts = s->nr_gparts * 1.2;
struct gpart *gparts_new = NULL;
diff --git a/src/gravity/Default/gravity.h b/src/gravity/Default/gravity.h
index da6b9aa01d7d382b1d6aa85f732f1fae1169ac34..d4249c46a3150a357aaecfb02f9251901d97a157 100644
--- a/src/gravity/Default/gravity.h
+++ b/src/gravity/Default/gravity.h
@@ -71,7 +71,7 @@ gravity_compute_timestep_self(const struct phys_const* const phys_const,
const float ac = (ac2 > 0.f) ? sqrtf(ac2) : FLT_MIN;
- const float dt = sqrt(2.f * const_gravity_eta * gp->epsilon / ac);
+ const float dt = sqrtf(2.f * const_gravity_eta * gp->epsilon / ac);
return dt;
}
@@ -115,10 +115,10 @@ __attribute__((always_inline)) INLINE static void gravity_init_gpart(
* Multiplies the forces and accelerations by the appropiate constants
*
* @param gp The particle to act upon
- * @param const_G Newton's constant
+ * @param const_G Newton's constant in internal units
*/
__attribute__((always_inline)) INLINE static void gravity_end_force(
- struct gpart* gp, double const_G) {
+ struct gpart* gp, float const_G) {
/* Let's get physical... */
gp->a_grav[0] *= const_G;
diff --git a/src/hydro.h b/src/hydro.h
index b2ae9d57c399ecea818e9f3dc7db238e01487a9a..4a2b0bd029d494e2091b9081d22b7949cec5648c 100644
--- a/src/hydro.h
+++ b/src/hydro.h
@@ -38,6 +38,10 @@
#include "./hydro/Default/hydro.h"
#include "./hydro/Default/hydro_iact.h"
#define SPH_IMPLEMENTATION "Default version of SPH"
+#elif defined(GIZMO_SPH)
+#include "./hydro/Gizmo/hydro.h"
+#include "./hydro/Gizmo/hydro_iact.h"
+#define SPH_IMPLEMENTATION "GIZMO (Hopkins 2015)"
#else
#error "Invalid choice of SPH variant"
#endif
diff --git a/src/hydro/Default/hydro.h b/src/hydro/Default/hydro.h
index 021599cd2daf61ff35e5f29e3f13b2ad61c8947a..f61bff55821809fe1f5da27c95d75afbecbc04cc 100644
--- a/src/hydro/Default/hydro.h
+++ b/src/hydro/Default/hydro.h
@@ -73,6 +73,28 @@ __attribute__((always_inline)) INLINE static float hydro_get_soundspeed(
return p->force.soundspeed;
}
+/**
+ * @brief Returns the density of a particle
+ *
+ * @param p The particle of interest
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_density(
+ const struct part *restrict p) {
+
+ return p->rho;
+}
+
+/**
+ * @brief Returns the mass of a particle
+ *
+ * @param p The particle of interest
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_mass(
+ const struct part *restrict p) {
+
+ return p->mass;
+}
+
/**
* @brief Modifies the thermal state of a particle to the imposed internal
* energy
@@ -288,16 +310,31 @@ __attribute__((always_inline)) INLINE static void hydro_reset_acceleration(
*
* @param p The particle
* @param xp The extended data of the particle
+ * @param dt The drift time-step.
* @param t0 The time at the start of the drift
* @param t1 The time at the end of the drift
* @param timeBase The minimal time-step size
*/
__attribute__((always_inline)) INLINE static void hydro_predict_extra(
- struct part *restrict p, struct xpart *restrict xp, int t0, int t1,
- double timeBase) {
+ struct part *restrict p, struct xpart *restrict xp, float dt, int t0,
+ int t1, double timeBase) {
float u, w;
- const float dt = (t1 - t0) * timeBase;
+ const float h_inv = 1.f / p->h;
+
+ /* Predict smoothing length */
+ const float w1 = p->force.h_dt * h_inv * dt;
+ if (fabsf(w1) < 0.2f)
+ p->h *= approx_expf(w1); /* 4th order expansion of exp(w) */
+ else
+ p->h *= expf(w1);
+
+ /* Predict density */
+ const float w2 = -hydro_dimension * w1;
+ if (fabsf(w2) < 0.2f)
+ p->rho *= approx_expf(w2); /* 4th order expansion of exp(w) */
+ else
+ p->rho *= expf(w2);
/* Predict internal energy */
w = p->force.u_dt / p->u * dt;
diff --git a/src/hydro/Default/hydro_part.h b/src/hydro/Default/hydro_part.h
index a2f4453dc69ed06ca4f315b6be29844c177d0435..f42c3dc886ae1ab8f472ffdf5ff508f6735d1bb1 100644
--- a/src/hydro/Default/hydro_part.h
+++ b/src/hydro/Default/hydro_part.h
@@ -28,6 +28,8 @@ struct xpart {
/* Velocity at the last full step. */
float v_full[3];
+ float u_full;
+
/* Old density. */
float omega;
diff --git a/src/hydro/Gadget2/hydro.h b/src/hydro/Gadget2/hydro.h
index e9d626cb8c147c0cf4fa8d27f8bab31d2471beae..09a8f50d5b2e9abd43c3b9bd43a12fad8a347258 100644
--- a/src/hydro/Gadget2/hydro.h
+++ b/src/hydro/Gadget2/hydro.h
@@ -20,6 +20,7 @@
#define SWIFT_GADGET2_HYDRO_H
#include "adiabatic_index.h"
+#include "approx_math.h"
#include "dimension.h"
#include "equation_of_state.h"
#include "hydro_properties.h"
@@ -77,6 +78,28 @@ __attribute__((always_inline)) INLINE static float hydro_get_soundspeed(
return p->force.soundspeed;
}
+/**
+ * @brief Returns the density of a particle
+ *
+ * @param p The particle of interest
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_density(
+ const struct part *restrict p) {
+
+ return p->rho;
+}
+
+/**
+ * @brief Returns the mass of a particle
+ *
+ * @param p The particle of interest
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_mass(
+ const struct part *restrict p) {
+
+ return p->mass;
+}
+
/**
* @brief Modifies the thermal state of a particle to the imposed internal
* energy
@@ -285,13 +308,30 @@ __attribute__((always_inline)) INLINE static void hydro_reset_acceleration(
*
* @param p The particle
* @param xp The extended data of the particle
+ * @param dt The drift time-step.
* @param t0 The time at the start of the drift
* @param t1 The time at the end of the drift
* @param timeBase The minimal time-step size
*/
__attribute__((always_inline)) INLINE static void hydro_predict_extra(
- struct part *restrict p, const struct xpart *restrict xp, int t0, int t1,
- double timeBase) {
+ struct part *restrict p, const struct xpart *restrict xp, float dt, int t0,
+ int t1, double timeBase) {
+
+ const float h_inv = 1.f / p->h;
+
+ /* Predict smoothing length */
+ const float w1 = p->force.h_dt * h_inv * dt;
+ if (fabsf(w1) < 0.2f)
+ p->h *= approx_expf(w1); /* 4th order expansion of exp(w) */
+ else
+ p->h *= expf(w1);
+
+ /* Predict density */
+ const float w2 = -hydro_dimension * w1;
+ if (fabsf(w2) < 0.2f)
+ p->rho *= approx_expf(w2); /* 4th order expansion of exp(w) */
+ else
+ p->rho *= expf(w2);
/* Drift the pressure */
const float dt_entr = (t1 - (p->ti_begin + p->ti_end) / 2) * timeBase;
diff --git a/src/hydro/Gizmo/hydro.h b/src/hydro/Gizmo/hydro.h
index f69dc3f1798f014e895c4a63760805b1739cec94..e24a44529dc1907c9ceadeedfbfc9e49c308bcda 100644
--- a/src/hydro/Gizmo/hydro.h
+++ b/src/hydro/Gizmo/hydro.h
@@ -17,17 +17,25 @@
*
******************************************************************************/
+#include <float.h>
+#include "adiabatic_index.h"
+#include "approx_math.h"
+#include "hydro_gradients.h"
+
/**
* @brief Computes the hydro time-step of a given particle
*
- * @param p Pointer to the particle data
- * @param xp Pointer to the extended particle data
- *
+ * @param p Pointer to the particle data.
+ * @param xp Pointer to the extended particle data.
+ * @param hydro_properties Pointer to the hydro parameters.
*/
__attribute__((always_inline)) INLINE static float hydro_compute_timestep(
- struct part* p, struct xpart* xp) {
+ const struct part* restrict p, const struct xpart* restrict xp,
+ const struct hydro_props* restrict hydro_properties) {
- return const_cfl * p->h / fabs(p->timestepvars.vmax);
+ const float CFL_condition = hydro_properties->CFL_condition;
+
+ return CFL_condition * p->h / fabsf(p->timestepvars.vmax);
}
/**
@@ -36,39 +44,37 @@ __attribute__((always_inline)) INLINE static float hydro_compute_timestep(
* This function is called only once just after the ICs have been
* read in to do some conversions.
*
+ * In this case, we copy the particle velocities into the corresponding
+ * primitive variable field. We do this because the particle velocities in GIZMO
+ * can be independent of the actual fluid velocity. The latter is stored as a
+ * primitive variable and integrated using the linear momentum, a conserved
+ * variable.
+ *
* @param p The particle to act upon
* @param xp The extended particle data to act upon
*/
__attribute__((always_inline)) INLINE static void hydro_first_init_part(
- struct part* p, struct xpart* xp) {}
+ struct part* p, struct xpart* xp) {
+
+ xp->v_full[0] = p->v[0];
+ xp->v_full[1] = p->v[1];
+ xp->v_full[2] = p->v[2];
+
+ p->primitives.v[0] = p->v[0];
+ p->primitives.v[1] = p->v[1];
+ p->primitives.v[2] = p->v[2];
+}
/**
* @brief Prepares a particle for the volume calculation.
*
+ * Simply makes sure all necessary variables are initialized to zero.
+ *
* @param p The particle to act upon
*/
__attribute__((always_inline)) INLINE static void hydro_init_part(
struct part* p) {
-#ifdef SPH_GRADIENTS
- /* use the old volumes to estimate new primitive variables to be used for the
- gradient calculation */
- if (p->conserved.mass) {
- p->primitives.rho = p->conserved.mass / p->geometry.volume;
- p->primitives.v[0] = p->conserved.momentum[0] / p->conserved.mass;
- p->primitives.v[1] = p->conserved.momentum[1] / p->conserved.mass;
- p->primitives.v[2] = p->conserved.momentum[2] / p->conserved.mass;
- p->primitives.P =
- (const_hydro_gamma - 1.) *
- (p->conserved.energy -
- 0.5 * (p->conserved.momentum[0] * p->conserved.momentum[0] +
- p->conserved.momentum[1] * p->conserved.momentum[1] +
- p->conserved.momentum[2] * p->conserved.momentum[2]) /
- p->conserved.mass) /
- p->geometry.volume;
- }
-#endif
-
p->density.wcount = 0.0f;
p->density.wcount_dh = 0.0f;
p->geometry.volume = 0.0f;
@@ -81,540 +87,417 @@ __attribute__((always_inline)) INLINE static void hydro_init_part(
p->geometry.matrix_E[2][0] = 0.0f;
p->geometry.matrix_E[2][1] = 0.0f;
p->geometry.matrix_E[2][2] = 0.0f;
-
-#ifdef SPH_GRADIENTS
- p->primitives.gradients.rho[0] = 0.0f;
- p->primitives.gradients.rho[1] = 0.0f;
- p->primitives.gradients.rho[2] = 0.0f;
-
- p->primitives.gradients.v[0][0] = 0.0f;
- p->primitives.gradients.v[0][1] = 0.0f;
- p->primitives.gradients.v[0][2] = 0.0f;
-
- p->primitives.gradients.v[1][0] = 0.0f;
- p->primitives.gradients.v[1][1] = 0.0f;
- p->primitives.gradients.v[1][2] = 0.0f;
-
- p->primitives.gradients.v[2][0] = 0.0f;
- p->primitives.gradients.v[2][1] = 0.0f;
- p->primitives.gradients.v[2][2] = 0.0f;
-
- p->primitives.gradients.P[0] = 0.0f;
- p->primitives.gradients.P[1] = 0.0f;
- p->primitives.gradients.P[2] = 0.0f;
-
- p->primitives.limiter.rho[0] = FLT_MAX;
- p->primitives.limiter.rho[1] = -FLT_MAX;
- p->primitives.limiter.v[0][0] = FLT_MAX;
- p->primitives.limiter.v[0][1] = -FLT_MAX;
- p->primitives.limiter.v[1][0] = FLT_MAX;
- p->primitives.limiter.v[1][1] = -FLT_MAX;
- p->primitives.limiter.v[2][0] = FLT_MAX;
- p->primitives.limiter.v[2][1] = -FLT_MAX;
- p->primitives.limiter.P[0] = FLT_MAX;
- p->primitives.limiter.P[1] = -FLT_MAX;
-
- p->primitives.limiter.maxr = -FLT_MAX;
-#endif
}
/**
- * @brief Finishes the density calculation.
+ * @brief Finishes the volume calculation.
*
* Multiplies the density and number of neighbours by the appropiate constants
- * and add the self-contribution term.
+ * and adds the self-contribution term. Calculates the volume and uses it to
+ * update the primitive variables (based on the conserved variables). The latter
+ * should only be done for active particles. This is okay, since this method is
+ * only called for active particles.
*
- * @param p The particle to act upon
- */
-__attribute__((always_inline)) INLINE static void hydro_end_volume(
- struct part* p) {
-
- /* Some smoothing length multiples. */
- const float h = p->h;
- const float ih = 1.0f / h;
-
- /* Final operation on the density. */
- p->density.wcount =
- (p->density.wcount + kernel_root) * (4.0f / 3.0 * M_PI * kernel_gamma3);
- p->density.wcount_dh =
- p->density.wcount_dh * ih * (4.0f / 3.0 * M_PI * kernel_gamma3);
-}
-
-/**
- * @brief Prepare a particle for the force calculation.
+ * Multiplies the components of the matrix E with the appropriate constants and
+ * inverts it. Initializes the variables used during the gradient loop. This
+ * cannot be done in hydro_prepare_force, since that method is called for all
+ * particles, and not just the active ones. If we would initialize the
+ * variables there, gradients for passive particles would be zero, while we
+ * actually use the old gradients in the flux calculation between active and
+ * passive particles.
*
- * Computes viscosity term, conduction term and smoothing length gradient terms.
- *
- * @param p The particle to act upon
- * @param xp The extended particle data to act upon
+ * @param p The particle to act upon.
+ * @param The current physical time.
*/
-__attribute__((always_inline)) INLINE static void hydro_prepare_gradient(
- struct part* p, struct xpart* xp) {
+__attribute__((always_inline)) INLINE static void hydro_end_density(
+ struct part* restrict p, float time) {
/* Some smoothing length multiples. */
const float h = p->h;
const float ih = 1.0f / h;
- const float ih2 = ih * ih;
- float detE, volume;
- float E[3][3];
- GFLOAT m, momentum[3], energy;
+ /* Final operation on the density. */
+ p->density.wcount += kernel_root;
+ p->density.wcount *= kernel_norm;
-#ifndef THERMAL_ENERGY
- GFLOAT momentum2;
-#endif
+ p->density.wcount_dh *= ih * kernel_gamma * kernel_norm;
-#if defined(SPH_GRADIENTS) && defined(SLOPE_LIMITER)
- GFLOAT gradrho[3], gradv[3][3], gradP[3];
- GFLOAT gradtrue, gradmax, gradmin, alpha;
-#endif
+ const float ihdim = pow_dimension(ih);
/* Final operation on the geometry. */
/* we multiply with the smoothing kernel normalization ih3 and calculate the
* volume */
- volume = ih * ih2 * (p->geometry.volume + kernel_root);
- p->geometry.volume = volume = 1. / volume;
- /* we multiply with the smoothing kernel normalization */
- p->geometry.matrix_E[0][0] = E[0][0] = ih * ih2 * p->geometry.matrix_E[0][0];
- p->geometry.matrix_E[0][1] = E[0][1] = ih * ih2 * p->geometry.matrix_E[0][1];
- p->geometry.matrix_E[0][2] = E[0][2] = ih * ih2 * p->geometry.matrix_E[0][2];
- p->geometry.matrix_E[1][0] = E[1][0] = ih * ih2 * p->geometry.matrix_E[1][0];
- p->geometry.matrix_E[1][1] = E[1][1] = ih * ih2 * p->geometry.matrix_E[1][1];
- p->geometry.matrix_E[1][2] = E[1][2] = ih * ih2 * p->geometry.matrix_E[1][2];
- p->geometry.matrix_E[2][0] = E[2][0] = ih * ih2 * p->geometry.matrix_E[2][0];
- p->geometry.matrix_E[2][1] = E[2][1] = ih * ih2 * p->geometry.matrix_E[2][1];
- p->geometry.matrix_E[2][2] = E[2][2] = ih * ih2 * p->geometry.matrix_E[2][2];
-
- /* invert the E-matrix */
- /* code shamelessly stolen from the public version of GIZMO */
- /* But since we should never invert a matrix, this code has to be replaced */
- detE = E[0][0] * E[1][1] * E[2][2] + E[0][1] * E[1][2] * E[2][0] +
- E[0][2] * E[1][0] * E[2][1] - E[0][2] * E[1][1] * E[2][0] -
- E[0][1] * E[1][0] * E[2][2] - E[0][0] * E[1][2] * E[2][1];
- /* check for zero determinant */
- if ((detE != 0) && !isnan(detE)) {
- p->geometry.matrix_E[0][0] = (E[1][1] * E[2][2] - E[1][2] * E[2][1]) / detE;
- p->geometry.matrix_E[0][1] = (E[0][2] * E[2][1] - E[0][1] * E[2][2]) / detE;
- p->geometry.matrix_E[0][2] = (E[0][1] * E[1][2] - E[0][2] * E[1][1]) / detE;
- p->geometry.matrix_E[1][0] = (E[1][2] * E[2][0] - E[1][0] * E[2][2]) / detE;
- p->geometry.matrix_E[1][1] = (E[0][0] * E[2][2] - E[0][2] * E[2][0]) / detE;
- p->geometry.matrix_E[1][2] = (E[0][2] * E[1][0] - E[0][0] * E[1][2]) / detE;
- p->geometry.matrix_E[2][0] = (E[1][0] * E[2][1] - E[1][1] * E[2][0]) / detE;
- p->geometry.matrix_E[2][1] = (E[0][1] * E[2][0] - E[0][0] * E[2][1]) / detE;
- p->geometry.matrix_E[2][2] = (E[0][0] * E[1][1] - E[0][1] * E[1][0]) / detE;
- } else {
- /* if the E-matrix is not well behaved, we cannot use it */
- p->geometry.matrix_E[0][0] = 0.0f;
- p->geometry.matrix_E[0][1] = 0.0f;
- p->geometry.matrix_E[0][2] = 0.0f;
- p->geometry.matrix_E[1][0] = 0.0f;
- p->geometry.matrix_E[1][1] = 0.0f;
- p->geometry.matrix_E[1][2] = 0.0f;
- p->geometry.matrix_E[2][0] = 0.0f;
- p->geometry.matrix_E[2][1] = 0.0f;
- p->geometry.matrix_E[2][2] = 0.0f;
- }
-
-#ifdef SPH_GRADIENTS
- /* finalize gradients by multiplying with volume */
- p->primitives.gradients.rho[0] *= ih2 * ih2 * volume;
- p->primitives.gradients.rho[1] *= ih2 * ih2 * volume;
- p->primitives.gradients.rho[2] *= ih2 * ih2 * volume;
-
- p->primitives.gradients.v[0][0] *= ih2 * ih2 * volume;
- p->primitives.gradients.v[0][1] *= ih2 * ih2 * volume;
- p->primitives.gradients.v[0][2] *= ih2 * ih2 * volume;
-
- p->primitives.gradients.v[1][0] *= ih2 * ih2 * volume;
- p->primitives.gradients.v[1][1] *= ih2 * ih2 * volume;
- p->primitives.gradients.v[1][2] *= ih2 * ih2 * volume;
-
- p->primitives.gradients.v[2][0] *= ih2 * ih2 * volume;
- p->primitives.gradients.v[2][1] *= ih2 * ih2 * volume;
- p->primitives.gradients.v[2][2] *= ih2 * ih2 * volume;
-
- p->primitives.gradients.P[0] *= ih2 * ih2 * volume;
- p->primitives.gradients.P[1] *= ih2 * ih2 * volume;
- p->primitives.gradients.P[2] *= ih2 * ih2 * volume;
-
-/* slope limiter */
-#ifdef SLOPE_LIMITER
- gradrho[0] = p->primitives.gradients.rho[0];
- gradrho[1] = p->primitives.gradients.rho[1];
- gradrho[2] = p->primitives.gradients.rho[2];
-
- gradv[0][0] = p->primitives.gradients.v[0][0];
- gradv[0][1] = p->primitives.gradients.v[0][1];
- gradv[0][2] = p->primitives.gradients.v[0][2];
-
- gradv[1][0] = p->primitives.gradients.v[1][0];
- gradv[1][1] = p->primitives.gradients.v[1][1];
- gradv[1][2] = p->primitives.gradients.v[1][2];
-
- gradv[2][0] = p->primitives.gradients.v[2][0];
- gradv[2][1] = p->primitives.gradients.v[2][1];
- gradv[2][2] = p->primitives.gradients.v[2][2];
-
- gradP[0] = p->primitives.gradients.P[0];
- gradP[1] = p->primitives.gradients.P[1];
- gradP[2] = p->primitives.gradients.P[2];
-
- gradtrue = sqrtf(gradrho[0] * gradrho[0] + gradrho[1] * gradrho[1] +
- gradrho[2] * gradrho[2]);
- /* gradtrue might be zero. In this case, there is no gradient and we don't
- need to slope limit anything... */
- if (gradtrue) {
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.rho[1] - p->primitives.rho;
- gradmin = p->primitives.rho - p->primitives.limiter.rho[0];
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.rho[0] *= alpha;
- p->primitives.gradients.rho[1] *= alpha;
- p->primitives.gradients.rho[2] *= alpha;
- }
+ const float volume = 1.f / (ihdim * (p->geometry.volume + kernel_root));
+ p->geometry.volume = volume;
- gradtrue = sqrtf(gradv[0][0] * gradv[0][0] + gradv[0][1] * gradv[0][1] +
- gradv[0][2] * gradv[0][2]);
- if (gradtrue) {
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.v[0][1] - p->primitives.v[0];
- gradmin = p->primitives.v[0] - p->primitives.limiter.v[0][0];
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.v[0][0] *= alpha;
- p->primitives.gradients.v[0][1] *= alpha;
- p->primitives.gradients.v[0][2] *= alpha;
- }
-
- gradtrue = sqrtf(gradv[1][0] * gradv[1][0] + gradv[1][1] * gradv[1][1] +
- gradv[1][2] * gradv[1][2]);
- if (gradtrue) {
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.v[1][1] - p->primitives.v[1];
- gradmin = p->primitives.v[1] - p->primitives.limiter.v[1][0];
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.v[1][0] *= alpha;
- p->primitives.gradients.v[1][1] *= alpha;
- p->primitives.gradients.v[1][2] *= alpha;
- }
+ /* we multiply with the smoothing kernel normalization */
+ p->geometry.matrix_E[0][0] = ihdim * p->geometry.matrix_E[0][0];
+ p->geometry.matrix_E[0][1] = ihdim * p->geometry.matrix_E[0][1];
+ p->geometry.matrix_E[0][2] = ihdim * p->geometry.matrix_E[0][2];
+ p->geometry.matrix_E[1][0] = ihdim * p->geometry.matrix_E[1][0];
+ p->geometry.matrix_E[1][1] = ihdim * p->geometry.matrix_E[1][1];
+ p->geometry.matrix_E[1][2] = ihdim * p->geometry.matrix_E[1][2];
+ p->geometry.matrix_E[2][0] = ihdim * p->geometry.matrix_E[2][0];
+ p->geometry.matrix_E[2][1] = ihdim * p->geometry.matrix_E[2][1];
+ p->geometry.matrix_E[2][2] = ihdim * p->geometry.matrix_E[2][2];
- gradtrue = sqrtf(gradv[2][0] * gradv[2][0] + gradv[2][1] * gradv[2][1] +
- gradv[2][2] * gradv[2][2]);
- if (gradtrue) {
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.v[2][1] - p->primitives.v[2];
- gradmin = p->primitives.v[2] - p->primitives.limiter.v[2][0];
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.v[2][0] *= alpha;
- p->primitives.gradients.v[2][1] *= alpha;
- p->primitives.gradients.v[2][2] *= alpha;
- }
+ invert_dimension_by_dimension_matrix(p->geometry.matrix_E);
- gradtrue =
- sqrtf(gradP[0] * gradP[0] + gradP[1] * gradP[1] + gradP[2] * gradP[2]);
- if (gradtrue) {
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.P[1] - p->primitives.P;
- gradmin = p->primitives.P - p->primitives.limiter.P[0];
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.P[0] *= alpha;
- p->primitives.gradients.P[1] *= alpha;
- p->primitives.gradients.P[2] *= alpha;
- }
-#endif // SLOPE_LIMITER
-#else // SPH_GRADIENTS
- p->primitives.gradients.rho[0] = 0.0f;
- p->primitives.gradients.rho[1] = 0.0f;
- p->primitives.gradients.rho[2] = 0.0f;
-
- p->primitives.gradients.v[0][0] = 0.0f;
- p->primitives.gradients.v[0][1] = 0.0f;
- p->primitives.gradients.v[0][2] = 0.0f;
-
- p->primitives.gradients.v[1][0] = 0.0f;
- p->primitives.gradients.v[1][1] = 0.0f;
- p->primitives.gradients.v[1][2] = 0.0f;
-
- p->primitives.gradients.v[2][0] = 0.0f;
- p->primitives.gradients.v[2][1] = 0.0f;
- p->primitives.gradients.v[2][2] = 0.0f;
-
- p->primitives.gradients.P[0] = 0.0f;
- p->primitives.gradients.P[1] = 0.0f;
- p->primitives.gradients.P[2] = 0.0f;
-
- p->primitives.limiter.rho[0] = FLT_MAX;
- p->primitives.limiter.rho[1] = -FLT_MAX;
- p->primitives.limiter.v[0][0] = FLT_MAX;
- p->primitives.limiter.v[0][1] = -FLT_MAX;
- p->primitives.limiter.v[1][0] = FLT_MAX;
- p->primitives.limiter.v[1][1] = -FLT_MAX;
- p->primitives.limiter.v[2][0] = FLT_MAX;
- p->primitives.limiter.v[2][1] = -FLT_MAX;
- p->primitives.limiter.P[0] = FLT_MAX;
- p->primitives.limiter.P[1] = -FLT_MAX;
-
- p->primitives.limiter.maxr = -FLT_MAX;
-#endif // SPH_GRADIENTS
+ hydro_gradients_init(p);
/* compute primitive variables */
/* eqns (3)-(5) */
- m = p->conserved.mass;
- if (m) {
+ const float m = p->conserved.mass;
+ if (m > 0.f) {
+ float momentum[3];
momentum[0] = p->conserved.momentum[0];
momentum[1] = p->conserved.momentum[1];
momentum[2] = p->conserved.momentum[2];
-#ifndef THERMAL_ENERGY
- momentum2 = (momentum[0] * momentum[0] + momentum[1] * momentum[1] +
- momentum[2] * momentum[2]);
-#endif
- energy = p->conserved.energy;
p->primitives.rho = m / volume;
p->primitives.v[0] = momentum[0] / m;
p->primitives.v[1] = momentum[1] / m;
p->primitives.v[2] = momentum[2] / m;
-#ifndef THERMAL_ENERGY
- p->primitives.P =
- (const_hydro_gamma - 1.) * (energy - 0.5 * momentum2 / m) / volume;
-#else
- p->primitives.P = (const_hydro_gamma - 1.) * energy / volume;
-#endif
+ const float energy = p->conserved.energy;
+ p->primitives.P = hydro_gamma_minus_one * energy / volume;
}
}
/**
- * @brief Finishes the gradient calculation.
+ * @brief Prepare a particle for the gradient calculation.
*
- * @param p The particle to act upon
+ * The name of this method is confusing, as this method is really called after
+ * the density loop and before the gradient loop.
+ *
+ * We use it to set the physical timestep for the particle and to copy the
+ * actual velocities, which we need to boost our interfaces during the flux
+ * calculation. We also initialize the variables used for the time step
+ * calculation.
+ *
+ * @param p The particle to act upon.
+ * @param xp The extended particle data to act upon.
+ * @param ti_current Current integer time.
+ * @param timeBase Conversion factor between integer time and physical time.
*/
-__attribute__((always_inline)) INLINE static void hydro_end_gradient(
- struct part* p) {
-
-#ifndef SPH_GRADIENTS
- float h, ih, ih2, ih3;
-#ifdef SLOPE_LIMITER
- GFLOAT gradrho[3], gradv[3][3], gradP[3];
- GFLOAT gradtrue, gradmax, gradmin, alpha;
-#endif
-
- /* add kernel normalization to gradients */
- h = p->h;
- ih = 1.0f / h;
- ih2 = ih * ih;
- ih3 = ih * ih2;
-
- p->primitives.gradients.rho[0] *= ih3;
- p->primitives.gradients.rho[1] *= ih3;
- p->primitives.gradients.rho[2] *= ih3;
-
- p->primitives.gradients.v[0][0] *= ih3;
- p->primitives.gradients.v[0][1] *= ih3;
- p->primitives.gradients.v[0][2] *= ih3;
- p->primitives.gradients.v[1][0] *= ih3;
- p->primitives.gradients.v[1][1] *= ih3;
- p->primitives.gradients.v[1][2] *= ih3;
- p->primitives.gradients.v[2][0] *= ih3;
- p->primitives.gradients.v[2][1] *= ih3;
- p->primitives.gradients.v[2][2] *= ih3;
-
- p->primitives.gradients.P[0] *= ih3;
- p->primitives.gradients.P[1] *= ih3;
- p->primitives.gradients.P[2] *= ih3;
-
-/* slope limiter */
-#ifdef SLOPE_LIMITER
- gradrho[0] = p->primitives.gradients.rho[0];
- gradrho[1] = p->primitives.gradients.rho[1];
- gradrho[2] = p->primitives.gradients.rho[2];
-
- gradv[0][0] = p->primitives.gradients.v[0][0];
- gradv[0][1] = p->primitives.gradients.v[0][1];
- gradv[0][2] = p->primitives.gradients.v[0][2];
-
- gradv[1][0] = p->primitives.gradients.v[1][0];
- gradv[1][1] = p->primitives.gradients.v[1][1];
- gradv[1][2] = p->primitives.gradients.v[1][2];
-
- gradv[2][0] = p->primitives.gradients.v[2][0];
- gradv[2][1] = p->primitives.gradients.v[2][1];
- gradv[2][2] = p->primitives.gradients.v[2][2];
-
- gradP[0] = p->primitives.gradients.P[0];
- gradP[1] = p->primitives.gradients.P[1];
- gradP[2] = p->primitives.gradients.P[2];
-
- gradtrue = gradrho[0] * gradrho[0] + gradrho[1] * gradrho[1] +
- gradrho[2] * gradrho[2];
- /* gradtrue might be zero. In this case, there is no gradient and we don't
- need to slope limit anything... */
- if (gradtrue) {
- gradtrue = sqrtf(gradtrue);
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.rho[1] - p->primitives.rho;
- gradmin = p->primitives.rho - p->primitives.limiter.rho[0];
- /* gradmin and gradmax might be negative if the value of the current
- particle is larger/smaller than all neighbouring values */
- gradmax = fabs(gradmax);
- gradmin = fabs(gradmin);
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.rho[0] *= alpha;
- p->primitives.gradients.rho[1] *= alpha;
- p->primitives.gradients.rho[2] *= alpha;
- }
-
- gradtrue = gradv[0][0] * gradv[0][0] + gradv[0][1] * gradv[0][1] +
- gradv[0][2] * gradv[0][2];
- if (gradtrue) {
- gradtrue = sqrtf(gradtrue);
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.v[0][1] - p->primitives.v[0];
- gradmin = p->primitives.v[0] - p->primitives.limiter.v[0][0];
- gradmax = fabs(gradmax);
- gradmin = fabs(gradmin);
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.v[0][0] *= alpha;
- p->primitives.gradients.v[0][1] *= alpha;
- p->primitives.gradients.v[0][2] *= alpha;
- }
-
- gradtrue = gradv[1][0] * gradv[1][0] + gradv[1][1] * gradv[1][1] +
- gradv[1][2] * gradv[1][2];
- if (gradtrue) {
- gradtrue = sqrtf(gradtrue);
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.v[1][1] - p->primitives.v[1];
- gradmin = p->primitives.v[1] - p->primitives.limiter.v[1][0];
- gradmax = fabs(gradmax);
- gradmin = fabs(gradmin);
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.v[1][0] *= alpha;
- p->primitives.gradients.v[1][1] *= alpha;
- p->primitives.gradients.v[1][2] *= alpha;
- }
+__attribute__((always_inline)) INLINE static void hydro_prepare_force(
+ struct part* restrict p, struct xpart* restrict xp, int ti_current,
+ double timeBase) {
- gradtrue = gradv[2][0] * gradv[2][0] + gradv[2][1] * gradv[2][1] +
- gradv[2][2] * gradv[2][2];
- if (gradtrue) {
- gradtrue = sqrtf(gradtrue);
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.v[2][1] - p->primitives.v[2];
- gradmin = p->primitives.v[2] - p->primitives.limiter.v[2][0];
- gradmax = fabs(gradmax);
- gradmin = fabs(gradmin);
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.v[2][0] *= alpha;
- p->primitives.gradients.v[2][1] *= alpha;
- p->primitives.gradients.v[2][2] *= alpha;
- }
+ /* Set the physical time step */
+ p->force.dt = (p->ti_end - p->ti_begin) * timeBase;
- gradtrue = gradP[0] * gradP[0] + gradP[1] * gradP[1] + gradP[2] * gradP[2];
- if (gradtrue) {
- gradtrue = sqrtf(gradtrue);
- gradtrue *= p->primitives.limiter.maxr;
- gradmax = p->primitives.limiter.P[1] - p->primitives.P;
- gradmin = p->primitives.P - p->primitives.limiter.P[0];
- gradmax = fabs(gradmax);
- gradmin = fabs(gradmin);
- alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
- p->primitives.gradients.P[0] *= alpha;
- p->primitives.gradients.P[1] *= alpha;
- p->primitives.gradients.P[2] *= alpha;
- }
-#endif // SLOPE_LIMITER
+ /* Initialize time step criterion variables */
+ p->timestepvars.vmax = 0.0f;
-#endif // SPH_GRADIENTS
+ /* Set the actual velocity of the particle */
+ p->force.v_full[0] = xp->v_full[0];
+ p->force.v_full[1] = xp->v_full[1];
+ p->force.v_full[2] = xp->v_full[2];
}
/**
- * @brief Prepare a particle for the fluxes calculation.
+ * @brief Finishes the gradient calculation.
*
- * @param p The particle to act upon
- * @param xp The extended particle data to act upon
+ * Just a wrapper around hydro_gradients_finalize, which can be an empty method,
+ * in which case no gradients are used.
+ *
+ * This method also initializes the force loop variables.
+ *
+ * @param p The particle to act upon.
*/
-__attribute__((always_inline)) INLINE static void hydro_prepare_fluxes(
- struct part* p, struct xpart* xp) {
+__attribute__((always_inline)) INLINE static void hydro_end_gradient(
+ struct part* p) {
- /* initialize variables used for timestep calculation */
- p->timestepvars.vmax = 0.0f;
+ hydro_gradients_finalize(p);
+
+ p->gravity.mflux[0] = 0.0f;
+ p->gravity.mflux[1] = 0.0f;
+ p->gravity.mflux[2] = 0.0f;
}
/**
* @brief Reset acceleration fields of a particle
*
- * Resets all hydro acceleration and time derivative fields in preparation
- * for the sums taking place in the variaous force tasks
+ * This is actually not necessary for GIZMO, since we just set the accelerations
+ * after the flux calculation.
*
- * @param p The particle to act upon
+ * @param p The particle to act upon.
*/
__attribute__((always_inline)) INLINE static void hydro_reset_acceleration(
struct part* p) {
- /* figure out what to put here */
+ /* Reset the acceleration. */
+ p->a_hydro[0] = 0.0f;
+ p->a_hydro[1] = 0.0f;
+ p->a_hydro[2] = 0.0f;
+
+ /* Reset the time derivatives. */
+ p->force.h_dt = 0.0f;
}
/**
- * @brief Finishes the fluxes calculation.
+ * @brief Converts the hydrodynamic variables from the initial condition file to
+ * conserved variables that can be used during the integration
*
- * Multiplies the forces and accelerationsby the appropiate constants
+ * Requires the volume to be known.
*
- * @param p The particle to act upon
+ * The initial condition file contains a mixture of primitive and conserved
+ * variables. Mass is a conserved variable, and we just copy the particle
+ * mass into the corresponding conserved quantity. We need the volume to
+ * also derive a density, which is then used to convert the internal energy
+ * to a pressure. However, we do not actually use these variables anymore.
+ * We do need to initialize the linear momentum, based on the mass and the
+ * velocity of the particle.
+ *
+ * @param p The particle to act upon.
*/
-__attribute__((always_inline)) INLINE static void hydro_end_fluxes(
+__attribute__((always_inline)) INLINE static void hydro_convert_quantities(
struct part* p) {
- /* do nothing */
+ const float volume = p->geometry.volume;
+ const float m = p->conserved.mass;
+ p->primitives.rho = m / volume;
+
+ p->conserved.momentum[0] = m * p->primitives.v[0];
+ p->conserved.momentum[1] = m * p->primitives.v[1];
+ p->conserved.momentum[2] = m * p->primitives.v[2];
+
+ p->primitives.P =
+ hydro_gamma_minus_one * p->conserved.energy * p->primitives.rho;
+
+ p->conserved.energy *= m;
}
/**
- * @brief Converts hydro quantity of a particle
+ * @brief Extra operations to be done during the drift
*
- * Requires the volume to be known
+ * @param p Particle to act upon.
+ * @param xp The extended particle data to act upon.
+ * @param dt The drift time-step.
+ * @param t0 Integer start time of the drift interval.
+ * @param t1 Integer end time of the drift interval.
+ * @param timeBase Conversion factor between integer and physical time.
+ */
+__attribute__((always_inline)) INLINE static void hydro_predict_extra(
+ struct part* p, struct xpart* xp, float dt, int t0, int t1,
+ double timeBase) {
+
+ const float h_inv = 1.0f / p->h;
+
+ /* Predict smoothing length */
+ const float w1 = p->force.h_dt * h_inv * dt;
+ if (fabsf(w1) < 0.2f)
+ p->h *= approx_expf(w1); /* 4th order expansion of exp(w) */
+ else
+ p->h *= expf(w1);
+
+ const float w2 = -hydro_dimension * w1;
+ if (fabsf(w2) < 0.2f) {
+ p->primitives.rho *= approx_expf(w2);
+ } else {
+ p->primitives.rho *= expf(w2);
+ }
+
+ p->primitives.v[0] += (p->a_hydro[0] + p->gravity.old_a[0]) * dt;
+ p->primitives.v[1] += (p->a_hydro[1] + p->gravity.old_a[1]) * dt;
+ p->primitives.v[2] += (p->a_hydro[2] + p->gravity.old_a[2]) * dt;
+ const float u = p->conserved.energy + p->du_dt * dt;
+ p->primitives.P =
+ hydro_gamma_minus_one * u * p->primitives.rho / p->conserved.mass;
+}
+
+/**
+ * @brief Set the particle acceleration after the flux loop
*
- * @param p The particle to act upon
+ * We use the new conserved variables to calculate the new velocity of the
+ * particle, and use that to derive the change of the velocity over the particle
+ * time step.
+ *
+ * If the particle time step is zero, we set the accelerations to zero. This
+ * should only happen at the start of the simulation.
+ *
+ * @param p Particle to act upon.
