Merge branch 'develop' into 'develop'

authorize external threads to traverse a map + thread safe SphericalHarmonics

See merge request !72

CGoGN/include/Topology/generic/dart.h

@@ -34,7 +34,7 @@ namespace CGoGN
 const unsigned int EMBNULL = 0xffffffff;
 const unsigned int MRNULL = 0xffffffff;
 
-const unsigned int NB_THREAD = 16;
+const unsigned int NB_THREADS = 16;
 
 // DO NOT MODIFY (ORBIT_IN_PARENT function in Map classes)
 

CGoGN/include/Topology/generic/genericmap.h

@@ -104,24 +104,42 @@ protected:
 	// protected copy constructor to prevent the copy of map
 	GenericMap(const GenericMap& ) {}
 
+	/**
+	 * @brief m_thread_ids
+	 * vector of known thread ids, i.e. threads for which a mark vector,
+	 * a Dart buffer or a uint buffer will be given when asked
+	 */
+	mutable std::vector<std::thread::id> m_thread_ids;
 
 	/**
-	 *
+	 * @brief m_authorizeExternalThreads
+	 * if true, getCurrentThreadIndex will give an index to an unknown thread
 	 */
-	std::vector<std::thread::id> m_thread_ids;
+	bool m_authorizeExternalThreads;
+
 public:
 	/// compute thread index in the table of thread
 	inline unsigned int getCurrentThreadIndex() const;
 
-	/// add place for n new threads in the table of thread return index of first
-	inline unsigned int addEmptyThreadIds(unsigned int n);
+//	/// register the given thread ID for access to resources on this map
+//	inline void addThreadId(const std::thread::id id);
+
+	/// unregister the given thread to access resources on this map
+	inline void removeThreadId(const std::thread::id id);
 
-	/// remove  the n last added threads from table
-	inline void popThreadIds(unsigned int nb);
+	/// add room for a new thread ID (will be set on thread start)
+	inline std::thread::id& addEmptyThreadId();
 
-	/// get ref to jth threadId for updating (in thread)
-	inline std::thread::id& getThreadId(unsigned int j);
+	/// get a threadId based on its index
+	inline std::thread::id getThreadId(unsigned int index) const;
 
+	/**
+	 * @brief setExternalThreadsAuthorization
+	 * @param b
+	 * if true, threads that did not created the map will be able to traverse it
+	 * if false, traversal of the map by other threads will fail
+	 */
+	inline void setExternalThreadsAuthorization(bool b);
 
 protected:
 	/**
@@ -132,14 +150,14 @@ protected:
 	static std::map<std::string, RegisteredBaseAttribute*>* m_attributes_registry_map;
 
 	/// buffer for less memory allocation
-	static  std::vector< std::vector<Dart>* >* s_vdartsBuffers;
-	static  std::vector< std::vector<unsigned int>* >* s_vintsBuffers;
+	mutable std::vector< std::vector<Dart>* > s_vdartsBuffers[NB_THREADS];
+	mutable std::vector< std::vector<unsigned int>* > s_vintsBuffers[NB_THREADS];
 
 public:
 	/// table of instancied maps for Dart/CellMarker release
 	static std::vector<GenericMap*>* s_instances;
-protected:
 
+protected:
 	/**
 	 * Direct access to the Dart attributes that store the orbits embeddings
 	 * (only initialized when necessary, i.e. addEmbedding function)
@@ -154,7 +172,7 @@ protected:
 	AttributeMultiVector<NoTypeNameAttribute<std::vector<Dart> > >* m_quickLocalIncidentTraversal[NB_ORBITS][NB_ORBITS] ;
 	AttributeMultiVector<NoTypeNameAttribute<std::vector<Dart> > >* m_quickLocalAdjacentTraversal[NB_ORBITS][NB_ORBITS] ;
 
-	std::vector< AttributeMultiVector<MarkerBool>* > m_markVectors_free[NB_ORBITS][NB_THREAD] ;
+	std::vector< AttributeMultiVector<MarkerBool>* > m_markVectors_free[NB_ORBITS][NB_THREADS] ;
 	std::mutex m_MarkerStorageMutex[NB_ORBITS];
 
 	unsigned int m_nextMarkerId;

CGoGN/include/Topology/generic/genericmap.hpp

@@ -26,43 +26,79 @@
 
 namespace CGoGN
 {
+
 /****************************************
  *         THREAD ID MANAGEMENT         *
  ****************************************/
+
 inline unsigned int GenericMap::getCurrentThreadIndex() const
 {
 	std::thread::id id = std::this_thread::get_id();
-	unsigned int i=0;
-	while (id != m_thread_ids[i])
+	for (unsigned int i = 0; i < m_thread_ids.size(); ++i)
+	{
+		if (id == m_thread_ids[i])
+			return i;
+	}
+	assert(m_thread_ids.size() < NB_THREADS + 1);
+	if (m_authorizeExternalThreads)
+	{
+		m_thread_ids.push_back(id);
+		return m_thread_ids.size() - 1;
+	}
+	else
 	{
-		i++;
-		assert(i<m_thread_ids.size());
+		return -1;
 	}
-	return i;
 }
 
-//inline void GenericMap::addThreadId(const std::thread::id& id)
+//inline void GenericMap::addThreadId(const std::thread::id id)
 //{
-//	m_thread_ids.push_back(id);
+//	for (unsigned int i = 0; i < m_thread_ids.size(); ++i)
+//	{
+//		if (m_thread_ids[i] == id)
+//			return;
+//	}
+//	assert(m_thread_ids.size() < NB_THREADS + 1);
+//	if (m_authorizeExternalThreads)
+//		m_thread_ids.push_back(id);
 //}
 
-inline unsigned int GenericMap::addEmptyThreadIds(unsigned int n)
+inline void GenericMap::removeThreadId(const std::thread::id id)
+{
+	for (unsigned int i = 0; i < m_thread_ids.size(); ++i)
+	{
+		if (m_thread_ids[i] == id)
+		{
+			m_thread_ids[i] = m_thread_ids.back();
+			m_thread_ids.pop_back();
+			break;
+		}
+	}
+}
+
+inline std::thread::id& GenericMap::addEmptyThreadId()
 {
-	unsigned int nb = uint32(m_thread_ids.size());
-	m_thread_ids.resize(nb + n);
-	return nb;
+	assert(m_thread_ids.size() < NB_THREADS + 1);
+	unsigned int size = uint32(m_thread_ids.size());
+	m_thread_ids.resize(size + 1);
+	return m_thread_ids.back();
 }
 
-inline void GenericMap::popThreadIds(unsigned int nb)
+inline std::thread::id GenericMap::getThreadId(unsigned int index) const
 {
-	assert(nb<m_thread_ids.size());
-	for ( unsigned int i=0; i<nb; ++i)
-		m_thread_ids.pop_back();
+	assert(index < m_thread_ids.size());
+	return m_thread_ids[index];
 }
 
-inline std::thread::id& GenericMap::getThreadId(unsigned int j)
+inline void GenericMap::setExternalThreadsAuthorization(bool b)
 {
-	return m_thread_ids[j];
+	m_authorizeExternalThreads = b;
+	if (!m_authorizeExternalThreads)
+	{
+		// keep only the thread that created the map
+		while (m_thread_ids.size() > 1)
+			m_thread_ids.pop_back();
+	}
 }
 
 
@@ -90,7 +126,7 @@ inline void GenericMap::releaseDartBuffer(std::vector<Dart>* vd) const
 {
 	unsigned int thread = getCurrentThreadIndex();
 
-	if (vd->capacity()>1024)
+	if (vd->capacity() > 1024)
 	{
 		std::vector<Dart> v;
 		vd->swap(v);
@@ -100,7 +136,6 @@ inline void GenericMap::releaseDartBuffer(std::vector<Dart>* vd) const
 	s_vdartsBuffers[thread].push_back(vd);
 }
 
-
 inline std::vector<unsigned int>* GenericMap::askUIntBuffer() const
 {
 	unsigned int thread = getCurrentThreadIndex();
@@ -121,7 +156,7 @@ inline void GenericMap::releaseUIntBuffer(std::vector<unsigned int>* vui) const
 {
 	unsigned int thread = getCurrentThreadIndex();
 
