Solid Container

.github/workflows/other_tests.yml

@@ -0,0 +1,46 @@
+name: TESTS
+
+on: [ push, pull_request ]
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        test: 
+          - solid
+    steps:
+    - name: Git checkout
+      uses: actions/checkout@v3
+      with:
+        submodules: true
+    - name: Install dependencies
+      uses: ./.github/actions/install
+      with:
+        essentials: true
+        r: false
+        rdep: false
+        cuda: false
+        hip: false
+        openmpi: false
+        lcov: true
+    - name: Compile
+      shell: bash
+      run: |
+        cd tests/${{ matrix.test }}
+        make ADD_FLAGS="-fprofile-arcs -ftest-coverage"
+    - name: Run
+      shell: bash
+      run: |
+        cd tests/${{ matrix.test }}
+        make run
+    - name: Gather coverage data
+      uses: ./.github/actions/coverage
+      id: coverage
+    - name: Send coverage data
+      uses: codecov/codecov-action@v3
+      if: steps.coverage.outputs.reports != ''
+      with:
+        files: ${{ steps.coverage.outputs.reports }}
+        flags: ${{ matrix.model }}

.gitmodules

@@ -1,3 +1,3 @@
 [submodule "TCLB_tests"]
-	path = tests
+	path = tests/external
 	url =  ../../CFD-GO/TCLB_tests.git

models/flow/d3q27_cumulant_part/Dynamics.c.Rt

@@ -21,7 +21,7 @@
 ?>
 
 CudaDeviceFunction void CalcF(){
-	for (SyncParticleIterator p(X,Y,Z); p; ++p) {
+	for (auto p : SyncParticleIterator(X,Y,Z)) {
 	if ((NodeType & NODE_BOUNDARY) == 0) {
 	real_t d = getRho();
         vector_t u;

models/particles/d3q27_PSM/Dynamics.c.Rt

@@ -450,7 +450,7 @@ CudaDeviceFunction void CalcF() {
 	C( uP, 0)
 	C( sol, 0)
 ?>
-	for (SyncParticleIterator p(X,Y,Z); p; ++p) {
+	for (auto p : SyncParticleIterator(X,Y,Z)) {
 
         real_t dist = sqrt(p.diff.x*p.diff.x + p.diff.y*p.diff.y + p.diff.z*p.diff.z);
 

src/Consts.h.Rt

@@ -22,6 +22,9 @@
   #define X_BLOCK 32
 #endif
 
+#define PART_MAR <?%f PartMargin ?>
+#define PART_MAR_BOX <?%f max(0,PartMargin-0.5) ?>
+
 <?R
   big_hex = function(x,bits=16) {
     ret = ""

src/Handlers/acRemoteForceInterface.cpp

@@ -62,12 +62,12 @@ int acRemoteForceInterface::ConnectRemoteForceInterface(std::string integrator_)
           double py = solver->lattice->py;
           double pz = solver->lattice->pz;
           solver->lattice->RFI.DeclareSimpleBox(
-            px + reg.dx,
-            px + reg.dx + reg.nx,
-            py + reg.dy,
-            py + reg.dy + reg.ny,
-            pz + reg.dz,
-            pz + reg.dz + reg.nz);
+            px + reg.dx - PART_MAR_BOX,
+            px + reg.dx + reg.nx + PART_MAR_BOX,
+            py + reg.dy - PART_MAR_BOX,
+            py + reg.dy + reg.ny + PART_MAR_BOX,
+            pz + reg.dz - PART_MAR_BOX,
+            pz + reg.dz + reg.nz + PART_MAR_BOX);
         }
         MPI_Barrier(MPMD.local);
         solver->lattice->RFI.Connect(MPMD.work,inter.work);

src/Handlers/cbRunR.cpp.Rt

@@ -486,6 +486,7 @@ class rInfo: public rWrapper {
 			return ret;
 		}
 		ERROR("R: Not implemented!");
+		return ret;
 	}
 
 	Rcpp::CharacterVector Names() {

src/Lattice.cu.Rt

@@ -16,7 +16,8 @@
 #include "Lattice.h"
 #include <mpi.h>
 #include <assert.h>
-#include "BallTree.hpp"
+#include "SolidTree.hpp"
+#include "SolidGrid.hpp"
 
 #ifdef ENABLE_NVPROF
 	#include <nvToolsExt.h>
@@ -147,11 +148,10 @@ Lattice::Lattice(lbRegion _region, MPIInfo mpi_, int ns):region(_region), mpi(mp
 <?R } ?>
         ZoneIter = 0;
 	particle_data_size_max = 0;
+	SC.InitFinder(container->solidfinder);
 	container->particle_data = NULL;
 	container->particle_data_size = 0;
-	container->balltree_data = NULL;
-	balltree_data_size_max = 0;
-	BT.balls = &RFI;
+	SC.balls = &RFI;
 	RFI.name = "TCLB";
 }
 
@@ -451,13 +451,7 @@ void Lattice::<?%s FunName ?>(int tab0, int tab1, int iter_type)
 	for (stage in rows(action_stages)) {
 ?>
 	DEBUG_PROF_PUSH("<?%s stage$name ?>");
-//---------------- STAGE: <?%s stage$name ?> ----------------------- <?R
-                 if (old_stage_level > 0) { 
-                	for (m in NonEmptyMargin) { ?>
-	container->in.<?%s m$name ?> = Snaps[tab1].<?%s m$name ?>; <?R
-                        } 
-                 }
-                 old_stage_level = old_stage_level + 1 ?>
+//---------------- STAGE: <?%s stage$name ?> -----------------------
 
