The main program here is `looptest". There are some minor variants
of this in the `others' subdirectory, but I don't see why they'd be
useful to anyone except me...

This directory contains parameter files for testing FTAP's performance;
you should probably be somewhat familiar with FTAP (like being able to interpret
the output files) before doing this. The main configuration for doing this 
is connecting the MIDI output port back to the MIDI input port with a cable; 
then FTAP output messages will be immediately interpreted as keystroke messages,
and FTAP can be tested under a very high data throughput situation. This will 
show up potential problems with your MIDI card and/or driver; if those are 
not a problem, and you have a reasonably fast machine, and you run FTAP 
with root privileges, these benchmarks should verify to you that FTAP 
performs with millisecond precision up to the bandwidth allowed by MIDI 
(1 message/millisecond).  These benchmark programs can also be used to get 
some feel for the effect on performance of various changes: running Linux 
as single or multi-user, running an X-server, running FTAP with and without 
root privileges...

If your MIDI interface has female connectors, you can set up the loop with a 
standard MIDI cable; if the MIDI interface has male connectors, you will need 
a female-female MIDI connector which is harder to find (I had someone make 
one for me).

Evaluating performance is best done by inspection of the output file. Since
you might mistrust whether the output file is an accurate reflection of actual
time, some of these benchmarks are designed to be compared with externally 
measured elapsed time (e.g., using the Linux "time" program or shell 
command). This actual elapsed time will include FTAP startup overhead and 
cleanup and file writing at the end, something which is really not part of 
the benchmark time per se.  It is also worth looking at the scheduling
diagnostics that FTAP prints to the screen (and also write to the file).

The elapsed time benchmarks can be made longer by increasing the number of
events until termination (or increasing the delay, in cases where it's used),
but too many events (5000 or so) will overflow FTAP's static storage buffers (all
events to be written to the output file are stored in memory until the trial
is over).