*/
-__attribute__((always_inline)) INLINE static void hydro_convert_quantities(
+__attribute__((always_inline)) INLINE static void hydro_end_force(
struct part* p) {
- float volume;
- GFLOAT m;
- GFLOAT momentum[3];
-#ifndef THERMAL_ENERGY
- GFLOAT momentum2;
-#endif
- volume = p->geometry.volume;
+ /* Add normalization to h_dt. */
+ p->force.h_dt *= p->h * hydro_dimension_inv;
- /* set hydro velocities */
- p->primitives.v[0] = p->v[0];
- p->primitives.v[1] = p->v[1];
- p->primitives.v[2] = p->v[2];
- /* P actually contains internal energy at this point */
- p->primitives.P *= (const_hydro_gamma - 1.) * p->primitives.rho;
-
- p->conserved.mass = m = p->primitives.rho * volume;
- p->conserved.momentum[0] = momentum[0] = m * p->primitives.v[0];
- p->conserved.momentum[1] = momentum[1] = m * p->primitives.v[1];
- p->conserved.momentum[2] = momentum[2] = m * p->primitives.v[2];
-#ifndef THERMAL_ENERGY
- momentum2 = momentum[0] * momentum[0] + momentum[1] * momentum[1] +
- momentum[2] * momentum[2];
- p->conserved.energy =
- p->primitives.P / (const_hydro_gamma - 1.) * volume + 0.5 * momentum2 / m;
-#else
- p->conserved.energy = p->primitives.P / (const_hydro_gamma - 1.) * volume;
-#endif
+ /* Set the hydro acceleration, based on the new momentum and mass */
+ /* NOTE: the momentum and mass are only correct for active particles, since
+ only active particles have received flux contributions from all their
+ neighbours. Since this method is only called for active particles,
+ this is indeed the case. */
+ if (p->force.dt) {
+ float mnew;
+ float vnew[3];
+
+ mnew = p->conserved.mass + p->conserved.flux.mass;
+ vnew[0] = (p->conserved.momentum[0] + p->conserved.flux.momentum[0]) / mnew;
+ vnew[1] = (p->conserved.momentum[1] + p->conserved.flux.momentum[1]) / mnew;
+ vnew[2] = (p->conserved.momentum[2] + p->conserved.flux.momentum[2]) / mnew;
+
+ p->a_hydro[0] = (vnew[0] - p->force.v_full[0]) / p->force.dt;
+ p->a_hydro[1] = (vnew[1] - p->force.v_full[1]) / p->force.dt;
+ p->a_hydro[2] = (vnew[2] - p->force.v_full[2]) / p->force.dt;
+
+ p->du_dt = p->conserved.flux.energy / p->force.dt;
+ } else {
+ p->a_hydro[0] = 0.0f;
+ p->a_hydro[1] = 0.0f;
+ p->a_hydro[2] = 0.0f;
+
+ p->du_dt = 0.0f;
+ }
}
-// MATTHIEU
-__attribute__((always_inline)) INLINE static void hydro_end_density(
- struct part* p, float time) {}
-__attribute__((always_inline)) INLINE static void hydro_prepare_force(
- struct part* p, struct xpart* xp, int ti_current, double timeBase) {}
-__attribute__((always_inline)) INLINE static void hydro_predict_extra(
- struct part* p, struct xpart* xp, int t0, int t1, double timeBase) {}
-__attribute__((always_inline)) INLINE static void hydro_end_force(
- struct part* p) {}
+/**
+ * @brief Extra operations done during the kick
+ *
+ * Not used for GIZMO.
+ *
+ * @param p Particle to act upon.
+ * @param xp Extended particle data to act upon.
+ * @param dt Physical time step.
+ * @param half_dt Half the physical time step.
+ */
__attribute__((always_inline)) INLINE static void hydro_kick_extra(
- struct part* p, struct xpart* xp, float dt, float half_dt) {}
+ struct part* p, struct xpart* xp, float dt, float half_dt) {
+
+ float oldm, oldp[3], anew[3];
+
+ /* Retrieve the current value of the gravitational acceleration from the
+ gpart. We are only allowed to do this because this is the kick. We still
+ need to check whether gpart exists though.*/
+ if (p->gpart) {
+ anew[0] = p->gpart->a_grav[0];
+ anew[1] = p->gpart->a_grav[1];
+ anew[2] = p->gpart->a_grav[2];
+
+ /* Copy the old mass and momentum before updating the conserved variables */
+ oldm = p->conserved.mass;
+ oldp[0] = p->conserved.momentum[0];
+ oldp[1] = p->conserved.momentum[1];
+ oldp[2] = p->conserved.momentum[2];
+ }
+
+ /* Update conserved variables. */
+ p->conserved.mass += p->conserved.flux.mass;
+ p->conserved.momentum[0] += p->conserved.flux.momentum[0];
+ p->conserved.momentum[1] += p->conserved.flux.momentum[1];
+ p->conserved.momentum[2] += p->conserved.flux.momentum[2];
+ p->conserved.energy += p->conserved.flux.energy;
+
+ /* Add gravity. We only do this if we have gravity activated. */
+ if (p->gpart) {
+ p->conserved.momentum[0] +=
+ half_dt * (oldm * p->gravity.old_a[0] + p->conserved.mass * anew[0]);
+ p->conserved.momentum[1] +=
+ half_dt * (oldm * p->gravity.old_a[1] + p->conserved.mass * anew[1]);
+ p->conserved.momentum[2] +=
+ half_dt * (oldm * p->gravity.old_a[2] + p->conserved.mass * anew[2]);
+
+ float paold, panew;
+ paold = oldp[0] * p->gravity.old_a[0] + oldp[1] * p->gravity.old_a[1] +
+ oldp[2] * p->gravity.old_a[2];
+ panew = p->conserved.momentum[0] * anew[0] +
+ p->conserved.momentum[1] * anew[1] +
+ p->conserved.momentum[2] * anew[2];
+ p->conserved.energy += half_dt * (paold + panew);
+
+ float fluxaold, fluxanew;
+ fluxaold = p->gravity.old_a[0] * p->gravity.old_mflux[0] +
+ p->gravity.old_a[1] * p->gravity.old_mflux[1] +
+ p->gravity.old_a[2] * p->gravity.old_mflux[2];
+ fluxanew = anew[0] * p->gravity.mflux[0] + anew[1] * p->gravity.mflux[1] +
+ anew[2] * p->gravity.mflux[2];
+ p->conserved.energy += half_dt * (fluxaold + fluxanew);
+
+ /* Store gravitational acceleration and mass flux for next step */
+ p->gravity.old_a[0] = anew[0];
+ p->gravity.old_a[1] = anew[1];
+ p->gravity.old_a[2] = anew[2];
+ p->gravity.old_mflux[0] = p->gravity.mflux[0];
+ p->gravity.old_mflux[1] = p->gravity.mflux[1];
+ p->gravity.old_mflux[2] = p->gravity.mflux[2];
+ }
+
+ /* reset fluxes */
+ /* we can only do this here, since we need to keep the fluxes for inactive
+ particles */
+ p->conserved.flux.mass = 0.0f;
+ p->conserved.flux.momentum[0] = 0.0f;
+ p->conserved.flux.momentum[1] = 0.0f;
+ p->conserved.flux.momentum[2] = 0.0f;
+ p->conserved.flux.energy = 0.0f;
+}
+
+/**
+ * @brief Returns the internal energy of a particle
+ *
+ * @param p The particle of interest.
+ * @param dt Time since the last kick.
+ */
__attribute__((always_inline)) INLINE static float hydro_get_internal_energy(
- struct part* p) {
- return 0.f;
+ const struct part* restrict p, float dt) {
+
+ return p->primitives.P / hydro_gamma_minus_one / p->primitives.rho;
+}
+
+/**
+ * @brief Returns the entropy of a particle
+ *
+ * @param p The particle of interest.
+ * @param dt Time since the last kick.
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_entropy(
+ const struct part* restrict p, float dt) {
+
+ return p->primitives.P / pow_gamma(p->primitives.rho);
+}
+
+/**
+ * @brief Returns the sound speed of a particle
+ *
+ * @param p The particle of interest.
+ * @param dt Time since the last kick.
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_soundspeed(
+ const struct part* restrict p, float dt) {
+
+ return sqrtf(hydro_gamma * p->primitives.P / p->primitives.rho);
+}
+
+/**
+ * @brief Returns the pressure of a particle
+ *
+ * @param p The particle of interest
+ * @param dt Time since the last kick
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_pressure(
+ const struct part* restrict p, float dt) {
+
+ return p->primitives.P;
+}
+
+/**
+ * @brief Returns the mass of a particle
+ *
+ * @param p The particle of interest
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_mass(
+ const struct part* restrict p) {
+
+ return p->conserved.mass;
+}
+
+/**
+ * @brief Returns the density of a particle
+ *
+ * @param p The particle of interest
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_density(
+ const struct part* restrict p) {
+
+ return p->primitives.rho;
}
diff --git a/src/hydro/Gizmo/hydro_debug.h b/src/hydro/Gizmo/hydro_debug.h
index 365d85a2f651cf98b0713e8d82f11ae70fa9beaa..f4c071023a627b177fd06373856f25611fc9485d 100644
--- a/src/hydro/Gizmo/hydro_debug.h
+++ b/src/hydro/Gizmo/hydro_debug.h
@@ -18,10 +18,65 @@
******************************************************************************/
__attribute__((always_inline)) INLINE static void hydro_debug_particle(
- struct part* p, struct xpart* xp) {
+ const struct part* p, const struct xpart* xp) {
printf(
"x=[%.16e,%.16e,%.16e], "
- "v=[%.3e,%.3e,%.3e], a=[%.3e,%.3e,%.3e], volume=%.3e\n",
+ "v=[%.3e,%.3e,%.3e], "
+ "a=[%.3e,%.3e,%.3e], "
+ "h=%.3e, "
+ "ti_begin=%d, "
+ "ti_end=%d, "
+ "primitives={"
+ "v=[%.3e,%.3e,%.3e], "
+ "rho=%.3e, "
+ "P=%.3e, "
+ "gradients={"
+ "rho=[%.3e,%.3e,%.3e], "
+ "v=[[%.3e,%.3e,%.3e],[%.3e,%.3e,%.3e],[%.3e,%.3e,%.3e]], "
+ "P=[%.3e,%.3e,%.3e]}, "
+ "limiter={"
+ "rho=[%.3e,%.3e], "
+ "v=[[%.3e,%.3e],[%.3e,%.3e],[%.3e,%.3e]], "
+ "P=[%.3e,%.3e], "
+ "maxr=%.3e}}, "
+ "conserved={"
+ "momentum=[%.3e,%.3e,%.3e], "
+ "mass=%.3e, "
+ "energy=%.3e}, "
+ "geometry={"
+ "volume=%.3e, "
+ "matrix_E=[[%.3e,%.3e,%.3e],[%.3e,%.3e,%.3e],[%.3e,%.3e,%.3e]]}, "
+ "timestepvars={"
+ "vmax=%.3e}, "
+ "density={"
+ "div_v=%.3e, "
+ "wcount_dh=%.3e, "
+ "curl_v=[%.3e,%.3e,%.3e], "
+ "wcount=%.3e}\n",
p->x[0], p->x[1], p->x[2], p->v[0], p->v[1], p->v[2], p->a_hydro[0],
- p->a_hydro[1], p->a_hydro[2], p->geometry.volume);
+ p->a_hydro[1], p->a_hydro[2], p->h, p->ti_begin, p->ti_end,
+ p->primitives.v[0], p->primitives.v[1], p->primitives.v[2],
+ p->primitives.rho, p->primitives.P, p->primitives.gradients.rho[0],
+ p->primitives.gradients.rho[1], p->primitives.gradients.rho[2],
+ p->primitives.gradients.v[0][0], p->primitives.gradients.v[0][1],
+ p->primitives.gradients.v[0][2], p->primitives.gradients.v[1][0],
+ p->primitives.gradients.v[1][1], p->primitives.gradients.v[1][2],
+ p->primitives.gradients.v[2][0], p->primitives.gradients.v[2][1],
+ p->primitives.gradients.v[2][2], p->primitives.gradients.P[0],
+ p->primitives.gradients.P[1], p->primitives.gradients.P[2],
+ p->primitives.limiter.rho[0], p->primitives.limiter.rho[1],
+ p->primitives.limiter.v[0][0], p->primitives.limiter.v[0][1],
+ p->primitives.limiter.v[1][0], p->primitives.limiter.v[1][1],
+ p->primitives.limiter.v[2][0], p->primitives.limiter.v[2][1],
+ p->primitives.limiter.P[0], p->primitives.limiter.P[1],
+ p->primitives.limiter.maxr, p->conserved.momentum[0],
+ p->conserved.momentum[1], p->conserved.momentum[2], p->conserved.mass,
+ p->conserved.energy, p->geometry.volume, p->geometry.matrix_E[0][0],
+ p->geometry.matrix_E[0][1], p->geometry.matrix_E[0][2],
+ p->geometry.matrix_E[1][0], p->geometry.matrix_E[1][1],
+ p->geometry.matrix_E[1][2], p->geometry.matrix_E[2][0],
+ p->geometry.matrix_E[2][1], p->geometry.matrix_E[2][2],
+ p->timestepvars.vmax, p->density.div_v, p->density.wcount_dh,
+ p->density.curl_v[0], p->density.curl_v[1], p->density.curl_v[2],
+ p->density.wcount);
}
diff --git a/src/hydro/Gizmo/hydro_gradients.h b/src/hydro/Gizmo/hydro_gradients.h
new file mode 100644
index 0000000000000000000000000000000000000000..90448efc7adb8ccecaaa98c7388f89eaa8d16bcd
--- /dev/null
+++ b/src/hydro/Gizmo/hydro_gradients.h
@@ -0,0 +1,208 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (c) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com)
+ *
+ * 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_HYDRO_GRADIENTS_H
+#define SWIFT_HYDRO_GRADIENTS_H
+
+#include "hydro_slope_limiters.h"
+
+#if defined(GRADIENTS_SPH)
+
+#define HYDRO_GRADIENT_IMPLEMENTATION "SPH gradients (Price 2012)"
+#include "hydro_gradients_sph.h"
+
+#elif defined(GRADIENTS_GIZMO)
+
+#define HYDRO_GRADIENT_IMPLEMENTATION "GIZMO gradients (Hopkins 2015)"
+#include "hydro_gradients_gizmo.h"
+
+#else
+
+/* No gradients. Perfectly acceptable, but we have to provide empty functions */
+#define HYDRO_GRADIENT_IMPLEMENTATION "No gradients (first order scheme)"
+
+/**
+ * @brief Initialize gradient variables
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_init(
+ struct part* p) {}
+
+/**
+ * @brief Gradient calculations done during the neighbour loop
+ *
+ * @param r2 Squared distance between the two particles.
+ * @param dx Distance vector (pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_collect(
+ float r2, float* dx, float hi, float hj, struct part* pi, struct part* pj) {
+}
+
+/**
+ * @brief Gradient calculations done during the neighbour loop: non-symmetric
+ * version
+ *
+ * @param r2 Squared distance between the two particles.
+ * @param dx Distance vector (pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void
+hydro_gradients_nonsym_collect(float r2, float* dx, float hi, float hj,
+ struct part* pi, struct part* pj) {}
+
+/**
+ * @brief Finalize the gradient variables after all data have been collected
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_finalize(
+ struct part* p) {}
+
+#endif
+
+/**
+ * @brief Gradients reconstruction. Is the same for all gradient types (although
+ * gradients_none does nothing, since all gradients are zero -- are they?).
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_predict(
+ struct part* pi, struct part* pj, float hi, float hj, float* dx, float r,
+ float* xij_i, float* Wi, float* Wj, float mindt) {
+
+ float dWi[5], dWj[5];
+ float xij_j[3];
+ int k;
+ float xfac;
+
+ /* perform gradient reconstruction in space and time */
+ /* space */
+ /* Compute interface position (relative to pj, since we don't need the actual
+ * position) */
+ /* eqn. (8) */
+ xfac = hj / (hi + hj);
+ for (k = 0; k < 3; k++) xij_j[k] = xfac * dx[k];
+
+ dWi[0] = pi->primitives.gradients.rho[0] * xij_i[0] +
+ pi->primitives.gradients.rho[1] * xij_i[1] +
+ pi->primitives.gradients.rho[2] * xij_i[2];
+ dWi[1] = pi->primitives.gradients.v[0][0] * xij_i[0] +
+ pi->primitives.gradients.v[0][1] * xij_i[1] +
+ pi->primitives.gradients.v[0][2] * xij_i[2];
+ dWi[2] = pi->primitives.gradients.v[1][0] * xij_i[0] +
+ pi->primitives.gradients.v[1][1] * xij_i[1] +
+ pi->primitives.gradients.v[1][2] * xij_i[2];
+ dWi[3] = pi->primitives.gradients.v[2][0] * xij_i[0] +
+ pi->primitives.gradients.v[2][1] * xij_i[1] +
+ pi->primitives.gradients.v[2][2] * xij_i[2];
+ dWi[4] = pi->primitives.gradients.P[0] * xij_i[0] +
+ pi->primitives.gradients.P[1] * xij_i[1] +
+ pi->primitives.gradients.P[2] * xij_i[2];
+
+ dWj[0] = pj->primitives.gradients.rho[0] * xij_j[0] +
+ pj->primitives.gradients.rho[1] * xij_j[1] +
+ pj->primitives.gradients.rho[2] * xij_j[2];
+ dWj[1] = pj->primitives.gradients.v[0][0] * xij_j[0] +
+ pj->primitives.gradients.v[0][1] * xij_j[1] +
+ pj->primitives.gradients.v[0][2] * xij_j[2];
+ dWj[2] = pj->primitives.gradients.v[1][0] * xij_j[0] +
+ pj->primitives.gradients.v[1][1] * xij_j[1] +
+ pj->primitives.gradients.v[1][2] * xij_j[2];
+ dWj[3] = pj->primitives.gradients.v[2][0] * xij_j[0] +
+ pj->primitives.gradients.v[2][1] * xij_j[1] +
+ pj->primitives.gradients.v[2][2] * xij_j[2];
+ dWj[4] = pj->primitives.gradients.P[0] * xij_j[0] +
+ pj->primitives.gradients.P[1] * xij_j[1] +
+ pj->primitives.gradients.P[2] * xij_j[2];
+
+ hydro_slope_limit_face(Wi, Wj, dWi, dWj, xij_i, xij_j, r);
+
+ /* time */
+ dWi[0] -= 0.5 * mindt * (Wi[1] * pi->primitives.gradients.rho[0] +
+ Wi[2] * pi->primitives.gradients.rho[1] +
+ Wi[3] * pi->primitives.gradients.rho[2] +
+ Wi[0] * (pi->primitives.gradients.v[0][0] +
+ pi->primitives.gradients.v[1][1] +
+ pi->primitives.gradients.v[2][2]));
+ dWi[1] -= 0.5 * mindt * (Wi[1] * pi->primitives.gradients.v[0][0] +
+ Wi[2] * pi->primitives.gradients.v[0][1] +
+ Wi[3] * pi->primitives.gradients.v[0][2] +
+ pi->primitives.gradients.P[0] / Wi[0]);
+ dWi[2] -= 0.5 * mindt * (Wi[1] * pi->primitives.gradients.v[1][0] +
+ Wi[2] * pi->primitives.gradients.v[1][1] +
+ Wi[3] * pi->primitives.gradients.v[1][2] +
+ pi->primitives.gradients.P[1] / Wi[0]);
+ dWi[3] -= 0.5 * mindt * (Wi[1] * pi->primitives.gradients.v[2][0] +
+ Wi[2] * pi->primitives.gradients.v[2][1] +
+ Wi[3] * pi->primitives.gradients.v[2][2] +
+ pi->primitives.gradients.P[2] / Wi[0]);
+ dWi[4] -=
+ 0.5 * mindt * (Wi[1] * pi->primitives.gradients.P[0] +
+ Wi[2] * pi->primitives.gradients.P[1] +
+ Wi[3] * pi->primitives.gradients.P[2] +
+ hydro_gamma * Wi[4] * (pi->primitives.gradients.v[0][0] +
+ pi->primitives.gradients.v[1][1] +
+ pi->primitives.gradients.v[2][2]));
+
+ dWj[0] -= 0.5 * mindt * (Wj[1] * pj->primitives.gradients.rho[0] +
+ Wj[2] * pj->primitives.gradients.rho[1] +
+ Wj[3] * pj->primitives.gradients.rho[2] +
+ Wj[0] * (pj->primitives.gradients.v[0][0] +
+ pj->primitives.gradients.v[1][1] +
+ pj->primitives.gradients.v[2][2]));
+ dWj[1] -= 0.5 * mindt * (Wj[1] * pj->primitives.gradients.v[0][0] +
+ Wj[2] * pj->primitives.gradients.v[0][1] +
+ Wj[3] * pj->primitives.gradients.v[0][2] +
+ pj->primitives.gradients.P[0] / Wj[0]);
+ dWj[2] -= 0.5 * mindt * (Wj[1] * pj->primitives.gradients.v[1][0] +
+ Wj[2] * pj->primitives.gradients.v[1][1] +
+ Wj[3] * pj->primitives.gradients.v[1][2] +
+ pj->primitives.gradients.P[1] / Wj[0]);
+ dWj[3] -= 0.5 * mindt * (Wj[1] * pj->primitives.gradients.v[2][0] +
+ Wj[2] * pj->primitives.gradients.v[2][1] +
+ Wj[3] * pj->primitives.gradients.v[2][2] +
+ pj->primitives.gradients.P[2] / Wj[0]);
+ dWj[4] -=
+ 0.5 * mindt * (Wj[1] * pj->primitives.gradients.P[0] +
+ Wj[2] * pj->primitives.gradients.P[1] +
+ Wj[3] * pj->primitives.gradients.P[2] +
+ hydro_gamma * Wj[4] * (pj->primitives.gradients.v[0][0] +
+ pj->primitives.gradients.v[1][1] +
+ pj->primitives.gradients.v[2][2]));
+
+ Wi[0] += dWi[0];
+ Wi[1] += dWi[1];
+ Wi[2] += dWi[2];
+ Wi[3] += dWi[3];
+ Wi[4] += dWi[4];
+
+ Wj[0] += dWj[0];
+ Wj[1] += dWj[1];
+ Wj[2] += dWj[2];
+ Wj[3] += dWj[3];
+ Wj[4] += dWj[4];
+}
+
+#endif // SWIFT_HYDRO_GRADIENTS_H
diff --git a/src/hydro/Gizmo/hydro_gradients_gizmo.h b/src/hydro/Gizmo/hydro_gradients_gizmo.h
new file mode 100644
index 0000000000000000000000000000000000000000..aa6e4406b94e7a5cafcd0ca556162476003477de
--- /dev/null
+++ b/src/hydro/Gizmo/hydro_gradients_gizmo.h
@@ -0,0 +1,341 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (c) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com)
+ *
+ * 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/>.
+ *
+ ******************************************************************************/
+
+/**
+ * @brief Initialize gradient variables
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_init(
+ struct part *p) {
+
+ p->primitives.gradients.rho[0] = 0.0f;
+ p->primitives.gradients.rho[1] = 0.0f;
+ p->primitives.gradients.rho[2] = 0.0f;
+
+ p->primitives.gradients.v[0][0] = 0.0f;
+ p->primitives.gradients.v[0][1] = 0.0f;
+ p->primitives.gradients.v[0][2] = 0.0f;
+
+ p->primitives.gradients.v[1][0] = 0.0f;
+ p->primitives.gradients.v[1][1] = 0.0f;
+ p->primitives.gradients.v[1][2] = 0.0f;
+
+ p->primitives.gradients.v[2][0] = 0.0f;
+ p->primitives.gradients.v[2][1] = 0.0f;
+ p->primitives.gradients.v[2][2] = 0.0f;
+
+ p->primitives.gradients.P[0] = 0.0f;
+ p->primitives.gradients.P[1] = 0.0f;
+ p->primitives.gradients.P[2] = 0.0f;
+
+ hydro_slope_limit_cell_init(p);
+}
+
+/**
+ * @brief Gradient calculations done during the neighbour loop
+ *
+ * @param r2 Squared distance between the two particles.
+ * @param dx Distance vector (pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_collect(
+ float r2, float *dx, float hi, float hj, struct part *pi, struct part *pj) {
+
+ float r = sqrtf(r2);
+ float xi, xj;
+ float hi_inv, hj_inv;
+ float wi, wj, wi_dx, wj_dx;
+ int k, l;
+ float Bi[3][3];
+ float Bj[3][3];
+ float Wi[5], Wj[5];
+
+ /* Initialize local variables */
+ for (k = 0; k < 3; k++) {
+ for (l = 0; l < 3; l++) {
+ Bi[k][l] = pi->geometry.matrix_E[k][l];
+ Bj[k][l] = pj->geometry.matrix_E[k][l];
+ }
+ }
+ Wi[0] = pi->primitives.rho;
+ Wi[1] = pi->primitives.v[0];
+ Wi[2] = pi->primitives.v[1];
+ Wi[3] = pi->primitives.v[2];
+ Wi[4] = pi->primitives.P;
+ Wj[0] = pj->primitives.rho;
+ Wj[1] = pj->primitives.v[0];
+ Wj[2] = pj->primitives.v[1];
+ Wj[3] = pj->primitives.v[2];
+ Wj[4] = pj->primitives.P;
+
+ /* Compute kernel of pi. */
+ hi_inv = 1.0 / hi;
+ xi = r * hi_inv;
+ kernel_deval(xi, &wi, &wi_dx);
+
+ /* Compute gradients for pi */
+ /* there is a sign difference w.r.t. eqn. (6) because of the inverse
+ * definition of dx */
+ pi->primitives.gradients.rho[0] +=
+ (Wi[0] - Wj[0]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.rho[1] +=
+ (Wi[0] - Wj[0]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.rho[2] +=
+ (Wi[0] - Wj[0]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+
+ pi->primitives.gradients.v[0][0] +=
+ (Wi[1] - Wj[1]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.v[0][1] +=
+ (Wi[1] - Wj[1]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.v[0][2] +=
+ (Wi[1] - Wj[1]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+ pi->primitives.gradients.v[1][0] +=
+ (Wi[2] - Wj[2]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.v[1][1] +=
+ (Wi[2] - Wj[2]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.v[1][2] +=
+ (Wi[2] - Wj[2]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+ pi->primitives.gradients.v[2][0] +=
+ (Wi[3] - Wj[3]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.v[2][1] +=
+ (Wi[3] - Wj[3]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.v[2][2] +=
+ (Wi[3] - Wj[3]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+
+ pi->primitives.gradients.P[0] +=
+ (Wi[4] - Wj[4]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.P[1] +=
+ (Wi[4] - Wj[4]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.P[2] +=
+ (Wi[4] - Wj[4]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+
+ hydro_slope_limit_cell_collect(pi, pj, r);
+
+ /* Compute kernel of pj. */
+ hj_inv = 1.0 / hj;
+ xj = r * hj_inv;
+ kernel_deval(xj, &wj, &wj_dx);
+
+ /* Compute gradients for pj */
+ /* there is no sign difference w.r.t. eqn. (6) because dx is now what we
+ * want
+ * it to be */
+ pj->primitives.gradients.rho[0] +=
+ (Wi[0] - Wj[0]) * wj *
+ (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
+ pj->primitives.gradients.rho[1] +=
+ (Wi[0] - Wj[0]) * wj *
+ (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
+ pj->primitives.gradients.rho[2] +=
+ (Wi[0] - Wj[0]) * wj *
+ (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
+
+ pj->primitives.gradients.v[0][0] +=
+ (Wi[1] - Wj[1]) * wj *
+ (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
+ pj->primitives.gradients.v[0][1] +=
+ (Wi[1] - Wj[1]) * wj *
+ (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
+ pj->primitives.gradients.v[0][2] +=
+ (Wi[1] - Wj[1]) * wj *
+ (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
+ pj->primitives.gradients.v[1][0] +=
+ (Wi[2] - Wj[2]) * wj *
+ (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
+ pj->primitives.gradients.v[1][1] +=
+ (Wi[2] - Wj[2]) * wj *
+ (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
+ pj->primitives.gradients.v[1][2] +=
+ (Wi[2] - Wj[2]) * wj *
+ (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
+ pj->primitives.gradients.v[2][0] +=
+ (Wi[3] - Wj[3]) * wj *
+ (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
+ pj->primitives.gradients.v[2][1] +=
+ (Wi[3] - Wj[3]) * wj *
+ (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
+ pj->primitives.gradients.v[2][2] +=
+ (Wi[3] - Wj[3]) * wj *
+ (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
+
+ pj->primitives.gradients.P[0] +=
+ (Wi[4] - Wj[4]) * wj *
+ (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
+ pj->primitives.gradients.P[1] +=
+ (Wi[4] - Wj[4]) * wj *
+ (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
+ pj->primitives.gradients.P[2] +=
+ (Wi[4] - Wj[4]) * wj *
+ (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
+
+ hydro_slope_limit_cell_collect(pj, pi, r);
+}
+
+/**
+ * @brief Gradient calculations done during the neighbour loop
+ *
+ * @param r2 Squared distance between the two particles.
+ * @param dx Distance vector (pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void
+hydro_gradients_nonsym_collect(float r2, float *dx, float hi, float hj,
+ struct part *pi, struct part *pj) {
+
+ float r = sqrtf(r2);
+ float xi;
+ float hi_inv;
+ float wi, wi_dx;
+ int k, l;
+ float Bi[3][3];
+ float Wi[5], Wj[5];
+
+ /* Initialize local variables */
+ for (k = 0; k < 3; k++) {
+ for (l = 0; l < 3; l++) {
+ Bi[k][l] = pi->geometry.matrix_E[k][l];
+ }
+ }
+ Wi[0] = pi->primitives.rho;
+ Wi[1] = pi->primitives.v[0];
+ Wi[2] = pi->primitives.v[1];
+ Wi[3] = pi->primitives.v[2];
+ Wi[4] = pi->primitives.P;
+ Wj[0] = pj->primitives.rho;
+ Wj[1] = pj->primitives.v[0];
+ Wj[2] = pj->primitives.v[1];
+ Wj[3] = pj->primitives.v[2];
+ Wj[4] = pj->primitives.P;
+
+ /* Compute kernel of pi. */
+ hi_inv = 1.0 / hi;
+ xi = r * hi_inv;
+ kernel_deval(xi, &wi, &wi_dx);
+
+ /* Compute gradients for pi */
+ /* there is a sign difference w.r.t. eqn. (6) because of the inverse
+ * definition of dx */
+ pi->primitives.gradients.rho[0] +=
+ (Wi[0] - Wj[0]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.rho[1] +=
+ (Wi[0] - Wj[0]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.rho[2] +=
+ (Wi[0] - Wj[0]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+
+ pi->primitives.gradients.v[0][0] +=
+ (Wi[1] - Wj[1]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.v[0][1] +=
+ (Wi[1] - Wj[1]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.v[0][2] +=
+ (Wi[1] - Wj[1]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+ pi->primitives.gradients.v[1][0] +=
+ (Wi[2] - Wj[2]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.v[1][1] +=
+ (Wi[2] - Wj[2]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.v[1][2] +=
+ (Wi[2] - Wj[2]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+ pi->primitives.gradients.v[2][0] +=
+ (Wi[3] - Wj[3]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.v[2][1] +=
+ (Wi[3] - Wj[3]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.v[2][2] +=
+ (Wi[3] - Wj[3]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+
+ pi->primitives.gradients.P[0] +=
+ (Wi[4] - Wj[4]) * wi *
+ (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
+ pi->primitives.gradients.P[1] +=
+ (Wi[4] - Wj[4]) * wi *
+ (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
+ pi->primitives.gradients.P[2] +=
+ (Wi[4] - Wj[4]) * wi *
+ (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
+
+ hydro_slope_limit_cell_collect(pi, pj, r);
+}
+
+/**
+ * @brief Finalize the gradient variables after all data have been collected
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_finalize(
+ struct part *p) {
+
+ float h, ih;
+
+ /* add kernel normalization to gradients */
+ h = p->h;
+ ih = 1.0f / h;
+ const float ihdim = pow_dimension(ih);
+
+ p->primitives.gradients.rho[0] *= ihdim;
+ p->primitives.gradients.rho[1] *= ihdim;
+ p->primitives.gradients.rho[2] *= ihdim;
+
+ p->primitives.gradients.v[0][0] *= ihdim;
+ p->primitives.gradients.v[0][1] *= ihdim;
+ p->primitives.gradients.v[0][2] *= ihdim;
+ p->primitives.gradients.v[1][0] *= ihdim;
+ p->primitives.gradients.v[1][1] *= ihdim;
+ p->primitives.gradients.v[1][2] *= ihdim;
+ p->primitives.gradients.v[2][0] *= ihdim;
+ p->primitives.gradients.v[2][1] *= ihdim;
+ p->primitives.gradients.v[2][2] *= ihdim;
+
+ p->primitives.gradients.P[0] *= ihdim;
+ p->primitives.gradients.P[1] *= ihdim;
+ p->primitives.gradients.P[2] *= ihdim;
+
+ hydro_slope_limit_cell(p);
+}
diff --git a/src/hydro/Gizmo/hydro_gradients_sph.h b/src/hydro/Gizmo/hydro_gradients_sph.h
new file mode 100644
index 0000000000000000000000000000000000000000..f635faecea549f7da280ade9b944021a5e4aeb4c
--- /dev/null
+++ b/src/hydro/Gizmo/hydro_gradients_sph.h
@@ -0,0 +1,248 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (c) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com)
+ *
+ * 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/>.
+ *
+ ******************************************************************************/
+
+/**
+ * @brief Initialize gradient variables
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_init(
+ struct part *p) {
+
+ p->primitives.gradients.rho[0] = 0.0f;
+ p->primitives.gradients.rho[1] = 0.0f;
+ p->primitives.gradients.rho[2] = 0.0f;
+
+ p->primitives.gradients.v[0][0] = 0.0f;
+ p->primitives.gradients.v[0][1] = 0.0f;
+ p->primitives.gradients.v[0][2] = 0.0f;
+
+ p->primitives.gradients.v[1][0] = 0.0f;
+ p->primitives.gradients.v[1][1] = 0.0f;
+ p->primitives.gradients.v[1][2] = 0.0f;
+ p->primitives.gradients.v[2][0] = 0.0f;
+ p->primitives.gradients.v[2][1] = 0.0f;
+ p->primitives.gradients.v[2][2] = 0.0f;
+
+ p->primitives.gradients.P[0] = 0.0f;
+ p->primitives.gradients.P[1] = 0.0f;
+ p->primitives.gradients.P[2] = 0.0f;
+
+ hydro_slope_limit_cell_init(p);
+}
+
+/**
+ * @brief Gradient calculations done during the neighbour loop
+ *
+ * @param r2 Squared distance between the two particles.
+ * @param dx Distance vector (pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_collect(
+ float r2, float *dx, float hi, float hj, struct part *pi, struct part *pj) {
+
+ float wi, wi_dx, xi, hi_inv;
+ float wj, wj_dx, xj, hj_inv;
+ float r = sqrtf(r2);
+
+ hi_inv = 1.0f / hi;
+ xi = r * hi_inv;
+ kernel_deval(xi, &wi, &wi_dx);
+
+ /* very basic gradient estimate */
+ pi->primitives.gradients.rho[0] -=
+ wi_dx * dx[0] * (pi->primitives.rho - pj->primitives.rho) / r;
+ pi->primitives.gradients.rho[1] -=
+ wi_dx * dx[1] * (pi->primitives.rho - pj->primitives.rho) / r;
+ pi->primitives.gradients.rho[2] -=
+ wi_dx * dx[2] * (pi->primitives.rho - pj->primitives.rho) / r;
+
+ pi->primitives.gradients.v[0][0] -=
+ wi_dx * dx[0] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
+ pi->primitives.gradients.v[0][1] -=
+ wi_dx * dx[1] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
+ pi->primitives.gradients.v[0][2] -=
+ wi_dx * dx[2] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
+
+ pi->primitives.gradients.v[1][0] -=
+ wi_dx * dx[0] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
+ pi->primitives.gradients.v[1][1] -=
+ wi_dx * dx[1] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
+ pi->primitives.gradients.v[1][2] -=
+ wi_dx * dx[2] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
+
+ pi->primitives.gradients.v[2][0] -=
+ wi_dx * dx[0] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
+ pi->primitives.gradients.v[2][1] -=
+ wi_dx * dx[1] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
+ pi->primitives.gradients.v[2][2] -=
+ wi_dx * dx[2] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
+
+ pi->primitives.gradients.P[0] -=
+ wi_dx * dx[0] * (pi->primitives.P - pj->primitives.P) / r;
+ pi->primitives.gradients.P[1] -=
+ wi_dx * dx[1] * (pi->primitives.P - pj->primitives.P) / r;
+ pi->primitives.gradients.P[2] -=
+ wi_dx * dx[2] * (pi->primitives.P - pj->primitives.P) / r;
+
+ hydro_slope_limit_cell_collect(pi, pj, r);
+
+ hj_inv = 1.0f / hj;
+ xj = r * hj_inv;
+ kernel_deval(xj, &wj, &wj_dx);
+
+ /* signs are the same as before, since we swap i and j twice */
+ pj->primitives.gradients.rho[0] -=
+ wj_dx * dx[0] * (pi->primitives.rho - pj->primitives.rho) / r;
+ pj->primitives.gradients.rho[1] -=
+ wj_dx * dx[1] * (pi->primitives.rho - pj->primitives.rho) / r;
+ pj->primitives.gradients.rho[2] -=
+ wj_dx * dx[2] * (pi->primitives.rho - pj->primitives.rho) / r;
+
+ pj->primitives.gradients.v[0][0] -=
+ wj_dx * dx[0] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
+ pj->primitives.gradients.v[0][1] -=
+ wj_dx * dx[1] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
+ pj->primitives.gradients.v[0][2] -=
+ wj_dx * dx[2] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
+
+ pj->primitives.gradients.v[1][0] -=
+ wj_dx * dx[0] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
+ pj->primitives.gradients.v[1][1] -=
+ wj_dx * dx[1] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
+ pj->primitives.gradients.v[1][2] -=
+ wj_dx * dx[2] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
+ pj->primitives.gradients.v[2][0] -=
+ wj_dx * dx[0] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
+ pj->primitives.gradients.v[2][1] -=
+ wj_dx * dx[1] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
+ pj->primitives.gradients.v[2][2] -=
+ wj_dx * dx[2] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
+
+ pj->primitives.gradients.P[0] -=
+ wj_dx * dx[0] * (pi->primitives.P - pj->primitives.P) / r;
+ pj->primitives.gradients.P[1] -=
+ wj_dx * dx[1] * (pi->primitives.P - pj->primitives.P) / r;
+ pj->primitives.gradients.P[2] -=
+ wj_dx * dx[2] * (pi->primitives.P - pj->primitives.P) / r;
+
+ hydro_slope_limit_cell_collect(pj, pi, r);
+}
+
+/**
+ * @brief Gradient calculations done during the neighbour loop: non-symmetric
+ * version
+ *
+ * @param r2 Squared distance between the two particles.
+ * @param dx Distance vector (pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void
+hydro_gradients_nonsym_collect(float r2, float *dx, float hi, float hj,
+ struct part *pi, struct part *pj) {
+
+ float wi, wi_dx, xi, hi_inv;
+ float r = sqrtf(r2);
+
+ hi_inv = 1.0f / hi;
+ xi = r * hi_inv;
+ kernel_deval(xi, &wi, &wi_dx);
+
+ /* very basic gradient estimate */
+ pi->primitives.gradients.rho[0] -=
+ wi_dx * dx[0] * (pi->primitives.rho - pj->primitives.rho) / r;
+ pi->primitives.gradients.rho[1] -=
+ wi_dx * dx[1] * (pi->primitives.rho - pj->primitives.rho) / r;
+ pi->primitives.gradients.rho[2] -=
+ wi_dx * dx[2] * (pi->primitives.rho - pj->primitives.rho) / r;
+
+ pi->primitives.gradients.v[0][0] -=
+ wi_dx * dx[0] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
+ pi->primitives.gradients.v[0][1] -=
+ wi_dx * dx[1] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
+ pi->primitives.gradients.v[0][2] -=
+ wi_dx * dx[2] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
+
+ pi->primitives.gradients.v[1][0] -=
+ wi_dx * dx[0] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
+ pi->primitives.gradients.v[1][1] -=
+ wi_dx * dx[1] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
+ pi->primitives.gradients.v[1][2] -=
+ wi_dx * dx[2] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
+
+ pi->primitives.gradients.v[2][0] -=
+ wi_dx * dx[0] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
+ pi->primitives.gradients.v[2][1] -=
+ wi_dx * dx[1] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
+ pi->primitives.gradients.v[2][2] -=
+ wi_dx * dx[2] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
+
+ pi->primitives.gradients.P[0] -=
+ wi_dx * dx[0] * (pi->primitives.P - pj->primitives.P) / r;
+ pi->primitives.gradients.P[1] -=
+ wi_dx * dx[1] * (pi->primitives.P - pj->primitives.P) / r;
+ pi->primitives.gradients.P[2] -=
+ wi_dx * dx[2] * (pi->primitives.P - pj->primitives.P) / r;
+
+ hydro_slope_limit_cell_collect(pi, pj, r);
+}
+
+/**
+ * @brief Finalize the gradient variables after all data have been collected
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_gradients_finalize(
+ struct part *p) {
+
+ const float h = p->h;
+ const float ih = 1.0f / h;
+ const float ihdimp1 = pow_dimension_plus_one(ih);
+
+ float volume = p->geometry.volume;
+
+ /* finalize gradients by multiplying with volume */
+ p->primitives.gradients.rho[0] *= ihdimp1 * volume;
+ p->primitives.gradients.rho[1] *= ihdimp1 * volume;
+ p->primitives.gradients.rho[2] *= ihdimp1 * volume;
+
+ p->primitives.gradients.v[0][0] *= ihdimp1 * volume;
+ p->primitives.gradients.v[0][1] *= ihdimp1 * volume;
+ p->primitives.gradients.v[0][2] *= ihdimp1 * volume;
+
+ p->primitives.gradients.v[1][0] *= ihdimp1 * volume;
+ p->primitives.gradients.v[1][1] *= ihdimp1 * volume;
+ p->primitives.gradients.v[1][2] *= ihdimp1 * volume;
+
+ p->primitives.gradients.v[2][0] *= ihdimp1 * volume;
+ p->primitives.gradients.v[2][1] *= ihdimp1 * volume;
+ p->primitives.gradients.v[2][2] *= ihdimp1 * volume;
+
+ p->primitives.gradients.P[0] *= ihdimp1 * volume;
+ p->primitives.gradients.P[1] *= ihdimp1 * volume;
+ p->primitives.gradients.P[2] *= ihdimp1 * volume;
+
+ hydro_slope_limit_cell(p);
+}
diff --git a/src/hydro/Gizmo/hydro_iact.h b/src/hydro/Gizmo/hydro_iact.h
index 30a8d6cbebc851b44a5ee2339950aec9e15057c0..79973364617bb04855115bff9bfbf3808f46d04f 100644
--- a/src/hydro/Gizmo/hydro_iact.h
+++ b/src/hydro/Gizmo/hydro_iact.h
@@ -19,14 +19,28 @@
*
******************************************************************************/
+#include "adiabatic_index.h"
+#include "hydro_gradients.h"
#include "riemann.h"
-#define USE_GRADIENTS
-#define PER_FACE_LIMITER
-/* #define PRINT_ID 0 */
-
-/* this corresponds to task_subtype_hydro_loop1 */
-__attribute__((always_inline)) INLINE static void runner_iact_hydro_loop1(
+/**
+ * @brief Calculate the volume interaction between particle i and particle j
+ *
+ * The volume is in essence the same as the weighted number of neighbours in a
+ * classical SPH density calculation.
