Merge remote-tracking branch 'origin/master' into simgrid

# Conflicts:
#	CMakeLists.txt
#	Src/GroupTree/Core/FGroupTaskStarpuAlgorithm.hpp
#	Src/GroupTree/StarPUUtils/FStarPUTaskNameParams.hpp
#	Tests/GroupTree/testBlockedUniform.cpp

.clang-format

+Language: Cpp
+AccessModifierOffset: -4
+AlignAfterOpenBracket: true
+AlignConsecutiveAssignments: true
+AlignConsecutiveDeclarations: false
+AlignEscapedNewlinesLeft: false
+AlignOperands: false
+AlignTrailingComments: true
+AllowAllParametersOfDeclarationOnNextLine: true
+AllowShortBlocksOnASingleLine: false
+AllowShortCaseLabelsOnASingleLine: false
+AllowShortFunctionsOnASingleLine: false
+AllowShortIfStatementsOnASingleLine: false
+AllowShortLoopsOnASingleLine: false
+AlwaysBreakAfterDefinitionReturnType: None
+AlwaysBreakAfterReturnType: None
+AlwaysBreakBeforeMultilineStrings: false
+AlwaysBreakTemplateDeclarations: false
+BinPackArguments: true
+BinPackParameters: true
+BraceWrapping:
+  AfterClass:      false
+  AfterControlStatement: false
+  AfterEnum:       false
+  AfterFunction:   false
+  AfterNamespace:  false
+  AfterObjCDeclaration: false
+  AfterStruct:     false
+  AfterUnion:      false
+  BeforeCatch:     false
+  BeforeElse:      false
+  IndentBraces:    false
+BreakBeforeBinaryOperators: None
+BreakBeforeBraces: Attach
+BreakBeforeTernaryOperators: false
+BreakConstructorInitializersBeforeComma: false
+ColumnLimit: 0
+CommentPragmas: '^ IWYU pragma:'
+ConstructorInitializerAllOnOneLineOrOnePerLine: false
+ConstructorInitializerIndentWidth: 0
+ContinuationIndentWidth: 4
+Cpp11BracedListStyle: false
+DerivePointerAlignment: false
+DisableFormat: false
+ExperimentalAutoDetectBinPacking: false
+ForEachMacros: [ foreach, Q_FOREACH, BOOST_FOREACH ]
+IndentCaseLabels: true
+IndentWidth: 4
+IndentWrappedFunctionNames: false
+KeepEmptyLinesAtTheStartOfBlocks: true
+MacroBlockBegin: ''
+MacroBlockEnd: ''
+MaxEmptyLinesToKeep: 2
+NamespaceIndentation: None
+PenaltyBreakBeforeFirstCallParameter: 1400
+PenaltyBreakComment: 1400
+PenaltyBreakFirstLessLess: 140000
+PenaltyBreakString: 1400
+PenaltyExcessCharacter: 1
+PointerAlignment: Left
+ReflowComments: true
+SortIncludes: false
+SpaceAfterCStyleCast: false
+SpaceBeforeAssignmentOperators: true
+SpaceBeforeParens: Never
+SpaceInEmptyParentheses: false
+SpacesBeforeTrailingComments: 1
+SpacesInAngles: true
+SpacesInContainerLiterals: false
+SpacesInCStyleCastParentheses: false
+SpacesInParentheses: false
+SpacesInSquareBrackets: false
+Standard: Cpp11
+TabWidth: 4
+UseTab: Never
+DisableFormat: false

Addons/CKernelApi/CMakeLists.txt

@@ -22,7 +22,7 @@ if(SCALFMM_ADDON_CKERNELAPI)
         file( GLOB_RECURSE source_lib_files Src/*.cpp )
 
         # Adding cpp files to project
-        add_library( scalfmmckernelapi STATIC ${source_lib_files} )
+        add_library( scalfmmckernelapi ${source_lib_files} )
 
         # Add blas library (even if it is set to off)
         target_link_libraries( scalfmmckernelapi scalfmm)
@@ -35,7 +35,7 @@ if(SCALFMM_ADDON_CKERNELAPI)
     )
 
         # Install lib
-        install( TARGETS scalfmmckernelapi ARCHIVE DESTINATION lib )
+        install( TARGETS scalfmmckernelapi ARCHIVE DESTINATION lib LIBRARY DESTINATION lib)
 
