Dear Yue-Wen, On 27/11/2012 13:38, Yue-Wen Fang wrote: > Does it supply us a prominent acceleration in pw.x, ph.x and so on? As Filippo has recently reported ph.x is still not supported. It might take advantage of the linear algebra libraries for GPU we have implemented for pw.x but we have still no record of people that have experienced on it. PW version for GPU can be considered a stable version. "Prominent acceleration" is a relative concept. It's definitely worth using it where NVIDIA GPUs (with a decent peak of double-precision) are available.
For further question/discussion related to GPU please refer to q-e-gpgpu at qe-forge.org. Regards, Ivan
