Closed sjsprecious closed 3 months ago
According to @mwaxmonsky's suggestion, it may be worth the efforts to try Cublas API for the AlphaMinusJacobian CUDA function (https://github.com/NCAR/micm/blob/main/include/micm/solver/cuda_rosenbrock.hpp#L67C30-L67C48) and see if it brings additional computational benefits compared to the CUDA implementation.
AlphaMinusJacobian
This should be done after we resolve the data transfer issues from https://github.com/NCAR/micm/issues/431.
It is not straightforward to implement an efficient version so drop it as not planned.
According to @mwaxmonsky's suggestion, it may be worth the efforts to try Cublas API for the
AlphaMinusJacobian
CUDA function (https://github.com/NCAR/micm/blob/main/include/micm/solver/cuda_rosenbrock.hpp#L67C30-L67C48) and see if it brings additional computational benefits compared to the CUDA implementation.This should be done after we resolve the data transfer issues from https://github.com/NCAR/micm/issues/431.