         # Install header
         SET(my_include_dirs "Src")

Addons/CKernelApi/Src/CScalfmmApi.h

@@ -145,10 +145,10 @@ typedef void (*Callback_free_cell)(void*);
 /**
  * @brief Callback used to know the size of userData.
  * @param level current level of current cell
- * @param userData Datas that will be serialize
  * @param morton_index of the current cell
+ * @param userData : ptr to user's kernel
  */
-typedef FSize (*Callback_get_cell_size)(int level, void * userDatas, long long morton_index);
+typedef FSize (*Callback_get_cell_size)(int level, long long morton_index, void * userData);
 
 /**
  * @brief Callback used to serialize userdata inside an array of size
@@ -159,6 +159,39 @@ typedef FSize (*Callback_get_cell_size)(int level, void * userDatas, long long m
  */
 typedef void (*Callback_copy_cell)(void * userDatas, FSize size, void * memoryAllocated);
 
+/**
+ * @brief Callback used by scalfmm to know the size needed by the user
+ * to store its local datas linked to a leaf.
+ *
+ * @param nbParts : number of point inside current Leaf
+ * @return Size (in bytes) needed to serialize the user datas
+ */
+typedef FSize (*Callback_get_leaf_size)(FSize nbParts);
+
+/**
+ * @brief Callback used by scalfmm to serialize the leaf user data.
+ *
+ * @param Morton MortonIndex of current Leaf
+ * @param nbParts : number of point inside current Leaf
+ * @param userdata : leaf User data
+ * @param memAllocated : Ptr to an array of size =
+ * Callback_get_leaf_size(...), to be filled by the user.
+ */
+typedef void (*Callback_copy_leaf)(FSize nbParts, void * userDatas, void * memAllocated);
+
+
+/**
+ * @brief Callback used by scalfmm to retreive leaf data from an array
+ * (previously filled by the user through Callback_copy_leaf)
+ *
+ * @param Morton MortonIndex of current Leaf
+ * @param nbParts : number of point inside current Leaf
+ * @param memAllocated : Ptr to an array of size =
+ * Callback_get_leaf_size(...), to be read by the user in order to
+ * re-build its leaf.
+ * @return new userdata local to the leaf
+ */
+typedef void * (*Callback_restore_leaf)(FSize nbParts, void * memAllocated);
 
 /**
  * @brief Callback called if scalfmm_finalize_cell is called.
@@ -214,10 +247,18 @@ typedef struct User_Scalfmm_Cell_Descriptor{
     Callback_get_cell_size user_get_size;
     Callback_copy_cell user_copy_cell;
     Callback_restore_cell user_restore_cell;
-    Callback_init_leaf user_init_leaf;
-    Callback_free_leaf user_free_leaf;
 }Scalfmm_Cell_Descriptor;
 
+/**
+ * @brief This struct contains all the callback affecting leaves
+ */
+typedef struct User_Scalfmm_Leaf_Descriptor{
+    Callback_init_leaf user_init_leaf;
+    Callback_free_leaf user_free_leaf;
+    Callback_get_leaf_size user_get_size;
+    Callback_copy_leaf user_copy_leaf;
+    Callback_restore_leaf user_restore_leaf;
+}Scalfmm_Leaf_Descriptor;
 
 /**
  * @brief This function build the tree. If scalfmm_init has been
@@ -228,7 +269,10 @@ typedef struct User_Scalfmm_Cell_Descriptor{
  * @param BoxWidth Width of the entire simulation box.
  * @param BoxCenter Coordinate of the center of the box (ie array)
  */
-void scalfmm_build_tree(scalfmm_handle handle,int TreeHeight,double BoxWidth,double* BoxCenter,Scalfmm_Cell_Descriptor user_cell_descriptor);
+void scalfmm_build_tree(scalfmm_handle handle,int TreeHeight,double BoxWidth,
+                        double* BoxCenter,
+                        Scalfmm_Cell_Descriptor user_cell_descriptor,
+                        Scalfmm_Leaf_Descriptor user_leaf_descriptor);
 
 
 /**
@@ -488,6 +532,15 @@ typedef void (*Callback_P2M)(void* cellData, void * leafData, FSize nbParticles,
  */
 typedef void (*Callback_M2M)(int level, void* parentCell, int childPosition, void* childCell, void* userData);
 
+/**
+ * @brief Function to be filled by user's M2M
+ * @param level current level in the tree
+ * @prama parentCell cell to be filled
+ * @param userData datas specific to the user's kernel
+ * @param childCell array of cells to be read
+ */
+typedef void (*Callback_M2M_Full)(int level, void* parentCell, void* childCell[8], void* userData);
+
 /**
  * @brief Function to be filled by user's M2L
  * @param level current level in the tree
@@ -529,6 +582,15 @@ typedef void (*Callback_M2LFull)(int level, void* targetCell, const int * neighb
  */
 typedef void (*Callback_L2L)(int level, void* parentCell, int childPosition, void* childCell, void* userData);
 
+/**
+ * @brief Function to be filled by user's L2L
+ * @param level current level in the tree
+ * @prama parentCell cell to be filled
+ * @param userData datas specific to the user's kernel
+ * @param childCell array of cells to be read
+ */
+typedef void (*Callback_L2L_Full)(int level, void* parentCell, void* childCell[8], void* userData);
+
 /**
  * @brief Function to be filled by user's L2P
  * @param cellData cell to be read
@@ -599,6 +661,29 @@ typedef void (*Callback_P2PInner)(void * targetLeaf,FSize nbParticles, const FSi
 typedef void (*Callback_P2PSym)(void * targetLeaf, FSize nbParticles, const FSize* particleIndexes,
                                 void * sourceLeaf, FSize nbSourceParticles, const FSize* sourceParticleIndexes, void* userData);
 
+
+/**
+ * @brief Function to be filled by user's P2P Remote, the sources will
+ * be erased after the call, so no need to modify them
+ * @param targetLeaf ptr to user target leaf
+ * @param nbParticles number of particle in current leaf
+ * @param particleIndexes indexes of particles currently computed
+ * @praram sourceLeaves array of user source target
+ * @param sourceParticleIndexes array of indexes for each source
+ * @param sourceNbPart : array containing the number of part for each source
+ * @param size Number of direct neighbors
+ * @param userData datas specific to the user's kernel
+
+ * @attention This function will be called in distributed scalfmm
+ * only, between leaves belonging to diffenrent MPI processus. On ly
+ * the target leaf should be modified, since we don't send back the
+ * sources.
+ */
+typedef void (*Callback_P2PRemote)(void * targetLeaf,FSize nbParticles, const FSize* particleIndexes,
+                                   void ** sourceLeaves,
+                                   const FSize ** sourceParticleIndexes,FSize * sourceNbPart,
+                                   const int * sourcePosition, const int size, void* userData);
+
 /**
  * @brief Function to be filled by user's method to reset a user's cell
  * @param level  level of the cell.
@@ -618,15 +703,18 @@ typedef void (*Callback_apply_on_cell)(int level, long long morton_index, int* t
 typedef struct User_Scalfmm_Kernel_Descriptor {
     Callback_P2M p2m;
     Callback_M2M m2m;
+    Callback_M2M_Full m2m_full;
     Callback_M2L m2l;
     Callback_M2L_Ext m2l_ext;
     Callback_M2LFull m2l_full;
     Callback_L2L l2l;
+    Callback_L2L_Full l2l_full;
     Callback_L2P l2p;
     Callback_P2P p2p;
     Callback_P2PFull p2p_full;
     Callback_P2PInner p2pinner;
     Callback_P2PSym p2p_sym;
+    Callback_P2PRemote p2p_remote;
 }Scalfmm_Kernel_Descriptor;
 
 
@@ -662,6 +750,14 @@ void scalfmm_user_kernel_config(scalfmm_handle Handle, Scalfmm_Kernel_Descriptor
  */
 void scalfmm_execute_fmm(scalfmm_handle Handle);
 
+/**
+ * @brief This function launch the fmm on the parameters given, but
+ * only the far field will be evaluated
+ * @param Handle scalfmm_handle provided by scalfmm_init
+ */
+void scalfmm_execute_fmm_far_field(scalfmm_handle Handle);
+
+
 /**
  * @brief This function apply the call back on each leaf. Should be
  * called after insert_parts.
@@ -758,8 +854,7 @@ scalfmm_handle scalfmm_init_distributed( scalfmm_kernel_type KernelType,scalfmm_
  * once the partitionning done. Can be inserted with no changes.
  * @param indexesFilled Array that store the global index of each part
  * in the localArrayFilled.
- * @param stride stride between two attributes inside attr.
- * @param attr array of attribute to be distributed alongside the positions.
+ * @param outputNbPoint number of points dedicated to this proc
  */
 void scalfmm_create_local_partition(scalfmm_handle handle, int nbPoints, double * particleXYZ, double ** localArrayFilled,
                                     FSize ** indexesFilled, FSize * outputNbPoint);

Addons/CKernelApi/Src/FInterEngine.hpp

@@ -83,7 +83,8 @@ public:
         FScalFMMEngine<FReal>::Algorithm = algo;
     }
 
