Update to gpu_mod@1.1.2 and update faust_gpu_mod consequently

[skip ci].

gpu_mod @ cf5f672d

-Subproject commit 530b868724952d3417ac3aa14aeebc95dcda9bb8
+Subproject commit cf5f672d2d9921d077b800458b03618403a7b96c

src/faust_linear_operator/CPU/faust_TransformHelper.hpp

@@ -304,7 +304,7 @@ namespace Faust {
 #ifdef USE_GPU_MOD
 				case 10:
 					if(gpu_faust != nullptr)
-						M = gpu_faust->multiply(&A);
+						M = gpu_faust->multiply(&A, is_transposed, is_conjugate);
 					break;
 #endif
 				default:

src/faust_linear_operator/CPU/faust_gpu_mod_gen.hpp.in

@@ -74,11 +74,22 @@ namespace Faust
 		{
 			const MatSparse<@FAUST_SCALAR_FOR_GM@, Cpu>* sp_mat;
 			const MatDense<@FAUST_SCALAR_FOR_GM@, Cpu>* ds_mat;
+			gm_DenseMat_t gpu_prod_mat_dense;
+			@GM_SCALAR@ one;
+			set_one<@GM_SCALAR@>(&one);
+			gm_Op op;
+			if(transpose && conjugate)
+				op = OP_CONJTRANSP;
+			else if(transpose)
+				op = OP_TRANSP;
+			else
+				op = OP_NOTRANSP;
 
 			gm_MatArrayFunc_@GM_SCALAR@* marr_funcs = (gm_MatArrayFunc_@GM_SCALAR@*) this->marr_funcs;
 			if(sp_mat = dynamic_cast<const MatSparse<@FAUST_SCALAR_FOR_GM@,Cpu>*>(A))
 			{
 
+				throw std::runtime_error("FaustGPU::multiply() by MatSparse isn't yet impl.");
 				//		std::cout << "FaustGPU::multiply(MatSparse): " << sp_mat->getNbRow() << " " << sp_mat->getNbCol()<< " " << sp_mat->getNonZeros()<< std::endl;
 				marr_funcs->togpu_spm(gpu_mat_arr, sp_mat->getNbRow(), sp_mat->getNbCol(), sp_mat->getNonZeros(), (int32_t *) sp_mat->getOuterIndexPtr(), (int32_t*) sp_mat->getInnerIndexPtr(), (@GM_SCALAR@*) reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(const_cast<@FAUST_SCALAR_FOR_GM@*>(sp_mat->getValuePtr())));
 			}
@@ -86,13 +97,17 @@ namespace Faust
 			{
 
 				//		std::cout << "FaustGPU::multiply(MatDense): " << ds_mat->getNbRow() << " " << ds_mat->getNbCol()<< " " << ds_mat->getNonZeros()<< std::endl;
-				marr_funcs->togpu_dsm(gpu_mat_arr, ds_mat->getNbRow(), ds_mat->getNbCol(), (@GM_SCALAR@*) reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(const_cast<@FAUST_SCALAR_FOR_GM@*>(ds_mat->getData())));
+				gpu_prod_mat_dense = marr_funcs->chain_matmul_by_cpu_dsm(gpu_mat_arr, one, op, (@GM_SCALAR@*) reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(const_cast<@FAUST_SCALAR_FOR_GM@*>(ds_mat->getData())), ds_mat->getNbRow(), ds_mat->getNbCol());
 			}
 
-			auto M = this->get_product();
-
-			marr_funcs->erase_at(gpu_mat_arr, size);
-			return M;
+			//TODO: add to GPU mod a chain_matmul overload to directly get cpu mat result from here (I mean an argument to data of pre-allocated DenseMat)
+			int32_t nrows, ncols;
+			gm_DenseMatFunc_@GM_SCALAR@* dsm_funcs = (gm_DenseMatFunc_@GM_SCALAR@*) this->dsm_funcs;
+			dsm_funcs->info(gpu_prod_mat_dense, &nrows, &ncols);
+			MatDense<@FAUST_SCALAR_FOR_GM@, Cpu> gpu2cpu_mat(nrows, ncols);
+			dsm_funcs->tocpu(gpu_prod_mat_dense, (@GM_SCALAR@*) reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(gpu2cpu_mat.getData()));
+			dsm_funcs->free(gpu_prod_mat_dense);
+			return gpu2cpu_mat;
 		}
 
 	template <>