+ *
+ * We also calculate the components of the matrix E, which is used for second
+ * order accurate gradient calculations and for the calculation of the interface
+ * surface areas.
+ *
+ * @param r2 Squared distance between particle i and particle j.
+ * @param dx Distance vector between the particles (dx = pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void runner_iact_density(
float r2, float *dx, float hi, float hj, struct part *pi, struct part *pj) {
float r = sqrtf(r2);
@@ -48,71 +62,6 @@ __attribute__((always_inline)) INLINE static void runner_iact_hydro_loop1(
for (k = 0; k < 3; k++)
for (l = 0; l < 3; l++) pi->geometry.matrix_E[k][l] += dx[k] * dx[l] * wi;
-#ifdef SPH_GRADIENTS
- /* very basic gradient estimate */
- pi->primitives.gradients.rho[0] -=
- wi_dx * dx[0] * (pi->primitives.rho - pj->primitives.rho) / r;
- pi->primitives.gradients.rho[1] -=
- wi_dx * dx[1] * (pi->primitives.rho - pj->primitives.rho) / r;
- pi->primitives.gradients.rho[2] -=
- wi_dx * dx[2] * (pi->primitives.rho - pj->primitives.rho) / r;
-
- pi->primitives.gradients.v[0][0] -=
- wi_dx * dx[0] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
- pi->primitives.gradients.v[0][1] -=
- wi_dx * dx[1] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
- pi->primitives.gradients.v[0][2] -=
- wi_dx * dx[2] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
-
- pi->primitives.gradients.v[1][0] -=
- wi_dx * dx[0] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
- pi->primitives.gradients.v[1][1] -=
- wi_dx * dx[1] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
- pi->primitives.gradients.v[1][2] -=
- wi_dx * dx[2] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
-
- pi->primitives.gradients.v[2][0] -=
- wi_dx * dx[0] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
- pi->primitives.gradients.v[2][1] -=
- wi_dx * dx[1] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
- pi->primitives.gradients.v[2][2] -=
- wi_dx * dx[2] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
-
- pi->primitives.gradients.P[0] -=
- wi_dx * dx[0] * (pi->primitives.P - pj->primitives.P) / r;
- pi->primitives.gradients.P[1] -=
- wi_dx * dx[1] * (pi->primitives.P - pj->primitives.P) / r;
- pi->primitives.gradients.P[2] -=
- wi_dx * dx[2] * (pi->primitives.P - pj->primitives.P) / r;
-
- /* basic slope limiter: collect the maximal and the minimal value for the
- * primitive variables among the ngbs */
- pi->primitives.limiter.rho[0] =
- fmin(pj->primitives.rho, pi->primitives.limiter.rho[0]);
- pi->primitives.limiter.rho[1] =
- fmax(pj->primitives.rho, pi->primitives.limiter.rho[1]);
-
- pi->primitives.limiter.v[0][0] =
- fmin(pj->primitives.v[0], pi->primitives.limiter.v[0][0]);
- pi->primitives.limiter.v[0][1] =
- fmax(pj->primitives.v[0], pi->primitives.limiter.v[0][1]);
- pi->primitives.limiter.v[1][0] =
- fmin(pj->primitives.v[1], pi->primitives.limiter.v[1][0]);
- pi->primitives.limiter.v[1][1] =
- fmax(pj->primitives.v[1], pi->primitives.limiter.v[1][1]);
- pi->primitives.limiter.v[2][0] =
- fmin(pj->primitives.v[2], pi->primitives.limiter.v[2][0]);
- pi->primitives.limiter.v[2][1] =
- fmax(pj->primitives.v[2], pi->primitives.limiter.v[2][1]);
-
- pi->primitives.limiter.P[0] =
- fmin(pj->primitives.P, pi->primitives.limiter.P[0]);
- pi->primitives.limiter.P[1] =
- fmax(pj->primitives.P, pi->primitives.limiter.P[1]);
-
- pi->primitives.limiter.maxr = fmax(r, pi->primitives.limiter.maxr);
-#endif
-
/* Compute density of pj. */
h_inv = 1.0 / hj;
xj = r * h_inv;
@@ -125,78 +74,28 @@ __attribute__((always_inline)) INLINE static void runner_iact_hydro_loop1(
pj->geometry.volume += wj;
for (k = 0; k < 3; k++)
for (l = 0; l < 3; l++) pj->geometry.matrix_E[k][l] += dx[k] * dx[l] * wj;
-
-#ifdef SPH_GRADIENTS
- /* very basic gradient estimate */
- /* signs are the same as before, since we swap i and j twice */
- pj->primitives.gradients.rho[0] -=
- wj_dx * dx[0] * (pi->primitives.rho - pj->primitives.rho) / r;
- pj->primitives.gradients.rho[1] -=
- wj_dx * dx[1] * (pi->primitives.rho - pj->primitives.rho) / r;
- pj->primitives.gradients.rho[2] -=
- wj_dx * dx[2] * (pi->primitives.rho - pj->primitives.rho) / r;
-
- pj->primitives.gradients.v[0][0] -=
- wj_dx * dx[0] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
- pj->primitives.gradients.v[0][1] -=
- wj_dx * dx[1] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
- pj->primitives.gradients.v[0][2] -=
- wj_dx * dx[2] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
-
- pj->primitives.gradients.v[1][0] -=
- wj_dx * dx[0] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
- pj->primitives.gradients.v[1][1] -=
- wj_dx * dx[1] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
- pj->primitives.gradients.v[1][2] -=
- wj_dx * dx[2] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
-
- pj->primitives.gradients.v[2][0] -=
- wj_dx * dx[0] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
- pj->primitives.gradients.v[2][1] -=
- wj_dx * dx[1] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
- pj->primitives.gradients.v[2][2] -=
- wj_dx * dx[2] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
-
- pj->primitives.gradients.P[0] -=
- wj_dx * dx[0] * (pi->primitives.P - pj->primitives.P) / r;
- pj->primitives.gradients.P[1] -=
- wj_dx * dx[1] * (pi->primitives.P - pj->primitives.P) / r;
- pj->primitives.gradients.P[2] -=
- wj_dx * dx[2] * (pi->primitives.P - pj->primitives.P) / r;
-
- /* basic slope limiter: collect the maximal and the minimal value for the
- * primitive variables among the ngbs */
- pj->primitives.limiter.rho[0] =
- fmin(pi->primitives.rho, pj->primitives.limiter.rho[0]);
- pj->primitives.limiter.rho[1] =
- fmax(pi->primitives.rho, pj->primitives.limiter.rho[1]);
-
- pj->primitives.limiter.v[0][0] =
- fmin(pi->primitives.v[0], pj->primitives.limiter.v[0][0]);
- pj->primitives.limiter.v[0][1] =
- fmax(pi->primitives.v[0], pj->primitives.limiter.v[0][1]);
- pj->primitives.limiter.v[1][0] =
- fmin(pi->primitives.v[1], pj->primitives.limiter.v[1][0]);
- pj->primitives.limiter.v[1][1] =
- fmax(pi->primitives.v[1], pj->primitives.limiter.v[1][1]);
- pj->primitives.limiter.v[2][0] =
- fmin(pi->primitives.v[2], pj->primitives.limiter.v[2][0]);
- pj->primitives.limiter.v[2][1] =
- fmax(pi->primitives.v[2], pj->primitives.limiter.v[2][1]);
-
- pj->primitives.limiter.P[0] =
- fmin(pi->primitives.P, pj->primitives.limiter.P[0]);
- pj->primitives.limiter.P[1] =
- fmax(pi->primitives.P, pj->primitives.limiter.P[1]);
-
- pj->primitives.limiter.maxr = fmax(r, pj->primitives.limiter.maxr);
-#endif
}
-/* this corresponds to task_subtype_hydro_loop1 */
-__attribute__((always_inline)) INLINE static void
-runner_iact_nonsym_hydro_loop1(float r2, float *dx, float hi, float hj,
- struct part *pi, struct part *pj) {
+/**
+ * @brief Calculate the volume interaction between particle i and particle j:
+ * non-symmetric version
+ *
+ * The volume is in essence the same as the weighted number of neighbours in a
+ * classical SPH density calculation.
+ *
+ * We also calculate the components of the matrix E, which is used for second
+ * order accurate gradient calculations and for the calculation of the interface
+ * surface areas.
+ *
+ * @param r2 Squared distance between particle i and particle j.
+ * @param dx Distance vector between the particles (dx = pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void runner_iact_nonsym_density(
+ float r2, float *dx, float hi, float hj, struct part *pi, struct part *pj) {
float r;
float xi;
@@ -218,397 +117,80 @@ runner_iact_nonsym_hydro_loop1(float r2, float *dx, float hi, float hj,
pi->geometry.volume += wi;
for (k = 0; k < 3; k++)
for (l = 0; l < 3; l++) pi->geometry.matrix_E[k][l] += dx[k] * dx[l] * wi;
-
-#ifdef SPH_GRADIENTS
- /* very basic gradient estimate */
- pi->primitives.gradients.rho[0] -=
- wi_dx * dx[0] * (pi->primitives.rho - pj->primitives.rho) / r;
- pi->primitives.gradients.rho[1] -=
- wi_dx * dx[1] * (pi->primitives.rho - pj->primitives.rho) / r;
- pi->primitives.gradients.rho[2] -=
- wi_dx * dx[2] * (pi->primitives.rho - pj->primitives.rho) / r;
-
- pi->primitives.gradients.v[0][0] -=
- wi_dx * dx[0] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
- pi->primitives.gradients.v[0][1] -=
- wi_dx * dx[1] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
- pi->primitives.gradients.v[0][2] -=
- wi_dx * dx[2] * (pi->primitives.v[0] - pj->primitives.v[0]) / r;
-
- pi->primitives.gradients.v[1][0] -=
- wi_dx * dx[0] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
- pi->primitives.gradients.v[1][1] -=
- wi_dx * dx[1] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
- pi->primitives.gradients.v[1][2] -=
- wi_dx * dx[2] * (pi->primitives.v[1] - pj->primitives.v[1]) / r;
-
- pi->primitives.gradients.v[2][0] -=
- wi_dx * dx[0] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
- pi->primitives.gradients.v[2][1] -=
- wi_dx * dx[1] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
- pi->primitives.gradients.v[2][2] -=
- wi_dx * dx[2] * (pi->primitives.v[2] - pj->primitives.v[2]) / r;
-
- pi->primitives.gradients.P[0] -=
- wi_dx * dx[0] * (pi->primitives.P - pj->primitives.P) / r;
- pi->primitives.gradients.P[1] -=
- wi_dx * dx[1] * (pi->primitives.P - pj->primitives.P) / r;
- pi->primitives.gradients.P[2] -=
- wi_dx * dx[2] * (pi->primitives.P - pj->primitives.P) / r;
-
- /* slope limiter */
- pi->primitives.limiter.rho[0] =
- fmin(pj->primitives.rho, pi->primitives.limiter.rho[0]);
- pi->primitives.limiter.rho[1] =
- fmax(pj->primitives.rho, pi->primitives.limiter.rho[1]);
-
- pi->primitives.limiter.v[0][0] =
- fmin(pj->primitives.v[0], pi->primitives.limiter.v[0][0]);
- pi->primitives.limiter.v[0][1] =
- fmax(pj->primitives.v[0], pi->primitives.limiter.v[0][1]);
- pi->primitives.limiter.v[1][0] =
- fmin(pj->primitives.v[1], pi->primitives.limiter.v[1][0]);
- pi->primitives.limiter.v[1][1] =
- fmax(pj->primitives.v[1], pi->primitives.limiter.v[1][1]);
- pi->primitives.limiter.v[2][0] =
- fmin(pj->primitives.v[2], pi->primitives.limiter.v[2][0]);
- pi->primitives.limiter.v[2][1] =
- fmax(pj->primitives.v[2], pi->primitives.limiter.v[2][1]);
-
- pi->primitives.limiter.P[0] =
- fmin(pj->primitives.P, pi->primitives.limiter.P[0]);
- pi->primitives.limiter.P[1] =
- fmax(pj->primitives.P, pi->primitives.limiter.P[1]);
-
- pi->primitives.limiter.maxr = fmax(r, pi->primitives.limiter.maxr);
-#endif
}
-__attribute__((always_inline)) INLINE static void runner_iact_hydro_loop2(
+/**
+ * @brief Calculate the gradient interaction between particle i and particle j
+ *
+ * This method wraps around hydro_gradients_collect, which can be an empty
+ * method, in which case no gradients are used.
+ *
+ * @param r2 Squared distance between particle i and particle j.
+ * @param dx Distance vector between the particles (dx = pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void runner_iact_gradient(
float r2, float *dx, float hi, float hj, struct part *pi, struct part *pj) {
-#ifndef SPH_GRADIENTS
-
- float r = sqrtf(r2);
- float xi, xj;
- float hi_inv, hj_inv;
- float wi, wj, wi_dx, wj_dx;
- int k, l;
- float Bi[3][3];
- float Bj[3][3];
- GFLOAT Wi[5], Wj[5];
-
- /* Initialize local variables */
- for (k = 0; k < 3; k++) {
- for (l = 0; l < 3; l++) {
- Bi[k][l] = pi->geometry.matrix_E[k][l];
- Bj[k][l] = pj->geometry.matrix_E[k][l];
- }
- }
- Wi[0] = pi->primitives.rho;
- Wi[1] = pi->primitives.v[0];
- Wi[2] = pi->primitives.v[1];
- Wi[3] = pi->primitives.v[2];
- Wi[4] = pi->primitives.P;
- Wj[0] = pj->primitives.rho;
- Wj[1] = pj->primitives.v[0];
- Wj[2] = pj->primitives.v[1];
- Wj[3] = pj->primitives.v[2];
- Wj[4] = pj->primitives.P;
-
- /* Compute kernel of pi. */
- hi_inv = 1.0 / hi;
- xi = r * hi_inv;
- kernel_deval(xi, &wi, &wi_dx);
-
- /* Compute gradients for pi */
- /* there is a sign difference w.r.t. eqn. (6) because of the inverse
- * definition of dx */
- pi->primitives.gradients.rho[0] +=
- (Wi[0] - Wj[0]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.rho[1] +=
- (Wi[0] - Wj[0]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.rho[2] +=
- (Wi[0] - Wj[0]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
-
- pi->primitives.gradients.v[0][0] +=
- (Wi[1] - Wj[1]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.v[0][1] +=
- (Wi[1] - Wj[1]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.v[0][2] +=
- (Wi[1] - Wj[1]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
- pi->primitives.gradients.v[1][0] +=
- (Wi[2] - Wj[2]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.v[1][1] +=
- (Wi[2] - Wj[2]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.v[1][2] +=
- (Wi[2] - Wj[2]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
- pi->primitives.gradients.v[2][0] +=
- (Wi[3] - Wj[3]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.v[2][1] +=
- (Wi[3] - Wj[3]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.v[2][2] +=
- (Wi[3] - Wj[3]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
-
- pi->primitives.gradients.P[0] +=
- (Wi[4] - Wj[4]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.P[1] +=
- (Wi[4] - Wj[4]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.P[2] +=
- (Wi[4] - Wj[4]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
-
- /* basic slope limiter: collect the maximal and the minimal value for the
- * primitive variables among the ngbs */
- pi->primitives.limiter.rho[0] =
- fmin(pj->primitives.rho, pi->primitives.limiter.rho[0]);
- pi->primitives.limiter.rho[1] =
- fmax(pj->primitives.rho, pi->primitives.limiter.rho[1]);
-
- pi->primitives.limiter.v[0][0] =
- fmin(pj->primitives.v[0], pi->primitives.limiter.v[0][0]);
- pi->primitives.limiter.v[0][1] =
- fmax(pj->primitives.v[0], pi->primitives.limiter.v[0][1]);
- pi->primitives.limiter.v[1][0] =
- fmin(pj->primitives.v[1], pi->primitives.limiter.v[1][0]);
- pi->primitives.limiter.v[1][1] =
- fmax(pj->primitives.v[1], pi->primitives.limiter.v[1][1]);
- pi->primitives.limiter.v[2][0] =
- fmin(pj->primitives.v[2], pi->primitives.limiter.v[2][0]);
- pi->primitives.limiter.v[2][1] =
- fmax(pj->primitives.v[2], pi->primitives.limiter.v[2][1]);
-
- pi->primitives.limiter.P[0] =
- fmin(pj->primitives.P, pi->primitives.limiter.P[0]);
- pi->primitives.limiter.P[1] =
- fmax(pj->primitives.P, pi->primitives.limiter.P[1]);
-
- pi->primitives.limiter.maxr = fmax(r, pi->primitives.limiter.maxr);
-
- /* Compute kernel of pj. */
- hj_inv = 1.0 / hj;
- xj = r * hj_inv;
- kernel_deval(xj, &wj, &wj_dx);
-
- /* Compute gradients for pj */
- /* there is no sign difference w.r.t. eqn. (6) because dx is now what we want
- * it to be */
- pj->primitives.gradients.rho[0] +=
- (Wi[0] - Wj[0]) * wj *
- (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
- pj->primitives.gradients.rho[1] +=
- (Wi[0] - Wj[0]) * wj *
- (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
- pj->primitives.gradients.rho[2] +=
- (Wi[0] - Wj[0]) * wj *
- (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
-
- pj->primitives.gradients.v[0][0] +=
- (Wi[1] - Wj[1]) * wj *
- (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
- pj->primitives.gradients.v[0][1] +=
- (Wi[1] - Wj[1]) * wj *
- (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
- pj->primitives.gradients.v[0][2] +=
- (Wi[1] - Wj[1]) * wj *
- (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
- pj->primitives.gradients.v[1][0] +=
- (Wi[2] - Wj[2]) * wj *
- (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
- pj->primitives.gradients.v[1][1] +=
- (Wi[2] - Wj[2]) * wj *
- (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
- pj->primitives.gradients.v[1][2] +=
- (Wi[2] - Wj[2]) * wj *
- (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
- pj->primitives.gradients.v[2][0] +=
- (Wi[3] - Wj[3]) * wj *
- (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
- pj->primitives.gradients.v[2][1] +=
- (Wi[3] - Wj[3]) * wj *
- (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
- pj->primitives.gradients.v[2][2] +=
- (Wi[3] - Wj[3]) * wj *
- (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
-
- pj->primitives.gradients.P[0] +=
- (Wi[4] - Wj[4]) * wj *
- (Bj[0][0] * dx[0] + Bj[0][1] * dx[1] + Bj[0][2] * dx[2]);
- pj->primitives.gradients.P[1] +=
- (Wi[4] - Wj[4]) * wj *
- (Bj[1][0] * dx[0] + Bj[1][1] * dx[1] + Bj[1][2] * dx[2]);
- pj->primitives.gradients.P[2] +=
- (Wi[4] - Wj[4]) * wj *
- (Bj[2][0] * dx[0] + Bj[2][1] * dx[1] + Bj[2][2] * dx[2]);
-
- /* basic slope limiter: collect the maximal and the minimal value for the
- * primitive variables among the ngbs */
- pj->primitives.limiter.rho[0] =
- fmin(pi->primitives.rho, pj->primitives.limiter.rho[0]);
- pj->primitives.limiter.rho[1] =
- fmax(pi->primitives.rho, pj->primitives.limiter.rho[1]);
-
- pj->primitives.limiter.v[0][0] =
- fmin(pi->primitives.v[0], pj->primitives.limiter.v[0][0]);
- pj->primitives.limiter.v[0][1] =
- fmax(pi->primitives.v[0], pj->primitives.limiter.v[0][1]);
- pj->primitives.limiter.v[1][0] =
- fmin(pi->primitives.v[1], pj->primitives.limiter.v[1][0]);
- pj->primitives.limiter.v[1][1] =
- fmax(pi->primitives.v[1], pj->primitives.limiter.v[1][1]);
- pj->primitives.limiter.v[2][0] =
- fmin(pi->primitives.v[2], pj->primitives.limiter.v[2][0]);
- pj->primitives.limiter.v[2][1] =
- fmax(pi->primitives.v[2], pj->primitives.limiter.v[2][1]);
-
- pj->primitives.limiter.P[0] =
- fmin(pi->primitives.P, pj->primitives.limiter.P[0]);
- pj->primitives.limiter.P[1] =
- fmax(pi->primitives.P, pj->primitives.limiter.P[1]);
-
- pj->primitives.limiter.maxr = fmax(r, pj->primitives.limiter.maxr);
-
-#endif
+ hydro_gradients_collect(r2, dx, hi, hj, pi, pj);
}
-__attribute__((always_inline)) INLINE static void
-runner_iact_nonsym_hydro_loop2(float r2, float *dx, float hi, float hj,
- struct part *pi, struct part *pj) {
-
-#ifndef SPH_GRADIENTS
-
- float r = sqrtf(r2);
- float xi;
- float hi_inv;
- float wi, wi_dx;
- int k, l;
- float Bi[3][3];
- GFLOAT Wi[5], Wj[5];
-
- /* Initialize local variables */
- for (k = 0; k < 3; k++) {
- for (l = 0; l < 3; l++) {
- Bi[k][l] = pi->geometry.matrix_E[k][l];
- }
- }
- Wi[0] = pi->primitives.rho;
- Wi[1] = pi->primitives.v[0];
- Wi[2] = pi->primitives.v[1];
- Wi[3] = pi->primitives.v[2];
- Wi[4] = pi->primitives.P;
- Wj[0] = pj->primitives.rho;
- Wj[1] = pj->primitives.v[0];
- Wj[2] = pj->primitives.v[1];
- Wj[3] = pj->primitives.v[2];
- Wj[4] = pj->primitives.P;
-
- /* Compute kernel of pi. */
- hi_inv = 1.0 / hi;
- xi = r * hi_inv;
- kernel_deval(xi, &wi, &wi_dx);
-
- /* Compute gradients for pi */
- /* there is a sign difference w.r.t. eqn. (6) because of the inverse
- * definition of dx */
- pi->primitives.gradients.rho[0] +=
- (Wi[0] - Wj[0]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.rho[1] +=
- (Wi[0] - Wj[0]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.rho[2] +=
- (Wi[0] - Wj[0]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
-
- pi->primitives.gradients.v[0][0] +=
- (Wi[1] - Wj[1]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.v[0][1] +=
- (Wi[1] - Wj[1]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.v[0][2] +=
- (Wi[1] - Wj[1]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
- pi->primitives.gradients.v[1][0] +=
- (Wi[2] - Wj[2]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.v[1][1] +=
- (Wi[2] - Wj[2]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.v[1][2] +=
- (Wi[2] - Wj[2]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
- pi->primitives.gradients.v[2][0] +=
- (Wi[3] - Wj[3]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.v[2][1] +=
- (Wi[3] - Wj[3]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.v[2][2] +=
- (Wi[3] - Wj[3]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
-
- pi->primitives.gradients.P[0] +=
- (Wi[4] - Wj[4]) * wi *
- (Bi[0][0] * dx[0] + Bi[0][1] * dx[1] + Bi[0][2] * dx[2]);
- pi->primitives.gradients.P[1] +=
- (Wi[4] - Wj[4]) * wi *
- (Bi[1][0] * dx[0] + Bi[1][1] * dx[1] + Bi[1][2] * dx[2]);
- pi->primitives.gradients.P[2] +=
- (Wi[4] - Wj[4]) * wi *
- (Bi[2][0] * dx[0] + Bi[2][1] * dx[1] + Bi[2][2] * dx[2]);
-
- /* slope limiter */
- pi->primitives.limiter.rho[0] =
- fmin(pj->primitives.rho, pi->primitives.limiter.rho[0]);
- pi->primitives.limiter.rho[1] =
- fmax(pj->primitives.rho, pi->primitives.limiter.rho[1]);
-
- pi->primitives.limiter.v[0][0] =
- fmin(pj->primitives.v[0], pi->primitives.limiter.v[0][0]);
- pi->primitives.limiter.v[0][1] =
- fmax(pj->primitives.v[0], pi->primitives.limiter.v[0][1]);
- pi->primitives.limiter.v[1][0] =
- fmin(pj->primitives.v[1], pi->primitives.limiter.v[1][0]);
- pi->primitives.limiter.v[1][1] =
- fmax(pj->primitives.v[1], pi->primitives.limiter.v[1][1]);
- pi->primitives.limiter.v[2][0] =
- fmin(pj->primitives.v[2], pi->primitives.limiter.v[2][0]);
- pi->primitives.limiter.v[2][1] =
- fmax(pj->primitives.v[2], pi->primitives.limiter.v[2][1]);
-
- pi->primitives.limiter.P[0] =
- fmin(pj->primitives.P, pi->primitives.limiter.P[0]);
- pi->primitives.limiter.P[1] =
- fmax(pj->primitives.P, pi->primitives.limiter.P[1]);
-
- pi->primitives.limiter.maxr = fmax(r, pi->primitives.limiter.maxr);
+/**
+ * @brief Calculate the gradient interaction between particle i and particle j:
+ * non-symmetric version
+ *
+ * This method wraps around hydro_gradients_nonsym_collect, which can be an
+ * empty method, in which case no gradients are used.
+ *
+ * @param r2 Squared distance between particle i and particle j.
+ * @param dx Distance vector between the particles (dx = pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void runner_iact_nonsym_gradient(
+ float r2, float *dx, float hi, float hj, struct part *pi, struct part *pj) {
-#endif
+ hydro_gradients_nonsym_collect(r2, dx, hi, hj, pi, pj);
}
+/**
+ * @brief Common part of the flux calculation between particle i and j
+ *
+ * Since the only difference between the symmetric and non-symmetric version
+ * of the flux calculation is in the update of the conserved variables at the
+ * very end (which is not done for particle j if mode is 0 and particle j is
+ * active), both runner_iact_force and runner_iact_nonsym_force call this
+ * method, with an appropriate mode.
+ *
+ * This method calculates the surface area of the interface between particle i
+ * and particle j, as well as the interface position and velocity. These are
+ * then used to reconstruct and predict the primitive variables, which are then
+ * fed to a Riemann solver that calculates a flux. This flux is used to update
+ * the conserved variables of particle i or both particles.
+ *
+ * This method also calculates the maximal velocity used to calculate the time
+ * step.
+ *
+ * @param r2 Squared distance between particle i and particle j.
+ * @param dx Distance vector between the particles (dx = pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
__attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
float r2, float *dx, float hi, float hj, struct part *pi, struct part *pj,
int mode) {
float r = sqrtf(r2);
float xi, xj;
- float hi_inv, hi_inv2;
- float hj_inv, hj_inv2;
+ float hi_inv, hi_inv_dim;
+ float hj_inv, hj_inv_dim;
float wi, wj, wi_dx, wj_dx;
int k, l;
float A[3];
@@ -619,13 +201,7 @@ __attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
float xij_i[3], xfac, xijdotdx;
float vmax, dvdotdx;
float vi[3], vj[3], vij[3];
- GFLOAT Wi[5], Wj[5]; //, Whalf[5];
-#ifdef USE_GRADIENTS
- GFLOAT dWi[5], dWj[5];
- float xij_j[3];
-#endif
- // GFLOAT rhoe;
- // GFLOAT flux[5][3];
+ float Wi[5], Wj[5];
float dti, dtj, mindt;
float n_unit[3];
@@ -635,8 +211,8 @@ __attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
Bi[k][l] = pi->geometry.matrix_E[k][l];
Bj[k][l] = pj->geometry.matrix_E[k][l];
}
- vi[k] = pi->v[k]; /* particle velocities */
- vj[k] = pj->v[k];
+ vi[k] = pi->force.v_full[k]; /* particle velocities */
+ vj[k] = pj->force.v_full[k];
}
Vi = pi->geometry.volume;
Vj = pj->geometry.volume;
@@ -650,16 +226,17 @@ __attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
Wj[2] = pj->primitives.v[1];
Wj[3] = pj->primitives.v[2];
Wj[4] = pj->primitives.P;
- dti = pi->ti_end - pi->ti_begin; // MATTHIEU
- dtj = pj->ti_end - pj->ti_begin;
- // if(dti > 1.e-7 || dtj > 1.e-7){
- // message("Timestep too large: %g %g!", dti, dtj);
- // }
+ dti = pi->force.dt;
+ dtj = pj->force.dt;
/* calculate the maximal signal velocity */
- vmax = sqrtf(const_hydro_gamma * Wi[4] / Wi[0]) +
- sqrtf(const_hydro_gamma * Wj[4] / Wj[0]);
+ if (Wi[0] && Wj[0]) {
+ vmax =
+ sqrtf(hydro_gamma * Wi[4] / Wi[0]) + sqrtf(hydro_gamma * Wj[4] / Wj[0]);
+ } else {
+ vmax = 0.0f;
+ }
dvdotdx = (Wi[1] - Wj[1]) * dx[0] + (Wi[2] - Wj[2]) * dx[1] +
(Wi[3] - Wj[3]) * dx[2];
if (dvdotdx > 0.) {
@@ -673,28 +250,44 @@ __attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
/* The flux will be exchanged using the smallest time step of the two
* particles */
mindt = fminf(dti, dtj);
+ dti = mindt;
+ dtj = mindt;
/* Compute kernel of pi. */
hi_inv = 1.0 / hi;
- hi_inv2 = hi_inv * hi_inv;
+ hi_inv_dim = pow_dimension(hi_inv);
xi = r * hi_inv;
kernel_deval(xi, &wi, &wi_dx);
/* Compute kernel of pj. */
hj_inv = 1.0 / hj;
- hj_inv2 = hj_inv * hj_inv;
+ hj_inv_dim = pow_dimension(hj_inv);
xj = r * hj_inv;
kernel_deval(xj, &wj, &wj_dx);
+ /* Compute h_dt. We are going to use an SPH-like estimate of div_v for that */
+ float dvdr = (pi->v[0] - pj->v[0]) * dx[0] + (pi->v[1] - pj->v[1]) * dx[1] +
+ (pi->v[2] - pj->v[2]) * dx[2];
+ float ri = 1.0f / r;
+ float hidp1 = pow_dimension_plus_one(hi_inv);
+ float hjdp1 = pow_dimension_plus_one(hj_inv);
+ float wi_dr = hidp1 * wi_dx;
+ float wj_dr = hjdp1 * wj_dx;
+ dvdr *= ri;
+ pi->force.h_dt -= pj->conserved.mass * dvdr / pj->primitives.rho * wi_dr;
+ if (mode == 1) {
+ pj->force.h_dt -= pi->conserved.mass * dvdr / pi->primitives.rho * wj_dr;
+ }
+
/* Compute area */
/* eqn. (7) */
Anorm = 0.0f;
for (k = 0; k < 3; k++) {
/* we add a minus sign since dx is pi->x - pj->x */
A[k] = -Vi * (Bi[k][0] * dx[0] + Bi[k][1] * dx[1] + Bi[k][2] * dx[2]) * wi *
- hi_inv * hi_inv2 -
+ hi_inv_dim -
Vj * (Bj[k][0] * dx[0] + Bj[k][1] * dx[1] + Bj[k][2] * dx[2]) * wj *
- hj_inv * hj_inv2;
+ hj_inv_dim;
Anorm += A[k] * A[k];
}
@@ -708,13 +301,6 @@ __attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
Anorm = sqrtf(Anorm);
for (k = 0; k < 3; k++) n_unit[k] = A[k] / Anorm;
-#ifdef PRINT_ID
- if (pi->id == PRINT_ID || pj->id == PRINT_ID) {
- printf("pi: %g %g %g\npj: %g %g %g\nA = %g %g %g\n", pi->x[0], pi->x[1],
- pi->x[2], pj->x[0], pj->x[1], pj->x[2], A[0], A[1], A[2]);
- }
-#endif
-
/* Compute interface position (relative to pi, since we don't need the actual
* position) */
/* eqn. (8) */
@@ -746,192 +332,7 @@ __attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
Wj[2] -= vij[1];
Wj[3] -= vij[2];
-#ifdef USE_GRADIENTS
- /* perform gradient reconstruction in space and time */
- /* space */
- /* Compute interface position (relative to pj, since we don't need the actual
- * position) */
- /* eqn. (8) */
- xfac = hj / (hi + hj);
- for (k = 0; k < 3; k++) xij_j[k] = xfac * dx[k];
-
- dWi[0] = pi->primitives.gradients.rho[0] * xij_i[0] +
- pi->primitives.gradients.rho[1] * xij_i[1] +
- pi->primitives.gradients.rho[2] * xij_i[2];
- dWi[1] = pi->primitives.gradients.v[0][0] * xij_i[0] +
- pi->primitives.gradients.v[0][1] * xij_i[1] +
- pi->primitives.gradients.v[0][2] * xij_i[2];
- dWi[2] = pi->primitives.gradients.v[1][0] * xij_i[0] +
- pi->primitives.gradients.v[1][1] * xij_i[1] +
- pi->primitives.gradients.v[1][2] * xij_i[2];
- dWi[3] = pi->primitives.gradients.v[2][0] * xij_i[0] +
- pi->primitives.gradients.v[2][1] * xij_i[1] +
- pi->primitives.gradients.v[2][2] * xij_i[2];
- dWi[4] = pi->primitives.gradients.P[0] * xij_i[0] +
- pi->primitives.gradients.P[1] * xij_i[1] +
- pi->primitives.gradients.P[2] * xij_i[2];
-
- dWj[0] = pj->primitives.gradients.rho[0] * xij_j[0] +
- pj->primitives.gradients.rho[1] * xij_j[1] +
- pj->primitives.gradients.rho[2] * xij_j[2];
- dWj[1] = pj->primitives.gradients.v[0][0] * xij_j[0] +
- pj->primitives.gradients.v[0][1] * xij_j[1] +
- pj->primitives.gradients.v[0][2] * xij_j[2];
- dWj[2] = pj->primitives.gradients.v[1][0] * xij_j[0] +
- pj->primitives.gradients.v[1][1] * xij_j[1] +
- pj->primitives.gradients.v[1][2] * xij_j[2];
- dWj[3] = pj->primitives.gradients.v[2][0] * xij_j[0] +
- pj->primitives.gradients.v[2][1] * xij_j[1] +
- pj->primitives.gradients.v[2][2] * xij_j[2];
- dWj[4] = pj->primitives.gradients.P[0] * xij_j[0] +
- pj->primitives.gradients.P[1] * xij_j[1] +
- pj->primitives.gradients.P[2] * xij_j[2];
-
-#ifdef PER_FACE_LIMITER
-
- float xij_i_norm;
- GFLOAT phi_i, phi_j;
- GFLOAT delta1, delta2;
- GFLOAT phiminus, phiplus;
- GFLOAT phimin, phimax;
- GFLOAT phibar;
- /* free parameters, values from Hopkins */
- GFLOAT psi1 = 0.5, psi2 = 0.25;
- GFLOAT phi_mid0, phi_mid;
-
- for (k = 0; k < 10; k++) {
- if (k < 5) {
- phi_i = Wi[k];
- phi_j = Wj[k];
- phi_mid0 = Wi[k] + dWi[k];
- xij_i_norm = sqrtf(xij_i[0] * xij_i[0] + xij_i[1] * xij_i[1] +
- xij_i[2] * xij_i[2]);
- } else {
- phi_i = Wj[k - 5];
- phi_j = Wi[k - 5];
- phi_mid0 = Wj[k - 5] + dWj[k - 5];
- xij_i_norm = sqrtf(xij_j[0] * xij_j[0] + xij_j[1] * xij_j[1] +
- xij_j[2] * xij_j[2]);
- }
-
- delta1 = psi1 * fabs(phi_i - phi_j);
- delta2 = psi2 * fabs(phi_i - phi_j);
-
- phimin = fmin(phi_i, phi_j);
- phimax = fmax(phi_i, phi_j);
-
- phibar = phi_i + xij_i_norm / r * (phi_j - phi_i);
-
- /* if sign(phimax+delta1) == sign(phimax) */
- if ((phimax + delta1) * phimax > 0.0f) {
- phiplus = phimax + delta1;
- } else {
- phiplus = phimax / (1.0f + delta1 / fabs(phimax));
- }
-
- /* if sign(phimin-delta1) == sign(phimin) */
- if ((phimin - delta1) * phimin > 0.0f) {
- phiminus = phimin - delta1;
- } else {
- phiminus = phimin / (1.0f + delta1 / fabs(phimin));
- }
-
- if (phi_i == phi_j) {
- phi_mid = phi_i;
- } else {
- if (phi_i < phi_j) {
- phi_mid = fmax(phiminus, fmin(phibar + delta2, phi_mid0));
- } else {
- phi_mid = fmin(phiplus, fmax(phibar - delta2, phi_mid0));
- }
- }
-
- if (k < 5) {
- dWi[k] = phi_mid - phi_i;
- } else {
- dWj[k - 5] = phi_mid - phi_i;
- }
- }
-
-#endif
-
- // printf("dWL: %g %g %g %g %g\n", dWi[0], dWi[1], dWi[2], dWi[3], dWi[4]);
- // printf("dWR: %g %g %g %g %g\n", dWj[0], dWj[1], dWj[2], dWj[3], dWj[4]);
-
- /* time */
- dWi[0] -= 0.5 * mindt * (Wi[1] * pi->primitives.gradients.rho[0] +
- Wi[2] * pi->primitives.gradients.rho[1] +
- Wi[3] * pi->primitives.gradients.rho[2] +
- Wi[0] * (pi->primitives.gradients.v[0][0] +
- pi->primitives.gradients.v[1][1] +
- pi->primitives.gradients.v[2][2]));
- dWi[1] -= 0.5 * mindt * (Wi[1] * pi->primitives.gradients.v[0][0] +
- Wi[2] * pi->primitives.gradients.v[0][1] +
- Wi[3] * pi->primitives.gradients.v[0][2] +
- pi->primitives.gradients.P[0] / Wi[0]);
- dWi[2] -= 0.5 * mindt * (Wi[1] * pi->primitives.gradients.v[1][0] +
- Wi[2] * pi->primitives.gradients.v[1][1] +
- Wi[3] * pi->primitives.gradients.v[1][2] +
- pi->primitives.gradients.P[1] / Wi[0]);
- dWi[3] -= 0.5 * mindt * (Wi[1] * pi->primitives.gradients.v[2][0] +
- Wi[2] * pi->primitives.gradients.v[2][1] +
- Wi[3] * pi->primitives.gradients.v[2][2] +
- pi->primitives.gradients.P[2] / Wi[0]);
- dWi[4] -= 0.5 * mindt *
- (Wi[1] * pi->primitives.gradients.P[0] +
- Wi[2] * pi->primitives.gradients.P[1] +
- Wi[3] * pi->primitives.gradients.P[2] +
- const_hydro_gamma * Wi[4] * (pi->primitives.gradients.v[0][0] +
- pi->primitives.gradients.v[1][1] +
- pi->primitives.gradients.v[2][2]));
-
- dWj[0] -= 0.5 * mindt * (Wj[1] * pj->primitives.gradients.rho[0] +
- Wj[2] * pj->primitives.gradients.rho[1] +
- Wj[3] * pj->primitives.gradients.rho[2] +
- Wj[0] * (pj->primitives.gradients.v[0][0] +
- pj->primitives.gradients.v[1][1] +
- pj->primitives.gradients.v[2][2]));
- dWj[1] -= 0.5 * mindt * (Wj[1] * pj->primitives.gradients.v[0][0] +
- Wj[2] * pj->primitives.gradients.v[0][1] +
- Wj[3] * pj->primitives.gradients.v[0][2] +
- pj->primitives.gradients.P[0] / Wj[0]);
- dWj[2] -= 0.5 * mindt * (Wj[1] * pj->primitives.gradients.v[1][0] +
- Wj[2] * pj->primitives.gradients.v[1][1] +
- Wj[3] * pj->primitives.gradients.v[1][2] +
- pj->primitives.gradients.P[1] / Wj[0]);
- dWj[3] -= 0.5 * mindt * (Wj[1] * pj->primitives.gradients.v[2][0] +
- Wj[2] * pj->primitives.gradients.v[2][1] +
- Wj[3] * pj->primitives.gradients.v[2][2] +
- pj->primitives.gradients.P[2] / Wj[0]);
- dWj[4] -= 0.5 * mindt *
- (Wj[1] * pj->primitives.gradients.P[0] +
- Wj[2] * pj->primitives.gradients.P[1] +
- Wj[3] * pj->primitives.gradients.P[2] +
- const_hydro_gamma * Wj[4] * (pj->primitives.gradients.v[0][0] +
- pj->primitives.gradients.v[1][1] +
- pj->primitives.gradients.v[2][2]));
-
- // printf("WL: %g %g %g %g %g\n", Wi[0], Wi[1], Wi[2], Wi[3], Wi[4]);
- // printf("WR: %g %g %g %g %g\n", Wj[0], Wj[1], Wj[2], Wj[3], Wj[4]);
-
- // printf("dWL: %g %g %g %g %g\n", dWi[0], dWi[1], dWi[2], dWi[3], dWi[4]);
- // printf("dWR: %g %g %g %g %g\n", dWj[0], dWj[1], dWj[2], dWj[3], dWj[4]);
-
- Wi[0] += dWi[0];
- Wi[1] += dWi[1];
- Wi[2] += dWi[2];
- Wi[3] += dWi[3];
- Wi[4] += dWi[4];
-
- Wj[0] += dWj[0];
- Wj[1] += dWj[1];
- Wj[2] += dWj[2];
- Wj[3] += dWj[3];
- Wj[4] += dWj[4];
-#endif
-
- /* apply slope limiter interface by interface */
- /* ... to be done ... */
+ hydro_gradients_predict(pi, pj, hi, hj, dx, r, xij_i, Wi, Wj, mindt);
/* we don't need to rotate, we can use the unit vector in the Riemann problem
* itself (see GIZMO) */
@@ -940,7 +341,9 @@ __attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
printf("mindt: %g\n", mindt);
printf("WL: %g %g %g %g %g\n", pi->primitives.rho, pi->primitives.v[0],
pi->primitives.v[1], pi->primitives.v[2], pi->primitives.P);
+#ifdef USE_GRADIENTS
printf("dWL: %g %g %g %g %g\n", dWi[0], dWi[1], dWi[2], dWi[3], dWi[4]);
+#endif
printf("gradWL[0]: %g %g %g\n", pi->primitives.gradients.rho[0],
pi->primitives.gradients.rho[1], pi->primitives.gradients.rho[2]);
printf("gradWL[1]: %g %g %g\n", pi->primitives.gradients.v[0][0],
@@ -954,7 +357,9 @@ __attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
printf("WL': %g %g %g %g %g\n", Wi[0], Wi[1], Wi[2], Wi[3], Wi[4]);
printf("WR: %g %g %g %g %g\n", pj->primitives.rho, pj->primitives.v[0],
pj->primitives.v[1], pj->primitives.v[2], pj->primitives.P);
+#ifdef USE_GRADIENTS
printf("dWR: %g %g %g %g %g\n", dWj[0], dWj[1], dWj[2], dWj[3], dWj[4]);
+#endif
printf("gradWR[0]: %g %g %g\n", pj->primitives.gradients.rho[0],
pj->primitives.gradients.rho[1], pj->primitives.gradients.rho[2]);
printf("gradWR[1]: %g %g %g\n", pj->primitives.gradients.v[0][0],
@@ -969,61 +374,133 @@ __attribute__((always_inline)) INLINE static void runner_iact_fluxes_common(
error("Negative density or pressure!\n");
}
- GFLOAT totflux[5];
+ float totflux[5];
riemann_solve_for_flux(Wi, Wj, n_unit, vij, totflux);
+ /* Store mass flux */
+ float mflux = dti * Anorm * totflux[0];
+ pi->gravity.mflux[0] += mflux * dx[0];
+ pi->gravity.mflux[1] += mflux * dx[1];
+ pi->gravity.mflux[2] += mflux * dx[2];
+
/* Update conserved variables */
/* eqn. (16) */
- pi->conserved.mass -= mindt * Anorm * totflux[0];
- pi->conserved.momentum[0] -= mindt * Anorm * totflux[1];
- pi->conserved.momentum[1] -= mindt * Anorm * totflux[2];
- pi->conserved.momentum[2] -= mindt * Anorm * totflux[3];
- pi->conserved.energy -= mindt * Anorm * totflux[4];
-
-#ifdef THERMAL_ENERGY
- float ekin = 0.5 * (pi->primitives.v[0] * pi->primitives.v[0] +
- pi->primitives.v[1] * pi->primitives.v[1] +
- pi->primitives.v[2] * pi->primitives.v[2]);
- pi->conserved.energy += mindt * Anorm * totflux[1] * pi->primitives.v[0];
- pi->conserved.energy += mindt * Anorm * totflux[2] * pi->primitives.v[1];
- pi->conserved.energy += mindt * Anorm * totflux[3] * pi->primitives.v[2];
- pi->conserved.energy -= mindt * Anorm * totflux[0] * ekin;
-#endif
-
- /* the non symmetric version is never called when using mindt, whether this
- * piece of code
- * should always be executed or only in the symmetric case is currently
- * unclear */
- if (mode == 1) {
- pj->conserved.mass += mindt * Anorm * totflux[0];
- pj->conserved.momentum[0] += mindt * Anorm * totflux[1];
- pj->conserved.momentum[1] += mindt * Anorm * totflux[2];
- pj->conserved.momentum[2] += mindt * Anorm * totflux[3];
- pj->conserved.energy += mindt * Anorm * totflux[4];
-
-#ifdef THERMAL_ENERGY
- ekin = 0.5 * (pj->primitives.v[0] * pj->primitives.v[0] +
- pj->primitives.v[1] * pj->primitives.v[1] +
- pj->primitives.v[2] * pj->primitives.v[2]);
- pj->conserved.energy -= mindt * Anorm * totflux[1] * pj->primitives.v[0];
- pj->conserved.energy -= mindt * Anorm * totflux[2] * pj->primitives.v[1];
- pj->conserved.energy -= mindt * Anorm * totflux[3] * pj->primitives.v[2];
- pj->conserved.energy += mindt * Anorm * totflux[0] * ekin;
-#endif
+ pi->conserved.flux.mass -= dti * Anorm * totflux[0];
+ pi->conserved.flux.momentum[0] -= dti * Anorm * totflux[1];
+ pi->conserved.flux.momentum[1] -= dti * Anorm * totflux[2];
+ pi->conserved.flux.momentum[2] -= dti * Anorm * totflux[3];
+ pi->conserved.flux.energy -= dti * Anorm * totflux[4];
+
+ float ekin = 0.5f * (pi->primitives.v[0] * pi->primitives.v[0] +
+ pi->primitives.v[1] * pi->primitives.v[1] +
+ pi->primitives.v[2] * pi->primitives.v[2]);
+ pi->conserved.flux.energy += dti * Anorm * totflux[1] * pi->primitives.v[0];
+ pi->conserved.flux.energy += dti * Anorm * totflux[2] * pi->primitives.v[1];
+ pi->conserved.flux.energy += dti * Anorm * totflux[3] * pi->primitives.v[2];
+ pi->conserved.flux.energy -= dti * Anorm * totflux[0] * ekin;
+
+ /* here is how it works:
+ Mode will only be 1 if both particles are ACTIVE and they are in the same
+ cell. In this case, this method IS the flux calculation for particle j, and
+ we HAVE TO UPDATE it.