-	if (vui->capacity()>1024)
+	if (vui->capacity() > 1024)
 	{
 		std::vector<unsigned int> v;
 		vui->swap(v);
@@ -132,7 +167,6 @@ inline void GenericMap::releaseUIntBuffer(std::vector<unsigned int>* vui) const
 }
 
 
-
 /****************************************
  *           DARTS MANAGEMENT           *
  ****************************************/
@@ -280,27 +314,25 @@ AttributeMultiVector<MarkerBool>* GenericMap::askMarkVector()
 	{
 		std::lock_guard<std::mutex> lockMV(m_MarkerStorageMutex[ORBIT]);
 
-		unsigned int x=m_nextMarkerId++;
+		unsigned int x = m_nextMarkerId++;
 		std::string number("___");
-		number[2]= '0'+x%10;
+		number[2] = '0'+x%10;
 		x = x/10;
-		number[1]= '0'+x%10;
+		number[1] = '0'+x%10;
 		x = x/10;
-		number[0]= '0'+x%10;
+		number[0] = '0'+x%10;
 
 		AttributeMultiVector<MarkerBool>* amv = m_attribs[ORBIT].addMarkerAttribute("marker_" + orbitName(ORBIT) + number);
 		return amv;
 	}
 }
 
-
 template <unsigned int ORBIT>
 inline void GenericMap::releaseMarkVector(AttributeMultiVector<MarkerBool>* amv)
 {
 	assert(isOrbitEmbedded<ORBIT>() || !"Invalid parameter: orbit not embedded") ;
 
 	unsigned int thread = getCurrentThreadIndex();
-
 	m_markVectors_free[ORBIT][thread].push_back(amv);
 }
 

CGoGN/include/Topology/generic/traversor/traversorCell.hpp

@@ -660,7 +660,8 @@ protected:
 	bool& m_finished;
 	unsigned int m_id;
 	FUNC m_lambda;
-	std::thread::id& m_threadId;	// ref on thread::id in table of threads in genericMap for init at operator()
+	std::thread::id& m_threadId; // ref on thread::id in table of threads in genericMap for init at operator()
+
 public:
 	ThreadFunction(FUNC func, std::vector<CELL>& vd, Utils::Barrier& s1, Utils::Barrier& s2, bool& finished, unsigned int id, std::thread::id& threadId) :
 		m_cells(vd), m_sync1(s1), m_sync2(s2), m_finished(finished), m_id(id), m_lambda(func), m_threadId(threadId)
@@ -670,6 +671,8 @@ public:
 	ThreadFunction(const ThreadFunction<ORBIT, FUNC>& tf):
 		m_cells(tf.m_cells), m_sync1(tf.m_sync1), m_sync2(tf.m_sync2), m_finished(tf.m_finished), m_id(tf.m_id), m_lambda(tf.m_lambda){}
 
+	std::thread::id& getThreadId() { return m_threadId; }
+
 	void operator()()
 	{
 		// first thing to do set the thread id in genericMap
@@ -714,10 +717,12 @@ void foreach_cell_tmpl(MAP& map, FUNC func, unsigned int nbth)
 	ThreadFunction<ORBIT,FUNC>** tfs = new ThreadFunction<ORBIT,FUNC>*[nbth];
 
 	// add place for nbth new threads in the table of threadId in genericmap
-	unsigned int firstThread = map.addEmptyThreadIds(nbth);
+//	unsigned int firstThread = map.addEmptyThreadIds(nbth);
+
 	for (unsigned int i = 0; i < nbth; ++i)
 	{
-		tfs[i] = new ThreadFunction<ORBIT,FUNC>(func, vd[i],sync1,sync2, finished,1+i,map.getThreadId(firstThread+i));
+		std::thread::id& threadId = map.addEmptyThreadId();
+		tfs[i] = new ThreadFunction<ORBIT,FUNC>(func, vd[i], sync1, sync2, finished, 1+i, threadId);
 		threads[i] = new std::thread( std::ref( *(tfs[i]) ) );
 	}
 
@@ -744,21 +749,19 @@ void foreach_cell_tmpl(MAP& map, FUNC func, unsigned int nbth)
 		sync2.wait();// everybody refilled then go
 	}
 
-	sync1.wait();// wait for all thread to finish its vector
-	finished = true;// say finsih to everyone
+	sync1.wait(); // wait for all thread to finish its vector
+	finished = true; // say finsih to everyone
 	sync2.wait(); // just wait for last barrier wait !
 
