Merge pull request #292 from ludwig-cf/develop

Release 0.21.0

.github/workflows/codeql-analysis.yml

@@ -36,7 +36,7 @@ jobs:
 
     steps:
     - name: Checkout repository
-      uses: actions/checkout@v2
+      uses: actions/checkout@v3
 
     # Initializes the CodeQL tools for scanning.
     - name: Initialize CodeQL

.github/workflows/regression.yml

@@ -89,3 +89,4 @@ jobs:
     - name: Check
       run: |
         make unit
+        make -C tests d3q27

CHANGES.md

@@ -1,6 +1,45 @@
 
 ### Changes
 
+
+version 0.21.0
+
+- Ellipsoids. A feature which allows ellipsoidal particles to be
+  used has been added by Sumesh Thampi.
+
+- "colloid_type" has been split into a number of more general
+  properties including: boundary condition (bc) either bbl or
+  subgrid; shale either "disk", "sphere", or "ellipsoid";
+  active (logical); magnetic (logical).
+  The default is: bc = bbl, shape = sphere, active = no, magnetic = no.
+  There are some conditions, e.g., "disk" expects a 2D D2Q9 model.
+  This affects the way colloid details are specified in the input.
+  Older colloid state files should still work.
+  See https://ludwig.epcc.ed.ac.uk
+
+- There is no longer an option available for field halo swaps; the
+  scheme has been consolidated is always the same (for CPU). GPU
+  run time configuration is pending. Halo information is still
+  available https://ludwig.epcc.ed.ac.uk/inputs/parallel.html
+- If you use a 2-dimensional system (L_z = 1) with an order parameter, you
+  must use a 2d gradient calculation. This is related to the (non-) treatment
+  of halo swaps in the third dimension for fields.
+- A `colloid_buoyancy` option has been added which is similar to
+  `colloid_gravity` but computes a force proportional to colloid volume.
+  See https://ludwig.epcc.ed.ac.uk/inputs/colloid.html#external-forces
+
+Bug fixes:
+
+- Issue 270: a colloid with an initial position placed exactly at a sub-domain
+             boundary can fail to be attached to the cell list in some
+	     circumstances causing a crash. This has been fixed by adding
+	     a small position adjustment, or if this doesn't work, failing
+	     with a message.
+- Issue 268: if using wetting information read from porous media files,
+             the form of the key words in the input file has been
+	     adjusted. See https://ludwig.epcc.ed.ac.uk/inputs/porous.html
+
+
 version 0.20.1
 - Issue 271: missing stub prevents compilation at some compiler optimisation
              levels.
@@ -26,16 +65,16 @@ version 0.19.0
 
 - There has been a significant change to the way that i/o is undertaken.
   See https://ludwig.epcc.ed.ac.uk/outputs/fluid.html
-  It is the intension to replace completely the 'old' I/O mechanism
+  It is the intention to replace completely the 'old' I/O mechanism
   by release v0.21.0.
 - The extract_colloids.c utility has been updated so that it takes
-  only one comand line argument for new I/O metadata.
+  only one command line argument for new I/O metadata.
 
 - Input associated with the choice of force arising from the free energy
   sector has been made more general. Specifically, input keys
   `fd_force_divergence` and `fe_use_stress_relaxation` are replaced.
   - For scalar order parameters an extra version of the "phi_gradmu"
-    approach is avaialble" "phi_gradmu_correction".
+    approach is available" "phi_gradmu_correction".
   - See https://ludwig.epcc.ed.ac.uk/inputs/force.html for details.
 - Order parameter statistics are now reported as "phi" for scalars
   [Px,Py,Pz] for vectors, and [Qxx, Qxy, Qxz, Qyy, Qyz] for liquid
@@ -77,7 +116,7 @@ version 0.17.0
 - add liquid crystal anchoring "fd_gradient_calculation s7_anchoring"
   - this is a replcement for "3d_7pt_fluid" and does a slightly better
     job at the edges and corners by using a consistent surface normal.
-    The anchoring properties are now specifed in a slightly different
+    The anchoring properties are now specified in a slightly different
     way.
   - For walls, see https://ludwig.epcc.ed.ac.uk/inputs/walls.html
   - For colloids, see https://ludwig.epcc.ed.ac.uk/inputs/colloid.html
@@ -161,7 +200,7 @@ version 0.13.0
 version 0.12.0
 
 - Allow user to specify a linear combination of slip and no-slip for
-  plane walls. This was originally implementated by Katrin Wolff when
+  plane walls. This was originally implemented by Katrin Wolff when
   at Edinburgh, and has been resurrected with the help of Ryan Keogh
   and Tyler Shendruk. See https://ludwig.epcc.ed.ac.uk/inputs/walls.html
 - Various minor code quality improvements
@@ -217,7 +256,7 @@ version 0.9.0
 - Regression tests have been re-organised into different directories
   and are run on a per-directory basis (see tests/Makefile)
 
