We know parallel versions for basic algorithms currently being used in most kernels
-
There are some counter examples: 1) Parallelize incomplete LU factorization for irregular sparse matrices 2) Develop new neural network to predict and hence avoid hurricanes or avoid computer failures.... 3) Optimal preconditioning for 3D partial differential equation is not known in many applications --- to be expected as little experience.
-
Can't use a computer of performance 10P without substantial experience on a computer with performance P
|