+ Mode 0 can mean several things: it can mean that particle j is INACTIVE, in
+ which case we NEED TO UPDATE it, since otherwise the flux is lost from the
+ system and the conserved variable is not conserved.
+ It can also mean that particle j sits in another cell and is ACTIVE. In
+ this case, the flux exchange for particle j is done TWICE and we SHOULD NOT
+ UPDATE particle j.
+ ==> we update particle j if (MODE IS 1) OR (j IS INACTIVE)
+ */
+ if (mode == 1 || pj->ti_end > pi->ti_end) {
+ /* Store mass flux */
+ mflux = dtj * Anorm * totflux[0];
+ pj->gravity.mflux[0] -= mflux * dx[0];
+ pj->gravity.mflux[1] -= mflux * dx[1];
+ pj->gravity.mflux[2] -= mflux * dx[2];
+
+ pj->conserved.flux.mass += dtj * Anorm * totflux[0];
+ pj->conserved.flux.momentum[0] += dtj * Anorm * totflux[1];
+ pj->conserved.flux.momentum[1] += dtj * Anorm * totflux[2];
+ pj->conserved.flux.momentum[2] += dtj * Anorm * totflux[3];
+ pj->conserved.flux.energy += dtj * Anorm * totflux[4];
+
+ ekin = 0.5f * (pj->primitives.v[0] * pj->primitives.v[0] +
+ pj->primitives.v[1] * pj->primitives.v[1] +
+ pj->primitives.v[2] * pj->primitives.v[2]);
+ pj->conserved.flux.energy -= dtj * Anorm * totflux[1] * pj->primitives.v[0];
+ pj->conserved.flux.energy -= dtj * Anorm * totflux[2] * pj->primitives.v[1];
+ pj->conserved.flux.energy -= dtj * Anorm * totflux[3] * pj->primitives.v[2];
+ pj->conserved.flux.energy += dtj * Anorm * totflux[0] * ekin;
}
}
-/* this corresponds to task_subtype_fluxes */
-__attribute__((always_inline)) INLINE static void runner_iact_hydro_loop3(
+/**
+ * @brief Flux calculation between particle i and particle j
+ *
+ * This method calls runner_iact_fluxes_common with mode 1.
+ *
+ * @param r2 Squared distance between particle i and particle j.
+ * @param dx Distance vector between the particles (dx = pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void runner_iact_force(
float r2, float *dx, float hi, float hj, struct part *pi, struct part *pj) {
runner_iact_fluxes_common(r2, dx, hi, hj, pi, pj, 1);
}
-/* this corresponds to task_subtype_fluxes */
-__attribute__((always_inline)) INLINE static void
-runner_iact_nonsym_hydro_loop3(float r2, float *dx, float hi, float hj,
- struct part *pi, struct part *pj) {
+/**
+ * @brief Flux calculation between particle i and particle j: non-symmetric
+ * version
+ *
+ * This method calls runner_iact_fluxes_common with mode 0.
+ *
+ * @param r2 Squared distance between particle i and particle j.
+ * @param dx Distance vector between the particles (dx = pi->x - pj->x).
+ * @param hi Smoothing length of particle i.
+ * @param hj Smoothing length of particle j.
+ * @param pi Particle i.
+ * @param pj Particle j.
+ */
+__attribute__((always_inline)) INLINE static void runner_iact_nonsym_force(
+ float r2, float *dx, float hi, float hj, struct part *pi, struct part *pj) {
runner_iact_fluxes_common(r2, dx, hi, hj, pi, pj, 0);
}
+
+//// EMPTY VECTORIZED VERSIONS (gradients methods are missing...)
+
+__attribute__((always_inline)) INLINE static void runner_iact_vec_density(
+ float *R2, float *Dx, float *Hi, float *Hj, struct part **pi,
+ struct part **pj) {
+ error(
+ "Vectorised versions of the Gizmo interaction functions do not exist "
+ "yet!");
+}
+
+__attribute__((always_inline)) INLINE static void
+runner_iact_nonsym_vec_density(float *R2, float *Dx, float *Hi, float *Hj,
+ struct part **pi, struct part **pj) {
+ error(
+ "Vectorised versions of the Gizmo interaction functions do not exist "
+ "yet!");
+}
+
+__attribute__((always_inline)) INLINE static void runner_iact_vec_force(
+ float *R2, float *Dx, float *Hi, float *Hj, struct part **pi,
+ struct part **pj) {
+ error(
+ "Vectorised versions of the Gizmo interaction functions do not exist "
+ "yet!");
+}
+
+__attribute__((always_inline)) INLINE static void runner_iact_nonsym_vec_force(
+ float *R2, float *Dx, float *Hi, float *Hj, struct part **pi,
+ struct part **pj) {
+ error(
+ "Vectorised versions of the Gizmo interaction functions do not exist "
+ "yet!");
+}
diff --git a/src/hydro/Gizmo/hydro_io.h b/src/hydro/Gizmo/hydro_io.h
index 3c51653d994bd9f01864bcc24c6886eba25d1d05..e5f221ae4345dc519a50d332131ecf296f318338 100644
--- a/src/hydro/Gizmo/hydro_io.h
+++ b/src/hydro/Gizmo/hydro_io.h
@@ -1,6 +1,6 @@
/*******************************************************************************
* This file is part of SWIFT.
- * Coypright (c) 2016 Matthieu Schaller (matthieu.schaller@durham.ac.uk)
+ * Coypright (c) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com)
*
* 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
@@ -17,77 +17,124 @@
*
******************************************************************************/
+#include "adiabatic_index.h"
+#include "hydro_gradients.h"
+#include "hydro_slope_limiters.h"
+#include "io_properties.h"
+#include "riemann.h"
+
/**
- * @brief Reads the different particles to the HDF5 file
+ * @brief Specifies which particle fields to read from a dataset
*
- * @param h_grp The HDF5 group in which to read the arrays.
- * @param N The number of particles on that MPI rank.
- * @param N_total The total number of particles (only used in MPI mode)
- * @param offset The offset of the particles for this MPI rank (only used in MPI
- *mode)
- * @param parts The particle array
+ * @param parts The particle array.
+ * @param list The list of i/o properties to read.
+ * @param num_fields The number of i/o fields to read.
+ */
+void hydro_read_particles(struct part* parts, struct io_props* list,
+ int* num_fields) {
+
+ *num_fields = 8;
+
+ /* List what we want to read */
+ list[0] = io_make_input_field("Coordinates", DOUBLE, 3, COMPULSORY,
+ UNIT_CONV_LENGTH, parts, x);
+ list[1] = io_make_input_field("Velocities", FLOAT, 3, COMPULSORY,
+ UNIT_CONV_SPEED, parts, v);
+ list[2] = io_make_input_field("Masses", FLOAT, 1, COMPULSORY, UNIT_CONV_MASS,
+ parts, conserved.mass);
+ list[3] = io_make_input_field("SmoothingLength", FLOAT, 1, COMPULSORY,
+ UNIT_CONV_LENGTH, parts, h);
+ list[4] = io_make_input_field("InternalEnergy", FLOAT, 1, COMPULSORY,
+ UNIT_CONV_ENERGY_PER_UNIT_MASS, parts,
+ conserved.energy);
+ list[5] = io_make_input_field("ParticleIDs", ULONGLONG, 1, COMPULSORY,
+ UNIT_CONV_NO_UNITS, parts, id);
+ list[6] = io_make_input_field("Accelerations", FLOAT, 3, OPTIONAL,
+ UNIT_CONV_ACCELERATION, parts, a_hydro);
+ list[7] = io_make_input_field("Density", FLOAT, 1, OPTIONAL,
+ UNIT_CONV_DENSITY, parts, primitives.rho);
+}
+
+/**
+ * @brief Get the internal energy of a particle
*
+ * @param e #engine.
+ * @param p Particle.
+ * @return Internal energy of the particle
*/
-__attribute__((always_inline)) INLINE static void hydro_read_particles(
- hid_t h_grp, int N, long long N_total, long long offset,
- struct part* parts) {
-
- /* Read arrays */
- readArray(h_grp, "Coordinates", DOUBLE, N, 3, parts, N_total, offset, x,
- COMPULSORY);
- readArray(h_grp, "Velocities", FLOAT, N, 3, parts, N_total, offset, v,
- COMPULSORY);
- readArray(h_grp, "Masses", FLOAT, N, 1, parts, N_total, offset,
- conserved.mass, COMPULSORY);
- readArray(h_grp, "SmoothingLength", FLOAT, N, 1, parts, N_total, offset, h,
- COMPULSORY);
- readArray(h_grp, "InternalEnergy", FLOAT, N, 1, parts, N_total, offset,
- primitives.P, COMPULSORY);
- readArray(h_grp, "ParticleIDs", ULONGLONG, N, 1, parts, N_total, offset, id,
- COMPULSORY);
- readArray(h_grp, "Acceleration", FLOAT, N, 3, parts, N_total, offset, a_hydro,
- OPTIONAL);
- readArray(h_grp, "Density", FLOAT, N, 1, parts, N_total, offset,
- primitives.rho, OPTIONAL);
+float convert_u(struct engine* e, struct part* p) {
+ return p->primitives.P / hydro_gamma_minus_one / p->primitives.rho;
}
/**
- * @brief Writes the different particles to the HDF5 file
+ * @brief Get the entropic function of a particle
*
- * @param h_grp The HDF5 group in which to write the arrays.
- * @param fileName The name of the file (unsued in MPI mode).
- * @param xmfFile The XMF file to write to (unused in MPI mode).
- * @param N The number of particles on that MPI rank.
- * @param N_total The total number of particles (only used in MPI mode)
- * @param mpi_rank The MPI rank of this node (only used in MPI mode)
- * @param offset The offset of the particles for this MPI rank (only used in MPI
- *mode)
- * @param parts The particle array
- * @param us The unit system to use
+ * @param e #engine.
+ * @param p Particle.
+ * @return Entropic function of the particle
+ */
+float convert_A(struct engine* e, struct part* p) {
+ return p->primitives.P / pow_gamma(p->primitives.rho);
+}
+
+/**
+ * @brief Get the total energy of a particle
*
+ * @param e #engine.
+ * @param p Particle.
+ * @return Total energy of the particle
*/
-__attribute__((always_inline)) INLINE static void hydro_write_particles(
- hid_t h_grp, char* fileName, FILE* xmfFile, int N, long long N_total,
- int mpi_rank, long long offset, struct part* parts, struct UnitSystem* us) {
-
- /* Write arrays */
- writeArray(h_grp, fileName, xmfFile, "Coordinates", DOUBLE, N, 3, parts,
- N_total, mpi_rank, offset, x, us, UNIT_CONV_LENGTH);
- writeArray(h_grp, fileName, xmfFile, "Velocities", FLOAT, N, 3, parts,
- N_total, mpi_rank, offset, v, us, UNIT_CONV_SPEED);
- writeArray(h_grp, fileName, xmfFile, "Masses", FLOAT, N, 1, parts, N_total,
- mpi_rank, offset, conserved.mass, us, UNIT_CONV_MASS);
- writeArray(h_grp, fileName, xmfFile, "SmoothingLength", FLOAT, N, 1, parts,
- N_total, mpi_rank, offset, h, us, UNIT_CONV_LENGTH);
- writeArray(h_grp, fileName, xmfFile, "InternalEnergy", FLOAT, N, 1, parts,
- N_total, mpi_rank, offset, primitives.P, us,
- UNIT_CONV_ENTROPY_PER_UNIT_MASS);
- writeArray(h_grp, fileName, xmfFile, "ParticleIDs", ULONGLONG, N, 1, parts,
- N_total, mpi_rank, offset, id, us, UNIT_CONV_NO_UNITS);
- writeArray(h_grp, fileName, xmfFile, "Acceleration", FLOAT, N, 3, parts,
- N_total, mpi_rank, offset, a_hydro, us, UNIT_CONV_ACCELERATION);
- writeArray(h_grp, fileName, xmfFile, "Density", FLOAT, N, 1, parts, N_total,
- mpi_rank, offset, primitives.rho, us, UNIT_CONV_DENSITY);
+float convert_Etot(struct engine* e, struct part* p) {
+ float momentum2;
+
+ momentum2 = p->conserved.momentum[0] * p->conserved.momentum[0] +
+ p->conserved.momentum[1] * p->conserved.momentum[1] +
+ p->conserved.momentum[2] * p->conserved.momentum[2];
+
+ return p->conserved.energy + 0.5f * momentum2 / p->conserved.mass;
+}
+
+/**
+ * @brief Specifies which particle fields to write to a dataset
+ *
+ * @param parts The particle array.
+ * @param list The list of i/o properties to write.
+ * @param num_fields The number of i/o fields to write.
+ */
+void hydro_write_particles(struct part* parts, struct io_props* list,
+ int* num_fields) {
+
+ *num_fields = 13;
+
+ /* List what we want to write */
+ list[0] = io_make_output_field("Coordinates", DOUBLE, 3, UNIT_CONV_LENGTH,
+ parts, x);
+ list[1] = io_make_output_field("Velocities", FLOAT, 3, UNIT_CONV_SPEED, parts,
+ primitives.v);
+ list[2] = io_make_output_field("Masses", FLOAT, 1, UNIT_CONV_MASS, parts,
+ conserved.mass);
+ list[3] = io_make_output_field("SmoothingLength", FLOAT, 1, UNIT_CONV_LENGTH,
+ parts, h);
+ list[4] = io_make_output_field_convert_part("InternalEnergy", FLOAT, 1,
+ UNIT_CONV_ENERGY_PER_UNIT_MASS,
+ parts, primitives.P, convert_u);
+ list[5] = io_make_output_field("ParticleIDs", ULONGLONG, 1,
+ UNIT_CONV_NO_UNITS, parts, id);
+ list[6] = io_make_output_field("Acceleration", FLOAT, 3,
+ UNIT_CONV_ACCELERATION, parts, a_hydro);
+ list[7] = io_make_output_field("Density", FLOAT, 1, UNIT_CONV_DENSITY, parts,
+ primitives.rho);
+ list[8] = io_make_output_field("Volume", FLOAT, 1, UNIT_CONV_VOLUME, parts,
+ geometry.volume);
+ list[9] = io_make_output_field("GradDensity", FLOAT, 3, UNIT_CONV_DENSITY,
+ parts, primitives.gradients.rho);
+ list[10] = io_make_output_field_convert_part(
+ "Entropy", FLOAT, 1, UNIT_CONV_ENTROPY, parts, primitives.P, convert_A);
+ list[11] = io_make_output_field("Pressure", FLOAT, 1, UNIT_CONV_PRESSURE,
+ parts, primitives.P);
+ list[12] =
+ io_make_output_field_convert_part("TotEnergy", FLOAT, 1, UNIT_CONV_ENERGY,
+ parts, conserved.energy, convert_Etot);
}
/**
@@ -95,26 +142,24 @@ __attribute__((always_inline)) INLINE static void hydro_write_particles(
* @param h_grpsph The HDF5 group in which to write
*/
void writeSPHflavour(hid_t h_grpsph) {
+ /* Gradient information */
+ writeAttribute_s(h_grpsph, "Gradient reconstruction model",
+ HYDRO_GRADIENT_IMPLEMENTATION);
+
+ /* Slope limiter information */
+ writeAttribute_s(h_grpsph, "Cell wide slope limiter model",
+ HYDRO_SLOPE_LIMITER_CELL_IMPLEMENTATION);
+ writeAttribute_s(h_grpsph, "Piecewise slope limiter model",
+ HYDRO_SLOPE_LIMITER_FACE_IMPLEMENTATION);
- /* Kernel function description */
- writeAttribute_s(h_grpsph, "Kernel", kernel_name);
- writeAttribute_f(h_grpsph, "Kernel eta", const_eta_kernel);
- writeAttribute_f(h_grpsph, "Weighted N_ngb", kernel_nwneigh);
- writeAttribute_f(h_grpsph, "Delta N_ngb", const_delta_nwneigh);
- writeAttribute_f(h_grpsph, "Hydro gamma", const_hydro_gamma);
-
- /* Viscosity and thermal conduction */
- writeAttribute_s(h_grpsph, "Thermal Conductivity Model",
- "(No treatment) Legacy Gadget-2 as in Springel (2005)");
- writeAttribute_s(h_grpsph, "Viscosity Model",
- "Legacy Gadget-2 as in Springel (2005)");
- writeAttribute_f(h_grpsph, "Viscosity alpha", const_viscosity_alpha);
- writeAttribute_f(h_grpsph, "Viscosity beta", 3.f);
-
- /* Time integration properties */
- writeAttribute_f(h_grpsph, "CFL parameter", const_cfl);
- writeAttribute_f(h_grpsph, "Maximal ln(Delta h) change over dt",
- const_ln_max_h_change);
- writeAttribute_f(h_grpsph, "Maximal Delta h change over dt",
- exp(const_ln_max_h_change));
+ /* Riemann solver information */
+ writeAttribute_s(h_grpsph, "Riemann solver type",
+ RIEMANN_SOLVER_IMPLEMENTATION);
}
+
+/**
+ * @brief Are we writing entropy in the internal energy field ?
+ *
+ * @return 1 if entropy is in 'internal energy', 0 otherwise.
+ */
+int writeEntropyFlag() { return 0; }
diff --git a/src/hydro/Gizmo/hydro_part.h b/src/hydro/Gizmo/hydro_part.h
index 9e5f32f758248d1d1616f4556c81fc8e0b52e83b..d425294671d4bc172f45c928c2290f8cfa8e093c 100644
--- a/src/hydro/Gizmo/hydro_part.h
+++ b/src/hydro/Gizmo/hydro_part.h
@@ -1,8 +1,6 @@
/*******************************************************************************
* This file is part of SWIFT.
- * Coypright (c) 2012 Pedro Gonnet (pedro.gonnet@durham.ac.uk)
- * Matthieu Schaller (matthieu.schaller@durham.ac.uk)
- * Bert Vandenbroucke (bert.vandenbroucke@ugent.be)
+ * Coypright (c) 2014 Bert Vandenbroucke (bert.vandenbroucke@ugent.be)
*
* 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
@@ -19,26 +17,15 @@
*
******************************************************************************/
-/* Some standard headers. */
-#include <stdlib.h>
-
-#define GFLOAT float
-
/* Extra particle data not needed during the computation. */
struct xpart {
- /* Old position, at last tree rebuild. */
- double x_old[3];
+ /* Offset between current position and position at last tree rebuild. */
+ float x_diff[3];
/* Velocity at the last full step. */
float v_full[3];
- /* Entropy at the half-step. */
- float u_hdt;
-
- /* Old density. */
- float omega;
-
} __attribute__((aligned(xpart_align)));
/* Data of a single particle. */
@@ -47,18 +34,13 @@ struct part {
/* Particle position. */
double x[3];
- /* Particle velocity. */
+ /* Particle predicted velocity. */
float v[3];
/* Particle acceleration. */
float a_hydro[3];
- float mass; // MATTHIEU
- float h_dt;
- float rho;
- float rho_dh;
-
- /* Particle cutoff radius. */
+ /* Particle smoothing length. */
float h;
/* Particle time of beginning of time-step. */
@@ -67,76 +49,103 @@ struct part {
/* Particle time of end of time-step. */
int ti_end;
- /* The primitive hydrodynamical variables */
+ /* Old internal energy flux */
+ float du_dt;
+
+ /* The primitive hydrodynamical variables. */
struct {
- /* fluid velocity */
- GFLOAT v[3];
+ /* Fluid velocity. */
+ float v[3];
- /* density */
- GFLOAT rho;
+ /* Density. */
+ float rho;
- /* pressure */
- GFLOAT P;
+ /* Pressure. */
+ float P;
+ /* Gradients of the primitive variables. */
struct {
- GFLOAT rho[3];
+ /* Density gradients. */
+ float rho[3];
- GFLOAT v[3][3];
+ /* Fluid velocity gradients. */
+ float v[3][3];
- GFLOAT P[3];
+ /* Pressure gradients. */
+ float P[3];
} gradients;
+ /* Quantities needed by the slope limiter. */
struct {
- /* extreme values among the neighbours */
- GFLOAT rho[2];
+ /* Extreme values of the density among the neighbours. */
+ float rho[2];
- GFLOAT v[3][2];
+ /* Extreme values of the fluid velocity among the neighbours. */
+ float v[3][2];
- GFLOAT P[2];
+ /* Extreme values of the pressure among the neighbours. */
+ float P[2];
- /* maximal distance to all neighbouring faces */
+ /* Maximal distance to all neighbouring faces. */
float maxr;
} limiter;
} primitives;
- /* The conserved hydrodynamical variables */
+ /* The conserved hydrodynamical variables. */
struct {
- /* fluid momentum */
- GFLOAT momentum[3];
+ /* Fluid momentum. */
+ float momentum[3];
+
+ /* Fluid mass */
+ float mass;
+
+ /* Fluid thermal energy (not per unit mass!). */
+ float energy;
+
+ /* Fluxes. */
+ struct {
+
+ /* Mass flux. */
+ float mass;
+
+ /* Momentum flux. */
+ float momentum[3];
- /* fluid mass */
- GFLOAT mass;
+ /* Energy flux. */
+ float energy;
- /* fluid energy */
- GFLOAT energy;
+ } flux;
} conserved;
- /* Geometrical quantities used for hydro */
+ /* Geometrical quantities used for hydro. */
struct {
- /* volume of the particle */
+ /* Volume of the particle. */
float volume;
- /* gradient matrix */
+ /* Geometrical shear matrix used to calculate second order accurate
+ gradients */
float matrix_E[3][3];
} geometry;
+ /* Variables used for timestep calculation (currently not used). */
struct {
+ /* Maximum fluid velocity among all neighbours. */
float vmax;
} timestepvars;
- /* Quantities used during the density loop */
+ /* Quantities used during the volume (=density) loop. */
struct {
/* Particle velocity divergence. */
@@ -153,8 +162,36 @@ struct part {
} density;
+ /* Quantities used during the force loop. */
+ struct {
+
+ /* Needed to drift the primitive variables. */
+ float h_dt;
+
+ /* Physical time step of the particle. */
+ float dt;
+
+ /* Actual velocity of the particle. */
+ float v_full[3];
+
+ } force;
+
+ /* Specific stuff for the gravity-hydro coupling. */
+ struct {
+
+ /* Previous value of the gravitational acceleration. */
+ float old_a[3];
+
+ /* Previous value of the mass flux vector. */
+ float old_mflux[3];
+
+ /* Current value of the mass flux vector. */
+ float mflux[3];
+
+ } gravity;
+
/* Particle ID. */
- unsigned long long id;
+ long long id;
/* Associated gravitas. */
struct gpart *gpart;
diff --git a/src/hydro/Gizmo/hydro_slope_limiters.h b/src/hydro/Gizmo/hydro_slope_limiters.h
new file mode 100644
index 0000000000000000000000000000000000000000..cd66f05ac9eb9d51744723d93f899b0c8c668e2e
--- /dev/null
+++ b/src/hydro/Gizmo/hydro_slope_limiters.h
@@ -0,0 +1,94 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (c) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com)
+ *
+ * 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_HYDRO_SLOPE_LIMITERS_H
+#define SWIFT_HYDRO_SLOPE_LIMITERS_H
+
+#include "dimension.h"
+#include "kernel_hydro.h"
+
+#ifdef SLOPE_LIMITER_PER_FACE
+
+#define HYDRO_SLOPE_LIMITER_FACE_IMPLEMENTATION \
+ "GIZMO piecewise slope limiter (Hopkins 2015)"
+#include "hydro_slope_limiters_face.h"
+
+#else
+
+#define HYDRO_SLOPE_LIMITER_FACE_IMPLEMENTATION "No piecewise slope limiter"
+
+/**
+ * @brief Slope limit the slopes at the interface between two particles
+ *
+ * @param Wi Hydrodynamic variables of particle i.
+ * @param Wj Hydrodynamic variables of particle j.
+ * @param dWi Difference between the hydrodynamic variables of particle i at the
+ * position of particle i and at the interface position.
+ * @param dWj Difference between the hydrodynamic variables of particle j at the
+ * position of particle j and at the interface position.
+ * @param xij_i Relative position vector of the interface w.r.t. particle i.
+ * @param xij_j Relative position vector of the interface w.r.t. partilce j.
+ * @param r Distance between particle i and particle j.
+ */
+__attribute__((always_inline)) INLINE static void hydro_slope_limit_face(
+ float *Wi, float *Wj, float *dWi, float *dWj, float *xij_i, float *xij_j,
+ float r) {}
+
+#endif
+
+#ifdef SLOPE_LIMITER_CELL_WIDE
+
+#define HYDRO_SLOPE_LIMITER_CELL_IMPLEMENTATION \
+ "Cell wide slope limiter (Springel 2010)"
+#include "hydro_slope_limiters_cell.h"
+
+#else
+
+#define HYDRO_SLOPE_LIMITER_CELL_IMPLEMENTATION "No cell wide slope limiter"
+
+/**
+ * @brief Initialize variables for the cell wide slope limiter
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_slope_limit_cell_init(
+ struct part *p) {}
+
+/**
+ * @brief Collect information for the cell wide slope limiter during the
+ * neighbour loop
+ *
+ * @param pi Particle i.
+ * @param pj Particle j.
+ * @param r Distance between particle i and particle j.
+ */
+__attribute__((always_inline)) INLINE static void
+hydro_slope_limit_cell_collect(struct part *pi, struct part *pj, float r) {}
+
+/**
+ * @brief Slope limit cell gradients
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_slope_limit_cell(
+ struct part *p) {}
+
+#endif
+
+#endif // SWIFT_HYDRO_SLOPE_LIMITERS_H
diff --git a/src/hydro/Gizmo/hydro_slope_limiters_cell.h b/src/hydro/Gizmo/hydro_slope_limiters_cell.h
new file mode 100644
index 0000000000000000000000000000000000000000..aa99b43721f669f47a7888a5da0b1933ca1ebd62
--- /dev/null
+++ b/src/hydro/Gizmo/hydro_slope_limiters_cell.h
@@ -0,0 +1,173 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (c) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com)
+ *
+ * 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/>.
+ *
+ ******************************************************************************/
+
+#include <float.h>
+
+/**
+ * @brief Initialize variables for the cell wide slope limiter
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_slope_limit_cell_init(
+ struct part* p) {
+
+ p->primitives.limiter.rho[0] = FLT_MAX;
+ p->primitives.limiter.rho[1] = -FLT_MAX;
+ p->primitives.limiter.v[0][0] = FLT_MAX;
+ p->primitives.limiter.v[0][1] = -FLT_MAX;
+ p->primitives.limiter.v[1][0] = FLT_MAX;
+ p->primitives.limiter.v[1][1] = -FLT_MAX;
+ p->primitives.limiter.v[2][0] = FLT_MAX;
+ p->primitives.limiter.v[2][1] = -FLT_MAX;
+ p->primitives.limiter.P[0] = FLT_MAX;
+ p->primitives.limiter.P[1] = -FLT_MAX;
+
+ p->primitives.limiter.maxr = -FLT_MAX;
+}
+
+/**
+ * @brief Collect information for the cell wide slope limiter during the
+ * neighbour loop
+ *
+ * @param pi Particle i.
+ * @param pj Particle j.
+ * @param r Distance between particle i and particle j.
+ */
+__attribute__((always_inline)) INLINE static void
+hydro_slope_limit_cell_collect(struct part* pi, struct part* pj, float r) {
+
+ /* basic slope limiter: collect the maximal and the minimal value for the
+ * primitive variables among the ngbs */
+ pi->primitives.limiter.rho[0] =
+ fmin(pj->primitives.rho, pi->primitives.limiter.rho[0]);
+ pi->primitives.limiter.rho[1] =
+ fmax(pj->primitives.rho, pi->primitives.limiter.rho[1]);
+
+ pi->primitives.limiter.v[0][0] =
+ fmin(pj->primitives.v[0], pi->primitives.limiter.v[0][0]);
+ pi->primitives.limiter.v[0][1] =
+ fmax(pj->primitives.v[0], pi->primitives.limiter.v[0][1]);
+ pi->primitives.limiter.v[1][0] =
+ fmin(pj->primitives.v[1], pi->primitives.limiter.v[1][0]);
+ pi->primitives.limiter.v[1][1] =
+ fmax(pj->primitives.v[1], pi->primitives.limiter.v[1][1]);
+ pi->primitives.limiter.v[2][0] =
+ fmin(pj->primitives.v[2], pi->primitives.limiter.v[2][0]);
+ pi->primitives.limiter.v[2][1] =
+ fmax(pj->primitives.v[2], pi->primitives.limiter.v[2][1]);
+
+ pi->primitives.limiter.P[0] =
+ fmin(pj->primitives.P, pi->primitives.limiter.P[0]);
+ pi->primitives.limiter.P[1] =
+ fmax(pj->primitives.P, pi->primitives.limiter.P[1]);
+
+ pi->primitives.limiter.maxr = fmax(r, pi->primitives.limiter.maxr);
+}
+
+/**
+ * @brief Slope limit cell gradients
+ *
+ * @param p Particle.
+ */
+__attribute__((always_inline)) INLINE static void hydro_slope_limit_cell(
+ struct part* p) {
+
+ float gradrho[3], gradv[3][3], gradP[3];
+ float gradtrue, gradmax, gradmin, alpha;
+
+ gradrho[0] = p->primitives.gradients.rho[0];
+ gradrho[1] = p->primitives.gradients.rho[1];
+ gradrho[2] = p->primitives.gradients.rho[2];
+
+ gradv[0][0] = p->primitives.gradients.v[0][0];
+ gradv[0][1] = p->primitives.gradients.v[0][1];
+ gradv[0][2] = p->primitives.gradients.v[0][2];
+
+ gradv[1][0] = p->primitives.gradients.v[1][0];
+ gradv[1][1] = p->primitives.gradients.v[1][1];
+ gradv[1][2] = p->primitives.gradients.v[1][2];
+
+ gradv[2][0] = p->primitives.gradients.v[2][0];
+ gradv[2][1] = p->primitives.gradients.v[2][1];
+ gradv[2][2] = p->primitives.gradients.v[2][2];
+
+ gradP[0] = p->primitives.gradients.P[0];
+ gradP[1] = p->primitives.gradients.P[1];
+ gradP[2] = p->primitives.gradients.P[2];
+
+ gradtrue = sqrtf(gradrho[0] * gradrho[0] + gradrho[1] * gradrho[1] +
+ gradrho[2] * gradrho[2]);
+ if (gradtrue) {
+ gradtrue *= p->primitives.limiter.maxr;
+ gradmax = p->primitives.limiter.rho[1] - p->primitives.rho;
+ gradmin = p->primitives.rho - p->primitives.limiter.rho[0];
+ alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
+ p->primitives.gradients.rho[0] *= alpha;
+ p->primitives.gradients.rho[1] *= alpha;
+ p->primitives.gradients.rho[2] *= alpha;
+ }
+
+ gradtrue = sqrtf(gradv[0][0] * gradv[0][0] + gradv[0][1] * gradv[0][1] +
+ gradv[0][2] * gradv[0][2]);
+ if (gradtrue) {
+ gradtrue *= p->primitives.limiter.maxr;
+ gradmax = p->primitives.limiter.v[0][1] - p->primitives.v[0];
+ gradmin = p->primitives.v[0] - p->primitives.limiter.v[0][0];
+ alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
+ p->primitives.gradients.v[0][0] *= alpha;
+ p->primitives.gradients.v[0][1] *= alpha;
+ p->primitives.gradients.v[0][2] *= alpha;
+ }
+
+ gradtrue = sqrtf(gradv[1][0] * gradv[1][0] + gradv[1][1] * gradv[1][1] +
+ gradv[1][2] * gradv[1][2]);
+ if (gradtrue) {
+ gradtrue *= p->primitives.limiter.maxr;
+ gradmax = p->primitives.limiter.v[1][1] - p->primitives.v[1];
+ gradmin = p->primitives.v[1] - p->primitives.limiter.v[1][0];
+ alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
+ p->primitives.gradients.v[1][0] *= alpha;
+ p->primitives.gradients.v[1][1] *= alpha;
+ p->primitives.gradients.v[1][2] *= alpha;
+ }
+
+ gradtrue = sqrtf(gradv[2][0] * gradv[2][0] + gradv[2][1] * gradv[2][1] +
+ gradv[2][2] * gradv[2][2]);
+ if (gradtrue) {
+ gradtrue *= p->primitives.limiter.maxr;
+ gradmax = p->primitives.limiter.v[2][1] - p->primitives.v[2];
+ gradmin = p->primitives.v[2] - p->primitives.limiter.v[2][0];
+ alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
+ p->primitives.gradients.v[2][0] *= alpha;
+ p->primitives.gradients.v[2][1] *= alpha;
+ p->primitives.gradients.v[2][2] *= alpha;
+ }
+
+ gradtrue =
+ sqrtf(gradP[0] * gradP[0] + gradP[1] * gradP[1] + gradP[2] * gradP[2]);
+ if (gradtrue) {
+ gradtrue *= p->primitives.limiter.maxr;
+ gradmax = p->primitives.limiter.P[1] - p->primitives.P;
+ gradmin = p->primitives.P - p->primitives.limiter.P[0];
+ alpha = fmin(1.0f, fmin(gradmax / gradtrue, gradmin / gradtrue));
+ p->primitives.gradients.P[0] *= alpha;
+ p->primitives.gradients.P[1] *= alpha;
+ p->primitives.gradients.P[2] *= alpha;
+ }
+}
diff --git a/src/hydro/Gizmo/hydro_slope_limiters_face.h b/src/hydro/Gizmo/hydro_slope_limiters_face.h
new file mode 100644
index 0000000000000000000000000000000000000000..7ae5dd2eb073d9aae8ab6f2efffdf8df15b4bb4a
--- /dev/null
+++ b/src/hydro/Gizmo/hydro_slope_limiters_face.h
@@ -0,0 +1,121 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (c) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com)
+ *
+ * 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/>.