-- The default test is regression/d3q19-short 
+- The default test is regression/d3q19-short
 
 - A link to new build and test instructions has been made available
   from the README

codecov.yml

@@ -0,0 +1,9 @@
+coverage:
+  status:
+    project:
+      default:
+        target: 33%
+        threshold: 2%
+    patch:
+      default:
+        informational: true

config/epcc-cirrus-nvcc.mk

@@ -2,30 +2,35 @@
 #
 #  nvcc build
 #
-#  module load intel-mpi-17
-#  module load intel-comiplers-17
-#  module load cuda/9.1
+#  There are a number of ways forward here... see further comments below
 #
-#  Host   As epcc-cirrus-intel.mk
-#  Device NVIDIA Tesla V100-SXM2
+#  E.g., serial with stub MPI
+#  module load nvidia/nvhpc-nompi/22.11
+#
+#  E.g., parallel using NVHPC with MPI ...
+#  module load nvidia/nvhpc/22.11
 #
 ###############################################################################
 
-BUILD   = parallel
-MODEL   = -D_D3Q19_
+BUILD  = parallel
+MODEL  = -D_D3Q19_
 
 CC     = nvcc
-CFLAGS = -ccbin=icpc -DADDR_SOA -DNDEBUG -arch=sm_70 -x cu -dc -Xcompiler -fast
+CFLAGS = -g -DADDR_SOA -O2 -arch=sm_70 -x cu -dc
+
+# PTX assembler extra information:  -Xptxas -v
+# Alternative compiler, e.g., Intel: -ccbin=icpc -Xcompiler -fast
 
 AR = ar
 ARFLAGS = -cr
-LDFLAGS= -ccbin=icpc -arch=sm_70
+LDFLAGS = -arch=sm_70
 
-MPI_HOME     = /lustre/sw/intel/compilers_and_libraries_2017.2.174/linux/mpi
-MPI_INC_PATH = -I$(MPI_HOME)/include64
-MPI_LIB_PATH = -L$(MPI_HOME)/lib64 -lmpi
+# nvhpc (mpicc is present but drives nvc not nvcc) so use nvcc still ... but
+MPI_HOME     = ${NVHPC_ROOT}/comm_libs/mpi
+MPI_INC_PATH = -I$(MPI_HOME)/include
+MPI_LIB_PATH = -L$(MPI_HOME)/lib -lmpi
 
+# NVHPC bundled MPI must use mpirun supplied ...
 LAUNCH_SERIAL_CMD =
 LAUNCH_MPIRUN_CMD = mpirun
 MPIRUN_NTASK_FLAG = -np
-

config/unix-nvcc-default.mk

@@ -0,0 +1,25 @@
+##############################################################################
+#
+#  unix-nvcc-default.mk
+#
+#  This example uses "nvcc -ccbin=mpicc" to drive the compilation.
+#  Further include/library information may be required...
+#
+##############################################################################
+
+BUILD   = parallel
+MODEL   = -D_D3Q19_
+
+CC     = nvcc
+CFLAGS = -ccbin=mpicc -O2 -DADDR_SOA -arch=sm_61 -x cu -dc
+
+AR = ar
+ARFLAGS = -cr
+LDFLAGS= -arch=sm_61
+
+MPI_INC_PATH=-I/usr/lib/x86_64-linux-gnu/openmpi/include
+MPI_LIB_PATH=-L/usr/lib/x86_64-linux-gnu/openmpi/lib -lmpi
+
+LAUNCH_SERIAL_CMD =
+LAUNCH_MPIRUN_CMD = mpirun
+MPIRUN_NTASK_FLAG = -np

src/advection.c

@@ -22,7 +22,7 @@
  *  Edinburgh Soft Matter and Statistical Physics Group and
  *  Edinburgh Parallel Computing Centre
  *
- *  (c) 2010-2021  The University of Edinburgh
+ *  (c) 2010-2023  The University of Edinburgh
  *
  *  Contributing authors:
  *  Kevin Stratford ([email protected])
@@ -31,6 +31,7 @@
  *****************************************************************************/
 
 #include <assert.h>
+#include <limits.h>
 #include <stdlib.h>
 
 #include "advection_s.h"
@@ -171,6 +172,11 @@ __host__ int advflux_create(pe_t * pe, cs_t * cs, lees_edw_t * le, int nf,
   obj->nf = nf;
   obj->nsite = nsites;
 
+  if (nf < 1 || nsites < 1 || INT_MAX/nf < nsites) {
+    pe_info(pe,  "advflux_create: failure in int32_t indexing\n");
+    return -1;
+  }
+
   if (obj->le == NULL) {
     /* If no Lees Edwards, we only require an fx = fw */