-
-	//wait for all theads to be finished
+	// wait for all theads to be finished
 	for (unsigned int i = 0; i < nbth; ++i)
 	{
 		threads[i]->join();
 		delete threads[i];
+		map.removeThreadId(tfs[i]->getThreadId());
 		delete tfs[i];
 	}
 
-	map.popThreadIds(nbth);
-
 	delete[] tfs;
 	delete[] threads;
 	delete[] vd;

CGoGN/include/Topology/generic/traversor/traversorDoO.h

@@ -39,7 +39,7 @@ private:
 	std::vector<Dart>* m_vd ;
 	const GenericMap* m_map;
 
-	TraversorDartsOfOrbit( const TraversorDartsOfOrbit<MAP,ORBIT>& /*tr*/){}
+	TraversorDartsOfOrbit(const TraversorDartsOfOrbit<MAP,ORBIT>& /*tr*/){}
 
 public:
 	TraversorDartsOfOrbit(const MAP& map, Cell<ORBIT> c) ;
@@ -66,7 +66,7 @@ private:
 	std::vector<Dart>* m_vd ;
 	const GenericMap* m_map;
 
-	VTraversorDartsOfOrbit( const VTraversorDartsOfOrbit<MAP,ORBIT>& /*tr*/){}
+	VTraversorDartsOfOrbit(const VTraversorDartsOfOrbit<MAP,ORBIT>& /*tr*/){}
 
 public:
 	VTraversorDartsOfOrbit(const MAP& map, Cell<ORBIT> c) ;

CGoGN/include/Utils/sphericalHarmonics.h

@@ -31,7 +31,10 @@ private :
 	static int nb_coefs ;                                    // number of coefs = (resolution + 1) * (resolution + 1)
 	static unsigned long cpt_instances;                      // number of instances of the class
 	static Tscalar K_tab [(max_resolution+1)*(max_resolution+1)];   // table containing constants K
-	static Tscalar F_tab [(max_resolution+1)*(max_resolution+1)];   // table for computing the functions : P or Y or y
+
+//	static Tscalar F_tab [(max_resolution+1)*(max_resolution+1)];   // table for computing the functions : P or Y or y
+
+	static Tscalar F_tab [32][(max_resolution+1)*(max_resolution+1)];
 
 	Tcoef* coefs;                                            // table of coefficients
 
@@ -48,12 +51,12 @@ public :
 
 	// evaluation
 
-	static void set_eval_direction (Tscalar theta, Tscalar phi) ;       // fix the direction in which the SH has to be evaluated
-	static void set_eval_direction (Tscalar x, Tscalar y, Tscalar z) ;  // fix the direction in which the SH has to be evaluated
-	Tcoef evaluate () const;				  						    // evaluates at a fixed direction
+	static void set_eval_direction (Tscalar theta, Tscalar phi, unsigned int threadId = 0) ;       // fix the direction in which the SH has to be evaluated
+	static void set_eval_direction (Tscalar x, Tscalar y, Tscalar z, unsigned int threadId = 0) ;  // fix the direction in which the SH has to be evaluated
+	Tcoef evaluate (unsigned int threadId = 0) const;				  						    // evaluates at a fixed direction
 
-	Tcoef evaluate_at (Tscalar theta, Tscalar phi) const;               // eval spherical coordinates
-	Tcoef evaluate_at (Tscalar x, Tscalar y, Tscalar z) const;          // eval cartesian coordinates
+	Tcoef evaluate_at (Tscalar theta, Tscalar phi, unsigned int threadId = 0) const;               // eval spherical coordinates
+	Tcoef evaluate_at (Tscalar x, Tscalar y, Tscalar z, unsigned int threadId = 0) const;          // eval cartesian coordinates
 
 	// I/O
 	const Tcoef& get_coef (int l, int m) const {assert ((l>=0 && l <=resolution) || !" maybe you forgot to call set_level()"); assert (m >= (-l) && m <= l); return get_coef(index(l,m));}
@@ -77,17 +80,17 @@ public :
 
