faq:cuda_support
This is an old revision of the document!
Which parts of CP2K are CUDA-accelerated?
- FFTs, when compiled with
-D__PW_CUDA
. - Anything that uses
dbcsr_multiply
, i.e. sparse matrix multiplication, when compiled with-D__DBCSR_CUDA
. This benfits in particular the linear scaling DFT code. See also the DBCSR project. - If linked against an accelerated scalapack/blas library (in particular pdgemm/dgemm) that executes these calls on the GPU. The impact of this is most visible for MP2 and RPA calculations. On the hybrid Cray XC30 linking against libsci_acc makes this happen.
faq/cuda_support.1396684573.txt.gz · Last modified: 2020/08/21 10:15 (external edit)