Distributed power control is crucial for ensuring reliable communication and minimizing energy consumption in wireless communication networks, particularly under random switching topologies and uncertain network dynamics. Existing methods primarily rely on traditional norms or control paradigms, often facing computational challenges due to inherent nonconvexities. In this paper, we propose a novel distributed optimization framework for wireless networks using DC programming to minimize the L1/L∞-Hankel norm of the system. Unlike conventional approaches that may yield suboptimal solutions or suffer from computational inefficiency, the DC programming approach systematically decomposes the nonconvex Hankel norm minimization problem into tractable convex subproblems, ensuring rapid convergence and globally optimal or near-optimal solutions. Our method specifically addresses dynamic topologies modeled by random switching and parameter uncertainties, significantly improving robustness and interference management. Simulation results demonstrate that the proposed framework effectively maintains the desired SINR levels and accelerates convergence of global solution.

Modern mobile applications have grown rapidly in binary size, which restricts user growth and hinders updates for existing users. Thus, reducing the binary size is important for application developers. Recent studies have shown the possibility of using link-time code size optimizations by re-invoking certain compiler optimizations on the linked intermediate representation of the program. However, such methods often incur significant build time overhead and require intrusive changes to the existing build pipeline.

We design, implement and evaluate a parallel radix sort that simultaneously utilizes the CPU and GPU devices on the AMD Fusion APU. The parallel sort, referred to as Fusion Sort, partitions the sort keys between the CPU and GPU devices and utilizes the integrated memory system of the APU to avoid data copying between the devices. We identify three design issues that impact overhead and performance: the granularity of sharing between the two devices, the scheme of data partitioning and the allocation of data in memory regions accessible by each device. We present three variants of Fusion Sort that share data at coarse and fine granularities and with fixed and variable data partitioning schemes. In each variant, data is allocated to minimize the overhead of non-preferred memory accesses of each device. Our evaluation shows that fine-grain sharing with variable data partitioning performs the best. Further, Fusion Sort outperforms CPU-only and GPU-only parallel radix sorts by up to 1.8X and 1.9X respectively. These results demonstrate the viability of the integrated memory system of the APU in the context of sorting.