On Wed, May 25, 2016 at 4:34 PM, Henry Rich <[email protected]> wrote: > We would like a benchmark that uses scalar code heavily. This is the kind > of application that you might say is 'not good for J'. We'd like to make J > better on those applications. If you have some old FORTRAN-style code > written in J, we'd like to see it.
A quick look through rosettacode finds some plausible possibilities. Note, however, that these are not necessarily "purely scalar". For example, when indexing a matrix you need to use a boxed pair. This is something that would be "scalar" in fortran, but goes a different route in J. Or maybe we're using a rank operator to represent an inner loop. For example: http://rosettacode.org/wiki/Solve_a_Holy_Knight's_tour#J So, anyways, I do not actually know how many of these are actually in the ballpark of what you need. Some of these might also need to use different data to really exercise the machine. But if you could pick out a "top 3" for me, and maybe identify any obvious data deficiencies, I might be able to see if I can find some more things that seem to fit that pattern: http://rosettacode.org/wiki/Cipolla's_algorithm#J http://rosettacode.org/wiki/Dijkstra's_algorithm#J http://rosettacode.org/wiki/Knight's_tour#J http://rosettacode.org/wiki/Levenshtein_distance#J http://rosettacode.org/wiki/Maze_solving#J http://rosettacode.org/wiki/Self-referential_sequence#J http://rosettacode.org/wiki/Superpermutation_minimisation#J http://rosettacode.org/wiki/Tonelli-Shanks_algorithm#J I hope this helps, -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
