Comment by krastanov | Hacker News Reader

Top New Best Ask Show Jobs

Comment by krastanov | Hacker News Reader

Parent Full thread

CUDA is great for big dense matrices, which is incredibly important. But there is a lot more to scientific computing than big dense matrices.

Also, CUDA is low-level a library by one vendor for one class of hardware. Not a language. Languages serious about scientific computing already have multiple interfaces for CUDA at various level of abstraction.

There is a Fortran version of CUDA: https://developer.nvidia.com/cuda-fortran

There is cuSparse for sparse matrices:

https://docs.nvidia.com/cuda/cusparse/

What are some example workloads in scientific computing that are not suitable for CUDA?

solving small and medium sized differential equations, most monte-carlo algorithms

Anything that's not linear algebra

One of the reasons why CUDA won over OpenCL is because it is polyglot.

Reply on news.ycombinator.com