+ *
+ ******************************************************************************/
+
+/**
+ * @brief Slope limit a single quantity at the interface
+ *
+ * @param phi_i Value of the quantity at the particle position.
+ * @param phi_j Value of the quantity at the neighbouring particle position.
+ * @param phi_mid0 Extrapolated value of the quantity at the interface position.
+ * @param xij_norm Distance between the particle position and the interface
+ * position.
+ * @param r Distance between the particle and its neighbour.
+ * @return The slope limited difference between the quantity at the particle
+ * position and the quantity at the interface position.
+ */
+__attribute__((always_inline)) INLINE static float
+hydro_slope_limit_face_quantity(float phi_i, float phi_j, float phi_mid0,
+ float xij_norm, float r) {
+
+ float delta1, delta2, phimin, phimax, phibar, phiplus, phiminus, phi_mid;
+ const float psi1 = 0.5f;
+ const float psi2 = 0.25f;
+
+ if (phi_i == phi_j) {
+ return 0.0f;
+ }
+
+ delta1 = psi1 * fabs(phi_i - phi_j);
+ delta2 = psi2 * fabs(phi_i - phi_j);
+
+ phimin = fmin(phi_i, phi_j);
+ phimax = fmax(phi_i, phi_j);
+
+ phibar = phi_i + xij_norm / r * (phi_j - phi_i);
+
+ /* if sign(phimax+delta1) == sign(phimax) */
+ if ((phimax + delta1) * phimax > 0.0f) {
+ phiplus = phimax + delta1;
+ } else {
+ phiplus = phimax / (1.0f + delta1 / fabs(phimax));
+ }
+
+ /* if sign(phimin-delta1) == sign(phimin) */
+ if ((phimin - delta1) * phimin > 0.0f) {
+ phiminus = phimin - delta1;
+ } else {
+ phiminus = phimin / (1.0f + delta1 / fabs(phimin));
+ }
+
+ if (phi_i < phi_j) {
+ phi_mid = fmax(phiminus, fmin(phibar + delta2, phi_mid0));
+ } else {
+ phi_mid = fmin(phiplus, fmax(phibar - delta2, phi_mid0));
+ }
+
+ return phi_mid - phi_i;
+}
+
+/**
+ * @brief Slope limit the slopes at the interface between two particles
+ *
+ * @param Wi Hydrodynamic variables of particle i.
+ * @param Wj Hydrodynamic variables of particle j.
+ * @param dWi Difference between the hydrodynamic variables of particle i at the
+ * position of particle i and at the interface position.
+ * @param dWj Difference between the hydrodynamic variables of particle j at the
+ * position of particle j and at the interface position.
+ * @param xij_i Relative position vector of the interface w.r.t. particle i.
+ * @param xij_j Relative position vector of the interface w.r.t. partilce j.
+ * @param r Distance between particle i and particle j.
+ */
+__attribute__((always_inline)) INLINE static void hydro_slope_limit_face(
+ float *Wi, float *Wj, float *dWi, float *dWj, float *xij_i, float *xij_j,
+ float r) {
+
+ float xij_i_norm, xij_j_norm;
+
+ xij_i_norm =
+ sqrtf(xij_i[0] * xij_i[0] + xij_i[1] * xij_i[1] + xij_i[2] * xij_i[2]);
+
+ xij_j_norm =
+ sqrtf(xij_j[0] * xij_j[0] + xij_j[1] * xij_j[1] + xij_j[2] * xij_j[2]);
+
+ dWi[0] = hydro_slope_limit_face_quantity(Wi[0], Wj[0], Wi[0] + dWi[0],
+ xij_i_norm, r);
+ dWi[1] = hydro_slope_limit_face_quantity(Wi[1], Wj[1], Wi[1] + dWi[1],
+ xij_i_norm, r);
+ dWi[2] = hydro_slope_limit_face_quantity(Wi[2], Wj[2], Wi[2] + dWi[2],
+ xij_i_norm, r);
+ dWi[3] = hydro_slope_limit_face_quantity(Wi[3], Wj[3], Wi[3] + dWi[3],
+ xij_i_norm, r);
+ dWi[4] = hydro_slope_limit_face_quantity(Wi[4], Wj[4], Wi[4] + dWi[4],
+ xij_i_norm, r);
+
+ dWj[0] = hydro_slope_limit_face_quantity(Wj[0], Wi[0], Wj[0] + dWj[0],
+ xij_j_norm, r);
+ dWj[1] = hydro_slope_limit_face_quantity(Wj[1], Wi[1], Wj[1] + dWj[1],
+ xij_j_norm, r);
+ dWj[2] = hydro_slope_limit_face_quantity(Wj[2], Wi[2], Wj[2] + dWj[2],
+ xij_j_norm, r);
+ dWj[3] = hydro_slope_limit_face_quantity(Wj[3], Wi[3], Wj[3] + dWj[3],
+ xij_j_norm, r);
+ dWj[4] = hydro_slope_limit_face_quantity(Wj[4], Wi[4], Wj[4] + dWj[4],
+ xij_j_norm, r);
+}
diff --git a/src/hydro/Minimal/hydro.h b/src/hydro/Minimal/hydro.h
index a5d73aad02372aa22b840c2cb0d3100cd439e75d..0bbc77f4a2384c79f7d3329c20b92990598d5c63 100644
--- a/src/hydro/Minimal/hydro.h
+++ b/src/hydro/Minimal/hydro.h
@@ -35,6 +35,7 @@
*/
#include "adiabatic_index.h"
+#include "approx_math.h"
#include "dimension.h"
#include "equation_of_state.h"
#include "hydro_properties.h"
@@ -102,6 +103,28 @@ __attribute__((always_inline)) INLINE static float hydro_get_soundspeed(
return gas_soundspeed_from_internal_energy(p->rho, u);
}
+/**
+ * @brief Returns the density of a particle
+ *
+ * @param p The particle of interest
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_density(
+ const struct part *restrict p) {
+
+ return p->rho;
+}
+
+/**
+ * @brief Returns the mass of a particle
+ *
+ * @param p The particle of interest
+ */
+__attribute__((always_inline)) INLINE static float hydro_get_mass(
+ const struct part *restrict p) {
+
+ return p->mass;
+}
+
/**
* @brief Modifies the thermal state of a particle to the imposed internal
* energy
@@ -286,15 +309,32 @@ __attribute__((always_inline)) INLINE static void hydro_reset_acceleration(
* Additional hydrodynamic quantites are drifted forward in time here. These
* include thermal quantities (thermal energy or total energy or entropy, ...).
*
- * @param p The particle
- * @param xp The extended data of the particle
- * @param t0 The time at the start of the drift (on the timeline)
- * @param t1 The time at the end of the drift (on the timeline)
- * @param timeBase The minimal time-step size
+ * @param p The particle.
+ * @param xp The extended data of the particle.
+ * @param dt The drift time-step.
+ * @param t0 The time at the start of the drift (on the timeline).
+ * @param t1 The time at the end of the drift (on the timeline).
+ * @param timeBase The minimal time-step size.
*/
__attribute__((always_inline)) INLINE static void hydro_predict_extra(
- struct part *restrict p, const struct xpart *restrict xp, int t0, int t1,
- double timeBase) {
+ struct part *restrict p, const struct xpart *restrict xp, float dt, int t0,
+ int t1, double timeBase) {
+
+ const float h_inv = 1.f / p->h;
+
+ /* Predict smoothing length */
+ const float w1 = p->force.h_dt * h_inv * dt;
+ if (fabsf(w1) < 0.2f)
+ p->h *= approx_expf(w1); /* 4th order expansion of exp(w) */
+ else
+ p->h *= expf(w1);
+
+ /* Predict density */
+ const float w2 = -hydro_dimension * w1;
+ if (fabsf(w2) < 0.2f)
+ p->rho *= approx_expf(w2); /* 4th order expansion of exp(w) */
+ else
+ p->rho *= expf(w2);
/* Drift the pressure */
const float dt_entr = (t1 - (p->ti_begin + p->ti_end) / 2) * timeBase;
diff --git a/src/hydro_io.h b/src/hydro_io.h
index 30d663f647c9b763e9b19177e9ba8ef374855768..f0a619b90b774574c434007b1c01a0e55e75e464 100644
--- a/src/hydro_io.h
+++ b/src/hydro_io.h
@@ -28,6 +28,8 @@
#include "./hydro/Gadget2/hydro_io.h"
#elif defined(DEFAULT_SPH)
#include "./hydro/Default/hydro_io.h"
+#elif defined(GIZMO_SPH)
+#include "./hydro/Gizmo/hydro_io.h"
#else
#error "Invalid choice of SPH variant"
#endif
diff --git a/src/kernel_gravity.h b/src/kernel_gravity.h
index b38feb5758debf87add2007ee3684d869f393f7e..a1e382a21d04b7354aaf215069e999627e56ee07 100644
--- a/src/kernel_gravity.h
+++ b/src/kernel_gravity.h
@@ -50,7 +50,7 @@ static const float
-10.66666667f,
38.4f,
-48.f,
- 21.3333333,
+ 21.3333333f,
0.f,
0.f,
-0.066666667f, /* 0.5 < u < 1 */
diff --git a/src/kernel_long_gravity.h b/src/kernel_long_gravity.h
index d247c7a461d4bd116f30ab106143f6c75e1b941e..6952681999f833bce7755a72aaee742a7fa0ed22 100644
--- a/src/kernel_long_gravity.h
+++ b/src/kernel_long_gravity.h
@@ -41,7 +41,7 @@ __attribute__((always_inline)) INLINE static void kernel_long_grav_eval(
const float arg2 = u * one_over_sqrt_pi;
const float arg3 = -arg1 * arg1;
- const float term1 = erfc(arg1);
+ const float term1 = erfcf(arg1);
const float term2 = arg2 * expf(arg3);
*W = term1 + term2;
diff --git a/src/kick.h b/src/kick.h
index b57e13d4ebf27d3a366d571e7fd4cd819653f726..e3fa3bf78c7da514abacf697a9d94212020e5a7b 100644
--- a/src/kick.h
+++ b/src/kick.h
@@ -39,8 +39,8 @@ __attribute__((always_inline)) INLINE static void kick_gpart(
/* Compute the time step for this kick */
const int ti_start = (gp->ti_begin + gp->ti_end) / 2;
const int ti_end = gp->ti_end + new_dti / 2;
- const double dt = (ti_end - ti_start) * timeBase;
- const double half_dt = (ti_end - gp->ti_end) * timeBase;
+ const float dt = (ti_end - ti_start) * timeBase;
+ const float half_dt = (ti_end - gp->ti_end) * timeBase;
/* Move particle forward in time */
gp->ti_begin = gp->ti_end;
@@ -70,8 +70,8 @@ __attribute__((always_inline)) INLINE static void kick_part(
/* Compute the time step for this kick */
const int ti_start = (p->ti_begin + p->ti_end) / 2;
const int ti_end = p->ti_end + new_dti / 2;
- const double dt = (ti_end - ti_start) * timeBase;
- const double half_dt = (ti_end - p->ti_end) * timeBase;
+ const float dt = (ti_end - ti_start) * timeBase;
+ const float half_dt = (ti_end - p->ti_end) * timeBase;
/* Move particle forward in time */
p->ti_begin = p->ti_end;
diff --git a/src/parallel_io.c b/src/parallel_io.c
index d543aac7e700dc008c439ffca7e45108c6c1904e..5d10cf03137937a4b89db68218bb723d3a282c59 100644
--- a/src/parallel_io.c
+++ b/src/parallel_io.c
@@ -528,27 +528,18 @@ void read_ic_parallel(char* fileName, const struct UnitSystem* internal_units,
int num_fields = 0;
struct io_props list[100];
- size_t N = 0;
+ size_t Nparticles = 0;
/* Read particle fields into the particle structure */
switch (ptype) {
case GAS:
- /* if (!dry_run) */
- /* hydro_read_particles(h_grp, N[ptype], N_total[ptype],
- * offset[ptype], */
- /* *parts); */
- /* break; */
- N = *Ngas;
+ Nparticles = *Ngas;
hydro_read_particles(*parts, list, &num_fields);
break;
case DM:
- /* if (!dry_run) */
- /* darkmatter_read_particles(h_grp, N[ptype], N_total[ptype], */
- /* offset[ptype], *gparts); */
- /* break; */
- N = Ndm;
+ Nparticles = Ndm;
darkmatter_read_particles(*gparts, list, &num_fields);
break;
@@ -559,7 +550,7 @@ void read_ic_parallel(char* fileName, const struct UnitSystem* internal_units,
/* Read everything */
if (!dry_run)
for (int i = 0; i < num_fields; ++i)
- readArray(h_grp, list[i], N, N_total[ptype], offset[ptype],
+ readArray(h_grp, list[i], Nparticles, N_total[ptype], offset[ptype],
internal_units, ic_units);
/* Close particle group */
@@ -793,13 +784,13 @@ void write_output_parallel(struct engine* e, const char* baseName,
int num_fields = 0;
struct io_props list[100];
- size_t N = 0;
+ size_t Nparticles = 0;
/* Write particle fields from the particle structure */
switch (ptype) {
case GAS:
- N = Ngas;
+ Nparticles = Ngas;
hydro_write_particles(parts, list, &num_fields);
break;
@@ -816,7 +807,7 @@ void write_output_parallel(struct engine* e, const char* baseName,
collect_dm_gparts(gparts, Ntot, dmparts, Ndm);
/* Write DM particles */
- N = Ndm;
+ Nparticles = Ndm;
darkmatter_write_particles(dmparts, list, &num_fields);
/* Free temporary array */
@@ -829,9 +820,9 @@ void write_output_parallel(struct engine* e, const char* baseName,
/* Write everything */
for (int i = 0; i < num_fields; ++i)
- writeArray(e, h_grp, fileName, xmfFile, partTypeGroupName, list[i], N,
- N_total[ptype], mpi_rank, offset[ptype], internal_units,
- snapshot_units);
+ writeArray(e, h_grp, fileName, xmfFile, partTypeGroupName, list[i],
+ Nparticles, N_total[ptype], mpi_rank, offset[ptype],
+ internal_units, snapshot_units);
/* Free temporary array */
free(dmparts);
diff --git a/src/part.h b/src/part.h
index efca7b6b5bef49f20df1e2c45b30f65ecbbf4960..ea895e6e0295d6a8b63309c7bd6855daa2cf7d64 100644
--- a/src/part.h
+++ b/src/part.h
@@ -45,6 +45,9 @@
#include "./hydro/Gadget2/hydro_part.h"
#elif defined(DEFAULT_SPH)
#include "./hydro/Default/hydro_part.h"
+#elif defined(GIZMO_SPH)
+#include "./hydro/Gizmo/hydro_part.h"
+#define EXTRA_HYDRO_LOOP
#else
#error "Invalid choice of SPH variant"
#endif
diff --git a/src/potentials.h b/src/potentials.h
index f0de5b4dc784be2e613c634bf687300d57f40ecd..74f0fd28566f355962c83e5d743aeae9afe09c59 100644
--- a/src/potentials.h
+++ b/src/potentials.h
@@ -78,13 +78,15 @@ struct external_potential {
*
* See Creasey, Theuns & Bower, 2013, MNRAS, Volume 429, Issue 3, p.1922-1948
*
- * @param phys_cont The physical constants in internal units.
- * @param gp Pointer to the g-particle data.
+ * @param potential The properties of the potential.
+ * @param phys_const The physical constants in internal units.
+ * @param g Pointer to the g-particle data.
*/
__attribute__((always_inline)) INLINE static float
-external_gravity_disk_patch_timestep(const struct external_potential* potential,
- const struct phys_const* const phys_const,
- const struct gpart* const g) {
+external_gravity_disk_patch_timestep(
+ const struct external_potential* restrict potential,
+ const struct phys_const* restrict phys_const,
+ const struct gpart* restrict g) {
/* initilize time step to disk dynamical time */
const float dt_dyn = potential->disk_patch_potential.dynamical_time;
@@ -136,13 +138,15 @@ external_gravity_disk_patch_timestep(const struct external_potential* potential,
*
* See Creasey, Theuns & Bower, 2013, MNRAS, Volume 429, Issue 3, p.1922-1948
*
- * @param phys_cont The physical constants in internal units.
+ * @param time The current time in internal units.
+ * @param potential The properties of the potential.
+ * @param phys_const The physical constants in internal units.
* @param g Pointer to the g-particle data.
*/
__attribute__((always_inline)) INLINE static void
external_gravity_disk_patch_potential(
- const double time, const struct external_potential* potential,
- const struct phys_const* const phys_const, struct gpart* g) {
+ double time, const struct external_potential* restrict potential,
+ const struct phys_const* restrict phys_const, struct gpart* restrict g) {
const float G_newton = phys_const->const_newton_G;
const float dz = g->x[2] - potential->disk_patch_potential.z_disk;
@@ -178,8 +182,9 @@ external_gravity_disk_patch_potential(
*/
__attribute__((always_inline)) INLINE static float
external_gravity_isothermalpotential_timestep(
- const struct external_potential* potential,
- const struct phys_const* const phys_const, const struct gpart* const g) {
+ const struct external_potential* restrict potential,
+ const struct phys_const* restrict phys_const,
+ const struct gpart* restrict g) {
const float dx = g->x[0] - potential->isothermal_potential.x;
const float dy = g->x[1] - potential->isothermal_potential.y;
@@ -247,9 +252,10 @@ external_gravity_isothermalpotential(const struct external_potential* potential,
* @param g Pointer to the g-particle data.
*/
__attribute__((always_inline)) INLINE static float
-external_gravity_pointmass_timestep(const struct external_potential* potential,
- const struct phys_const* const phys_const,
- const struct gpart* const g) {
+external_gravity_pointmass_timestep(
+ const struct external_potential* restrict potential,
+ const struct phys_const* restrict phys_const,
+ const struct gpart* restrict g) {
const float G_newton = phys_const->const_newton_G;
const float dx = g->x[0] - potential->point_mass.x;
@@ -284,8 +290,8 @@ external_gravity_pointmass_timestep(const struct external_potential* potential,
* @param g Pointer to the g-particle data.
*/
__attribute__((always_inline)) INLINE static void external_gravity_pointmass(
- const struct external_potential* potential,
- const struct phys_const* const phys_const, struct gpart* g) {
+ const struct external_potential* restrict potential,
+ const struct phys_const* restrict phys_const, struct gpart* restrict g) {
const float dx = g->x[0] - potential->point_mass.x;
const float dy = g->x[1] - potential->point_mass.y;
diff --git a/src/riemann.h b/src/riemann.h
index d647b021167317d14f4cd7316d09c247794f3d23..0191058e3ba856d7dcec42ba6185def3ee71d090 100644
--- a/src/riemann.h
+++ b/src/riemann.h
@@ -27,18 +27,27 @@
#include "stdio.h"
#include "stdlib.h"
-#define HLLC_SOLVER
+#if defined(RIEMANN_SOLVER_EXACT)
-#ifdef EXACT_SOLVER
+#define RIEMANN_SOLVER_IMPLEMENTATION "Exact Riemann solver (Toro 2009)"
#include "riemann/riemann_exact.h"
-#endif
-#ifdef TRRS_SOLVER
+#elif defined(RIEMANN_SOLVER_TRRS)
+
+#define RIEMANN_SOLVER_IMPLEMENTATION \
+ "Two Rarefaction Riemann Solver (Toro 2009)"
#include "riemann/riemann_trrs.h"
-#endif
-#ifdef HLLC_SOLVER
+#elif defined(RIEMANN_SOLVER_HLLC)
+
+#define RIEMANN_SOLVER_IMPLEMENTATION \
+ "Harten-Lax-van Leer-Contact Riemann solver (Toro 2009)"
#include "riemann/riemann_hllc.h"
+
+#else
+
+#error "Error: no Riemann solver selected!"
+
#endif
#endif /* SWIFT_RIEMANN_H */
diff --git a/src/riemann/riemann_exact.h b/src/riemann/riemann_exact.h
index a2f3c30fb1daf5d53bf35abe4ca7e73eafba6018..9763d9f0d12da32d9142c24481105b9f139be588 100644
--- a/src/riemann/riemann_exact.h
+++ b/src/riemann/riemann_exact.h
@@ -29,19 +29,9 @@
#ifndef SWIFT_RIEMANN_EXACT_H
#define SWIFT_RIEMANN_EXACT_H
-/* frequently used combinations of const_hydro_gamma */
-#define const_riemann_gp1d2g \
- (0.5f * (const_hydro_gamma + 1.0f) / const_hydro_gamma)
-#define const_riemann_gm1d2g \
- (0.5f * (const_hydro_gamma - 1.0f) / const_hydro_gamma)
-#define const_riemann_gm1dgp1 \
- ((const_hydro_gamma - 1.0f) / (const_hydro_gamma + 1.0f))
-#define const_riemann_tdgp1 (2.0f / (const_hydro_gamma + 1.0f))
-#define const_riemann_tdgm1 (2.0f / (const_hydro_gamma - 1.0f))
-#define const_riemann_gm1d2 (0.5f * (const_hydro_gamma - 1.0f))
-#define const_riemann_tgdgm1 \
- (2.0f * const_hydro_gamma / (const_hydro_gamma - 1.0f))
-#define const_riemann_ginv (1.0f / const_hydro_gamma)
+#include <float.h>
+#include "adiabatic_index.h"
+#include "riemann_vacuum.h"
/**
* @brief Functions (4.6) and (4.7) in Toro.
@@ -50,19 +40,18 @@
* @param W The left or right state vector
* @param a The left or right sound speed
*/
-__attribute__((always_inline)) INLINE static GFLOAT riemann_fb(GFLOAT p,
- GFLOAT* W,
- GFLOAT a) {
+__attribute__((always_inline)) INLINE static float riemann_fb(float p, float* W,
+ float a) {
- GFLOAT fval = 0.;
- GFLOAT A, B;
+ float fval = 0.;
+ float A, B;
if (p > W[4]) {
- A = const_riemann_tdgp1 / W[0];
- B = const_riemann_gm1dgp1 * W[4];
+ A = hydro_two_over_gamma_plus_one / W[0];
+ B = hydro_gamma_minus_one_over_gamma_plus_one * W[4];
fval = (p - W[4]) * sqrtf(A / (p + B));
} else {
- fval =
- const_riemann_tdgm1 * a * (powf(p / W[4], const_riemann_gm1d2g) - 1.0f);
+ fval = hydro_two_over_gamma_minus_one * a *
+ (pow_gamma_minus_one_over_two_gamma(p / W[4]) - 1.0f);
}
return fval;
}
@@ -78,9 +67,8 @@ __attribute__((always_inline)) INLINE static GFLOAT riemann_fb(GFLOAT p,
* @param aL The left sound speed
* @param aR The right sound speed
*/
-__attribute__((always_inline)) INLINE static GFLOAT riemann_f(
- GFLOAT p, GFLOAT* WL, GFLOAT* WR, GFLOAT vL, GFLOAT vR, GFLOAT aL,
- GFLOAT aR) {
+__attribute__((always_inline)) INLINE static float riemann_f(
+ float p, float* WL, float* WR, float vL, float vR, float aL, float aR) {
return riemann_fb(p, WL, aL) + riemann_fb(p, WR, aR) + (vR - vL);
}
@@ -92,18 +80,18 @@ __attribute__((always_inline)) INLINE static GFLOAT riemann_f(
* @param W The left or right state vector
* @param a The left or right sound speed
*/
-__attribute__((always_inline)) INLINE static GFLOAT riemann_fprimeb(GFLOAT p,
- GFLOAT* W,
- GFLOAT a) {
+__attribute__((always_inline)) INLINE static float riemann_fprimeb(float p,
+ float* W,
+ float a) {
- GFLOAT fval = 0.;
- GFLOAT A, B;
+ float fval = 0.;
+ float A, B;
if (p > W[4]) {
- A = const_riemann_tdgp1 / W[0];
- B = const_riemann_gm1dgp1 * W[4];
+ A = hydro_two_over_gamma_plus_one / W[0];
+ B = hydro_gamma_minus_one_over_gamma_plus_one * W[4];
fval = (1.0f - 0.5f * (p - W[4]) / (B + p)) * sqrtf(A / (p + B));
} else {
- fval = 1.0f / (W[0] * a) * powf(p / W[4], -const_riemann_gp1d2g);
+ fval = 1.0f / W[0] / a * pow_minus_gamma_plus_one_over_two_gamma(p / W[4]);
}
return fval;
}
@@ -117,8 +105,8 @@ __attribute__((always_inline)) INLINE static GFLOAT riemann_fprimeb(GFLOAT p,
* @param aL The left sound speed
* @param aR The right sound speed
*/
-__attribute__((always_inline)) INLINE static GFLOAT riemann_fprime(
- GFLOAT p, GFLOAT* WL, GFLOAT* WR, GFLOAT aL, GFLOAT aR) {
+__attribute__((always_inline)) INLINE static float riemann_fprime(
+ float p, float* WL, float* WR, float aL, float aR) {
return riemann_fprimeb(p, WL, aL) + riemann_fprimeb(p, WR, aR);
}
@@ -129,12 +117,12 @@ __attribute__((always_inline)) INLINE static GFLOAT riemann_fprime(
* @param p The current guess for the pressure
* @param W The left or right state vector
*/
-__attribute__((always_inline)) INLINE static GFLOAT riemann_gb(GFLOAT p,
- GFLOAT* W) {
+__attribute__((always_inline)) INLINE static float riemann_gb(float p,
+ float* W) {
- GFLOAT A, B;
- A = const_riemann_tdgp1 / W[0];
- B = const_riemann_gm1dgp1 * W[4];
+ float A, B;
+ A = hydro_two_over_gamma_plus_one / W[0];
+ B = hydro_gamma_minus_one_over_gamma_plus_one * W[4];
return sqrtf(A / (p + B));
}
@@ -151,11 +139,11 @@ __attribute__((always_inline)) INLINE static GFLOAT riemann_gb(GFLOAT p,
* @param aL The left sound speed
* @param aR The right sound speed
*/
-__attribute__((always_inline)) INLINE static GFLOAT riemann_guess_p(
- GFLOAT* WL, GFLOAT* WR, GFLOAT vL, GFLOAT vR, GFLOAT aL, GFLOAT aR) {
+__attribute__((always_inline)) INLINE static float riemann_guess_p(
+ float* WL, float* WR, float vL, float vR, float aL, float aR) {
- GFLOAT pguess, pmin, pmax, qmax;
- GFLOAT ppv;
+ float pguess, pmin, pmax, qmax;
+ float ppv;
pmin = fminf(WL[4], WR[4]);
pmax = fmaxf(WL[4], WR[4]);
@@ -168,10 +156,10 @@ __attribute__((always_inline)) INLINE static GFLOAT riemann_guess_p(
} else {
if (ppv < pmin) {
/* two rarefactions */
- pguess = powf((aL + aR - const_riemann_gm1d2 * (vR - vL)) /
- (aL / powf(WL[4], const_riemann_gm1d2g) +
- aR / powf(WR[4], const_riemann_gm1d2g)),
- const_riemann_tgdgm1);
+ pguess = pow_two_gamma_over_gamma_minus_one(
+ (aL + aR - hydro_gamma_minus_one_over_two * (vR - vL)) /
+ (aL / pow_gamma_minus_one_over_two_gamma(WL[4]) +
+ aR / pow_gamma_minus_one_over_two_gamma(WR[4])));
} else {
/* two shocks */
pguess = (riemann_gb(ppv, WL) * WL[4] + riemann_gb(ppv, WR) * WR[4] - vR +
@@ -202,14 +190,14 @@ __attribute__((always_inline)) INLINE static GFLOAT riemann_guess_p(
* @param aL The left sound speed
* @param aR The right sound speed
*/
-__attribute__((always_inline)) INLINE static GFLOAT riemann_solve_brent(
- GFLOAT lower_limit, GFLOAT upper_limit, GFLOAT lowf, GFLOAT upf,
- GFLOAT error_tol, GFLOAT* WL, GFLOAT* WR, GFLOAT vL, GFLOAT vR, GFLOAT aL,
- GFLOAT aR) {
-
- GFLOAT a, b, c, d, s;
- GFLOAT fa, fb, fc, fs;
- GFLOAT tmp, tmp2;
+__attribute__((always_inline)) INLINE static float riemann_solve_brent(
+ float lower_limit, float upper_limit, float lowf, float upf,
+ float error_tol, float* WL, float* WR, float vL, float vR, float aL,
+ float aR) {
+
+ float a, b, c, d, s;
+ float fa, fb, fc, fs;
+ float tmp, tmp2;
int mflag;
int i;
@@ -296,148 +284,6 @@ __attribute__((always_inline)) INLINE static GFLOAT riemann_solve_brent(
return b;
}
-/**
- * @brief Vacuum Riemann solver, based on section 4.6 in Toro
- *
- * @param WL The left state vector
- * @param WR The right state vector
- * @param vL The left velocity along the interface normal
- * @param vR The right velocity along the interface normal
- * @param aL The left sound speed
- * @param aR The right sound speed
- * @param Whalf Empty state vector to store the solution in
- * @param n_unit Normal vector of the interface
- */
-__attribute__((always_inline)) INLINE static void riemann_solve_vacuum(
- GFLOAT* WL, GFLOAT* WR, GFLOAT vL, GFLOAT vR, GFLOAT aL, GFLOAT aR,
- GFLOAT* Whalf, float* n_unit) {
-
- GFLOAT SL, SR;
- GFLOAT vhalf;
-
- if (!WR[0] && !WL[0]) {
- /* if both states are vacuum, the solution is also vacuum */
- Whalf[0] = 0.0f;
- Whalf[1] = 0.0f;
- Whalf[2] = 0.0f;
- Whalf[3] = 0.0f;
- Whalf[4] = 0.0f;
- return;
- }
- if (!WR[0]) {
- Whalf[1] = WL[1];
- Whalf[2] = WL[2];
- Whalf[3] = WL[3];
- /* vacuum right state */
- if (vL < aL) {
- SL = vL + const_riemann_tdgm1 * aL;
- if (SL > 0.0f) {
- Whalf[0] =
- WL[0] * powf(const_riemann_tdgp1 + const_riemann_gm1dgp1 / aL * vL,
- const_riemann_tdgm1);
- vhalf = const_riemann_tdgp1 * (aL + const_riemann_gm1d2 * vL) - vL;
- Whalf[4] =
- WL[4] * powf(const_riemann_tdgp1 + const_riemann_gm1dgp1 / aL * vL,
- const_riemann_tgdgm1);
- } else {
- Whalf[0] = 0.0f;
- Whalf[1] = 0.0f;
- Whalf[2] = 0.0f;
- Whalf[3] = 0.0f;
- Whalf[4] = 0.0f;
- return;
- }
- } else {
- Whalf[0] = WL[0];
- vhalf = 0.0f;
- Whalf[4] = WL[4];
- }
- } else {
- if (!WL[0]) {
- Whalf[1] = WR[1];
- Whalf[2] = WR[2];
- Whalf[3] = WR[3];
- /* vacuum left state */
- if (-vR < aR) {
- SR = vR - const_riemann_tdgm1 * aR;
- if (SR >= 0.0f) {
- Whalf[0] = 0.0f;
- Whalf[1] = 0.0f;
- Whalf[2] = 0.0f;
- Whalf[3] = 0.0f;
- Whalf[4] = 0.0f;
- return;
- } else {
- Whalf[0] = WR[0] *
- powf(const_riemann_tdgp1 - const_riemann_gm1dgp1 / aR * vR,
- const_riemann_tdgm1);
- vhalf = const_riemann_tdgp1 * (-aR + const_riemann_gm1d2 * vR) - vR;
- Whalf[4] = WR[4] *
- powf(const_riemann_tdgp1 - const_riemann_gm1dgp1 / aR * vR,
- const_riemann_tgdgm1);
- }
- } else {
- Whalf[0] = WR[0];
- vhalf = 0.0f;
- Whalf[4] = WR[4];
- }
- } else {
- /* vacuum generation */
- SR = vR - const_riemann_tdgm1 * aR;
- SL = vL + const_riemann_tdgm1 * aL;
- if (SR > 0.0f && SL < 0.0f) {
- Whalf[0] = 0.0f;
- Whalf[1] = 0.0f;
- Whalf[2] = 0.0f;
- Whalf[3] = 0.0f;
- Whalf[4] = 0.0f;
- return;
- } else {
- if (SL >= 0.0f) {
- Whalf[1] = WL[1];
- Whalf[2] = WL[2];
- Whalf[3] = WL[3];
- if (aL > vL) {
- Whalf[0] = WL[0] * powf(const_riemann_tdgp1 +
- const_riemann_gm1dgp1 / aL * vL,
- const_riemann_tdgm1);
- vhalf = const_riemann_tdgp1 * (aL + const_riemann_gm1d2 * vL) - vL;
- Whalf[4] = WL[4] * powf(const_riemann_tdgp1 +
- const_riemann_gm1dgp1 / aL * vL,
- const_riemann_tgdgm1);
- } else {
- Whalf[0] = WL[0];
- vhalf = 0.0f;
- Whalf[4] = WL[4];
- }
- } else {
- Whalf[1] = WR[1];
- Whalf[2] = WR[2];
- Whalf[3] = WR[3];
- if (-vR < aR) {
- Whalf[0] = WR[0] * powf(const_riemann_tdgp1 -
- const_riemann_gm1dgp1 / aR * vR,
- const_riemann_tdgm1);
- vhalf = const_riemann_tdgp1 * (-aR + const_riemann_gm1d2 * vR) - vR;
- Whalf[4] = WR[4] * powf(const_riemann_tdgp1 -
- const_riemann_gm1dgp1 / aR * vR,
- const_riemann_tgdgm1);
- } else {
- Whalf[0] = WR[0];
- vhalf = 0.0f;
- Whalf[4] = WR[4];
- }
- }
- }
- }
- }
-
- /* Add the velocity solution along the interface normal to the velocities */
- Whalf[1] += vhalf * n_unit[0];
- Whalf[2] += vhalf * n_unit[1];
- Whalf[3] += vhalf * n_unit[2];
-}
-
/* Solve the Riemann problem between the states WL and WR and store the result
* in Whalf
* The Riemann problem is solved in the x-direction; the velocities in the y-
@@ -456,20 +302,20 @@ __attribute__((always_inline)) INLINE static void riemann_solve_vacuum(
* @param n_unit Normal vector of the interface
*/
__attribute__((always_inline)) INLINE static void riemann_solver_solve(
- GFLOAT* WL, GFLOAT* WR, GFLOAT* Whalf, float* n_unit) {
+ float* WL, float* WR, float* Whalf, float* n_unit) {
/* velocity of the left and right state in a frame aligned with n_unit */
- GFLOAT vL, vR, vhalf;
+ float vL, vR, vhalf;
/* sound speeds */
- GFLOAT aL, aR;
+ float aL, aR;
/* variables used for finding pstar */
- GFLOAT p, pguess, fp, fpguess;
+ float p, pguess, fp, fpguess;
/* variables used for sampling the solution */
- GFLOAT u;
- GFLOAT pdpR, SR;
- GFLOAT SHR, STR;
- GFLOAT pdpL, SL;
- GFLOAT SHL, STL;
+ float u;
+ float pdpR, SR;
+ float SHR, STR;
+ float pdpL, SL;
+ float SHL, STL;
int errorFlag = 0;
/* sanity checks */
@@ -500,156 +346,155 @@ __attribute__((always_inline)) INLINE static void riemann_solver_solve(
vR = WR[1] * n_unit[0] + WR[2] * n_unit[1] + WR[3] * n_unit[2];
/* calculate sound speeds */
- aL = sqrtf(const_hydro_gamma * WL[4] / WL[0]);
- aR = sqrtf(const_hydro_gamma * WR[4] / WR[0]);
+ aL = sqrtf(hydro_gamma * WL[4] / WL[0]);
+ aR = sqrtf(hydro_gamma * WR[4] / WR[0]);
- if (!WL[0] || !WR[0]) {
- /* vacuum: we need a vacuum riemann solver */
+ /* check vacuum (generation) condition */
+ if (riemann_is_vacuum(WL, WR, vL, vR, aL, aR)) {
riemann_solve_vacuum(WL, WR, vL, vR, aL, aR, Whalf, n_unit);
return;
}
- /* check vacuum generation condition */
- if (2.0f * aL / (const_hydro_gamma - 1.0f) +
- 2.0f * aR / (const_hydro_gamma - 1.0f) <
- fabs(vL - vR)) {
- /* vacuum generation: need a vacuum riemann solver */
- riemann_solve_vacuum(WL, WR, vL, vR, aL, aR, Whalf, n_unit);
- return;
- } else {
- /* values are ok: let's find pstar (riemann_f(pstar) = 0)! */
- /* We normally use a Newton-Raphson iteration to find the zeropoint
- of riemann_f(p), but if pstar is close to 0, we risk negative p values.
- Since riemann_f(p) is undefined for negative pressures, we don't
- want this to happen.
- We therefore use Brent's method if riemann_f(0) is larger than some
- value. -5 makes the iteration fail safe while almost never invoking
- the expensive Brent solver. */
- p = 0.;
- /* obtain a first guess for p */
- pguess = riemann_guess_p(WL, WR, vL, vR, aL, aR);
- fp = riemann_f(p, WL, WR, vL, vR, aL, aR);
- fpguess = riemann_f(pguess, WL, WR, vL, vR, aL, aR);
- /* ok, pstar is close to 0, better use Brent's method... */
- /* we use Newton-Raphson until we find a suitable interval */
- if (fp * fpguess >= 0.0f) {
- /* Newton-Raphson until convergence or until suitable interval is found
- to use Brent's method */
- unsigned int counter = 0;
- while (fabs(p - pguess) > 1.e-6f * 0.5f * (p + pguess) &&
- fpguess < 0.0f) {
- p = pguess;
- pguess = pguess - fpguess / riemann_fprime(pguess, WL, WR, aL, aR);
- fpguess = riemann_f(pguess, WL, WR, vL, vR, aL, aR);
- counter++;
- if (counter > 1000) {
- error("Stuck in Newton-Raphson!\n");
- }
- }
- }
- /* As soon as there is a suitable interval: use Brent's method */
- if (1.e6 * fabs(p - pguess) > 0.5f * (p + pguess) && fpguess > 0.0f) {
- p = 0.0f;
- fp = riemann_f(p, WL, WR, vL, vR, aL, aR);
- /* use Brent's method to find the zeropoint */
- p = riemann_solve_brent(p, pguess, fp, fpguess, 1.e-6, WL, WR, vL, vR, aL,
- aR);
- } else {
+ /* values are ok: let's find pstar (riemann_f(pstar) = 0)! */
+ /* We normally use a Newton-Raphson iteration to find the zeropoint
+ of riemann_f(p), but if pstar is close to 0, we risk negative p values.
+ Since riemann_f(p) is undefined for negative pressures, we don't
+ want this to happen.
