Tesla C1060 and C1070 supports double precision. The next architecture "Fermi" (product will come out next year) has more efficient support of double precision. If I recall correctly, CUDA released CUDO_blas, but I am not sure about lapack. However, for FFT, it is a big boost.
Huolin On Sat, Oct 17, 2009 at 4:14 AM, Paolo Giannozzi <giannozz at democritos.it> wrote: > > On Oct 17, 2009, at 7:54 , vega lew wrote: > >> But the most important thing is that the computer center only has GPU >> with CUDA for calculation instead of CPU with Fortran. Is there a >> possibility >> now to transplant the current Q-E code easily to the GPU computer with >> single precession? Do you think Double precession is so important >> for Q-E codes? > > you mean: single/double "precision" ? I think double precision is > important. > Anyway, there is some ongoing work on porting Q-E to GPUs. > > P. > --- > Paolo Giannozzi, Dept of Physics, University of Udine > via delle Scienze 208, 33100 Udine, Italy > Phone +39-0432-558216, fax +39-0432-558222 > > > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum >
