On Jan 19, 2013, at 6:58 PM, ????? ?????????? <tenebrosus.scriptor at gmail.com> wrote:. > 2. There are copyright limitations on that Intel software so I am also > interested in alternative tools to avoid them.
I perfectly understand your point. I recently add the support to OpenBLAS (check README.GPU) because not everybody can use Intel compilers. QE-GPU builds automatically the library with multi-threading support. Easy to do, no need to modify make.sys! You may want to try it... >> In case of QE-GPU, having 4 MPI on a multi-core workstation means share the >> GPU among them and split by a factor of the number of the MPI processes the >> available RAM on the GPU board (in your case 4). I do not know what kind of >> GPU do you have but you would like to avoid this scenario. > I use GeForce GTX 460 with 1Gb of memory onboard and I definitely want to > avoid this scenario. I've just rebuilt QE. Yes, definitely. 1 GByte is a low amount of memory (2~3 GByte is quite good, high-end cards has up to 6 GByte but these cards are quite expensive). Your card has compute capability "20", CUDA 5.0 is the best choice (if you have a old version you might have to upgrade the NVIDIA driver too). > And, probably, I should leave the heaviest tasks for x86-64 system with more > memory... Remember that You can restart a run performed using QE-GPU with standard CPU-only code and vice-versa. Feel free to ask more if you need clarifications. Happy computing. Cheers, Filippo -- Filippo SPIGA ** Sent from my iPad, sorry if I am brief **