 <?R if (stage$fixedPoint) { ?> for (int fix=0; fix<100; fix++) { <?R } ?>
 <?R if (stage$first_particle) { ?>
@@ -486,18 +480,19 @@ void Lattice::<?%s FunName ?>(int tab0, int tab1, int iter_type)
 		CudaMemcpyAsync(container->particle_data, RFI.Particles(), RFI.mem_size(), CudaMemcpyHostToDevice, kernelStream);
 	}
 	DEBUG_PROF_PUSH("Tree Build");
-		BT.Build();
+		SC.Build();
 	DEBUG_PROF_POP();
-		if (BT.size() > balltree_data_size_max) {
-			if (container->balltree_data != NULL) CudaFree(container->balltree_data);
-			balltree_data_size_max = BT.size();
-			CudaMalloc(&container->balltree_data, BT.mem_size());
-		}
-		container->balltree_data_size = BT.size();
-		if (BT.mem_size() > 0) {
-			CudaMemcpyAsync(container->balltree_data, BT.Tree(), BT.mem_size(), CudaMemcpyHostToDevice, kernelStream);
-		}
+		SC.CopyToGPU(container->solidfinder, kernelStream);
 <?R } ?>
+<?R if (old_stage_level > 0) { ?>
+	MPIStream_B();
+	CudaDeviceSynchronize();
+<?R   for (m in NonEmptyMargin) { ?>
+	container->in.<?%s m$name ?> = Snaps[tab1].<?%s m$name ?>; <?R
+      } 
+    }
+    old_stage_level = old_stage_level + 1
+?>
 	container->CopyToConst();
 	DEBUG_PROF_PUSH("Calculation");
 	switch(iter_type & ITER_INTEG){
@@ -527,7 +522,6 @@ void Lattice::<?%s FunName ?>(int tab0, int tab1, int iter_type)
                 break;
 	}	
 
-	MPIStream_B();
 	DEBUG_PROF_POP();
 <?R if (stage$last_particle) { ?>
 	if (RFI.mem_size() > 0) {
@@ -549,10 +543,12 @@ void Lattice::<?%s FunName ?>(int tab0, int tab1, int iter_type)
 	DEBUG_PROF_POP();
 	RFI.SendForces();
 <?R } ?>
-        CudaDeviceSynchronize();
+
 <?R if (stage$fixedPoint) { ?> } // for(fix) <?R } ?>
 	DEBUG_PROF_POP();
 <?R } ?>
+	MPIStream_B();
+	CudaDeviceSynchronize();
 	Snap = tab1;
 	MarkIteration();
 	updateAllSamples();	

src/Lattice.h.Rt

@@ -10,9 +10,8 @@
 #include "ZoneSettings.h"
 #include "SyntheticTurbulence.h"
 #include "Sampler.h"
-#include "RemoteForceInterface.h"
-#include "BallTree.h"
-#include "pinned_allocator.hpp"
+#include "SolidContainer.h"
+
 
 class lbRegion;
 class LatticeContainer;
@@ -76,12 +75,9 @@ public:
   lbRegion region; ///< Local lattice region
   real_t px, py, pz; 
   MPIInfo mpi; ///< MPI information
-  typedef rfi::RemoteForceInterface< rfi::ForceCalculator, rfi::RotParticle, rfi::ArrayOfStructures, real_t, pinned_allocator<real_t> > rfi_t;
   rfi_t RFI;
-  typedef BallTree< rfi_t > balltree_t;
-  balltree_t BT;
+  solidcontainer_t SC;
   size_t particle_data_size_max;
-  size_t balltree_data_size_max;
   char snapFileName[STRING_LEN];
   Lattice (lbRegion region, MPIInfo, int);
   ~Lattice ();

src/LatticeAccess.inc.cpp.Rt

@@ -122,7 +122,7 @@ template <class N> CudaDeviceFunction void LatticeContainer::getType(N & node)
    } else if (access == "add") { ?>
         <?%s m ?>[<?R C(off,float=FALSE) ?>] += <?%s d ?>; <?R
    } else if (access == "atomicadd") { ?>
-        atomicAddP(&( <?%s m ?>[<?R C(off,float=FALSE) ?>] ),<?%s d ?>); <?R
+        CudaAtomicAdd(&( <?%s m ?>[<?R C(off,float=FALSE) ?>] ),<?%s d ?>); <?R
    } else stop("Unknown access type in field.access.one");
   }
 

src/LatticeContainer.h.Rt

@@ -6,7 +6,7 @@
   File defining LatticeContainer and some additional CUDA functions
 */
 #include "SyntheticTurbulence.h"
-#include "BallTree.h"
+#include "SolidContainer.h"
 
 #ifndef SETTINGS_H
 <?R
@@ -47,8 +47,7 @@ class LatticeContainer {
   cut_t* Q; 
   size_t particle_data_size;
   real_t* particle_data;
-  size_t balltree_data_size;
-  tr_elem* balltree_data;
+  solidcontainer_t::finder_t solidfinder;
   real_t * Globals; ///< Pointer to the GPU table to store the calculated values of Globals
   int dx, dy, dz; ///< Offset of the region to calculate in the interior kernel run
   int nx, ny, nz; ///< Size of the Lattice region
@@ -140,10 +139,8 @@ class LatticeContainer {
   void Color( uchar4 *optr );
   template<class N> inline void RunBorderT(CudaStream_t);
   template<class N> inline void RunInteriorT(CudaStream_t);
-  template<class N> inline void RunParticlesT(CudaStream_t);
   template < eOperationType I, eCalculateGlobals G, eStage S > void RunBorder(CudaStream_t);
   template < eOperationType I, eCalculateGlobals G, eStage S > void RunInterior(CudaStream_t);
-  template < eOperationType I, eCalculateGlobals G, eStage S > void RunParticles(CudaStream_t);
 
   void CopyToConst();
   void WaitAll();