I have been tasked with the 开发者_运维技巧maintenance of a set of old FORTRAN scientific codes, mostly written using MPI parallelization, to be run in thousand+ node clusters. New features are always difficult to retrofit, and my boss has allowed me to go for partial rewrites of specific components. I'd like to give Erlang a go, but my experience is limited and I was wondering if it would be difficult to call ScaLAPACK routines from Erlang. Any advice would be appreciated.
It could be done by integrating the routines as NIFs (or using a port driver), but you'd need to get down to some really low level details, which is probably not what you want for a first Erlang project.
Maybe the Disco Project could be something for you. It uses Erlang for distribution, but lets you script the individual jobs in Python, which already has good support for calling scientific code.
精彩评论