Dear All, I am new to GetFEM++ and want to use it for GPU culculations. I guess that the easiest way to do this is to link GetFEM++ with CUBLAS library. Does anyone tried to do this? Is it possible in theory? Can I expect some speedup (what is the ratio (make model)/(linear solve) in time units for big 3D models)?
I have no NVidia card on hand, but there is two Tesla cards in our computing center, so I am interested in any information before requesting for computer time... Regards, Kosty _______________________________________________ Getfem-users mailing list [email protected] https://mail.gna.org/listinfo/getfem-users