+ We therefore use Brent's method if riemann_f(0) is larger than some
+ value. -5 makes the iteration fail safe while almost never invoking
+ the expensive Brent solver. */
+ p = 0.;
+ /* obtain a first guess for p */
+ pguess = riemann_guess_p(WL, WR, vL, vR, aL, aR);
+ fp = riemann_f(p, WL, WR, vL, vR, aL, aR);
+ fpguess = riemann_f(pguess, WL, WR, vL, vR, aL, aR);
+ /* ok, pstar is close to 0, better use Brent's method... */
+ /* we use Newton-Raphson until we find a suitable interval */
+ if (fp * fpguess >= 0.0f) {
+ /* Newton-Raphson until convergence or until suitable interval is found
+ to use Brent's method */
+ unsigned int counter = 0;
+ while (fabs(p - pguess) > 1.e-6f * 0.5f * (p + pguess) && fpguess < 0.0f) {
p = pguess;
+ pguess = pguess - fpguess / riemann_fprime(pguess, WL, WR, aL, aR);
+ fpguess = riemann_f(pguess, WL, WR, vL, vR, aL, aR);
+ counter++;
+ if (counter > 1000) {
+ error("Stuck in Newton-Raphson!\n");
+ }
}
+ }
+ /* As soon as there is a suitable interval: use Brent's method */
+ if (1.e6 * fabs(p - pguess) > 0.5f * (p + pguess) && fpguess > 0.0f) {
+ p = 0.0f;
+ fp = riemann_f(p, WL, WR, vL, vR, aL, aR);
+ /* use Brent's method to find the zeropoint */
+ p = riemann_solve_brent(p, pguess, fp, fpguess, 1.e-6, WL, WR, vL, vR, aL,
+ aR);
+ } else {
+ p = pguess;
+ }
- /* calculate the velocity in the intermediate state */
- u = 0.5f * (vL + vR) +
- 0.5f * (riemann_fb(p, WR, aR) - riemann_fb(p, WL, aL));
-
- /* sample the solution */
- /* This corresponds to the flow chart in Fig. 4.14 in Toro */
- if (u < 0.0f) {
- /* advect velocity components */
- Whalf[1] = WR[1];
- Whalf[2] = WR[2];
- Whalf[3] = WR[3];
- pdpR = p / WR[4];
- if (p > WR[4]) {
- /* shockwave */
- SR =
- vR + aR * sqrtf(const_riemann_gp1d2g * pdpR + const_riemann_gm1d2g);
- if (SR > 0.0f) {
- Whalf[0] = WR[0] * (pdpR + const_riemann_gm1dgp1) /
- (const_riemann_gm1dgp1 * pdpR + 1.0f);
+ /* calculate the velocity in the intermediate state */
+ u = 0.5f * (vL + vR) + 0.5f * (riemann_fb(p, WR, aR) - riemann_fb(p, WL, aL));
+
+ /* sample the solution */
+ /* This corresponds to the flow chart in Fig. 4.14 in Toro */
+ if (u < 0.0f) {
+ /* advect velocity components */
+ Whalf[1] = WR[1];
+ Whalf[2] = WR[2];
+ Whalf[3] = WR[3];
+ pdpR = p / WR[4];
+ if (p > WR[4]) {
+ /* shockwave */
+ SR = vR +
+ aR * sqrtf(hydro_gamma_plus_one_over_two_gamma * pdpR +
+ hydro_gamma_minus_one_over_two_gamma);
+ if (SR > 0.0f) {
+ Whalf[0] = WR[0] * (pdpR + hydro_gamma_minus_one_over_gamma_plus_one) /
+ (hydro_gamma_minus_one_over_gamma_plus_one * pdpR + 1.0f);
+ vhalf = u - vR;
+ Whalf[4] = p;
+ } else {
+ Whalf[0] = WR[0];
+ vhalf = 0.0f;
+ Whalf[4] = WR[4];
+ }
+ } else {
+ /* rarefaction wave */
+ SHR = vR + aR;
+ if (SHR > 0.0f) {
+ STR = u + aR * pow_gamma_minus_one_over_two_gamma(pdpR);
+ if (STR <= 0.0f) {
+ Whalf[0] =
+ WR[0] * pow_two_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one -
+ hydro_gamma_minus_one_over_gamma_plus_one / aR * vR);
+ vhalf = hydro_two_over_gamma_plus_one *
+ (-aR + hydro_gamma_minus_one_over_two * vR) -
+ vR;
+ Whalf[4] =
+ WR[4] * pow_two_gamma_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one -
+ hydro_gamma_minus_one_over_gamma_plus_one / aR * vR);
+ } else {
+ Whalf[0] = WR[0] * pow_one_over_gamma(pdpR);
vhalf = u - vR;
Whalf[4] = p;
- } else {
- Whalf[0] = WR[0];
- vhalf = 0.0f;
- Whalf[4] = WR[4];
}
} else {
- /* rarefaction wave */
- SHR = vR + aR;
- if (SHR > 0.0f) {
- STR = u + aR * powf(pdpR, const_riemann_gm1d2g);
- if (STR <= 0.0f) {
- Whalf[0] = WR[0] * powf(const_riemann_tdgp1 -
- const_riemann_gm1dgp1 / aR * vR,
- const_riemann_tdgm1);
- vhalf = const_riemann_tdgp1 * (-aR + const_riemann_gm1d2 * vR) - vR;
- Whalf[4] = WR[4] * powf(const_riemann_tdgp1 -
- const_riemann_gm1dgp1 / aR * vR,
- const_riemann_tgdgm1);
- } else {
- Whalf[0] = WR[0] * powf(pdpR, const_riemann_ginv);
- vhalf = u - vR;
- Whalf[4] = p;
- }
- } else {
- Whalf[0] = WR[0];
- vhalf = 0.0f;
- Whalf[4] = WR[4];
- }
+ Whalf[0] = WR[0];
+ vhalf = 0.0f;
+ Whalf[4] = WR[4];
+ }
+ }
+ } else {
+ Whalf[1] = WL[1];
+ Whalf[2] = WL[2];
+ Whalf[3] = WL[3];
+ pdpL = p / WL[4];
+ if (p > WL[4]) {
+ /* shockwave */
+ SL = vL -
+ aL * sqrtf(hydro_gamma_plus_one_over_two_gamma * pdpL +
+ hydro_gamma_minus_one_over_two_gamma);
+ if (SL < 0.0f) {
+ Whalf[0] = WL[0] * (pdpL + hydro_gamma_minus_one_over_gamma_plus_one) /
+ (hydro_gamma_minus_one_over_gamma_plus_one * pdpL + 1.0f);
+ vhalf = u - vL;
+ Whalf[4] = p;
+ } else {
+ Whalf[0] = WL[0];
+ vhalf = 0.0f;
+ Whalf[4] = WL[4];
}
} else {
- Whalf[1] = WL[1];
- Whalf[2] = WL[2];
- Whalf[3] = WL[3];
- pdpL = p / WL[4];
- if (p > WL[4]) {
- /* shockwave */
- SL =
- vL - aL * sqrtf(const_riemann_gp1d2g * pdpL + const_riemann_gm1d2g);
- if (SL < 0.0f) {
- Whalf[0] = WL[0] * (pdpL + const_riemann_gm1dgp1) /
- (const_riemann_gm1dgp1 * pdpL + 1.0f);
+ /* rarefaction wave */
+ SHL = vL - aL;
+ if (SHL < 0.0f) {
+ STL = u - aL * pow_gamma_minus_one_over_two_gamma(pdpL);
+ if (STL > 0.0f) {
+ Whalf[0] =
+ WL[0] * pow_two_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one +
+ hydro_gamma_minus_one_over_gamma_plus_one / aL * vL);
+ vhalf = hydro_two_over_gamma_plus_one *
+ (aL + hydro_gamma_minus_one_over_two * vL) -
+ vL;
+ Whalf[4] =
+ WL[4] * pow_two_gamma_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one +
+ hydro_gamma_minus_one_over_gamma_plus_one / aL * vL);
+ } else {
+ Whalf[0] = WL[0] * pow_one_over_gamma(pdpL);
vhalf = u - vL;
Whalf[4] = p;
- } else {
- Whalf[0] = WL[0];
- vhalf = 0.0f;
- Whalf[4] = WL[4];
}
} else {
- /* rarefaction wave */
- SHL = vL - aL;
- if (SHL < 0.0f) {
- STL = u - aL * powf(pdpL, const_riemann_gm1d2g);
- if (STL > 0.0f) {
- Whalf[0] = WL[0] * powf(const_riemann_tdgp1 +
- const_riemann_gm1dgp1 / aL * vL,
- const_riemann_tdgm1);
- vhalf = const_riemann_tdgp1 * (aL + const_riemann_gm1d2 * vL) - vL;
- Whalf[4] = WL[4] * powf(const_riemann_tdgp1 +
- const_riemann_gm1dgp1 / aL * vL,
- const_riemann_tgdgm1);
- } else {
- Whalf[0] = WL[0] * powf(pdpL, const_riemann_ginv);
- vhalf = u - vL;
- Whalf[4] = p;
- }
- } else {
- Whalf[0] = WL[0];
- vhalf = 0.0f;
- Whalf[4] = WL[4];
- }
+ Whalf[0] = WL[0];
+ vhalf = 0.0f;
+ Whalf[4] = WL[4];
}
}
}
@@ -661,10 +506,10 @@ __attribute__((always_inline)) INLINE static void riemann_solver_solve(
}
__attribute__((always_inline)) INLINE static void riemann_solve_for_flux(
- GFLOAT* Wi, GFLOAT* Wj, float* n_unit, float* vij, GFLOAT* totflux) {
+ float* Wi, float* Wj, float* n_unit, float* vij, float* totflux) {
- GFLOAT Whalf[5];
- GFLOAT flux[5][3];
+ float Whalf[5];
+ float flux[5][3];
float vtot[3];
float rhoe;
@@ -690,7 +535,7 @@ __attribute__((always_inline)) INLINE static void riemann_solve_for_flux(
/* eqn. (15) */
/* F_P = \rho e ( \vec{v} - \vec{v_{ij}} ) + P \vec{v} */
/* \rho e = P / (\gamma-1) + 1/2 \rho \vec{v}^2 */
- rhoe = Whalf[4] / (const_hydro_gamma - 1.0f) +
+ rhoe = Whalf[4] / hydro_gamma_minus_one +
0.5f * Whalf[0] *
(vtot[0] * vtot[0] + vtot[1] * vtot[1] + vtot[2] * vtot[2]);
flux[4][0] = rhoe * Whalf[1] + Whalf[4] * vtot[0];
diff --git a/src/riemann/riemann_hllc.h b/src/riemann/riemann_hllc.h
index 6c583f6410f53ed64d630082926d816129768fab..fdc22ce05b8d63bdba66e530d1a5a968801a9f10 100644
--- a/src/riemann/riemann_hllc.h
+++ b/src/riemann/riemann_hllc.h
@@ -20,13 +20,16 @@
#ifndef SWIFT_RIEMANN_HLLC_H
#define SWIFT_RIEMANN_HLLC_H
+#include "adiabatic_index.h"
+#include "riemann_vacuum.h"
+
__attribute__((always_inline)) INLINE static void riemann_solve_for_flux(
- GFLOAT *WL, GFLOAT *WR, float *n, float *vij, GFLOAT *totflux) {
+ float *WL, float *WR, float *n, float *vij, float *totflux) {
- GFLOAT uL, uR, aL, aR;
- GFLOAT rhobar, abar, pPVRS, pstar, qL, qR, SL, SR, Sstar;
- GFLOAT v2, eL, eR;
- GFLOAT UstarL[5], UstarR[5];
+ float uL, uR, aL, aR;
+ float rhobar, abar, pPVRS, pstar, qL, qR, SL, SR, Sstar;
+ float v2, eL, eR;
+ float UstarL[5], UstarR[5];
/* Handle pure vacuum */
if (!WL[0] && !WR[0]) {
@@ -41,16 +44,13 @@ __attribute__((always_inline)) INLINE static void riemann_solve_for_flux(
/* STEP 0: obtain velocity in interface frame */
uL = WL[1] * n[0] + WL[2] * n[1] + WL[3] * n[2];
uR = WR[1] * n[0] + WR[2] * n[1] + WR[3] * n[2];
- aL = sqrtf(const_hydro_gamma * WL[4] / WL[0]);
- aR = sqrtf(const_hydro_gamma * WR[4] / WR[0]);
+ aL = sqrtf(hydro_gamma * WL[4] / WL[0]);
+ aR = sqrtf(hydro_gamma * WR[4] / WR[0]);
/* Handle vacuum: vacuum does not require iteration and is always exact */
- if (!WL[0] || !WR[0]) {
- error("Vacuum not yet supported");
- }
- if (2. * aL / (const_hydro_gamma - 1.) + 2. * aR / (const_hydro_gamma - 1.) <
- fabs(uL - uR)) {
- error("Vacuum not yet supported");
+ if (riemann_is_vacuum(WL, WR, uL, uR, aL, aR)) {
+ riemann_solve_vacuum_flux(WL, WR, uL, uR, aL, aR, n, vij, totflux);
+ return;
}
/* STEP 1: pressure estimate */
@@ -64,14 +64,12 @@ __attribute__((always_inline)) INLINE static void riemann_solve_for_flux(
qL = 1.;
if (pstar > WL[4]) {
qL = sqrtf(1. +
- 0.5 * (const_hydro_gamma + 1.) / const_hydro_gamma *
- (pstar / WL[4] - 1.));
+ 0.5 * (hydro_gamma + 1.) / hydro_gamma * (pstar / WL[4] - 1.));
}
qR = 1.;
if (pstar > WR[4]) {
qR = sqrtf(1. +
- 0.5 * (const_hydro_gamma + 1.) / const_hydro_gamma *
- (pstar / WR[4] - 1.));
+ 0.5 * (hydro_gamma + 1.) / hydro_gamma * (pstar / WR[4] - 1.));
}
SL = uL - aL * qL;
SR = uR + aR * qR;
@@ -86,9 +84,9 @@ __attribute__((always_inline)) INLINE static void riemann_solve_for_flux(
(not rotated to interface frame) */
totflux[1] = WL[0] * WL[1] * uL + WL[4] * n[0];
totflux[2] = WL[0] * WL[2] * uL + WL[4] * n[1];
- totflux[3] = WL[0] * WL[2] * uL + WL[4] * n[2];
+ totflux[3] = WL[0] * WL[3] * uL + WL[4] * n[2];
v2 = WL[1] * WL[1] + WL[2] * WL[2] + WL[3] * WL[3];
- eL = WL[4] / (const_hydro_gamma - 1.) / WL[0] + 0.5 * v2;
+ eL = WL[4] / hydro_gamma_minus_one / WL[0] + 0.5 * v2;
totflux[4] = WL[0] * eL * uL + WL[4] * uL;
if (SL < 0.) {
/* add flux FstarL */
@@ -118,7 +116,7 @@ __attribute__((always_inline)) INLINE static void riemann_solve_for_flux(
totflux[2] = WR[0] * WR[2] * uR + WR[4] * n[1];
totflux[3] = WR[0] * WR[3] * uR + WR[4] * n[2];
v2 = WR[1] * WR[1] + WR[2] * WR[2] + WR[3] * WR[3];
- eR = WR[4] / (const_hydro_gamma - 1.) / WR[0] + 0.5 * v2;
+ eR = WR[4] / hydro_gamma_minus_one / WR[0] + 0.5 * v2;
totflux[4] = WR[0] * eR * uR + WR[4] * uR;
if (SR > 0.) {
/* add flux FstarR */
diff --git a/src/riemann/riemann_trrs.h b/src/riemann/riemann_trrs.h
index efdbfb59877c09a59d535a4785ad74620c0f3651..b13a76b4c57af548497780e974e5c9ee3a721fac 100644
--- a/src/riemann/riemann_trrs.h
+++ b/src/riemann/riemann_trrs.h
@@ -20,19 +20,8 @@
#ifndef SWIFT_RIEMANN_TRRS_H
#define SWIFT_RIEMANN_TRRS_H
-/* frequently used combinations of const_hydro_gamma */
-#define const_riemann_gp1d2g \
- (0.5f * (const_hydro_gamma + 1.0f) / const_hydro_gamma)
-#define const_riemann_gm1d2g \
- (0.5f * (const_hydro_gamma - 1.0f) / const_hydro_gamma)
-#define const_riemann_gm1dgp1 \
- ((const_hydro_gamma - 1.0f) / (const_hydro_gamma + 1.0f))
-#define const_riemann_tdgp1 (2.0f / (const_hydro_gamma + 1.0f))
-#define const_riemann_tdgm1 (2.0f / (const_hydro_gamma - 1.0f))
-#define const_riemann_gm1d2 (0.5f * (const_hydro_gamma - 1.0f))
-#define const_riemann_tgdgm1 \
- (2.0f * const_hydro_gamma / (const_hydro_gamma - 1.0f))
-#define const_riemann_ginv (1.0f / const_hydro_gamma)
+#include "adiabatic_index.h"
+#include "riemann_vacuum.h"
/**
* @brief Solve the Riemann problem using the Two Rarefaction Riemann Solver
@@ -50,31 +39,39 @@
* @param n_unit Normal vector of the interface
*/
__attribute__((always_inline)) INLINE static void riemann_solver_solve(
- GFLOAT* WL, GFLOAT* WR, GFLOAT* Whalf, float* n_unit) {
- GFLOAT aL, aR;
- GFLOAT PLR;
- GFLOAT vL, vR;
- GFLOAT ustar, pstar;
- GFLOAT vhalf;
- GFLOAT pdpR, SHR, STR;
- GFLOAT pdpL, SHL, STL;
+ float* WL, float* WR, float* Whalf, float* n_unit) {
+ float aL, aR;
+ float PLR;
+ float vL, vR;
+ float ustar, pstar;
+ float vhalf;
+ float pdpR, SHR, STR;
+ float pdpL, SHL, STL;
/* calculate the velocities along the interface normal */
vL = WL[1] * n_unit[0] + WL[2] * n_unit[1] + WL[3] * n_unit[2];
vR = WR[1] * n_unit[0] + WR[2] * n_unit[1] + WR[3] * n_unit[2];
/* calculate the sound speeds */
- aL = sqrtf(const_hydro_gamma * WL[4] / WL[0]);
- aR = sqrtf(const_hydro_gamma * WR[4] / WR[0]);
+ aL = sqrtf(hydro_gamma * WL[4] / WL[0]);
+ aR = sqrtf(hydro_gamma * WR[4] / WR[0]);
+
+ if (riemann_is_vacuum(WL, WR, vL, vR, aL, aR)) {
+ riemann_solve_vacuum(WL, WR, vL, vR, aL, aR, Whalf, n_unit);
+ return;
+ }
/* calculate the velocity and pressure in the intermediate state */
- PLR = pow(WL[4] / WR[4], const_riemann_gm1d2g);
- ustar = (PLR * vL / aL + vR / aR + const_riemann_tdgm1 * (PLR - 1.0f)) /
+ PLR = pow_gamma_minus_one_over_two_gamma(WL[4] / WR[4]);
+ ustar = (PLR * vL / aL + vR / aR +
+ hydro_two_over_gamma_minus_one * (PLR - 1.0f)) /
(PLR / aL + 1.0f / aR);
- pstar = 0.5f * (WL[4] * pow(1.0f + const_riemann_gm1d2 / aL * (vL - ustar),
- const_riemann_tgdgm1) +
- WR[4] * pow(1.0f + const_riemann_gm1d2 / aR * (ustar - vR),
- const_riemann_tgdgm1));
+ pstar =
+ 0.5f *
+ (WL[4] * pow_two_gamma_over_gamma_minus_one(
+ 1.0f + hydro_gamma_minus_one_over_two / aL * (vL - ustar)) +
+ WR[4] * pow_two_gamma_over_gamma_minus_one(
+ 1.0f + hydro_gamma_minus_one_over_two / aR * (ustar - vR)));
/* sample the solution */
if (ustar < 0.0f) {
@@ -86,17 +83,21 @@ __attribute__((always_inline)) INLINE static void riemann_solver_solve(
/* always a rarefaction wave, that's the approximation */
SHR = vR + aR;
if (SHR > 0.0f) {
- STR = ustar + aR * pow(pdpR, const_riemann_gm1d2g);
+ STR = ustar + aR * pow_gamma_minus_one_over_two_gamma(pdpR);
if (STR <= 0.0f) {
Whalf[0] =
- WR[0] * pow(const_riemann_tdgp1 - const_riemann_gm1dgp1 / aR * vR,
- const_riemann_tdgm1);
- vhalf = const_riemann_tdgp1 * (-aR + const_riemann_gm1d2 * vR) - vR;
+ WR[0] * pow_two_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one -
+ hydro_gamma_minus_one_over_gamma_plus_one / aR * vR);
+ vhalf = hydro_two_over_gamma_plus_one *
+ (-aR + hydro_gamma_minus_one_over_two * vR) -
+ vR;
Whalf[4] =
- WR[4] * pow(const_riemann_tdgp1 - const_riemann_gm1dgp1 / aR * vR,
- const_riemann_tgdgm1);
+ WR[4] * pow_two_gamma_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one -
+ hydro_gamma_minus_one_over_gamma_plus_one / aR * vR);
} else {
- Whalf[0] = WR[0] * pow(pdpR, const_riemann_ginv);
+ Whalf[0] = WR[0] * pow_one_over_gamma(pdpR);
vhalf = ustar - vR;
Whalf[4] = pstar;
}
@@ -114,17 +115,21 @@ __attribute__((always_inline)) INLINE static void riemann_solver_solve(
/* rarefaction wave */
SHL = vL - aL;
if (SHL < 0.0f) {
- STL = ustar - aL * pow(pdpL, const_riemann_gm1d2g);
+ STL = ustar - aL * pow_gamma_minus_one_over_two_gamma(pdpL);
if (STL > 0.0f) {
Whalf[0] =
- WL[0] * pow(const_riemann_tdgp1 + const_riemann_gm1dgp1 / aL * vL,
- const_riemann_tdgm1);
- vhalf = const_riemann_tdgp1 * (aL + const_riemann_gm1d2 * vL) - vL;
+ WL[0] * pow_two_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one +
+ hydro_gamma_minus_one_over_gamma_plus_one / aL * vL);
+ vhalf = hydro_two_over_gamma_plus_one *
+ (aL + hydro_gamma_minus_one_over_two * vL) -
+ vL;
Whalf[4] =
- WL[4] * pow(const_riemann_tdgp1 + const_riemann_gm1dgp1 / aL * vL,
- const_riemann_tgdgm1);
+ WL[4] * pow_two_gamma_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one +
+ hydro_gamma_minus_one_over_gamma_plus_one / aL * vL);
} else {
- Whalf[0] = WL[0] * pow(pdpL, const_riemann_ginv);
+ Whalf[0] = WL[0] * pow_one_over_gamma(pdpL);
vhalf = ustar - vL;
Whalf[4] = pstar;
}
@@ -141,4 +146,53 @@ __attribute__((always_inline)) INLINE static void riemann_solver_solve(
Whalf[3] += vhalf * n_unit[2];
}
+__attribute__((always_inline)) INLINE static void riemann_solve_for_flux(
+ float* Wi, float* Wj, float* n_unit, float* vij, float* totflux) {
+
+ float Whalf[5];
+ float flux[5][3];
+ float vtot[3];
+ float rhoe;
+
+ riemann_solver_solve(Wi, Wj, Whalf, n_unit);
+
+ flux[0][0] = Whalf[0] * Whalf[1];
+ flux[0][1] = Whalf[0] * Whalf[2];
+ flux[0][2] = Whalf[0] * Whalf[3];
+
+ vtot[0] = Whalf[1] + vij[0];
+ vtot[1] = Whalf[2] + vij[1];
+ vtot[2] = Whalf[3] + vij[2];
+ flux[1][0] = Whalf[0] * vtot[0] * Whalf[1] + Whalf[4];
+ flux[1][1] = Whalf[0] * vtot[0] * Whalf[2];
+ flux[1][2] = Whalf[0] * vtot[0] * Whalf[3];
+ flux[2][0] = Whalf[0] * vtot[1] * Whalf[1];
+ flux[2][1] = Whalf[0] * vtot[1] * Whalf[2] + Whalf[4];
+ flux[2][2] = Whalf[0] * vtot[1] * Whalf[3];
+ flux[3][0] = Whalf[0] * vtot[2] * Whalf[1];
+ flux[3][1] = Whalf[0] * vtot[2] * Whalf[2];
+ flux[3][2] = Whalf[0] * vtot[2] * Whalf[3] + Whalf[4];
+
+ /* eqn. (15) */
+ /* F_P = \rho e ( \vec{v} - \vec{v_{ij}} ) + P \vec{v} */
+ /* \rho e = P / (\gamma-1) + 1/2 \rho \vec{v}^2 */
+ rhoe = Whalf[4] / hydro_gamma_minus_one +
+ 0.5f * Whalf[0] *
+ (vtot[0] * vtot[0] + vtot[1] * vtot[1] + vtot[2] * vtot[2]);
+ flux[4][0] = rhoe * Whalf[1] + Whalf[4] * vtot[0];
+ flux[4][1] = rhoe * Whalf[2] + Whalf[4] * vtot[1];
+ flux[4][2] = rhoe * Whalf[3] + Whalf[4] * vtot[2];
+
+ totflux[0] =
+ flux[0][0] * n_unit[0] + flux[0][1] * n_unit[1] + flux[0][2] * n_unit[2];
+ totflux[1] =
+ flux[1][0] * n_unit[0] + flux[1][1] * n_unit[1] + flux[1][2] * n_unit[2];
+ totflux[2] =
+ flux[2][0] * n_unit[0] + flux[2][1] * n_unit[1] + flux[2][2] * n_unit[2];
+ totflux[3] =
+ flux[3][0] * n_unit[0] + flux[3][1] * n_unit[1] + flux[3][2] * n_unit[2];
+ totflux[4] =
+ flux[4][0] * n_unit[0] + flux[4][1] * n_unit[1] + flux[4][2] * n_unit[2];
+}
+
#endif /* SWIFT_RIEMANN_TRRS_H */
diff --git a/src/riemann/riemann_vacuum.h b/src/riemann/riemann_vacuum.h
new file mode 100644
index 0000000000000000000000000000000000000000..743abb910193380793ccdf3d7eddbcedc4968691
--- /dev/null
+++ b/src/riemann/riemann_vacuum.h
@@ -0,0 +1,254 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Coypright (c) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com)
+ *
+ * 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_RIEMANN_VACUUM_H
+#define SWIFT_RIEMANN_VACUUM_H
+
+/**
+ * @brief Check if the given input states are vacuum or will generate vacuum
+ */
+__attribute__((always_inline)) INLINE static int riemann_is_vacuum(
+ float* WL, float* WR, float vL, float vR, float aL, float aR) {
+
+ /* vacuum */
+ if (!WL[0] || !WR[0]) {
+ return 1;
+ }
+ /* vacuum generation */
+ if (2.0f * aL / hydro_gamma_minus_one + 2.0f * aR / hydro_gamma_minus_one <=
+ vR - vL) {
+ return 1;
+ }
+
+ /* no vacuum */
+ return 0;
+}
+
+/**
+ * @brief Vacuum Riemann solver, based on section 4.6 in Toro
+ *
+ * @param WL The left state vector
+ * @param WR The right state vector
+ * @param vL The left velocity along the interface normal
+ * @param vR The right velocity along the interface normal
+ * @param aL The left sound speed
+ * @param aR The right sound speed
+ * @param Whalf Empty state vector to store the solution in
+ * @param n_unit Normal vector of the interface
+ */
+__attribute__((always_inline)) INLINE static void riemann_solve_vacuum(
+ float* WL, float* WR, float vL, float vR, float aL, float aR, float* Whalf,
+ float* n_unit) {
+
+ float SL, SR;
+ float vhalf;
+
+ if (!WR[0] && !WL[0]) {
+ /* if both states are vacuum, the solution is also vacuum */
+ Whalf[0] = 0.0f;
+ Whalf[1] = 0.0f;
+ Whalf[2] = 0.0f;
+ Whalf[3] = 0.0f;
+ Whalf[4] = 0.0f;
+ return;
+ }
+ if (!WR[0]) {
+ Whalf[1] = WL[1];
+ Whalf[2] = WL[2];
+ Whalf[3] = WL[3];
+ /* vacuum right state */
+ if (vL < aL) {
+ SL = vL + hydro_two_over_gamma_minus_one * aL;
+ if (SL > 0.0f) {
+ Whalf[0] =
+ WL[0] * pow_two_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one +
+ hydro_gamma_minus_one_over_gamma_plus_one / aL * vL);
+ vhalf = hydro_two_over_gamma_plus_one *
+ (aL + hydro_gamma_minus_one_over_two * vL) -
+ vL;
+ Whalf[4] =
+ WL[4] * pow_two_gamma_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one +
+ hydro_gamma_minus_one_over_gamma_plus_one / aL * vL);
+ } else {
+ Whalf[0] = 0.0f;
+ Whalf[1] = 0.0f;
+ Whalf[2] = 0.0f;
+ Whalf[3] = 0.0f;
+ Whalf[4] = 0.0f;
+ return;
+ }
+ } else {
+ Whalf[0] = WL[0];
+ vhalf = 0.0f;
+ Whalf[4] = WL[4];
+ }
+ } else {
+ if (!WL[0]) {
+ Whalf[1] = WR[1];
+ Whalf[2] = WR[2];
+ Whalf[3] = WR[3];
+ /* vacuum left state */
+ if (-aR < vR) {
+ SR = vR - hydro_two_over_gamma_minus_one * aR;
+ if (SR >= 0.0f) {
+ Whalf[0] = 0.0f;
+ Whalf[1] = 0.0f;
+ Whalf[2] = 0.0f;
+ Whalf[3] = 0.0f;
+ Whalf[4] = 0.0f;
+ return;
+ } else {
+ Whalf[0] =
+ WR[0] * pow_two_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one -
+ hydro_gamma_minus_one_over_gamma_plus_one / aR * vR);
+ vhalf = hydro_two_over_gamma_plus_one *
+ (-aR + hydro_gamma_minus_one_over_two * vR) -
+ vR;
+ Whalf[4] =
+ WR[4] * pow_two_gamma_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one -
+ hydro_gamma_minus_one_over_gamma_plus_one / aR * vR);
+ }
+ } else {
+ Whalf[0] = WR[0];
+ vhalf = 0.0f;
+ Whalf[4] = WR[4];
+ }
+ } else {
+ /* vacuum generation */
+ SR = vR - hydro_two_over_gamma_minus_one * aR;
+ SL = vL + hydro_two_over_gamma_minus_one * aL;
+ if (SR > 0.0f && SL < 0.0f) {
+ Whalf[0] = 0.0f;
+ Whalf[1] = 0.0f;
+ Whalf[2] = 0.0f;
+ Whalf[3] = 0.0f;
+ Whalf[4] = 0.0f;
+ return;
+ } else {
+ if (SL >= 0.0f) {
+ Whalf[1] = WL[1];
+ Whalf[2] = WL[2];
+ Whalf[3] = WL[3];
+ if (aL > vL) {
+ Whalf[0] = WL[0] *
+ pow_two_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one +
+ hydro_gamma_minus_one_over_gamma_plus_one / aL * vL);
+ vhalf = hydro_two_over_gamma_plus_one *
+ (aL + hydro_gamma_minus_one_over_two * vL) -
+ vL;
+ Whalf[4] = WL[4] *
+ pow_two_gamma_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one +
+ hydro_gamma_minus_one_over_gamma_plus_one / aL * vL);
+ } else {
+ Whalf[0] = WL[0];
+ vhalf = 0.0f;
+ Whalf[4] = WL[4];
+ }
+ } else {
+ Whalf[1] = WR[1];
+ Whalf[2] = WR[2];
+ Whalf[3] = WR[3];
+ if (-aR < vR) {
+ Whalf[0] = WR[0] *
+ pow_two_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one -
+ hydro_gamma_minus_one_over_gamma_plus_one / aR * vR);
+ vhalf = hydro_two_over_gamma_plus_one *
+ (-aR + hydro_gamma_minus_one_over_two * vR) -
+ vR;
+ Whalf[4] = WR[4] *
+ pow_two_gamma_over_gamma_minus_one(
+ hydro_two_over_gamma_plus_one -
+ hydro_gamma_minus_one_over_gamma_plus_one / aR * vR);
+ } else {
+ Whalf[0] = WR[0];
+ vhalf = 0.0f;
+ Whalf[4] = WR[4];
+ }
+ }
+ }
+ }
+ }
+
+ /* Add the velocity solution along the interface normal to the velocities */
+ Whalf[1] += vhalf * n_unit[0];
+ Whalf[2] += vhalf * n_unit[1];
+ Whalf[3] += vhalf * n_unit[2];
+}
+
+/**
+ * @brief Solve the vacuum Riemann problem and return the fluxes
+ */
+__attribute__((always_inline)) INLINE static void riemann_solve_vacuum_flux(
+ float* WL, float* WR, float vL, float vR, float aL, float aR, float* n_unit,
+ float* vij, float* totflux) {
+
+ float Whalf[5];
+ float flux[5][3];
+ float vtot[3];
+ float rhoe;
+
+ riemann_solve_vacuum(WL, WR, vL, vR, aL, aR, Whalf, n_unit);
+
+ flux[0][0] = Whalf[0] * Whalf[1];
+ flux[0][1] = Whalf[0] * Whalf[2];
+ flux[0][2] = Whalf[0] * Whalf[3];
+
+ vtot[0] = Whalf[1] + vij[0];
+ vtot[1] = Whalf[2] + vij[1];
+ vtot[2] = Whalf[3] + vij[2];
+ flux[1][0] = Whalf[0] * vtot[0] * Whalf[1] + Whalf[4];
+ flux[1][1] = Whalf[0] * vtot[0] * Whalf[2];
+ flux[1][2] = Whalf[0] * vtot[0] * Whalf[3];
+ flux[2][0] = Whalf[0] * vtot[1] * Whalf[1];
+ flux[2][1] = Whalf[0] * vtot[1] * Whalf[2] + Whalf[4];
+ flux[2][2] = Whalf[0] * vtot[1] * Whalf[3];
+ flux[3][0] = Whalf[0] * vtot[2] * Whalf[1];
+ flux[3][1] = Whalf[0] * vtot[2] * Whalf[2];
+ flux[3][2] = Whalf[0] * vtot[2] * Whalf[3] + Whalf[4];
+
+ /* eqn. (15) */
+ /* F_P = \rho e ( \vec{v} - \vec{v_{ij}} ) + P \vec{v} */
+ /* \rho e = P / (\gamma-1) + 1/2 \rho \vec{v}^2 */
+ rhoe = Whalf[4] / hydro_gamma_minus_one +
+ 0.5f * Whalf[0] *
+ (vtot[0] * vtot[0] + vtot[1] * vtot[1] + vtot[2] * vtot[2]);
+ flux[4][0] = rhoe * Whalf[1] + Whalf[4] * vtot[0];
+ flux[4][1] = rhoe * Whalf[2] + Whalf[4] * vtot[1];
+ flux[4][2] = rhoe * Whalf[3] + Whalf[4] * vtot[2];
+
+ totflux[0] =
+ flux[0][0] * n_unit[0] + flux[0][1] * n_unit[1] + flux[0][2] * n_unit[2];
+ totflux[1] =
+ flux[1][0] * n_unit[0] + flux[1][1] * n_unit[1] + flux[1][2] * n_unit[2];
+ totflux[2] =
+ flux[2][0] * n_unit[0] + flux[2][1] * n_unit[1] + flux[2][2] * n_unit[2];
+ totflux[3] =
+ flux[3][0] * n_unit[0] + flux[3][1] * n_unit[1] + flux[3][2] * n_unit[2];
+ totflux[4] =
+ flux[4][0] * n_unit[0] + flux[4][1] * n_unit[1] + flux[4][2] * n_unit[2];
+}
+
+#endif /* SWIFT_RIEMANN_VACUUM_H */
diff --git a/src/runner.c b/src/runner.c
index 8434b3445bf3cebf6d7f5d1110c54a9ccdefb5a1..57a6b9de7017ccb953eaaf05d9880cc440bdd974 100644
--- a/src/runner.c
+++ b/src/runner.c
@@ -82,6 +82,13 @@ const char runner_flip[27] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0,
#define FUNCTION density
#include "runner_doiact.h"
+/* Import the gradient loop functions (if required). */
+#ifdef EXTRA_HYDRO_LOOP
+#undef FUNCTION
+#define FUNCTION gradient
+#include "runner_doiact.h"
+#endif
+
/* Import the force loop functions. */
#undef FUNCTION
#define FUNCTION force
@@ -422,7 +429,49 @@ void runner_do_init(struct runner *r, struct cell *c, int timer) {
}
/**
- * @brief Intermediate task between density and force
+ * @brief Intermediate task after the gradient loop that does final operations
+ * on the gradient quantities and optionally slope limits the gradients
+ *
+ * @param r The runner thread.
+ * @param c The cell.
+ */
+void runner_do_extra_ghost(struct runner *r, struct cell *c) {
+
+#ifdef EXTRA_HYDRO_LOOP
+
+ struct part *restrict parts = c->parts;
+ const int count = c->count;
+ const int ti_current = r->e->ti_current;
+
+ /* Recurse? */
+ if (c->split) {
+ for (int k = 0; k < 8; k++)
+ if (c->progeny[k] != NULL) runner_do_extra_ghost(r, c->progeny[k]);
+ return;
+ } else {
+
+ /* Loop over the parts in this cell. */
+ for (int i = 0; i < count; i++) {
+
+ /* Get a direct pointer on the part. */
+ struct part *restrict p = &parts[i];
+
+ if (p->ti_end <= ti_current) {
+
+ /* Get ready for a force calculation */
+ hydro_end_gradient(p);
+ }
+ }
+ }
+
+#else
+ error("SWIFT was not compiled with the extra hydro loop activated.");
+#endif
+}
+
+/**
+ * @brief Intermediate task after the density to check that the smoothing
+ * lengths are correct.
*
* @param r The runner thread.
* @param c The cell.
@@ -613,7 +662,7 @@ static void runner_do_drift(struct cell *c, struct engine *e) {
if (!c->split) {
/* Loop over all the g-particles in the cell */
- const int nr_gparts = c->gcount;
+ const size_t nr_gparts = c->gcount;
for (size_t k = 0; k < nr_gparts; k++) {
/* Get a handle on the gpart. */
@@ -657,7 +706,8 @@ static void runner_do_drift(struct cell *c, struct engine *e) {
const float v[3] = {xp->v_full[0] + p->a_hydro[0] * half_dt,
xp->v_full[1] + p->a_hydro[1] * half_dt,
xp->v_full[2] + p->a_hydro[2] * half_dt};
- const float m = p->mass;
+
+ const float m = hydro_get_mass(p);
/* Collect mass */
mass += m;
@@ -1089,13 +1139,15 @@ void *runner_main(void *data) {
struct cell *ci = t->ci;
struct cell *cj = t->cj;
t->rid = r->cpuid;
- t->last_rid = r->cpuid;
/* Different types of tasks... */
switch (t->type) {
case task_type_self:
- if (t->subtype == task_subtype_density)
- runner_doself1_density(r, ci);
+ if (t->subtype == task_subtype_density) runner_doself1_density(r, ci);
+#ifdef EXTRA_HYDRO_LOOP
+ else if (t->subtype == task_subtype_gradient)
+ runner_doself1_gradient(r, ci);
+#endif
else if (t->subtype == task_subtype_force)
runner_doself2_force(r, ci);
else if (t->subtype == task_subtype_grav)
@@ -1106,6 +1158,10 @@ void *runner_main(void *data) {
case task_type_pair:
if (t->subtype == task_subtype_density)
runner_dopair1_density(r, ci, cj);
+#ifdef EXTRA_HYDRO_LOOP
+ else if (t->subtype == task_subtype_gradient)
+ runner_dopair1_gradient(r, ci, cj);
+#endif
else if (t->subtype == task_subtype_force)
runner_dopair2_force(r, ci, cj);
else if (t->subtype == task_subtype_grav)
@@ -1119,6 +1175,10 @@ void *runner_main(void *data) {
case task_type_sub_self:
if (t->subtype == task_subtype_density)
runner_dosub_self1_density(r, ci, 1);
+#ifdef EXTRA_HYDRO_LOOP
+ else if (t->subtype == task_subtype_gradient)
+ runner_dosub_self1_gradient(r, ci, 1);
+#endif
else if (t->subtype == task_subtype_force)
runner_dosub_self2_force(r, ci, 1);
else if (t->subtype == task_subtype_grav)
@@ -1129,6 +1189,10 @@ void *runner_main(void *data) {
case task_type_sub_pair:
if (t->subtype == task_subtype_density)
runner_dosub_pair1_density(r, ci, cj, t->flags, 1);
+#ifdef EXTRA_HYDRO_LOOP
+ else if (t->subtype == task_subtype_gradient)
+ runner_dosub_pair1_gradient(r, ci, cj, t->flags, 1);
+#endif
else if (t->subtype == task_subtype_force)
runner_dosub_pair2_force(r, ci, cj, t->flags, 1);
else if (t->subtype == task_subtype_grav)
@@ -1142,12 +1206,18 @@ void *runner_main(void *data) {
case task_type_ghost:
runner_do_ghost(r, ci);
break;
+#ifdef EXTRA_HYDRO_LOOP
+ case task_type_extra_ghost:
+ runner_do_extra_ghost(r, ci);
+ break;
+#endif
case task_type_kick:
runner_do_kick(r, ci, 1);
break;
case task_type_kick_fixdt:
runner_do_kick_fixdt(r, ci, 1);
break;
+#ifdef WITH_MPI
case task_type_send:
if (t->subtype == task_subtype_tend) {
free(t->buff);
@@ -1161,6 +1231,7 @@ void *runner_main(void *data) {
runner_do_recv_cell(r, ci, 1);
}
break;
+#endif
case task_type_grav_mm:
runner_do_grav_mm(r, t->ci, 1);
break;
diff --git a/src/runner_doiact.h b/src/runner_doiact.h
index db439671ff5fee56c086444ddaa8268571c80a15..376400926432a9e9b6b9c736260dc3e119c2c64d 100644
--- a/src/runner_doiact.h
+++ b/src/runner_doiact.h
@@ -1739,7 +1739,7 @@ void DOSUB_PAIR1(struct runner *r, struct cell *ci, struct cell *cj, int sid,
if (ci->ti_end_min > ti_current && cj->ti_end_min > ti_current) return;
/* Get the cell dimensions. */
- const float h = fmin(ci->width[0], fmin(ci->width[1], ci->width[2]));
+ const float h = fminf(ci->width[0], fminf(ci->width[1], ci->width[2]));
/* Get the type of pair if not specified explicitly. */
// if ( sid < 0 )
@@ -2023,7 +2023,7 @@ void DOSUB_PAIR2(struct runner *r, struct cell *ci, struct cell *cj, int sid,
if (ci->ti_end_min > ti_current && cj->ti_end_min > ti_current) return;
/* Get the cell dimensions. */
- const float h = fmin(ci->width[0], fmin(ci->width[1], ci->width[2]));
+ const float h = fminf(ci->width[0], fminf(ci->width[1], ci->width[2]));
/* Get the type of pair if not specified explicitly. */
// if ( sid < 0 )
@@ -2336,7 +2336,7 @@ void DOSUB_SUBSET(struct runner *r, struct cell *ci, struct part *parts,
else {
/* Get the cell dimensions. */
- const float h = fmin(ci->width[0], fmin(ci->width[1], ci->width[2]));
+ const float h = fminf(ci->width[0], fminf(ci->width[1], ci->width[2]));
/* Recurse? */
if (ci->split && cj->split &&
@@ -2862,16 +2862,17 @@ void DOSUB_SUBSET(struct runner *r, struct cell *ci, struct part *parts,
}
/* Get the sorting index. */
- int sid = 0;
+ int new_sid = 0;
for (int k = 0; k < 3; k++)
- sid =
- 3 * sid + ((cj->loc[k] - ci->loc[k] + shift[k] < 0)
- ? 0
- : (cj->loc[k] - ci->loc[k] + shift[k] > 0) ? 2 : 1);
- sid = sortlistID[sid];
+ new_sid = 3 * new_sid +
+ ((cj->loc[k] - ci->loc[k] + shift[k] < 0)
+ ? 0
+ : (cj->loc[k] - ci->loc[k] + shift[k] > 0) ? 2 : 1);
+ new_sid = sortlistID[new_sid];
/* Do any of the cells need to be sorted first? */
- if (!(cj->sorted & (1 << sid))) runner_do_sort(r, cj, (1 << sid), 1);
+ if (!(cj->sorted & (1 << new_sid)))
+ runner_do_sort(r, cj, (1 << new_sid), 1);
/* Compute the interactions. */
DOPAIR_SUBSET(r, ci, parts, ind, count, cj);
diff --git a/src/scheduler.c b/src/scheduler.c
index adc63e55092ff9bc69e6a2f98d6dfd5c399857f4..ef4d19107fb48684ca299f286436a155a6fe0151 100644
--- a/src/scheduler.c
+++ b/src/scheduler.c
@@ -111,14 +111,14 @@ void scheduler_addunlock(struct scheduler *s, struct task *ta,
static void scheduler_splittask(struct task *t, struct scheduler *s) {
/* Static constants. */
- const static int pts[7][8] = {
+ static const int pts[7][8] = {
{-1, 12, 10, 9, 4, 3, 1, 0}, {-1, -1, 11, 10, 5, 4, 2, 1},
{-1, -1, -1, 12, 7, 6, 4, 3}, {-1, -1, -1, -1, 8, 7, 5, 4},
{-1, -1, -1, -1, -1, 12, 10, 9}, {-1, -1, -1, -1, -1, -1, 11, 10},
{-1, -1, -1, -1, -1, -1, -1, 12}};
- const static float sid_scale[13] = {0.1897, 0.4025, 0.1897, 0.4025, 0.5788,
- 0.4025, 0.1897, 0.4025, 0.1897, 0.4025,
- 0.5788, 0.4025, 0.5788};
+ static const float sid_scale[13] = {
+ 0.1897f, 0.4025f, 0.1897f, 0.4025f, 0.5788f, 0.4025f, 0.1897f,
+ 0.4025f, 0.1897f, 0.4025f, 0.5788f, 0.4025f, 0.5788f};
/* Iterate on this task until we're done with it. */
int redo = 1;
@@ -704,7 +704,6 @@ struct task *scheduler_addtask(struct scheduler *s, enum task_types type,
t->toc = 0;
t->nr_unlock_tasks = 0;
t->rid = -1;
- t->last_rid = -1;
/* Add an index for it. */
// lock_lock( &s->lock );
diff --git a/src/serial_io.c b/src/serial_io.c
index d0618ea93e07168274dff02f31ae95c89503859b..a9213819d4816a08614b82264ab463cecc060419 100644
--- a/src/serial_io.c
+++ b/src/serial_io.c
@@ -179,7 +179,8 @@ void readArray(hid_t grp, const struct io_props props, size_t N,
void prepareArray(struct engine* e, hid_t grp, char* fileName, FILE* xmfFile,
char* partTypeGroupName, const struct io_props props,
- long long N_total, const struct UnitSystem* internal_units,
+ unsigned long long N_total,
+ const struct UnitSystem* internal_units,
const struct UnitSystem* snapshot_units) {
/* Create data space */
@@ -220,7 +221,7 @@ void prepareArray(struct engine* e, hid_t grp, char* fileName, FILE* xmfFile,
/* Set chunk size */
h_err = H5Pset_chunk(h_prop, rank, chunk_shape);
if (h_err < 0) {
- error("Error while setting chunk size (%lld, %lld) for field '%s'.",
+ error("Error while setting chunk size (%llu, %llu) for field '%s'.",
chunk_shape[0], chunk_shape[1], props.name);
}
@@ -590,26 +591,18 @@ void read_ic_serial(char* fileName, const struct UnitSystem* internal_units,
int num_fields = 0;
struct io_props list[100];
- size_t N = 0;
+ size_t Nparticles = 0;
/* Read particle fields into the particle structure */
switch (ptype) {
case GAS:
- /* if (!dry_run) */
- /* hydro_read_particles(h_grp, N[ptype], N_total[ptype], */
- /* offset[ptype], *parts); */
- /* break; */
- N = *Ngas;
+ Nparticles = *Ngas;
hydro_read_particles(*parts, list, &num_fields);
break;
case DM:
- /* if (!dry_run) */
- /* darkmatter_read_particles(h_grp, N[ptype], N_total[ptype], */
- /* offset[ptype], *gparts); */
- /* break; */
- N = Ndm;
+ Nparticles = Ndm;
darkmatter_read_particles(*gparts, list, &num_fields);
break;
@@ -621,7 +614,7 @@ void read_ic_serial(char* fileName, const struct UnitSystem* internal_units,
/* Read everything */
if (!dry_run)
for (int i = 0; i < num_fields; ++i)
- readArray(h_grp, list[i], N, N_total[ptype], offset[ptype],
+ readArray(h_grp, list[i], Nparticles, N_total[ptype], offset[ptype],
internal_units, ic_units);
/* Close particle group */
@@ -886,13 +879,13 @@ void write_output_serial(struct engine* e, const char* baseName,
int num_fields = 0;
struct io_props list[100];
- size_t N = 0;
+ size_t Nparticles = 0;
/* Write particle fields from the particle structure */
switch (ptype) {
case GAS:
- N = Ngas;
+ Nparticles = Ngas;
hydro_write_particles(parts, list, &num_fields);
break;
@@ -907,7 +900,7 @@ void write_output_serial(struct engine* e, const char* baseName,
collect_dm_gparts(gparts, Ntot, dmparts, Ndm);
/* Write DM particles */
- N = Ndm;
+ Nparticles = Ndm;
darkmatter_write_particles(dmparts, list, &num_fields);
break;
@@ -918,9 +911,9 @@ void write_output_serial(struct engine* e, const char* baseName,
/* Write everything */
for (int i = 0; i < num_fields; ++i)
- writeArray(e, h_grp, fileName, xmfFile, partTypeGroupName, list[i], N,
- N_total[ptype], mpi_rank, offset[ptype], internal_units,
- snapshot_units);
+ writeArray(e, h_grp, fileName, xmfFile, partTypeGroupName, list[i],
+ Nparticles, N_total[ptype], mpi_rank, offset[ptype],
+ internal_units, snapshot_units);
/* Free temporary array */
free(dmparts);
diff --git a/src/single_io.c b/src/single_io.c
index 2d95c2b32fc4b9b2e0e696deb75c9eab1acd9158..df741ac22335fc05ec1ac334636a897e6def3e96 100644
--- a/src/single_io.c
+++ b/src/single_io.c
@@ -257,7 +257,7 @@ void writeArray(struct engine* e, hid_t grp, char* fileName, FILE* xmfFile,
/* Set chunk size */
h_err = H5Pset_chunk(h_prop, rank, chunk_shape);
if (h_err < 0) {
- error("Error while setting chunk size (%lld, %lld) for field '%s'.",
+ error("Error while setting chunk size (%llu, %llu) for field '%s'.",
chunk_shape[0], chunk_shape[1], props.name);
}
@@ -471,18 +471,18 @@ void read_ic_single(char* fileName, const struct UnitSystem* internal_units,
int num_fields = 0;
struct io_props list[100];
- size_t N = 0;
+ size_t Nparticles = 0;
/* Read particle fields into the structure */
switch (ptype) {
case GAS:
- N = *Ngas;
+ Nparticles = *Ngas;
hydro_read_particles(*parts, list, &num_fields);
break;
case DM:
- N = Ndm;
+ Nparticles = Ndm;
darkmatter_read_particles(*gparts, list, &num_fields);
break;
@@ -493,7 +493,7 @@ void read_ic_single(char* fileName, const struct UnitSystem* internal_units,
/* Read everything */
if (!dry_run)
for (int i = 0; i < num_fields; ++i)
- readArray(h_grp, list[i], N, internal_units, ic_units);
+ readArray(h_grp, list[i], Nparticles, internal_units, ic_units);
/* Close particle group */
H5Gclose(h_grp);
diff --git a/src/space.c b/src/space.c
index def8adad3d023efe5daf552328d523d3cfc37429..2f4442c4729f69d8010278f7cf772b40d100d1f1 100644
--- a/src/space.c
+++ b/src/space.c
@@ -352,14 +352,17 @@ void space_regrid(struct space *s, double cell_max, int verbose) {
s->cells[k].sorts = NULL;
s->cells[k].nr_tasks = 0;
s->cells[k].nr_density = 0;
+ s->cells[k].nr_gradient = 0;
s->cells[k].nr_force = 0;
s->cells[k].density = NULL;
+ s->cells[k].gradient = NULL;
s->cells[k].force = NULL;
s->cells[k].dx_max = 0.0f;
s->cells[k].sorted = 0;
s->cells[k].count = 0;
s->cells[k].gcount = 0;
s->cells[k].init = NULL;
+ s->cells[k].extra_ghost = NULL;
s->cells[k].ghost = NULL;
s->cells[k].kick = NULL;
s->cells[k].super = &s->cells[k];
@@ -426,7 +429,7 @@ void space_rebuild(struct space *s, double cell_max, int verbose) {
int *gind;
if ((gind = (int *)malloc(sizeof(int) * gind_size)) == NULL)
error("Failed to allocate temporary g-particle indices.");
- for (int k = 0; k < nr_gparts; k++) {
+ for (size_t k = 0; k < nr_gparts; k++) {
struct gpart *restrict gp = &s->gparts[k];
for (int j = 0; j < 3; j++)
if (gp->x[j] < 0.0)
@@ -484,7 +487,7 @@ void space_rebuild(struct space *s, double cell_max, int verbose) {
#endif
/* Move non-local gparts to the end of the list. */
- for (int k = 0; k < nr_gparts;) {
+ for (size_t k = 0; k < nr_gparts;) {
if (cells[gind[k]].nodeID != local_nodeID) {
cells[gind[k]].gcount -= 1;
nr_gparts -= 1;
@@ -593,7 +596,7 @@ void space_rebuild(struct space *s, double cell_max, int verbose) {
}
/* Assign each particle to its cell. */
- for (int k = nr_gparts; k < s->nr_gparts; k++) {
+ for (size_t k = nr_gparts; k < s->nr_gparts; k++) {
const struct gpart *const p = &s->gparts[k];
gind[k] =
cell_getid(cdim, p->x[0] * ih[0], p->x[1] * ih[1], p->x[2] * ih[2]);
@@ -711,7 +714,7 @@ void space_parts_sort(struct space *s, int *ind, size_t N, int min, int max,
if ((sort_struct.stack =
malloc(sizeof(struct qstack) * sort_struct.stack_size)) == NULL)
error("Failed to allocate sorting stack.");
- for (int i = 0; i < sort_struct.stack_size; i++)
+ for (unsigned int i = 0; i < sort_struct.stack_size; i++)
sort_struct.stack[i].ready = 0;
/* Add the first interval. */
@@ -896,7 +899,7 @@ void space_gparts_sort(struct space *s, int *ind, size_t N, int min, int max,
if ((sort_struct.stack =
malloc(sizeof(struct qstack) * sort_struct.stack_size)) == NULL)
error("Failed to allocate sorting stack.");
- for (int i = 0; i < sort_struct.stack_size; i++)
+ for (unsigned int i = 0; i < sort_struct.stack_size; i++)
sort_struct.stack[i].ready = 0;
/* Add the first interval. */
@@ -1580,13 +1583,13 @@ void space_init(struct space *s, const struct swift_params *params,
/* Check that all the part positions are reasonable, wrap if periodic. */
if (periodic) {
- for (int k = 0; k < Npart; k++)
+ for (size_t k = 0; k < Npart; k++)
for (int j = 0; j < 3; j++) {
while (parts[k].x[j] < 0) parts[k].x[j] += dim[j];
while (parts[k].x[j] >= dim[j]) parts[k].x[j] -= dim[j];
}
} else {
- for (int k = 0; k < Npart; k++)
+ for (size_t k = 0; k < Npart; k++)
for (int j = 0; j < 3; j++)
if (parts[k].x[j] < 0 || parts[k].x[j] >= dim[j])
error("Not all particles are within the specified domain.");
@@ -1594,13 +1597,13 @@ void space_init(struct space *s, const struct swift_params *params,
/* Same for the gparts */
if (periodic) {
- for (int k = 0; k < Ngpart; k++)
+ for (size_t k = 0; k < Ngpart; k++)
for (int j = 0; j < 3; j++) {
while (gparts[k].x[j] < 0) gparts[k].x[j] += dim[j];
while (gparts[k].x[j] >= dim[j]) gparts[k].x[j] -= dim[j];
}
} else {
- for (int k = 0; k < Ngpart; k++)
+ for (size_t k = 0; k < Ngpart; k++)
for (int j = 0; j < 3; j++)
if (gparts[k].x[j] < 0 || gparts[k].x[j] >= dim[j])
error("Not all g-particles are within the specified domain.");
diff --git a/src/task.c b/src/task.c
index 9e48f8abeb9659ea2674f48620b58da977a6775a..2cecce9f066727c9059b08764df647bd8f0f3901 100644
--- a/src/task.c
+++ b/src/task.c
@@ -46,18 +46,20 @@
#include "inline.h"
#include "lock.h"
-/* Task type names. */
const char *taskID_names[task_type_count] = {
- "none", "sort", "self", "pair", "sub_self",
- "sub_pair", "init", "ghost", "kick", "kick_fixdt",
- "send", "recv", "grav_gather_m", "grav_fft", "grav_mm",
- "grav_up", "grav_external"};
+ "none", "sort", "self", "pair", "sub_self",
+ "sub_pair", "init", "ghost", "extra_ghost", "kick",
+ "kick_fixdt", "send", "recv", "grav_gather_m", "grav_fft",
+ "grav_mm", "grav_up", "grav_external"};
-const char *subtaskID_names[task_type_count] = {"none", "density", "force",
- "grav", "tend"};
+const char *subtaskID_names[task_subtype_count] = {
+ "none", "density", "gradient", "force", "grav", "tend"};
/**
* @brief Computes the overlap between the parts array of two given cells.
+ *
+ * @param ci The first #cell.
+ * @param cj The second #cell.
*/
__attribute__((always_inline)) INLINE static size_t task_cell_overlap_part(
const struct cell *ci, const struct cell *cj) {
@@ -77,6 +79,9 @@ __attribute__((always_inline)) INLINE static size_t task_cell_overlap_part(
/**
* @brief Computes the overlap between the gparts array of two given cells.
+ *
+ * @param ci The first #cell.
+ * @param cj The second #cell.
*/
__attribute__((always_inline)) INLINE static size_t task_cell_overlap_gpart(
const struct cell *ci, const struct cell *cj) {
@@ -110,6 +115,7 @@ __attribute__((always_inline)) INLINE static enum task_actions task_acts_on(
case task_type_sort:
case task_type_ghost:
+ case task_type_extra_ghost:
return task_action_part;
break;
@@ -120,6 +126,7 @@ __attribute__((always_inline)) INLINE static enum task_actions task_acts_on(
switch (t->subtype) {
case task_subtype_density:
+ case task_subtype_gradient:
case task_subtype_force:
return task_action_part;
break;
diff --git a/src/task.h b/src/task.h
index db74c296bb7988d12e474368ed9abffd97ede49e..c0c9e47ee4ca221f9f960256e9208c749ae523ea 100644
--- a/src/task.h
+++ b/src/task.h
@@ -23,15 +23,15 @@
#ifndef SWIFT_TASK_H
#define SWIFT_TASK_H
+#include "../config.h"
+
/* Includes. */
#include "cell.h"
#include "cycle.h"
-/* Some constants. */
-#define task_maxwait 3
-#define task_maxunlock 15
-
-/* The different task types. */
+/**
+ * @brief The different task types.
+ */
enum task_types {
task_type_none = 0,
task_type_sort,
@@ -41,6 +41,7 @@ enum task_types {
task_type_sub_pair,
task_type_init,
task_type_ghost,
+ task_type_extra_ghost,
task_type_kick,
task_type_kick_fixdt,
task_type_send,
@@ -53,19 +54,22 @@ enum task_types {
task_type_count
};
-extern const char *taskID_names[];
-
-/* The different task sub-types. */
+/**
+ * @brief The different task sub-types (for pairs, selfs and sub-tasks).
+ */
enum task_subtypes {
task_subtype_none = 0,
task_subtype_density,
+ task_subtype_gradient,
task_subtype_force,
task_subtype_grav,
task_subtype_tend,
task_subtype_count
};
-/* The kind of action the task perform */
+/**
+ * @brief The type of particles/objects this task acts upon in a given cell.
+ */
enum task_actions {
task_action_none,
task_action_part,
@@ -75,29 +79,72 @@ enum task_actions {
task_action_count
};
+/**
+ * @brief Names of the task types.
+ */
+extern const char *taskID_names[];
+
+/**
+ * @brief Names of the task sub-types.
+ */
extern const char *subtaskID_names[];
-/* Data of a task. */
+/**
+ * @brief A task to be run by the #scheduler.
+ */
struct task {
+ /*! Type of the task */
enum task_types type;
+
+ /*! Sub-type of the task (for the tasks that have one */
enum task_subtypes subtype;
- char skip, tight, implicit;
- int flags, wait, rank, weight;
+ /*! Flags used to carry additional information (e.g. sort directions) */
+ int flags;
+
+ /*! Number of unsatisfied dependencies */
+ int wait;
+
+ /*! Rank of a task in the order */
+ int rank;
+
+ /*! Weight of the task */
+ int weight;
+
+ /*! Pointers to the cells this task acts upon */
struct cell *ci, *cj;
- void *buff;
+ /*! ID of the queue or runner owning this task */
+ int rid;
+
+ /*! Number of tasks unlocked by this one */
+ int nr_unlock_tasks;
+
+ /*! List of tasks unlocked by this one */
+ struct task **unlock_tasks;
#ifdef WITH_MPI
+
+ /*! Buffer for this task's communications */
+ void *buff;
+
+ /*! MPI request corresponding to this task */
MPI_Request req;
+
#endif
- int rid, last_rid;
+ /*! Start and end time of this task */
ticks tic, toc;
- int nr_unlock_tasks;
- struct task **unlock_tasks;
+ /*! Should the scheduler skip this task ? */
+ char skip;
+
+ /*! Does this task require the particles to be tightly in the cell ? */
+ char tight;
+
+ /*! Is this task implicit (i.e. does not do anything) ? */
+ char implicit;
};
/* Function prototypes. */
diff --git a/src/timers.c b/src/timers.c
index b621d27c90902f06c3760cbef6a88237a2b3b95b..c2f3b35d75ea340082acf1a5ec334bb3543bab12 100644
--- a/src/timers.c
+++ b/src/timers.c
@@ -37,11 +37,11 @@ ticks timers[timer_count];
* To reset all timers, use the mask #timers_mask_all.
*/
-void timers_reset(unsigned int mask) {
+void timers_reset(unsigned long long mask) {
int k;
/* Loop over the timers and set the masked ones to zero. */
for (k = 0; k < timer_count; k++)
- if (mask & (1 << k)) timers[k] = 0;
+ if (mask & (1ull << k)) timers[k] = 0;
}
diff --git a/src/timers.h b/src/timers.h
index aa8455397daf1e88b709a8332e3ae63694991e94..0bf1f3f413d123c84cc30edff73ca9dfce4ea159 100644
--- a/src/timers.h
+++ b/src/timers.h
@@ -36,17 +36,21 @@ enum {
timer_kick,
timer_dosort,
timer_doself_density,
+ timer_doself_gradient,
timer_doself_force,
timer_doself_grav_pp,
timer_dopair_density,
+ timer_dopair_gradient,
timer_dopair_force,
timer_dopair_grav_pm,
timer_dopair_grav_pp,
timer_dograv_external,
timer_dosub_self_density,
+ timer_dosub_self_gradient,
timer_dosub_self_force,
timer_dosub_self_grav,
timer_dosub_pair_density,
+ timer_dosub_pair_gradient,
timer_dosub_pair_force,
timer_dosub_pair_grav,
timer_dopair_subset,
@@ -64,7 +68,7 @@ enum {
extern ticks timers[timer_count];
/* Mask for all timers. */
-#define timers_mask_all ((1 << timer_count) - 1)
+#define timers_mask_all ((1ull << timer_count) - 1)
/* Define the timer macros. */
#ifdef TIMER
@@ -74,7 +78,7 @@ extern ticks timers[timer_count];
#define TIMER_TOC(t) timers_toc(t, tic)
#define TIMER_TIC2 ticks tic2 = getticks();
#define TIMER_TOC2(t) timers_toc(t, tic2)
-INLINE static ticks timers_toc(int t, ticks tic) {
+INLINE static ticks timers_toc(unsigned int t, ticks tic) {
ticks d = (getticks() - tic);
atomic_add(&timers[t], d);
return d;
@@ -87,6 +91,6 @@ INLINE static ticks timers_toc(int t, ticks tic) {
#endif
/* Function prototypes. */
-void timers_reset(unsigned int mask);
+void timers_reset(unsigned long long mask);
#endif /* SWIFT_TIMERS_H */
diff --git a/src/tools.c b/src/tools.c
index b64e17849081994e1969d5f8de0636768dcb729a..060bf1439f30dc6237938c060bc4ddc8d9be822b 100644
--- a/src/tools.c
+++ b/src/tools.c
@@ -442,43 +442,6 @@ void pairs_single_grav(double *dim, long long int pid,
aabs[2]);
}
-/**
- * @brief Test the density function by dumping it for two random parts.
- *
- * @param N number of intervals in [0,1].
- */
-void density_dump(int N) {
-
- int k;
- float r2[4] = {0.0f, 0.0f, 0.0f, 0.0f}, hi[4], hj[4];
- struct part /**pi[4], *pj[4],*/ Pi[4], Pj[4];
-
- /* Init the interaction parameters. */
- for (k = 0; k < 4; k++) {
- Pi[k].mass = 1.0f;
- Pi[k].rho = 0.0f;
- Pi[k].density.wcount = 0.0f;
- Pi[k].id = k;
- Pj[k].mass = 1.0f;
- Pj[k].rho = 0.0f;
- Pj[k].density.wcount = 0.0f;
- Pj[k].id = k + 4;
- hi[k] = 1.0;
- hj[k] = 1.0;
- }
-
- for (k = 0; k <= N; k++) {
- r2[3] = r2[2];
- r2[2] = r2[1];
- r2[1] = r2[0];
- r2[0] = ((float)k) / N;
- Pi[0].density.wcount = 0;
- Pj[0].density.wcount = 0;
- runner_iact_density(r2[0], NULL, hi[0], hj[0], &Pi[0], &Pj[0]);
- printf(" %e %e %e", r2[0], Pi[0].density.wcount, Pj[0].density.wcount);
- }
-}
-
/**
* @brief Compute the force on a single particle brute-force.
*/
@@ -499,7 +462,7 @@ void engine_single_density(double *dim, long long int pid,
hydro_init_part(&p);
/* Loop over all particle pairs (force). */
- for (int k = 0; k < N; k++) {
+ for (k = 0; k < N; k++) {
if (parts[k].id == p.id) continue;
for (int i = 0; i < 3; i++) {
dx[i] = p.x[i] - parts[k].x[i];
@@ -520,7 +483,7 @@ void engine_single_density(double *dim, long long int pid,
/* Dump the result. */
hydro_end_density(&p, 0);
message("part %lli (h=%e) has wcount=%e, rho=%e.", p.id, p.h,
- p.density.wcount, p.rho);
+ p.density.wcount, hydro_get_density(&p));
fflush(stdout);
}
diff --git a/src/units.c b/src/units.c
index 5c262ae03639262dfa126b101402b8fbfe41259a..f598b5ddf0b1a4b165648d5378915cd6f10f0bba 100644
--- a/src/units.c
+++ b/src/units.c
@@ -316,6 +316,10 @@ void units_get_base_unit_exponants_array(float baseUnitsExp[5],
case UNIT_CONV_TEMPERATURE:
baseUnitsExp[UNIT_TEMPERATURE] = 1.f;
+ break;
+
+ case UNIT_CONV_VOLUME:
+ baseUnitsExp[UNIT_LENGTH] = -3.f;
}
}
diff --git a/src/units.h b/src/units.h
index c67f6ebbab324e3c90ae86eb1ea11dcf013c5dc7..9adaa7dc621edafe0c1f097b8c5202a2f017e41d 100644
--- a/src/units.h
+++ b/src/units.h
@@ -29,25 +29,26 @@
* @brief The unit system used internally.
*
* This structure contains the conversion factors to the 7 cgs base units to the
- *internal units.
- * It is used everytime a conversion is performed or an i/o function is called.
- *
+ * internal units. It is used everytime a conversion is performed or an i/o
+ * function is called.
**/
struct UnitSystem {
- double UnitMass_in_cgs; /*< Conversion factor from grams to internal mass
- units */
- double UnitLength_in_cgs; /*< Conversion factor from centimeters to internal
- length units. */
+ /*! Conversion factor from grams to internal mass units */
+ double UnitMass_in_cgs;
+
+ /*! Conversion factor from centimeters to internal length unit */
+ double UnitLength_in_cgs;
- double UnitTime_in_cgs; /*< Conversion factor from seconds to internal time
- units. */
+ /*! Conversion factor from seconds to internal time units */
+ double UnitTime_in_cgs;
- double UnitCurrent_in_cgs; /*< Conversion factor from Ampere to internal
- current units. */
+ /*! Conversion factor from Ampere to internal current units */
+ double UnitCurrent_in_cgs;
- double UnitTemperature_in_cgs; /*< Conversion factor from Kelvins to internal
- temperature units. */
+ /*! Conversion factor from Kelvins to internal temperature units. */
+ double UnitTemperature_in_cgs;
+
};
/**
@@ -89,7 +90,8 @@ enum UnitConversionFactor {
UNIT_CONV_MAGNETIC_FLUX,
UNIT_CONV_MAGNETIC_FIELD,
UNIT_CONV_MAGNETIC_INDUCTANCE,
- UNIT_CONV_TEMPERATURE
+ UNIT_CONV_TEMPERATURE,
+ UNIT_CONV_VOLUME
};
void units_init_cgs(struct UnitSystem*);
diff --git a/src/version.c b/src/version.c
index 68a051fa08c53c68319c1e785c0c6503afe18f4d..8bd94e5651dbc597fcd80bc585a47c6633ee3993 100644
--- a/src/version.c
+++ b/src/version.c
@@ -218,7 +218,7 @@ const char *hdf5_version(void) {
#ifdef HAVE_HDF5
unsigned int majnum, minnum, relnum;
H5get_libversion(&majnum, &minnum, &relnum);
- sprintf(version, "%i.%i.%i", majnum, minnum, relnum);
+ sprintf(version, "%u.%u.%u", majnum, minnum, relnum);
#else
sprintf(version, "Unknown version");
#endif
diff --git a/tests/Makefile.am b/tests/Makefile.am
index 952787823d08772b79b4d69d0c8b3728d79e4bfd..5648c9a3ca0c3f944982b31060cfe9a0e9c3e45d 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -24,13 +24,16 @@ AM_LDFLAGS = ../src/.libs/libswiftsim.a $(HDF5_LDFLAGS) $(HDF5_LIBS) $(FFTW_LIBS
TESTS = testGreetings testMaths testReading.sh testSingle testKernel testSymmetry \
testPair.sh testPairPerturbed.sh test27cells.sh test27cellsPerturbed.sh \
testParser.sh testSPHStep test125cells.sh testKernelGrav testFFT \
- testThreadpool
+ testAdiabaticIndex testRiemannExact testRiemannTRRS testRiemannHLLC \
+ testMatrixInversion testThreadpool
# List of test programs to compile
check_PROGRAMS = testGreetings testReading testSingle testTimeIntegration \
testSPHStep testPair test27cells test125cells testParser \
testKernel testKernelGrav testFFT testInteractions testMaths \
- testSymmetry testThreadpool
+ testSymmetry testThreadpool \
+ testAdiabaticIndex testRiemannExact testRiemannTRRS \
+ testRiemannHLLC testMatrixInversion
# Sources for the individual programs
testGreetings_SOURCES = testGreetings.c
@@ -63,6 +66,16 @@ testFFT_SOURCES = testFFT.c
testInteractions_SOURCES = testInteractions.c
+testAdiabaticIndex_SOURCES = testAdiabaticIndex.c
+
+testRiemannExact_SOURCES = testRiemannExact.c
+
+testRiemannTRRS_SOURCES = testRiemannTRRS.c
+
+testRiemannHLLC_SOURCES = testRiemannHLLC.c
+
+testMatrixInversion_SOURCES = testMatrixInversion.c
+
testThreadpool_SOURCES = testThreadpool.c
# Files necessary for distribution
diff --git a/tests/test125cells.c b/tests/test125cells.c
index dc15c5aa2f4e00862633807bec9603d77fa4bf30..c7e01693b45f76fa21ffd397289fc06ad36af03d 100644
--- a/tests/test125cells.c
+++ b/tests/test125cells.c
@@ -99,6 +99,8 @@ void set_energy_state(struct part *part, enum pressure_field press, float size,
part->u = pressure / (hydro_gamma_minus_one * density);
#elif defined(MINIMAL_SPH)
part->u = pressure / (hydro_gamma_minus_one * density);
+#elif defined(GIZMO_SPH)
+ part->primitives.P = pressure;
#else
error("Need to define pressure here !");
#endif
@@ -127,7 +129,7 @@ void get_solution(const struct cell *main_cell, struct solution_part *solution,
float density, enum velocity_field vel,
enum pressure_field press, float size) {
- for (size_t i = 0; i < main_cell->count; ++i) {
+ for (int i = 0; i < main_cell->count; ++i) {
solution[i].id = main_cell->parts[i].id;
@@ -187,12 +189,29 @@ void get_solution(const struct cell *main_cell, struct solution_part *solution,
void reset_particles(struct cell *c, enum velocity_field vel,
enum pressure_field press, float size, float density) {
- for (size_t i = 0; i < c->count; ++i) {
+ for (int i = 0; i < c->count; ++i) {
struct part *p = &c->parts[i];
set_velocity(p, vel, size);
set_energy_state(p, press, size, density);
+
+#if defined(GIZMO_SPH)
+ p->geometry.volume = p->conserved.mass / density;
+ p->primitives.rho = density;
+ p->primitives.v[0] = p->v[0];
+ p->primitives.v[1] = p->v[1];
+ p->primitives.v[2] = p->v[2];
+ p->conserved.momentum[0] = p->conserved.mass * p->v[0];
+ p->conserved.momentum[1] = p->conserved.mass * p->v[1];
+ p->conserved.momentum[2] = p->conserved.mass * p->v[2];
+ p->conserved.energy =
+ p->primitives.P / hydro_gamma_minus_one * p->geometry.volume +
+ 0.5f * (p->conserved.momentum[0] * p->conserved.momentum[0] +
+ p->conserved.momentum[1] * p->conserved.momentum[1] +
+ p->conserved.momentum[2] * p->conserved.momentum[2]) /
+ p->conserved.mass;
+#endif
}
}
@@ -238,7 +257,12 @@ struct cell *make_cell(size_t n, const double offset[3], double size, double h,
part->x[1] = offset[1] + size * (y + 0.5) / (float)n;
part->x[2] = offset[2] + size * (z + 0.5) / (float)n;
part->h = size * h / (float)n;
+
+#ifdef GIZMO_SPH
+ part->conserved.mass = density * volume / count;
+#else
part->mass = density * volume / count;
+#endif
set_velocity(part, vel, size);
set_energy_state(part, press, size, density);
@@ -248,6 +272,24 @@ struct cell *make_cell(size_t n, const double offset[3], double size, double h,
part->ti_end = 1;
hydro_first_init_part(part, xpart);
+
+#if defined(GIZMO_SPH)
+ part->geometry.volume = part->conserved.mass / density;
+ part->primitives.rho = density;
+ part->primitives.v[0] = part->v[0];
+ part->primitives.v[1] = part->v[1];
+ part->primitives.v[2] = part->v[2];
+ part->conserved.momentum[0] = part->conserved.mass * part->v[0];
+ part->conserved.momentum[1] = part->conserved.mass * part->v[1];
+ part->conserved.momentum[2] = part->conserved.mass * part->v[2];
+ part->conserved.energy =
+ part->primitives.P / hydro_gamma_minus_one * part->geometry.volume +
+ 0.5f * (part->conserved.momentum[0] * part->conserved.momentum[0] +
+ part->conserved.momentum[1] * part->conserved.momentum[1] +
+ part->conserved.momentum[2] * part->conserved.momentum[2]) /
+ part->conserved.mass;
+#endif
+
++part;
++xpart;
}
@@ -304,7 +346,7 @@ void dump_particle_fields(char *fileName, struct cell *main_cell,
fprintf(file, "# Main cell --------------------------------------------\n");
/* Write main cell */
- for (size_t pid = 0; pid < main_cell->count; pid++) {
+ for (int pid = 0; pid < main_cell->count; pid++) {
fprintf(file,
"%6llu %8.5f %8.5f %8.5f %8.5f %8.5f %8.5f %8.5f %8.5f %8.5f %8.5f "
"%8.5f "
@@ -313,7 +355,7 @@ void dump_particle_fields(char *fileName, struct cell *main_cell,
main_cell->parts[pid].x[1], main_cell->parts[pid].x[2],
main_cell->parts[pid].v[0], main_cell->parts[pid].v[1],
main_cell->parts[pid].v[2], main_cell->parts[pid].h,
- main_cell->parts[pid].rho,
+ hydro_get_density(&main_cell->parts[pid]),
#ifdef MINIMAL_SPH
0.f,
#else
@@ -343,7 +385,7 @@ void dump_particle_fields(char *fileName, struct cell *main_cell,
fprintf(file, "# Solution ---------------------------------------------\n");
- for (size_t pid = 0; pid < main_cell->count; pid++) {
+ for (int pid = 0; pid < main_cell->count; pid++) {
fprintf(file,
"%6llu %8.5f %8.5f %8.5f %8.5f %8.5f %8.5f %8.5f %8.5f %8.5f "
"%8.5f %8.5f "
diff --git a/tests/test27cells.c b/tests/test27cells.c
index 6d935c0a4a57b3fbbdbd9b378a4b8cb97234aa63..1a1ab88748d922b3e7fbb30a73a10809dca10863 100644
--- a/tests/test27cells.c
+++ b/tests/test27cells.c
@@ -104,7 +104,11 @@ struct cell *make_cell(size_t n, double *offset, double size, double h,
}
part->h = size * h / (float)n;
part->id = ++(*partId);
+#ifdef GIZMO_SPH
+ part->conserved.mass = density * volume / count;
+#else
part->mass = density * volume / count;
+#endif
part->ti_begin = 0;
part->ti_end = 1;
++part;
@@ -146,7 +150,7 @@ void clean_up(struct cell *ci) {
* @brief Initializes all particles field to be ready for a density calculation
*/
void zero_particle_fields(struct cell *c) {
- for (size_t pid = 0; pid < c->count; pid++) {
+ for (int pid = 0; pid < c->count; pid++) {
hydro_init_part(&c->parts[pid]);
}
}
@@ -155,7 +159,7 @@ void zero_particle_fields(struct cell *c) {
* @brief Ends the loop by adding the appropriate coefficients
*/
void end_calculation(struct cell *c) {
- for (size_t pid = 0; pid < c->count; pid++) {
+ for (int pid = 0; pid < c->count; pid++) {
hydro_end_density(&c->parts[pid], 1);
}
}
@@ -177,14 +181,15 @@ void dump_particle_fields(char *fileName, struct cell *main_cell,
fprintf(file, "# Main cell --------------------------------------------\n");
/* Write main cell */
- for (size_t pid = 0; pid < main_cell->count; pid++) {
+ for (int pid = 0; pid < main_cell->count; pid++) {
fprintf(file,
"%6llu %10f %10f %10f %10f %10f %10f %13e %13e %13e %13e %13e "
"%13e %13e %13e\n",
main_cell->parts[pid].id, main_cell->parts[pid].x[0],
main_cell->parts[pid].x[1], main_cell->parts[pid].x[2],
main_cell->parts[pid].v[0], main_cell->parts[pid].v[1],
- main_cell->parts[pid].v[2], main_cell->parts[pid].rho,
+ main_cell->parts[pid].v[2],
+ hydro_get_density(&main_cell->parts[pid]),
#if defined(GIZMO_SPH)
0.f,
#else
@@ -219,14 +224,14 @@ void dump_particle_fields(char *fileName, struct cell *main_cell,
"# Offset: [%2d %2d %2d] -----------------------------------\n",
i - 1, j - 1, k - 1);
- for (size_t pjd = 0; pjd < cj->count; pjd++) {
+ for (int pjd = 0; pjd < cj->count; pjd++) {
fprintf(
file,
"%6llu %10f %10f %10f %10f %10f %10f %13e %13e %13e %13e %13e "
"%13e %13e %13e\n",
cj->parts[pjd].id, cj->parts[pjd].x[0], cj->parts[pjd].x[1],
cj->parts[pjd].x[2], cj->parts[pjd].v[0], cj->parts[pjd].v[1],
- cj->parts[pjd].v[2], cj->parts[pjd].rho,
+ cj->parts[pjd].v[2], hydro_get_density(&cj->parts[pjd]),
#if defined(GIZMO_SPH)
0.f,
#else
diff --git a/tests/testAdiabaticIndex.c b/tests/testAdiabaticIndex.c
new file mode 100644
index 0000000000000000000000000000000000000000..e0c8c4f54bd2d6e5ddadb25bc44b96f1ca19aad2
--- /dev/null
+++ b/tests/testAdiabaticIndex.c
@@ -0,0 +1,116 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (C) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com).
+ *
+ * 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/>.
+ *
+ ******************************************************************************/
+
+#include "adiabatic_index.h"
+#include "error.h"
+
+/**
+ * @brief Check that a and b are consistent (up to some absolute error)
+ *
+ * @param a First value
+ * @param b Second value
+ * @param s String used to identify this check in messages
+ */
+void check_value(float a, float b, const char* s) {
+ if (fabsf(a - b) > 1.e-5f) {
+ error("Values are inconsistent: %g %g (%s)!", a, b, s);
+ } else {
+ message("Values are consistent: %g %g (%s).", a, b, s);
+ }
+}
+
+/**
+ * @brief Check that the pre-defined adiabatic index constants contain correct
+ * values
+ */
+void check_constants() {
+ float val;
+
+ val = 0.5 * (hydro_gamma + 1.0f) / hydro_gamma;
+ check_value(val, hydro_gamma_plus_one_over_two_gamma, "(gamma+1)/(2 gamma)");
+
+ val = 0.5 * (hydro_gamma - 1.0f) / hydro_gamma;
+ check_value(val, hydro_gamma_minus_one_over_two_gamma, "(gamma-1)/(2 gamma)");
+
+ val = (hydro_gamma - 1.0f) / (hydro_gamma + 1.0f);
+ check_value(val, hydro_gamma_minus_one_over_gamma_plus_one,
+ "(gamma-1)/(gamma+1)");
+
+ val = 2.0f / (hydro_gamma + 1.0f);
+ check_value(val, hydro_two_over_gamma_plus_one, "2/(gamma+1)");
+
+ val = 2.0f / (hydro_gamma - 1.0f);
+ check_value(val, hydro_two_over_gamma_minus_one, "2/(gamma-1)");
+
+ val = 0.5f * (hydro_gamma - 1.0f);
+ check_value(val, hydro_gamma_minus_one_over_two, "(gamma-1)/2");
+
+ val = 2.0f * hydro_gamma / (hydro_gamma - 1.0f);
+ check_value(val, hydro_two_gamma_over_gamma_minus_one, "(2 gamma)/(gamma-1)");
+
+ val = 1.0f / hydro_gamma;
+ check_value(val, hydro_one_over_gamma, "1/gamma");
+}
+
+/**
+ * @brief Check that the adiabatic index power functions return the correct
+ * values
+ */
+void check_functions() {
+ float val_a, val_b;
+ const float x = 0.4;
+
+ val_a = pow(x, -hydro_gamma);
+ val_b = pow_minus_gamma(x);
+ check_value(val_a, val_b, "x^(-gamma)");
+
+ val_a = pow(x, 2.0f / (hydro_gamma - 1.0f));
+ val_b = pow_two_over_gamma_minus_one(x);
+ check_value(val_a, val_b, "x^(2/(gamma-1))");
+
+ val_a = pow(x, 2.0f * hydro_gamma / (hydro_gamma - 1.0f));
+ val_b = pow_two_gamma_over_gamma_minus_one(x);
+ check_value(val_a, val_b, "x^((2 gamma)/(gamma-1))");
+
+ val_a = pow(x, 0.5f * (hydro_gamma - 1.0f) / hydro_gamma);
+ val_b = pow_gamma_minus_one_over_two_gamma(x);
+ check_value(val_a, val_b, "x^((gamma-1)/(2 gamma))");
+
+ val_a = pow(x, -0.5f * (hydro_gamma + 1.0f) / hydro_gamma);
+ val_b = pow_minus_gamma_plus_one_over_two_gamma(x);
+ check_value(val_a, val_b, "x^(-(gamma+1)/(2 gamma))");
+
+ val_a = pow(x, 1.0f / hydro_gamma);
+ val_b = pow_one_over_gamma(x);
+ check_value(val_a, val_b, "x^(1/gamma)");
+}
+
+/**
+ * @brief Check adiabatic index constants and power functions
+ */
+int main() {
+
+ /* check the values of the adiabatic index constants */
+ check_constants();
+
+ /* check the adiabatic index power functions */
+ check_functions();
+
+ return 0;
+}
diff --git a/tests/testMatrixInversion.c b/tests/testMatrixInversion.c
new file mode 100644
index 0000000000000000000000000000000000000000..9a45cd52d6f5d3ec96cc6d3f34fd683971f4cf19
--- /dev/null
+++ b/tests/testMatrixInversion.c
@@ -0,0 +1,125 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (C) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com).