-    void build_tree(int TreeHeight, FReal BoxWidth , FReal * BoxCenter,User_Scalfmm_Cell_Descriptor notUsedHere){
+    virtual void build_tree(int TreeHeight, FReal BoxWidth , FReal * BoxCenter,
+                            User_Scalfmm_Cell_Descriptor notUsedHere, Scalfmm_Leaf_Descriptor notUsedHereToo){
         octree = new OctreeClass(TreeHeight,FMath::Min(3,TreeHeight-1),BoxWidth,FPoint<FReal>(BoxCenter));
         this->matrix = new MatrixKernelClass();
         this->kernel = new InterKernel(TreeHeight,BoxWidth,FPoint<FReal>(BoxCenter),matrix);
@@ -115,12 +116,12 @@ public:
         }else{
             if(type==SOURCE){
                 for(FSize idPart = 0; idPart<NbPositions ; ++idPart){
-                    octree->insert(FPoint<FReal>(&XYZ[3*idPart]),FParticleTypeSource,idPart);
+                    octree->insert(FPoint<FReal>(&XYZ[3*idPart]),FParticleType::FParticleTypeSource,idPart);
                 }
                 FScalFMMEngine<FReal>::nbPart += NbPositions;
             }else{
                 for(FSize idPart = 0; idPart<NbPositions ; ++idPart){
-                    octree->insert(FPoint<FReal>(&XYZ[3*idPart]),FParticleTypeTarget,idPart);
+                    octree->insert(FPoint<FReal>(&XYZ[3*idPart]),FParticleType::FParticleTypeTarget,idPart);
                 }
                 FScalFMMEngine<FReal>::nbPart += NbPositions;
             }
@@ -138,12 +139,12 @@ public:
         }else{
             if(type==SOURCE){
                 for(FSize idPart = 0; idPart<NbPositions ; ++idPart){
-                    octree->insert(FPoint<FReal>(X[idPart],Y[idPart],Z[idPart]),FParticleTypeSource,idPart);
+                    octree->insert(FPoint<FReal>(X[idPart],Y[idPart],Z[idPart]),FParticleType::FParticleTypeSource,idPart);
                 }
                 FScalFMMEngine<FReal>::nbPart += NbPositions;
             }else{
                 for(FSize idPart = 0; idPart<NbPositions ; ++idPart){
-                    octree->insert(FPoint<FReal>(X[idPart],Y[idPart],Z[idPart]),FParticleTypeTarget,idPart);
+                    octree->insert(FPoint<FReal>(X[idPart],Y[idPart],Z[idPart]),FParticleType::FParticleTypeTarget,idPart);
                 }
                 FScalFMMEngine<FReal>::nbPart += NbPositions;
             }
@@ -738,6 +739,47 @@ public:
     //     }
     // }
 
+    void execute_fmm_far_field(){
+        switch(FScalFMMEngine<FReal>::Algorithm){
+        case 0:
+            {
+                typedef FFmmAlgorithm<OctreeClass,InterCell,ContainerClass,InterKernel,LeafClass> AlgoClassSeq;
+                AlgoClassSeq* algoSeq = new AlgoClassSeq(octree,kernel);
+                algoSeq->execute(FFmmP2M | FFmmM2M | FFmmM2L | FFmmL2L | FFmmL2P);
+                FScalFMMEngine<FReal>::algoTimer = algoSeq;
+                FScalFMMEngine<FReal>::abstrct = algoSeq;
+                break;
+            }
+        case 1:
+            {
+                typedef FFmmAlgorithmThread<OctreeClass,InterCell,ContainerClass,InterKernel,LeafClass> AlgoClassThread;
+                AlgoClassThread* algoThread = new AlgoClassThread(octree,kernel);
+                algoThread->execute(FFmmP2M | FFmmM2M | FFmmM2L | FFmmL2L | FFmmL2P);
+                FScalFMMEngine<FReal>::algoTimer = algoThread;
+                FScalFMMEngine<FReal>::abstrct = algoThread;
+                break;
+            }
+        case 2:
+            {
+                typedef FFmmAlgorithmPeriodic<FReal,OctreeClass,InterCell,ContainerClass,InterKernel,LeafClass> AlgoClassPeriodic;
+                AlgoClassPeriodic algoPeriod(octree,2);
+                algoPeriod.setKernel(kernel);
+                algoPeriod.execute(FFmmP2M | FFmmM2M | FFmmM2L | FFmmL2L | FFmmL2P);
+                break;
+            }
+        case 3:
+            {
+                typedef FFmmAlgorithmThreadTsm<OctreeClass,InterCell,ContainerClass,InterKernel,LeafClass> AlgoClassTargetSource;
+                AlgoClassTargetSource* algoTS = new AlgoClassTargetSource(octree,kernel);
+                algoTS->execute(FFmmP2M | FFmmM2M | FFmmM2L | FFmmL2L | FFmmL2P);
+                FScalFMMEngine<FReal>::algoTimer = algoTS;
+                FScalFMMEngine<FReal>::abstrct = algoTS;
+                break;
+            }
+        default :
+            std::cout<< "No algorithm found (probably for strange reasons) : "<< FScalFMMEngine<FReal>::Algorithm <<" exiting" << std::endl;
+        }
+    }
 
