Efficient implementations of parallel applications on hetero-geneous hybrid architectures require a careful balance between compu-tations and communications with accelerator devices. Even if most of the communication time can be overlapped by computations, it is es-sential to reduce the total volume of communicated data. The litera-ture therefore abounds with ad hoc methods to reach that balance, but these are architecture and application dependent. We propose here a generic mechanism to automatically optimize the scheduling between CPUs and GPUs, and compare two strategies within this mechanism: the classical Heterogeneous Earliest Finish Time (HEFT) algorithm and our new, parametrized, Distributed Affinity Dual Approximation algo-rithm (DADA), which consists in grouping the tasks by affinity before running a fast dual approximation. We ran experiments on a heteroge-neous parallel machine with twelve CPU cores and eight NVIDIA Fermi GPUs. Three standard dense linear algebra kernels from the PLASMA library have been ported on top of the XKaapi runtime system. We re-port their performances. It results that HEFT and DADA perform well for various experimental conditions, but that DADA performs better for larger systems and number of GPUs, and, in most cases, generates much lower data transfers than HEFT to achieve the same performance.