Surface Evolver Documentation

Back to top of Surface Evolver documentation.

Some miscellaneous topics


Command line options

The syntax for starting the Evolver from the system command prompt is:

evolver [-ffilename] [-a-] [-d] [-e] [-i] [-m] [-pn] [-q] [-w] [-x] [-y] [datafile]

The current directory and EVOLVERPATH will be searched for datafile. If the datafile is not found, then a new search with extension .fe is done. Wildcard matching is in effect on some systems (Windows, linux, maybe others), but be very careful when using wildcards since there can be unexpected matches. If the datafile still not found, or the datafile is not given on the command line, then the user will be prompted. Options:

-a-
Do not enable automatic conversion to named methods and quantities mode when a situation requiring it arises.
-d
Prints YACC debugging trace as datafile is parsed. May be helpful if you can't figure out why your datafile doesn't get read in properly AND you know YACC.
-ffilename
Specifies the name of a file to be used as command input after the datafile is read. At the end of this file, input reverts to standard input. The effect is the same as redirecting input from the file, except that -f will echo commands to the screen and revert to standard input at the end. Also note that errors will cause input to revert to standard input.
-e
Echo input. Meant for echoing commands of piped input to screen so the user can follow what is going on in case Evolver is being controlled by another process.
-i
Keeps elements numbers as listed in the datafile, instead of renumbering them consecutively. The same effect can be achieved by putting the keyword keep_originals in the top of the datafile.
-m
Turn memory debugging on at start of program. Same effect as Evolver memdebug command.
-pn
Forces use of n processes for an Evolver compiled in multi-processor mode. n may be larger or smaller than the physically available number of processors. The default is 1. This option should be regarded as experimental; there is still too much overhead for it to be useful usually.
-q
Convert everything to named quantities internally. There are a few things for which no quantities exist yet; they will produce error messages.
-w
Causes Evolver to exit whenever a warning occurs. Meant to be used when Evolver is run in a shell script.
-x
Causes Evolver to exit whenever an error occurs. Meant to be used when Evolver is run in a shell script, so that Evolver doesn't get hung up waiting for input in response to an error.
-y
Causes Evolver to cease execution of commands and return to command prompt after any warning message. Same effect as break_after_warning runtime toggle.


Bugs

Bug reports should be submitted by email to brakke@susqu.edu. Please include the Evolver version number, a description of the problem, the initial data file, and the sequence of commands necessary to reproduce the problem.

Evolver Newsletters

The group of Surface Evolver users has grown large enough that I have started a newsletter. Contents include announcement of new versions and latest features, bibliography, and anything else anybody would like to contribute. If you would like to be on the mailing list, send your email address to brakke@susqu.edu.

Back issues:


Error handling

When the Surface Evolver detects an error, it prints an error message and tries to take appropriate action. If the -x command line option was given when Evolver was started, then Evolver exits immediately with a nonzero error code. This is useful when running Evolver from shell scripts. There are several categories of errors:


Surface initialization

Whenever the Surface Evolver loads a new surface, either on startup or in response to the q or load commands, the following actions occur:


Interrupting execution

Evolver operation may be interrupted with the standard keyboard interrupt, CTRL-C usually (SIGINT for you unix gurus). During repeated operations, this will set a flag which is checked at the end of each loop. Repetition will cease after the current step and control will return to the main prompt. If you give a second interrupt before the loop has ended, Evolver will abort the command and return to the main prompt. Beware that this may leave the surface in an inconsistent state if surface topology changing operations were going on. An immediate abort will also happen if an interrupt is received outside a loop. If Evolver receives SIGTERM (say from the unix kill command), it will dump to the default dump file and exit. This is useful for stopping a background Evolver running a script. The same thing will happen with SIGHUP, so losing a modem connection will save the current surface.

Note: In Microsoft Windows, the second interrupt doesn't do anything much since Windows creates a separate thread to handle the interrupt, and I can't find any way to force the offending thread to stop and longjmp back to where it should. So if the Evolver is really, really stuck, you may just have to kill the whole program.


Parallel computation

Parallel versions of the Surface Evolver are available for several types of systems, but these are experimental and not very useful.

Silicon Graphics parallel version

Silicon Graphics systems with multiple processors may run a version that will use some or all processor by including -DSGI_MULTI in CFLAGS in Makefile. This version will run fine with one processor, also. Currently, the only calculations done in parallel are the "named quantities", so it is best to start Evolver with the -q option. The number n of processes actually done in parallel can be controlled with the -pn command line option. It will slow things down to use more processes than are physically available, but it is possible to run multiple processes on a single-processor system. SGI's use shared memory, so no message passing is necessary, and this version runs pretty well.

PVM parallel version

PVM (Parallel Virtual Machine) is a portable message-passing interface for distributed parallel processing. The PVM Evolver is not too sophisticated yet, passing all data to all processes. Hence it should be regarded as experimental rather than practical. Improvements are planned some day. The only practical use of it so far has been massive knot energy calculations, where every process does need to have all data. Interested users should contact brakke@susqu.edu.

MPI parallel version

MPI (Message Passing Interface) is a portable message-passing interface for distributed parallel processing. The MPI Evolver is not too sophisticated yet, passing all data to all processes. Hence it should be regarded as experimental rather than practical. Improvements are planned some day. The only practical use of it so far has been massive knot energy calculations, where every process does need to have all data. Interested users should contact brakke@susqu.edu.

Back to top of Surface Evolver documentation.
Author's home page.