 	// fitting
 	template <typename Tdirection, typename Tchannel>
-	void fit_to_data(int n, Tdirection* t_theta, Tdirection* t_phi, Tchannel* t_R, Tchannel* t_G, Tchannel* t_B, double lambda);
+	void fit_to_data(int n, Tdirection* t_theta, Tdirection* t_phi, Tchannel* t_R, Tchannel* t_G, Tchannel* t_B, double lambda, unsigned int threadId = 0);
 	template <typename Tdirection, typename Tchannel>
-	void fit_to_data(int n, Tdirection* t_x, Tdirection* t_y, Tdirection* t_z, Tchannel* t_R, Tchannel* t_G, Tchannel* t_B, double lambda);
+	void fit_to_data(int n, Tdirection* t_x, Tdirection* t_y, Tdirection* t_z, Tchannel* t_R, Tchannel* t_G, Tchannel* t_B, double lambda, unsigned int threadId = 0);
 
 private :
-	static int index (int l, int m) {return l*(l+1)+m;}
+	static inline int index (int l, int m) { return l*(l+1)+m; }
 
 	// evaluation :
 	static void init_K_tab (); // compute the normalization constants K_l^m and store them into K_tab
-	static void compute_P_tab (Tscalar t); // Compute Legendre Polynomials at parameter t and store them in F_tab (only for m>=0)
-	static void compute_y_tab (Tscalar phi); // Compute the real basis functions y_l^m at (theta, phi) and store them in F_tab (compute_P_tab must have been called before)
+	static void compute_P_tab (Tscalar t, unsigned int threadId); // Compute Legendre Polynomials at parameter t and store them in F_tab (only for m>=0)
+	static void compute_y_tab (Tscalar phi, unsigned int threadId); // Compute the real basis functions y_l^m at (theta, phi) and store them in F_tab (compute_P_tab must have been called before)
 
 	const Tcoef& get_coef (int i) const {assert ((i>=0 && i<nb_coefs ) || !" maybe you forgot to call set_level()"); return coefs[i];}
 	Tcoef& get_coef (int i) {assert ((i>=0 && i<nb_coefs ) || !" maybe you forgot to call set_level()"); return coefs[i];}

CGoGN/include/Utils/sphericalHarmonics.hpp

@@ -16,7 +16,7 @@ template <typename Tscalar,typename Tcoef> int SphericalHarmonics<Tscalar,Tcoef>
 template <typename Tscalar,typename Tcoef> unsigned long SphericalHarmonics<Tscalar,Tcoef>::cpt_instances = 0;
 
 template <typename Tscalar,typename Tcoef> Tscalar SphericalHarmonics<Tscalar,Tcoef>::K_tab[(max_resolution+1)*(max_resolution+1)];
-template <typename Tscalar,typename Tcoef> Tscalar SphericalHarmonics<Tscalar,Tcoef>::F_tab[(max_resolution+1)*(max_resolution+1)];
+template <typename Tscalar,typename Tcoef> Tscalar SphericalHarmonics<Tscalar,Tcoef>::F_tab[32][(max_resolution+1)*(max_resolution+1)];
 
 
 /*************************************************************************
@@ -77,47 +77,52 @@ evaluation
 **************************************************************************/
 
 template <typename Tscalar,typename Tcoef>
-void SphericalHarmonics<Tscalar,Tcoef>::set_eval_direction (Tscalar theta, Tscalar phi)
+void SphericalHarmonics<Tscalar,Tcoef>::set_eval_direction (Tscalar theta, Tscalar phi, unsigned int threadId)
 {
-	compute_P_tab(std::cos(theta));
-	compute_y_tab(phi);
+	assert(threadId < 32);
+	compute_P_tab(std::cos(theta), threadId);
+	compute_y_tab(phi, threadId);
 }
 
 template <typename Tscalar,typename Tcoef>
-void SphericalHarmonics<Tscalar,Tcoef>::set_eval_direction (Tscalar x, Tscalar y, Tscalar z)
+void SphericalHarmonics<Tscalar,Tcoef>::set_eval_direction (Tscalar x, Tscalar y, Tscalar z, unsigned int threadId)
 {
-	compute_P_tab(z);
+	assert(threadId < 32);
+	compute_P_tab(z, threadId);
 
 	Tscalar phi (0);
 	if ((x*x + y*y) > 0.0)
 		phi = atan2(y, x);
 
