Tracy Harms wrote: > ... The array-oriented language option was ignored, but it's still possible to > consider how J fits into the situation being discussed. > > http://plgrand.blogspot.com/2009/01/summary-of-panel-discussion.html > > "... this stuff really IS the fabric of future society." They may ignore Array Programming Languages, but some of their comments are relevant to J and friends. One was the comment to the effect that Symmetric Multiprocessing doesn't scale well for large numbers of processors; another was about the limitations of shared memory. This is true for threaded solutions and to a lesser extend Software Transaction Memory.
The situation is different for array languages, which incur a lot of overhead for SMP when there is a small number of processors, but the relative overhead declines with more processors. Asymmetric mulitprocessing with some general purpose cores and some vector processing cores offer interesting possibilities for array languages. If the multicore trend continues, hardware and software will co-evolve and array languages could be part of the solution. Unfortunately, their limited popularity discourages hardware developers from optimizing the hardware for them. ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