     void execute_fmm(){
         switch(FScalFMMEngine<FReal>::Algorithm){

Addons/CKernelApi/Src/FScalFMMEngine.hpp

@@ -95,7 +95,9 @@ public:
     //by specific Engine
 
     //Function about the tree
-    virtual void build_tree(int TreeHeight,FReal BoxWidth,FReal* BoxCenter,Scalfmm_Cell_Descriptor user_cell_descriptor){
+    virtual void build_tree(int TreeHeight,FReal BoxWidth,FReal* BoxCenter,
+                            Scalfmm_Cell_Descriptor user_cell_descriptor,
+                            Scalfmm_Leaf_Descriptor user_leaf_descriptor){
         FAssertLF(0,"Nothing has been done yet, exiting");
     }
 
@@ -178,37 +180,6 @@ public:
     }
 
 
-    /** Test ... */
-    struct RunContainer{
-        template< int nbAttributeToInsert,class ContainerClass,class LeafClass, class CellClass>
-        static void Run(FOctree<FReal,CellClass,ContainerClass,LeafClass> * octree,
-                        int NbPartToInsert,int * strideForEachAtt,
-                        FReal* rawDatas){
-            generic_tree_abstract_insert<ContainerClass,LeafClass,CellClass,nbAttributeToInsert>(octree,
-                                                                                                 NbPartToInsert,strideForEachAtt,rawDatas);
-        }
-    };
-
-    template<class ContainerClass,class LeafClass, class CellClass, int nbAttributeToInsert>
-    void generic_tree_abstract_insert(FOctree<FReal,CellClass,ContainerClass,LeafClass> * octree,
-                                      int NbPartToInsert,int * strideForEachAtt,
-                                      FReal* rawDatas){
-        for(FSize idxPart = 0; idxPart<NbPartToInsert ; ++idxPart){
-            FPoint<FReal> pos = FPoint<FReal>(rawDatas[0],rawDatas[1],rawDatas[2]);
-            MortonIndex index = octree->getMortonFromPosition(pos);
-            //Insert with how many attributes ???
-            octree->insert(pos,idxPart);
-            //Get again the container
-            ContainerClass * containerToFill = octree->getLeafSrc(index);//cannot be nullptr
-            std::array<FReal,nbAttributeToInsert> arrayOfAttribute;
-            for(int idxAtt = 0; idxAtt<nbAttributeToInsert ; ++idxAtt){
-                arrayOfAttribute[idxAtt] = rawDatas[3+ strideForEachAtt[idxAtt]];
-            }
-            int idxToRemove = containerToFill->getNbParticles();
-            containerToFill->remove(&idxToRemove,1);
-            containerToFill->push(pos,idxPart,arrayOfAttribute);
-        }
-    }
 
     template<class ContainerClass,class LeafClass,class CellClass>
     void generic_get_forces_xyz(FOctree<FReal,CellClass,ContainerClass,LeafClass> * octree,
@@ -806,6 +777,9 @@ public:
     virtual void execute_fmm(){
         FAssertLF(0,"No kernel set, cannot execute anything, exiting ...\n");
     }
+    virtual void execute_fmm_far_field(){
+        FAssertLF(0,"No kernel set, cannot execute anything, exiting ...\n");
+    }
 
     virtual void intern_dealloc_handle(Callback_free_cell userDeallocator){
         FAssertLF(0,"No kernel set, cannot execute anything, exiting ...\n");
@@ -820,15 +794,17 @@ public:
      * get the time spent in each operator.
      */
     virtual void get_timers(FReal * Timers){
-        const FTic * timers = algoTimer->getAllTimers();
-        int nbTimers = algoTimer->getNbOfTimerRecorded();
-        for(int idTimer = 0; idTimer<nbTimers ; ++idTimer){
-            Timers[idTimer] = timers[idTimer].elapsed();
-        }
+        Timers[0] = algoTimer->Timers["P2M"].elapsed();
+        Timers[1] = algoTimer->Timers["M2M"].elapsed();
+        Timers[2] = algoTimer->Timers["M2L"].elapsed();
+        Timers[3] = algoTimer->Timers["L2L"].elapsed();
+        <