-	compute_y_tab(phi);
+	compute_y_tab(phi, threadId);
 }
 
 template <typename Tscalar,typename Tcoef>
-Tcoef SphericalHarmonics<Tscalar,Tcoef>::evaluate () const
+Tcoef SphericalHarmonics<Tscalar,Tcoef>::evaluate (unsigned int threadId) const
 {
+	assert(threadId < 32);
 	Tcoef r (0); // (0.0,0.0,0.0); //  TODO : use Tcoef (0)
 	for (int i = 0; i < nb_coefs; i++)
 	{
-		r += coefs[i] * F_tab[i];
+		r += coefs[i] * F_tab[threadId][i];
 	}
 	return r;
 }
 
 template <typename Tscalar,typename Tcoef>
-Tcoef SphericalHarmonics<Tscalar,Tcoef>::evaluate_at (Tscalar theta, Tscalar phi) const
+Tcoef SphericalHarmonics<Tscalar,Tcoef>::evaluate_at (Tscalar theta, Tscalar phi, unsigned int threadId) const
 {
-	set_eval_direction(theta, phi);
-	return evaluate();
+	assert(threadId < 32);
+	set_eval_direction(theta, phi, threadId);
+	return evaluate(threadId);
 }
 
 template <typename Tscalar,typename Tcoef>
-Tcoef SphericalHarmonics<Tscalar,Tcoef>::evaluate_at (Tscalar x, Tscalar y, Tscalar z) const
+Tcoef SphericalHarmonics<Tscalar,Tcoef>::evaluate_at (Tscalar x, Tscalar y, Tscalar z, unsigned int threadId) const
 {
-	set_eval_direction(x, y, z);
-	return evaluate();
+	assert(threadId < 32);
+	set_eval_direction(x, y, z, threadId);
+	return evaluate(threadId);
 }
 
 template <typename Tscalar,typename Tcoef>
@@ -150,28 +155,28 @@ void SphericalHarmonics<Tscalar,Tcoef>::copy_K_tab (Tscalar tab[])
 */
 
 template <typename Tscalar,typename Tcoef>
-void SphericalHarmonics<Tscalar,Tcoef>::compute_P_tab (Tscalar t)
+void SphericalHarmonics<Tscalar,Tcoef>::compute_P_tab (Tscalar t, unsigned int threadId)
 {
 //	if (t<0) {t=-t*t;} else {t=t*t;} // for plotting only : expand the param near equator
 
-	F_tab[index(0,0)] = 1;
+	F_tab[threadId][index(0,0)] = 1;
 	for (int l = 1; l <= resolution; l++)
 	{
-		F_tab[index(l,l)] = (1-2*l) * sqrt(1-t*t) * F_tab[index(l-1,l-1)];  // first diago
-		F_tab[index(l,l-1)] = t * (2*l-1) * F_tab[index(l-1,l-1)];// second diago
+		F_tab[threadId][index(l,l)] = (1-2*l) * sqrt(1-t*t) * F_tab[threadId][index(l-1,l-1)];  // first diago
+		F_tab[threadId][index(l,l-1)] = t * (2*l-1) * F_tab[threadId][index(l-1,l-1)];// second diago
 		for (int m = 0; m <= l-2; m++)
 		{// remaining of the line under the 2 diago
-			F_tab[index(l,m)] = t * (2*l-1) / (float) (l-m) * F_tab[index(l-1,m)] - (l+m-1) / (float) (l-m) * F_tab[index(l-2,m)];
+			F_tab[threadId][index(l,m)] = t * (2*l-1) / (float) (l-m) * F_tab[threadId][index(l-1,m)] - (l+m-1) / (float) (l-m) * F_tab[threadId][index(l-2,m)];
 		}
 	}
 }
 
 template <typename Tscalar,typename Tcoef>
-void SphericalHarmonics<Tscalar,Tcoef>::compute_y_tab (Tscalar phi)
+void SphericalHarmonics<Tscalar,Tcoef>::compute_y_tab (Tscalar phi, unsigned int threadId)
 {
 	for (int l = 0; l <= resolution; l++)
 	{
-		F_tab[index(l,0)] *= K_tab[index(l,0)]; // remove for plotting
+		F_tab[threadId][index(l,0)] *= K_tab[index(l,0)]; // remove for plotting
 	}
 
 	for (int m = 1; m <= resolution; m++)
@@ -181,10 +186,10 @@ void SphericalHarmonics<Tscalar,Tcoef>::compute_y_tab (Tscalar phi)
 
 		for (int l = m; l <= resolution; l++)
 		{
-			F_tab[index(l,m)] *= std::sqrt(2.0); // remove for plotting
-			F_tab[index(l,m)] *= K_tab[index(l,m)]; // remove for plotting
-			F_tab[index(l,-m)] = F_tab[index(l,m)] * sin_m_phi ; // store the values for -m<0 in the upper triangle
-			F_tab[index(l,m)] *= cos_m_phi;
+			F_tab[threadId][index(l,m)] *= std::sqrt(2.0); // remove for plotting
+			F_tab[threadId][index(l,m)] *= K_tab[index(l,m)]; // remove for plotting
+			F_tab[threadId][index(l,-m)] = F_tab[threadId][index(l,m)] * sin_m_phi ; // store the values for -m<0 in the upper triangle
+			F_tab[threadId][index(l,m)] *= cos_m_phi;
 		}
 	}
 
@@ -287,16 +292,17 @@ void SphericalHarmonics<Tscalar,Tcoef>::fit_to_data(
 	int n,
 	Tdirection* t_theta, Tdirection* t_phi,
 	Tchannel* t_R, Tchannel* t_G, Tchannel* t_B,
-	double lambda)
+	double lambda,
+	unsigned int threadId)
 {
 	Eigen::MatrixXd mM (nb_coefs,n); // matrix with basis function values, evaluated for all directions
 	// compute mM
 	for (int p = 0; p < n; ++p)
 	{
-		set_eval_direction(t_theta[p], t_phi[p]);
+		set_eval_direction(t_theta[p], t_phi[p], threadId);
 		for (int i = 0; i < nb_coefs; ++i)
 		{
-			mM(i,p) = F_tab[i];
+			mM(i,p) = F_tab[threadId][i];
 		}
 	}
 	fit_to_data(n, mM, t_R, t_G, t_B, lambda);
@@ -308,16 +314,17 @@ void SphericalHarmonics<Tscalar,Tcoef>::fit_to_data(
 	int n,
 	Tdirection* t_x, Tdirection* t_y, Tdirection* t_z,
 	Tchannel* t_R, Tchannel* t_G, Tchannel* t_B,
-	double lambda)
+	double lambda,
+	unsigned int threadId)
 {
 	Eigen::MatrixXd mM (nb_coefs,n); // matrix with basis function values, evaluated for all directions
 	// compute mM
 	for (int p=0; p<n; ++p)
 	{
-		set_eval_direction(t_x[p],t_y[p],t_z[p]);
+		set_eval_direction(t_x[p],t_y[p],t_z[p], threadId);
 		for (int i=0; i<nb_coefs; ++i)
 		{
-			mM(i,p) = F_tab[i];
+			mM(i,p) = F_tab[threadId][i];
 		}
 	}
 	fit_to_data(n, mM, t_R, t_G, t_B, lambda);

CGoGN/src/Topology/generic/genericmap.cpp

@@ -42,20 +42,16 @@ namespace Parallel
 CGoGN_TOPO_API int NumberOfThreads = getSystemNumberOfCores();
 }
 
-
 std::map<std::string, RegisteredBaseAttribute*>* GenericMap::m_attributes_registry_map = NULL;
 
-std::vector< std::vector<Dart>* >* GenericMap::s_vdartsBuffers = NULL;
-
-std::vector< std::vector<unsigned int>* >* GenericMap::s_vintsBuffers = NULL;
-
 //std::mutex* GenericMap::s_DartBufferMutex = NULL;
 //std::mutex* GenericMap::s_IntBufferMutex = NULL;
 
-std::vector<GenericMap*>*  GenericMap::s_instances=NULL;
+std::vector<GenericMap*>*  GenericMap::s_instances = NULL;
 
 GenericMap::GenericMap():
 	m_nextMarkerId(0),
+	m_authorizeExternalThreads(false),
 	m_manipulator(NULL)
 {
 	if(m_attributes_registry_map == NULL)
@@ -63,8 +59,8 @@ GenericMap::GenericMap():
 
 	s_instances->push_back(this);
 
