Add GPU TransformHelper::power_iteration support.

Update to gpu_mod@04b59b9e.
Commit related to issue #192.

gpu_mod @ 04b59b9e

-Subproject commit 99364247691c1cb8bc469fb126812cff0462b476
+Subproject commit 04b59b9eec5ea2d28ed8d26c41a531e58f6b3ca7

src/faust_linear_operator/GPU2/faust_TransformHelper_gpu.hpp

@@ -200,8 +200,7 @@ namespace Faust
 	template<typename FPP>
 		FPP TransformHelper<FPP,GPU2>::power_iteration(const faust_unsigned_int nb_iter_max, const Real<FPP>& threshold, int& flag)
 		{
-			throw std::runtime_error("power_iteration not yet supported.");
-			//TODO: simple to add in gpu_mod
+			return this->transform->power_iteration(nb_iter_max, threshold, flag);
 		}
 
 	template<typename FPP>

src/faust_linear_operator/GPU2/faust_Transform_gpu.cpp.in

@@ -341,6 +341,17 @@ namespace Faust
 			return marr_funcs->spectral_norm(gpu_mat_arr, threshold, nb_iter_max);
 		}
 
+	template<>
+		@FAUST_SCALAR_FOR_GM@ Transform<@FAUST_SCALAR_FOR_GM@,GPU2>::power_iteration(int32_t nb_iter_max, float threshold, int& flag)
+		{
+			auto marr_funcs = GPUModHandler::get_singleton()->marr_funcs((@FAUST_SCALAR_FOR_GM@)(0));
+			if(gpu_mat_arr == nullptr) throw std::runtime_error("gpu_mat_arr is nullptr");
+			//TODO: update gpu_mod to handle output flag
+			@FAUST_SCALAR_FOR_GM@ lambda;
+			marr_funcs->power_iteration(gpu_mat_arr, threshold, nb_iter_max, reinterpret_cast<@GM_SCALAR@*>(&lambda));
+			return lambda;
+		}
+
 	template<>
 		void Faust::Transform<@FAUST_SCALAR_FOR_GM@,GPU2>::multiply(const Faust::Transform<@FAUST_SCALAR_FOR_GM@,GPU2> & A)
 		{

src/faust_linear_operator/GPU2/faust_Transform_gpu.h

@@ -69,6 +69,7 @@ namespace Faust
 			void multiply(const FPP& a, const int32_t id=-1);
 			Vect<FPP,GPU2> multiply(const Vect<FPP,GPU2>& x, const char opThis='N');
 			Real<FPP> spectralNorm(int32_t nb_iter_max, float threshold, int& flag);
+			FPP power_iteration(int32_t nb_iter_max, float threshold, int& flag);
 			Real<FPP> normL1(const bool transpose = false) const;
 			void tocpu(Transform<FPP, Cpu>& cpu_transf) const;
 			Transform<FPP, Cpu> tocpu() const;