Processer Elves are an interim implementation of the Elves multiprocessing prodedure. They will eventually, be merged with the Elves main program, to make multi-processing as easy as falling off a log. Processer Elves live in ./Elfsheim with the rest of the Elves, but only support a few simple options. In order for Processer to run you must have indexed at least one of your wedges, preferable in a subdirectory called "./index", as is done by the Elves main program. You can examine the top of the script to see what all the options of Processer Elves are, but, to be brief:
will run Wedger in any wedge directories that the Elves main program set up. Run Scaler once everything has been processed, and then run shelx, solve, and Phaser in each possible space group.
unix% ./Elfsheim/Processer no solve
will do the Wedger and Scaler runs, but will not try to run heavy-atom finding programs (useful for multiwedge native data)
unix% ./Elfsheim/Processer solve only
will skip the Wedger runs, and go straight on to the heavy-atom finding. Scaler will only be run if the data files for heavy-atom finding havn't been created yet.
unix% ./Elfsheim/Processer hurry up
tells Processer to "hurry" and makes all the Elves use looser convergence criteria, and fewer steps to try and arrive at a rudimentary structure solution as quickly as possible. (useful when using Elves on-site at a synchrotron)
To use several CPUs at once, just run several copies of Processer Elves in the same directory:
unix% cd /where/ever/processing
unix% ./Elfsheim/Processer new >&! process1.log &
unix% ./Elfsheim/Processer >&! process2.log &
unix% ssh othermachine
othermachine% cd /nfs/where/ever/processing
othermachine% ./Elfsheim/Processer >&! process3.log &
the word "new" tells Processer to clear any signal files that might have been left by old, killed Processer runs, so make sure you only use it for the first launch of a multi-Processer run.
Back to the Elves Manual Table of Contents.