+ *
+ * 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/>.
+ *
+ ******************************************************************************/
+
+#include <stdlib.h>
+#include <string.h>
+#include "const.h"
+#include "dimension.h"
+#include "error.h"
+#include "tools.h"
+
+void setup_matrix(float A[3][3]) {
+ A[0][0] = random_uniform(-1.0, 1.0);
+ A[0][1] = random_uniform(-1.0, 1.0);
+ A[0][2] = random_uniform(-1.0, 1.0);
+ A[1][0] = random_uniform(-1.0, 1.0);
+ A[1][1] = random_uniform(-1.0, 1.0);
+ A[1][2] = random_uniform(-1.0, 1.0);
+ A[2][0] = random_uniform(-1.0, 1.0);
+ A[2][1] = random_uniform(-1.0, 1.0);
+ A[2][2] = random_uniform(-1.0, 1.0);
+}
+
+int is_unit_matrix(float A[3][3]) {
+ int check = 1;
+
+ check &= (fabsf(A[0][0] - 1.0f) < 1.e-6f);
+
+#if defined(HYDRO_DIMENSION_2D) && defined(HYDRO_DIMENSION_3D)
+ check &= (fabsf(A[0][1]) < 1.e-6f);
+ check &= (fabsf(A[1][0]) < 1.e-6f);
+ check &= (fabsf(A[1][1] - 1.0f) < 1.e-6f);
+#if defined(HYDRO_DIMENSION_3D)
+ check &= (fabsf(A[0][2]) < 1.e-6f);
+ check &= (fabsf(A[1][2]) < 1.e-6f);
+ check &= (fabsf(A[2][0]) < 1.e-6f);
+ check &= (fabsf(A[2][1]) < 1.e-6f);
+ check &= (fabsf(A[2][2] - 1.0f) < 1.e-6f);
+#endif // 3D
+#endif // 2D and 3D
+
+ return check;
+}
+
+void print_matrix(float A[3][3], const char* s) {
+ message("Matrix %s:", s);
+#if defined(HYDRO_DIMENSION_1D)
+ message("[%.3e]", A[0][0]);
+#elif defined(HYDRO_DIMENSION_2D)
+ message("[%.3e, %.3e]", A[0][0], A[0][1]);
+ message("[%.3e, %.3e]", A[1][0], A[1][1]);
+#elif defined(HYDRO_DIMENSION_3D)
+ message("[%.8e, %.8e, %.8e]", A[0][0], A[0][1], A[0][2]);
+ message("[%.8e, %.8e, %.8e]", A[1][0], A[1][1], A[1][2]);
+ message("[%.8e, %.8e, %.8e]", A[2][0], A[2][1], A[2][2]);
+#endif
+}
+
+void multiply_matrices(float A[3][3], float B[3][3], float C[3][3]) {
+#if defined(HYDRO_DIMENSION_1D)
+ C[0][0] = A[0][0] * B[0][0];
+#elif defined(HYDRO_DIMENSION_2D)
+ for (int i = 0; i < 2; ++i) {
+ for (int j = 0; j < 2; ++j) {
+ C[i][j] = 0.0f;
+ for (int k = 0; k < 2; ++k) {
+ C[i][j] += A[i][k] * B[k][j];
+ }
+ }
+ }
+#elif defined(HYDRO_DIMENSION_3D)
+ for (int i = 0; i < 3; ++i) {
+ for (int j = 0; j < 3; ++j) {
+ C[i][j] = 0.0f;
+ for (int k = 0; k < 3; ++k) {
+ C[i][j] += A[i][k] * B[k][j];
+ }
+ }
+ }
+#endif
+}
+
+int main() {
+
+ float A[3][3], B[3][3], C[3][3];
+ setup_matrix(A);
+
+ memcpy(B, A, 9 * sizeof(float));
+
+ for (int i = 0; i < 3; ++i) {
+ for (int j = 0; j < 3; ++j) {
+ if (A[i][j] != B[i][j]) {
+ error("Matrices not equal after copy!");
+ }
+ }
+ }
+
+ invert_dimension_by_dimension_matrix(A);
+
+ multiply_matrices(A, B, C);
+
+ if (!is_unit_matrix(C)) {
+ print_matrix(A, "A");
+ print_matrix(B, "B");
+ print_matrix(C, "C");
+ error("Inverted matrix is wrong!");
+ }
+
+ return 0;
+}
diff --git a/tests/testPair.c b/tests/testPair.c
index f7b20faa3dc2aeb39e829ae634666e164c47b60a..efa1e628c2d57bf7922be8affdd5436ebca2f9cf 100644
--- a/tests/testPair.c
+++ b/tests/testPair.c
@@ -63,7 +63,11 @@ struct cell *make_cell(size_t n, double *offset, double size, double h,
part->v[2] = random_uniform(-0.05, 0.05);
part->h = size * h / (float)n;
part->id = ++(*partId);
+#ifdef GIZMO_SPH
+ part->conserved.mass = density * volume / count;
+#else
part->mass = density * volume / count;
+#endif
part->ti_begin = 0;
part->ti_end = 1;
++part;
@@ -105,7 +109,7 @@ void clean_up(struct cell *ci) {
* @brief Initializes all particles field to be ready for a density calculation
*/
void zero_particle_fields(struct cell *c) {
- for (size_t pid = 0; pid < c->count; pid++) {
+ for (int pid = 0; pid < c->count; pid++) {
hydro_init_part(&c->parts[pid]);
}
}
@@ -125,13 +129,13 @@ void dump_particle_fields(char *fileName, struct cell *ci, struct cell *cj) {
fprintf(file, "# ci --------------------------------------------\n");
- for (size_t pid = 0; pid < ci->count; pid++) {
+ for (int pid = 0; pid < ci->count; pid++) {
fprintf(file,
"%6llu %10f %10f %10f %10f %10f %10f %13e %13e %13e %13e %13e "
"%13e %13e %13e\n",
ci->parts[pid].id, ci->parts[pid].x[0], ci->parts[pid].x[1],
ci->parts[pid].x[2], ci->parts[pid].v[0], ci->parts[pid].v[1],
- ci->parts[pid].v[2], ci->parts[pid].rho,
+ ci->parts[pid].v[2], hydro_get_density(&ci->parts[pid]),
#if defined(GIZMO_SPH)
0.f,
#else
@@ -149,13 +153,13 @@ void dump_particle_fields(char *fileName, struct cell *ci, struct cell *cj) {
fprintf(file, "# cj --------------------------------------------\n");
- for (size_t pjd = 0; pjd < cj->count; pjd++) {
+ for (int pjd = 0; pjd < cj->count; pjd++) {
fprintf(file,
"%6llu %10f %10f %10f %10f %10f %10f %13e %13e %13e %13e %13e "
"%13e %13e %13e\n",
cj->parts[pjd].id, cj->parts[pjd].x[0], cj->parts[pjd].x[1],
cj->parts[pjd].x[2], cj->parts[pjd].v[0], cj->parts[pjd].v[1],
- cj->parts[pjd].v[2], cj->parts[pjd].rho,
+ cj->parts[pjd].v[2], hydro_get_density(&cj->parts[pjd]),
#if defined(GIZMO_SPH)
0.f,
#else
diff --git a/tests/testReading.c b/tests/testReading.c
index 5a9707d2705ed021996859f324310c4a4926730c..2ef32a5ef11c7e24a379ce5131df9cbea153fa7c 100644
--- a/tests/testReading.c
+++ b/tests/testReading.c
@@ -28,7 +28,7 @@ int main() {
size_t Ngas = 0, Ngpart = 0;
int periodic = -1;
int flag_entropy_ICs = -1;
- int i, j, k, n;
+ int i, j, k;
double dim[3];
struct part *parts = NULL;
struct gpart *gparts = NULL;
@@ -55,14 +55,14 @@ int main() {
assert(periodic == 1);
/* Check particles */
- for (n = 0; n < Ngas; ++n) {
+ for (size_t n = 0; n < Ngas; ++n) {
/* Check that indices are in a reasonable range */
unsigned long long index = parts[n].id;
assert(index < Ngas);
/* Check masses */
- float mass = parts[n].mass;
+ float mass = hydro_get_mass(&parts[n]);
float correct_mass = boxSize * boxSize * boxSize * rho / Ngas;
assert(mass == correct_mass);
diff --git a/tests/testRiemannExact.c b/tests/testRiemannExact.c
new file mode 100644
index 0000000000000000000000000000000000000000..1943820339ba2ac06d194a17d2d450157ded1a31
--- /dev/null
+++ b/tests/testRiemannExact.c
@@ -0,0 +1,338 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (C) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com).
+ *
+ * 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/>.
+ *
+ ******************************************************************************/
+
+#include <string.h>
+#include "error.h"
+#include "riemann/riemann_exact.h"
+#include "tools.h"
+
+int opposite(float a, float b) {
+ if ((a - b)) {
+ return fabs((a + b) / (a - b)) < 1.e-4;
+ } else {
+ return a == 0.0f;
+ }
+}
+
+int equal(float a, float b) {
+ if ((a + b)) {
+ return fabs((a - b) / (a + b)) < 1.e-4;
+ } else {
+ return a == 0.0f;
+ }
+}
+
+/**
+ * @brief Check that a and b are consistent (up to some error)
+ *
+ * @param a First value
+ * @param b Second value
+ * @param s String used to identify this check in messages
+ */
+void check_value(float a, float b, const char* s) {
+ if (fabsf(a - b) / fabsf(a + b) > 1.e-5f && fabsf(a - b) > 1.e-5f) {
+ error("Values are inconsistent: %g %g (%s)!", a, b, s);
+ } else {
+ message("Values are consistent: %g %g (%s).", a, b, s);
+ }
+}
+
+struct riemann_statevector {
+ /*! @brief Density */
+ float rho;
+
+ /*! @brief Fluid velocity */
+ float v;
+
+ /*! @brief Pressure */
+ float P;
+};
+
+/**
+ * @brief Check that the solution to the Riemann problem with given left and
+ * right state is consistent with the given expected solution
+ *
+ * @param WL Left state
+ * @param WR Right state
+ * @param Whalf Expected solution
+ * @param s String used to identify this check in messages
+ */
+void check_riemann_solution(struct riemann_statevector* WL,
+ struct riemann_statevector* WR,
+ struct riemann_statevector* Whalf, const char* s) {
+ float WLarr[5], WRarr[5], Whalfarr[5], n_unit[3];
+
+ n_unit[0] = 1.0f;
+ n_unit[1] = 0.0f;
+ n_unit[2] = 0.0f;
+
+ WLarr[0] = WL->rho;
+ WLarr[1] = WL->v;
+ WLarr[2] = 0.0f;
+ WLarr[3] = 0.0f;
+ WLarr[4] = WL->P;
+
+ WRarr[0] = WR->rho;
+ WRarr[1] = WR->v;
+ WRarr[2] = 0.0f;
+ WRarr[3] = 0.0f;
+ WRarr[4] = WR->P;
+
+ riemann_solver_solve(WLarr, WRarr, Whalfarr, n_unit);
+
+ message("Checking %s...", s);
+ check_value(Whalfarr[0], Whalf->rho, "rho");
+ check_value(Whalfarr[1], Whalf->v, "v");
+ check_value(Whalfarr[4], Whalf->P, "P");
+}
+
+/**
+ * @brief Check the exact Riemann solver on the Toro test problems
+ */
+void check_riemann_exact() {
+ struct riemann_statevector WL, WR, Whalf;
+
+ /* Test 1 */
+ WL.rho = 1.0f;
+ WL.v = 0.0f;
+ WL.P = 1.0f;
+ WR.rho = 0.125f;
+ WR.v = 0.0f;
+ WR.P = 0.1f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 0.47969f;
+ Whalf.v = 0.841194f;
+ Whalf.P = 0.293945f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 0.411437f;
+ Whalf.v = 0.953205f;
+ Whalf.P = 0.306011f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 0.534767f;
+ Whalf.v = 0.760062f;
+ Whalf.P = 0.285975f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 1");
+
+ /* Test 2 */
+ WL.rho = 1.0f;
+ WL.v = -2.0f;
+ WL.P = 0.4f;
+ WR.rho = 1.0f;
+ WR.v = 2.0f;
+ WR.P = 0.4f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 0.00617903f;
+ Whalf.v = 0.0f;
+ Whalf.P = 8.32249e-5f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 0.0257933f;
+ Whalf.v = 0.0f;
+ Whalf.P = 0.00304838f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 0.0f;
+ Whalf.v = 0.0f;
+ Whalf.P = 0.0f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 2");
+
+ /* Test 3 */
+ WL.rho = 1.0f;
+ WL.v = 0.0f;
+ WL.P = 1000.0f;
+ WR.rho = 1.0f;
+ WR.v = 0.0f;
+ WR.P = 0.01f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 0.615719f;
+ Whalf.v = 18.2812f;
+ Whalf.P = 445.626f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 0.563517f;
+ Whalf.v = 19.9735f;
+ Whalf.P = 465.453f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 0.656768f;
+ Whalf.v = 16.9572f;
+ Whalf.P = 431.345f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 3");
+
+ /* Test 4 */
+ WL.rho = 1.0f;
+ WL.v = 0.0f;
+ WL.P = 0.01f;
+ WR.rho = 1.0f;
+ WR.v = 0.0f;
+ WR.P = 100.0f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 0.61577f;
+ Whalf.v = -5.78022f;
+ Whalf.P = 44.5687f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 0.563567f;
+ Whalf.v = -6.31525f;
+ Whalf.P = 46.5508f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 0.656819f;
+ Whalf.v = -5.36146f;
+ Whalf.P = 43.1412f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 4");
+
+ /* Test 5 */
+ WL.rho = 5.99924f;
+ WL.v = 19.5975f;
+ WL.P = 460.894f;
+ WR.rho = 5.99242f;
+ WR.v = -6.19633f;
+ WR.P = 46.0950f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 12.743f;
+ Whalf.v = 8.56045f;
+ Whalf.P = 1841.82f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 5.99924f;
+ Whalf.v = 19.5975f;
+ Whalf.P = 460.894f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 11.5089f;
+ Whalf.v = 8.42099f;
+ Whalf.P = 2026.27f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 5");
+}
+
+/**
+ * @brief Check the symmetry of the TRRS Riemann solver
+ */
+void check_riemann_symmetry() {
+ float WL[5], WR[5], Whalf1[5], Whalf2[5], n_unit1[3], n_unit2[3], n_norm,
+ vij[3], totflux1[5], totflux2[5];
+
+ WL[0] = random_uniform(0.1f, 1.0f);
+ WL[1] = random_uniform(-10.0f, 10.0f);
+ WL[2] = random_uniform(-10.0f, 10.0f);
+ WL[3] = random_uniform(-10.0f, 10.0f);
+ WL[4] = random_uniform(0.1f, 1.0f);
+ WR[0] = random_uniform(0.1f, 1.0f);
+ WR[1] = random_uniform(-10.0f, 10.0f);
+ WR[2] = random_uniform(-10.0f, 10.0f);
+ WR[3] = random_uniform(-10.0f, 10.0f);
+ WR[4] = random_uniform(0.1f, 1.0f);
+
+ n_unit1[0] = random_uniform(-1.0f, 1.0f);
+ n_unit1[1] = random_uniform(-1.0f, 1.0f);
+ n_unit1[2] = random_uniform(-1.0f, 1.0f);
+
+ n_norm = sqrtf(n_unit1[0] * n_unit1[0] + n_unit1[1] * n_unit1[1] +
+ n_unit1[2] * n_unit1[2]);
+ n_unit1[0] /= n_norm;
+ n_unit1[1] /= n_norm;
+ n_unit1[2] /= n_norm;
+
+ n_unit2[0] = -n_unit1[0];
+ n_unit2[1] = -n_unit1[1];
+ n_unit2[2] = -n_unit1[2];
+
+ riemann_solver_solve(WL, WR, Whalf1, n_unit1);
+ riemann_solver_solve(WR, WL, Whalf2, n_unit2);
+
+ if (!equal(Whalf1[0], Whalf2[0]) || !equal(Whalf1[1], Whalf2[1]) ||
+ !equal(Whalf1[2], Whalf2[2]) || !equal(Whalf1[3], Whalf2[3]) ||
+ !equal(Whalf1[4], Whalf2[4])) {
+ message(
+ "Solver asymmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ Whalf1[0], Whalf1[1], Whalf1[2], Whalf1[3], Whalf1[4], Whalf2[0],
+ Whalf2[1], Whalf2[2], Whalf2[3], Whalf2[4]);
+ message("Asymmetry in solution!\n");
+ /* This asymmetry is to be expected, since we do an iteration. Are the
+ results at least consistent? */
+ check_value(Whalf1[0], Whalf2[0], "Rho solution");
+ check_value(Whalf1[1], Whalf2[1], "V[0] solution");
+ check_value(Whalf1[2], Whalf2[2], "V[1] solution");
+ check_value(Whalf1[3], Whalf2[3], "V[2] solution");
+ check_value(Whalf1[4], Whalf2[4], "Pressure solution");
+ } else {
+ message(
+ "Solver symmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ Whalf1[0], Whalf1[1], Whalf1[2], Whalf1[3], Whalf1[4], Whalf2[0],
+ Whalf2[1], Whalf2[2], Whalf2[3], Whalf2[4]);
+ }
+
+ vij[0] = random_uniform(-10.0f, 10.0f);
+ vij[1] = random_uniform(-10.0f, 10.0f);
+ vij[2] = random_uniform(-10.0f, 10.0f);
+
+ riemann_solve_for_flux(WL, WR, n_unit1, vij, totflux1);
+ riemann_solve_for_flux(WR, WL, n_unit2, vij, totflux2);
+
+ if (!opposite(totflux1[0], totflux2[0]) ||
+ !opposite(totflux1[1], totflux2[1]) ||
+ !opposite(totflux1[2], totflux2[2]) ||
+ !opposite(totflux1[3], totflux2[3]) ||
+ !opposite(totflux1[4], totflux2[4])) {
+ message(
+ "Flux solver asymmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ totflux1[0], totflux1[1], totflux1[2], totflux1[3], totflux1[4],
+ totflux2[0], totflux2[1], totflux2[2], totflux2[3], totflux2[4]);
+ message("Asymmetry in flux solution!");
+ /* This asymmetry is to be expected, since we do an iteration. Are the
+ results at least consistent? */
+ check_value(totflux1[0], totflux2[0], "Mass flux");
+ check_value(totflux1[1], totflux2[1], "Momentum[0] flux");
+ check_value(totflux1[2], totflux2[2], "Momentum[1] flux");
+ check_value(totflux1[3], totflux2[3], "Momentum[2] flux");
+ check_value(totflux1[4], totflux2[4], "Energy flux");
+ } else {
+ message(
+ "Flux solver symmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ totflux1[0], totflux1[1], totflux1[2], totflux1[3], totflux1[4],
+ totflux2[0], totflux2[1], totflux2[2], totflux2[3], totflux2[4]);
+ }
+}
+
+/**
+ * @brief Check the exact Riemann solver
+ */
+int main() {
+
+ /* check the exact Riemann solver */
+ check_riemann_exact();
+
+ /* symmetry test */
+ int i;
+ for (i = 0; i < 100; ++i) check_riemann_symmetry();
+
+ return 0;
+}
diff --git a/tests/testRiemannHLLC.c b/tests/testRiemannHLLC.c
new file mode 100644
index 0000000000000000000000000000000000000000..4cf883b68efbcfd795d0b7894adb9e7265b14d14
--- /dev/null
+++ b/tests/testRiemannHLLC.c
@@ -0,0 +1,96 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (C) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com).
+ *
+ * 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/>.
+ *
+ ******************************************************************************/
+
+#include <string.h>
+#include "error.h"
+#include "riemann/riemann_hllc.h"
+#include "tools.h"
+
+int consistent_with_zero(float val) { return fabs(val) < 1.e-4; }
+
+/**
+ * @brief Check the symmetry of the TRRS Riemann solver
+ */
+void check_riemann_symmetry() {
+ float WL[5], WR[5], n_unit1[3], n_unit2[3], n_norm, vij[3], totflux1[5],
+ totflux2[5];
+
+ WL[0] = random_uniform(0.1f, 1.0f);
+ WL[1] = random_uniform(-10.0f, 10.0f);
+ WL[2] = random_uniform(-10.0f, 10.0f);
+ WL[3] = random_uniform(-10.0f, 10.0f);
+ WL[4] = random_uniform(0.1f, 1.0f);
+ WR[0] = random_uniform(0.1f, 1.0f);
+ WR[1] = random_uniform(-10.0f, 10.0f);
+ WR[2] = random_uniform(-10.0f, 10.0f);
+ WR[3] = random_uniform(-10.0f, 10.0f);
+ WR[4] = random_uniform(0.1f, 1.0f);
+
+ n_unit1[0] = random_uniform(-1.0f, 1.0f);
+ n_unit1[1] = random_uniform(-1.0f, 1.0f);
+ n_unit1[2] = random_uniform(-1.0f, 1.0f);
+
+ n_norm = sqrtf(n_unit1[0] * n_unit1[0] + n_unit1[1] * n_unit1[1] +
+ n_unit1[2] * n_unit1[2]);
+ n_unit1[0] /= n_norm;
+ n_unit1[1] /= n_norm;
+ n_unit1[2] /= n_norm;
+
+ n_unit2[0] = -n_unit1[0];
+ n_unit2[1] = -n_unit1[1];
+ n_unit2[2] = -n_unit1[2];
+
+ vij[0] = random_uniform(-10.0f, 10.0f);
+ vij[1] = random_uniform(-10.0f, 10.0f);
+ vij[2] = random_uniform(-10.0f, 10.0f);
+
+ riemann_solve_for_flux(WL, WR, n_unit1, vij, totflux1);
+ riemann_solve_for_flux(WR, WL, n_unit2, vij, totflux2);
+
+ if (!consistent_with_zero(totflux1[0] + totflux2[0]) ||
+ !consistent_with_zero(totflux1[1] + totflux2[1]) ||
+ !consistent_with_zero(totflux1[2] + totflux2[2]) ||
+ !consistent_with_zero(totflux1[3] + totflux2[3]) ||
+ !consistent_with_zero(totflux1[4] + totflux2[4])) {
+ message(
+ "Flux solver asymmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ totflux1[0], totflux1[1], totflux1[2], totflux1[3], totflux1[4],
+ totflux2[0], totflux2[1], totflux2[2], totflux2[3], totflux2[4]);
+ error("Asymmetry in flux solution!");
+ } else {
+ message(
+ "Flux solver symmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ totflux1[0], totflux1[1], totflux1[2], totflux1[3], totflux1[4],
+ totflux2[0], totflux2[1], totflux2[2], totflux2[3], totflux2[4]);
+ }
+}
+
+/**
+ * @brief Check the HLLC Riemann solver
+ */
+int main() {
+
+ int i;
+ /* symmetry test */
+ for (i = 0; i < 100; i++) check_riemann_symmetry();
+
+ return 0;
+}
diff --git a/tests/testRiemannTRRS.c b/tests/testRiemannTRRS.c
new file mode 100644
index 0000000000000000000000000000000000000000..18ecbdce9173f43674a63b21231322cb01620d29
--- /dev/null
+++ b/tests/testRiemannTRRS.c
@@ -0,0 +1,324 @@
+/*******************************************************************************
+ * This file is part of SWIFT.
+ * Copyright (C) 2016 Bert Vandenbroucke (bert.vandenbroucke@gmail.com).
+ *
+ * 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/>.
+ *
+ ******************************************************************************/
+
+#include <string.h>
+#include "error.h"
+#include "riemann/riemann_trrs.h"
+#include "tools.h"
+
+int opposite(float a, float b) {
+ if ((a - b)) {
+ return fabs((a + b) / (a - b)) < 1.e-4;
+ } else {
+ return a == 0.0f;
+ }
+}
+
+int equal(float a, float b) {
+ if ((a + b)) {
+ return fabs((a - b) / (a + b)) < 1.e-4;
+ } else {
+ return a == 0.0f;
+ }
+}
+
+/**
+ * @brief Check that a and b are consistent (up to some error)
+ *
+ * @param a First value
+ * @param b Second value
+ * @param s String used to identify this check in messages
+ */
+void check_value(float a, float b, const char* s) {
+ if (fabsf(a - b) / fabsf(a + b) > 1.e-5f && fabsf(a - b) > 1.e-5f) {
+ error("Values are inconsistent: %g %g (%s)!", a, b, s);
+ } else {
+ message("Values are consistent: %g %g (%s).", a, b, s);
+ }
+}
+
+struct riemann_statevector {
+ /*! @brief Density */
+ float rho;
+
+ /*! @brief Fluid velocity */
+ float v;
+
+ /*! @brief Pressure */
+ float P;
+};
+
+/**
+ * @brief Check that the solution to the Riemann problem with given left and
+ * right state is consistent with the given expected solution
+ *
+ * @param WL Left state
+ * @param WR Right state
+ * @param Whalf Expected solution
+ * @param s String used to identify this check in messages
+ */
+void check_riemann_solution(struct riemann_statevector* WL,
+ struct riemann_statevector* WR,
+ struct riemann_statevector* Whalf, const char* s) {
+ float WLarr[5], WRarr[5], Whalfarr[5], n_unit[3];
+
+ n_unit[0] = 1.0f;
+ n_unit[1] = 0.0f;
+ n_unit[2] = 0.0f;
+
+ WLarr[0] = WL->rho;
+ WLarr[1] = WL->v;
+ WLarr[2] = 0.0f;
+ WLarr[3] = 0.0f;
+ WLarr[4] = WL->P;
+
+ WRarr[0] = WR->rho;
+ WRarr[1] = WR->v;
+ WRarr[2] = 0.0f;
+ WRarr[3] = 0.0f;
+ WRarr[4] = WR->P;
+
+ riemann_solver_solve(WLarr, WRarr, Whalfarr, n_unit);
+
+ message("Checking %s...", s);
+ check_value(Whalfarr[0], Whalf->rho, "rho");
+ check_value(Whalfarr[1], Whalf->v, "v");
+ check_value(Whalfarr[4], Whalf->P, "P");
+}
+
+/**
+ * @brief Check the TRRS Riemann solver on the Toro test problems
+ */
+void check_riemann_trrs() {
+ struct riemann_statevector WL, WR, Whalf;
+
+ /* Test 1 */
+ WL.rho = 1.0f;
+ WL.v = 0.0f;
+ WL.P = 1.0f;
+ WR.rho = 0.125f;
+ WR.v = 0.0f;
+ WR.P = 0.1f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 0.481167f;
+ Whalf.v = 0.838085f;
+ Whalf.P = 0.295456f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 0.41586f;
+ Whalf.v = 0.942546f;
+ Whalf.P = 0.310406f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 0.53478f;
+ Whalf.v = 0.760037f;
+ Whalf.P = 0.285989f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 1");
+
+ /* Test 2 */
+ WL.rho = 1.0f;
+ WL.v = -2.0f;
+ WL.P = 0.4f;
+ WR.rho = 1.0f;
+ WR.v = 2.0f;
+ WR.P = 0.4f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 0.00617903f;
+ Whalf.v = 0.0f;
+ Whalf.P = 8.32249e-5f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 0.0257933f;
+ Whalf.v = 0.0f;
+ Whalf.P = 0.00304838f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 0.013932f;
+ Whalf.v = 0.0f;
+ Whalf.P = 7.76405e-5f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 2");
+
+ /* Test 3 */
+ WL.rho = 1.0f;
+ WL.v = 0.0f;
+ WL.P = 1000.0f;
+ WR.rho = 1.0f;
+ WR.v = 0.0f;
+ WR.P = 0.01f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 0.919498f;
+ Whalf.v = 3.37884f;
+ Whalf.P = 869.464f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 0.941258f;
+ Whalf.v = 2.19945f;
+ Whalf.P = 922.454f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 0.902032f;
+ Whalf.v = 4.49417f;
+ Whalf.P = 813.662f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 3");
+
+ /* Test 4 */
+ WL.rho = 1.0f;
+ WL.v = 0.0f;
+ WL.P = 0.01f;
+ WR.rho = 1.0f;
+ WR.v = 0.0f;
+ WR.P = 100.0f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 0.857525f;
+ Whalf.v = -1.93434f;
+ Whalf.P = 77.4007f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 0.880649f;
+ Whalf.v = -1.45215f;
+ Whalf.P = 84.4119f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 0.843058f;
+ Whalf.v = -2.31417f;
+ Whalf.P = 71.0747f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 4");
+
+ /* Test 5 */
+ WL.rho = 5.99924f;
+ WL.v = 19.5975f;
+ WL.P = 460.894f;
+ WR.rho = 5.99242f;
+ WR.v = -6.19633f;
+ WR.P = 46.0950f;
+#if defined(HYDRO_GAMMA_5_3)
+ Whalf.rho = 5.99924f;
+ Whalf.v = 19.5975f;
+ Whalf.P = 460.894f;
+#elif defined(HYDRO_GAMMA_4_3)
+ Whalf.rho = 5.99924f;
+ Whalf.v = 19.5975f;
+ Whalf.P = 460.894f;
+#elif defined(HYDRO_GAMMA_2_1)
+ Whalf.rho = 5.99924f;
+ Whalf.v = 19.5975f;
+ Whalf.P = 460.894f;
+#else
+#error "Unsupported adiabatic index!"
+#endif
+ check_riemann_solution(&WL, &WR, &Whalf, "Test 5");
+}
+
+/**
+ * @brief Check the symmetry of the TRRS Riemann solver
+ */
+void check_riemann_symmetry() {
+ float WL[5], WR[5], Whalf1[5], Whalf2[5], n_unit1[3], n_unit2[3], n_norm,
+ vij[3], totflux1[5], totflux2[5];
+
+ WL[0] = random_uniform(0.1f, 1.0f);
+ WL[1] = random_uniform(-10.0f, 10.0f);
+ WL[2] = random_uniform(-10.0f, 10.0f);
+ WL[3] = random_uniform(-10.0f, 10.0f);
+ WL[4] = random_uniform(0.1f, 1.0f);
+ WR[0] = random_uniform(0.1f, 1.0f);
+ WR[1] = random_uniform(-10.0f, 10.0f);
+ WR[2] = random_uniform(-10.0f, 10.0f);
+ WR[3] = random_uniform(-10.0f, 10.0f);
+ WR[4] = random_uniform(0.1f, 1.0f);
+
+ n_unit1[0] = random_uniform(-1.0f, 1.0f);
+ n_unit1[1] = random_uniform(-1.0f, 1.0f);
+ n_unit1[2] = random_uniform(-1.0f, 1.0f);
+
+ n_norm = sqrtf(n_unit1[0] * n_unit1[0] + n_unit1[1] * n_unit1[1] +
+ n_unit1[2] * n_unit1[2]);
+ n_unit1[0] /= n_norm;
+ n_unit1[1] /= n_norm;
+ n_unit1[2] /= n_norm;
+
+ n_unit2[0] = -n_unit1[0];
+ n_unit2[1] = -n_unit1[1];
+ n_unit2[2] = -n_unit1[2];
+
+ riemann_solver_solve(WL, WR, Whalf1, n_unit1);
+ riemann_solver_solve(WR, WL, Whalf2, n_unit2);
+
+ if (!equal(Whalf1[0], Whalf2[0]) || !equal(Whalf1[1], Whalf2[1]) ||
+ !equal(Whalf1[2], Whalf2[2]) || !equal(Whalf1[3], Whalf2[3]) ||
+ !equal(Whalf1[4], Whalf2[4])) {
+ message(
+ "Solver asymmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ Whalf1[0], Whalf1[1], Whalf1[2], Whalf1[3], Whalf1[4], Whalf2[0],
+ Whalf2[1], Whalf2[2], Whalf2[3], Whalf2[4]);
+ error("Asymmetry in solution!");
+ } else {
+ message(
+ "Solver symmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ Whalf1[0], Whalf1[1], Whalf1[2], Whalf1[3], Whalf1[4], Whalf2[0],
+ Whalf2[1], Whalf2[2], Whalf2[3], Whalf2[4]);
+ }
+
+ vij[0] = random_uniform(-10.0f, 10.0f);
+ vij[1] = random_uniform(-10.0f, 10.0f);
+ vij[2] = random_uniform(-10.0f, 10.0f);
+
+ riemann_solve_for_flux(WL, WR, n_unit1, vij, totflux1);
+ riemann_solve_for_flux(WR, WL, n_unit2, vij, totflux2);
+
+ if (!opposite(totflux1[0], totflux2[0]) ||
+ !opposite(totflux1[1], totflux2[1]) ||
+ !opposite(totflux1[2], totflux2[2]) ||
+ !opposite(totflux1[3], totflux2[3]) ||
+ !opposite(totflux1[4], totflux2[4])) {
+ message(
+ "Solver asymmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ totflux1[0], totflux1[1], totflux1[2], totflux1[3], totflux1[4],
+ totflux2[0], totflux2[1], totflux2[2], totflux2[3], totflux2[4]);
+ error("Asymmetry in solution!");
+ } else {
+ message(
+ "Solver symmetric: [%.3e,%.3e,%.3e,%.3e,%.3e] == "
+ "[%.3e,%.3e,%.3e,%.3e,%.3e]\n",
+ totflux1[0], totflux1[1], totflux1[2], totflux1[3], totflux1[4],
+ totflux2[0], totflux2[1], totflux2[2], totflux2[3], totflux2[4]);
+ }
+}
+
+/**
+ * @brief Check the TRRS Riemann solver
+ */
+int main() {
+
+ /* check the TRRS Riemann solver */
+ check_riemann_trrs();
+
+ /* symmetry test */
+ int i;
+ for (i = 0; i < 100; i++) check_riemann_symmetry();
+
+ return 0;
+}
diff --git a/tests/testSymmetry.c b/tests/testSymmetry.c
index eb3fab6becca08e9ef87e7c60cc8c04bd2a0290c..6469d314fb8b1438cc2c9737669c1a13a97bd803 100644
--- a/tests/testSymmetry.c
+++ b/tests/testSymmetry.c
@@ -46,6 +46,55 @@ int main(int argc, char *argv[]) {
pi.id = 1;
pj.id = 2;
+#if defined(GIZMO_SPH)
+ /* Give the primitive variables sensible values, since the Riemann solver does
+ not like negative densities and pressures */
+ pi.primitives.rho = random_uniform(0.1f, 1.0f);
+ pi.primitives.v[0] = random_uniform(-10.0f, 10.0f);
+ pi.primitives.v[1] = random_uniform(-10.0f, 10.0f);
+ pi.primitives.v[2] = random_uniform(-10.0f, 10.0f);
+ pi.primitives.P = random_uniform(0.1f, 1.0f);
+ pj.primitives.rho = random_uniform(0.1f, 1.0f);
+ pj.primitives.v[0] = random_uniform(-10.0f, 10.0f);
+ pj.primitives.v[1] = random_uniform(-10.0f, 10.0f);
+ pj.primitives.v[2] = random_uniform(-10.0f, 10.0f);
+ pj.primitives.P = random_uniform(0.1f, 1.0f);
+ /* make gradients zero */
+ pi.primitives.gradients.rho[0] = 0.0f;
+ pi.primitives.gradients.rho[1] = 0.0f;
+ pi.primitives.gradients.rho[2] = 0.0f;
+ pi.primitives.gradients.v[0][0] = 0.0f;
+ pi.primitives.gradients.v[0][1] = 0.0f;
+ pi.primitives.gradients.v[0][2] = 0.0f;
+ pi.primitives.gradients.v[1][0] = 0.0f;
+ pi.primitives.gradients.v[1][1] = 0.0f;
+ pi.primitives.gradients.v[1][2] = 0.0f;
+ pi.primitives.gradients.v[2][0] = 0.0f;
+ pi.primitives.gradients.v[2][1] = 0.0f;
+ pi.primitives.gradients.v[2][2] = 0.0f;
+ pi.primitives.gradients.P[0] = 0.0f;
+ pi.primitives.gradients.P[1] = 0.0f;
+ pi.primitives.gradients.P[2] = 0.0f;
+ pj.primitives.gradients.rho[0] = 0.0f;
+ pj.primitives.gradients.rho[1] = 0.0f;
+ pj.primitives.gradients.rho[2] = 0.0f;
+ pj.primitives.gradients.v[0][0] = 0.0f;
+ pj.primitives.gradients.v[0][1] = 0.0f;
+ pj.primitives.gradients.v[0][2] = 0.0f;
+ pj.primitives.gradients.v[1][0] = 0.0f;
+ pj.primitives.gradients.v[1][1] = 0.0f;
+ pj.primitives.gradients.v[1][2] = 0.0f;
+ pj.primitives.gradients.v[2][0] = 0.0f;
+ pj.primitives.gradients.v[2][1] = 0.0f;
+ pj.primitives.gradients.v[2][2] = 0.0f;
+ pj.primitives.gradients.P[0] = 0.0f;
+ pj.primitives.gradients.P[1] = 0.0f;
+ pj.primitives.gradients.P[2] = 0.0f;
+ /* set time step to reasonable value */
+ pi.force.dt = 0.001;
+ pj.force.dt = 0.001;
+#endif
+
/* Make an xpart companion */
struct xpart xpi, xpj;
bzero(&xpi, sizeof(struct xpart));
@@ -100,12 +149,54 @@ int main(int argc, char *argv[]) {
dx[2] = -dx[2];
runner_iact_nonsym_force(r2, dx, pj2.h, pi2.h, &pj2, &pi2);
- /* Check that the particles are the same */
+/* Check that the particles are the same */
+#if defined(GIZMO_SPH)
+ i_ok = 0;
+ j_ok = 0;
+ for (size_t i = 0; i < sizeof(struct part) / sizeof(float); ++i) {
+ float a = *(((float *)&pi) + i);
+ float b = *(((float *)&pi2) + i);
+ float c = *(((float *)&pj) + i);
+ float d = *(((float *)&pj2) + i);
+
+ int a_is_b;
+ if ((a + b)) {
+ a_is_b = (fabs((a - b) / (a + b)) > 1.e-4);
+ } else {
+ a_is_b = !(a == 0.0f);
+ }
+ int c_is_d;
+ if ((c + d)) {
+ c_is_d = (fabs((c - d) / (c + d)) > 1.e-4);
+ } else {
+ c_is_d = !(c == 0.0f);
+ }
+
+ if (a_is_b) {
+ message("%.8e, %.8e, %lu", a, b, i);
+ }
+ if (c_is_d) {
+ message("%.8e, %.8e, %lu", c, d, i);
+ }
+
+ i_ok |= a_is_b;
+ j_ok |= c_is_d;
+ }
+#else
i_ok = memcmp(&pi, &pi2, sizeof(struct part));
j_ok = memcmp(&pj, &pj2, sizeof(struct part));
+#endif
- if (i_ok) error("Particles 'pi' do not match after force");
- if (j_ok) error("Particles 'pj' do not match after force");
+ if (i_ok) {
+ printParticle_single(&pi, &xpi);
+ printParticle_single(&pi2, &xpi);
+ error("Particles 'pi' do not match after force");
+ }
+ if (j_ok) {
+ printParticle_single(&pj, &xpj);
+ printParticle_single(&pj2, &xpj);
+ error("Particles 'pj' do not match after force");
+ }
return 0;
}
diff --git a/tests/testThreadpool.c b/tests/testThreadpool.c
index 90ec58e10b16d816b52b785c87e6604ad11f61d8..aa65d533a29afbe4e7e8384fb887281822a31e58 100644
--- a/tests/testThreadpool.c
+++ b/tests/testThreadpool.c
@@ -77,4 +77,9 @@ int main(int argc, char *argv[]) {
fflush(stdout);
threadpool_map(&tp, map_function_first, data, N, sizeof(int), 2, NULL);
}
+
+ /* Be clean */
+ threadpool_clean(&tp);
+
+ return 0;
}