-	m_thread_ids.reserve(NB_THREAD+1);
-	m_thread_ids.push_back( std::this_thread::get_id() );
+	m_thread_ids.reserve(NB_THREADS + 1);
+	m_thread_ids.push_back(std::this_thread::get_id());
 
 	for(unsigned int i = 0; i < NB_ORBITS; ++i)
 	{
@@ -72,6 +68,15 @@ GenericMap::GenericMap():
 		m_attribs[i].setRegistry(m_attributes_registry_map) ;
 	}
 
+	for(unsigned int i = 0; i < NB_THREADS; ++i)
+	{
+		s_vdartsBuffers[i].reserve(8);
+		s_vintsBuffers[i].reserve(8);
+
+//		for(unsigned int j = 0; j < NB_ORBITS; ++j)
+//			m_markVectors_free[i][j].reserve(8);
+	}
+
 	init();
 }
 
@@ -79,8 +84,6 @@ void GenericMap::copyAllStatics(const StaticPointers& sp)
 {
 	m_attributes_registry_map = sp.att_registry;
 	s_instances = sp.instances;
-	s_vdartsBuffers = sp.vdartsBuffers;
-	s_vintsBuffers = sp.vintsBuffers;
 }
 
 void GenericMap::initAllStatics(StaticPointers* sp)
@@ -123,33 +126,18 @@ void GenericMap::initAllStatics(StaticPointers* sp)
 		registerAttribute<MarkerBool>("MarkerBool");
 	}
 
-	if (s_instances==NULL)
-		s_instances= new std::vector<GenericMap*>;
+	if (s_instances == NULL)
+		s_instances = new std::vector<GenericMap*>;
 
-	if (s_vdartsBuffers == NULL)
-	{
-		s_vdartsBuffers = new std::vector< std::vector<Dart>* >[NB_THREAD];
-		s_vintsBuffers = new std::vector< std::vector<unsigned int>* >[NB_THREAD];
-
-		for(unsigned int i = 0; i < NB_THREAD; ++i)
-		{
-			s_vdartsBuffers[i].reserve(8);
-			s_vintsBuffers[i].reserve(8);
-				// prealloc ?
-		}
-	}
 	if (sp != NULL)
 	{
 		sp->att_registry = m_attributes_registry_map;
 		sp->instances = s_instances;
-		sp->vdartsBuffers = s_vdartsBuffers;
-		sp->vintsBuffers = s_vintsBuffers;
 	}
 }
 
 GenericMap::~GenericMap()
 {
-
 	for(unsigned int i = 0; i < NB_ORBITS; ++i)
 	{
 		if(isOrbitEmbedded(i))
@@ -175,7 +163,7 @@ GenericMap::~GenericMap()
 //		delete m_attributes_registry_map;
 //		m_attributes_registry_map = NULL;
 
-//		for(unsigned int i = 0; i < NB_THREAD; ++i)
+//		for(unsigned int i = 0; i < NB_THREADS; ++i)
 //		{
 //			for (auto it =s_vdartsBuffers[i].begin(); it != s_vdartsBuffers[i].end(); ++it)
 //				delete *it;
@@ -214,7 +202,6 @@ bool GenericMap::releaseManipulate(MapManipulator* ptr)
 	return true;
 }
 
-
 void GenericMap::init(bool addBoundaryMarkers)
 {
 	for(unsigned int i = 0; i < NB_ORBITS; ++i)
@@ -229,19 +216,17 @@ void GenericMap::init(bool addBoundaryMarkers)
 			m_quickLocalAdjacentTraversal[i][j] = NULL ;
 		}
 
-		for(unsigned int j = 0; j < NB_THREAD; ++j)
+		for(unsigned int j = 0; j < NB_THREADS; ++j)
 			m_markVectors_free[i][j].clear();
 	}
 
-
 	if (addBoundaryMarkers)
 	{
 		m_boundaryMarkers[0] = m_attribs[DART].addMarkerAttribute("BoundaryMark0") ;
 		m_boundaryMarkers[1] = m_attribs[DART].addMarkerAttribute("BoundaryMark1") ;
 	}