Add a new signature of multiply_order_opt_all_ends to work on MatGeneric.

src/faust_linear_operator/CPU/faust_prod_opt.h

@@ -38,6 +38,9 @@ namespace Faust
 	template<typename FPP, FDevice DEVICE>
 		void multiply_order_opt_all_ends(std::vector<MatDense<FPP,DEVICE>*>& facts, MatDense<FPP,DEVICE>& out, FPP alpha=1.0, FPP beta_out=.0, std::vector<char> transconj_flags = std::vector<char>({'N'}));
 
+	template<typename FPP, FDevice DEVICE>
+		void multiply_order_opt_all_ends(std::vector<MatGeneric<FPP,DEVICE>*>& facts, MatDense<FPP,DEVICE>& out, FPP alpha=1.0, FPP beta_out=.0, std::vector<char> transconj_flags= {'N'});
+
 	/**
 	 *
 	 * This function does the same as multiply_order_opt_all_ends but is capable to multiply not only on the ends of the matrix chain but also in the middle if a better complexity guides to this.

src/faust_linear_operator/CPU/faust_prod_opt.hpp

@@ -44,6 +44,46 @@ void Faust::multiply_order_opt_all_ends(std::vector<Faust::MatDense<FPP,DEVICE>*
 	facts.erase(facts.begin(), facts.end());
 }
 
+template<typename FPP, FDevice DEVICE>
+void Faust::multiply_order_opt_all_ends(std::vector<Faust::MatGeneric<FPP,DEVICE>*>& facts, Faust::MatDense<FPP,DEVICE>& out, FPP alpha/* =1.0*/, FPP beta_out/*=.0*/, std::vector<char> transconj_flags /* = {'N'}*/)
+{
+	Faust::MatDense<FPP, DEVICE> tmpr, tmpl;
+	int nfacts = facts.size();
+	int ri = 0, li = nfacts-1;
+	Faust::MatGeneric<FPP,DEVICE> *R1, *R2, *L1, *L2;
+	faust_unsigned_int R1nr, R1nc, L1nr, L1nc;
+	while(li-ri > 1)
+	{
+		R1 = facts[ri];
+		R2 = facts[ri+1];
+		L1 = facts[li-1];
+		L2 = facts[li];
+		R1nr = R1->getNbRow();
+		R1nc = R1->getNbCol();
+		L1nr = L1->getNbRow();
+		L1nc = L1->getNbCol();
+		if(R1nr * R1nc * R2->getNbCol() < L1nr * L1nc * L2->getNbCol())
+		{
+			gemm_gen(*R1, *R2, tmpr, (FPP)1.0, (FPP)0.0, transconj_flags[transconj_flags.size()>ri?ri:0], transconj_flags[transconj_flags.size()>ri+1?ri+1:0]);
+			ri++;
+			facts[ri] = &tmpr;
+			if(transconj_flags.size() > ri)
+				transconj_flags[ri] = 'N';
+		}
+		else
+		{
+			gemm_gen(*L1, *L2, tmpl, (FPP)1.0, (FPP)0.0, transconj_flags[transconj_flags.size()>li-1?li-1:0], transconj_flags[transconj_flags.size()>li?li:0]);
+			li--;
+			facts[li] = &tmpl;
+			if(transconj_flags.size() > li)
+				transconj_flags[li] = 'N';
+		}
+	}
+	// last mul
+	gemm_gen(*facts[ri], *facts[li], out, alpha, beta_out, ri==0?transconj_flags[0]:'N', li==nfacts-1&&transconj_flags.size()>li?transconj_flags[li]:'N');
+	facts.erase(facts.begin(), facts.end());
+}
+
 template<typename FPP, FDevice DEVICE>
 void Faust::multiply_order_opt_all_best(std::vector<Faust::MatDense<FPP,DEVICE>*>& facts, Faust::MatDense<FPP,DEVICE>& out, FPP alpha/* =1.0*/, FPP beta_out/*=.0*/, std::vector<char> transconj_flags /* = {'N'}*/)
 {