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, when compiled with -D__DBCSR_CUDA. This includes in particular the linear scaling DFT code.
Anything that uses dgemm (not pdgem), when linked against Cray's accelerated LibSci. Currently, this is most useful for the MP2 method.