kcl speed problems

["CTL::GATEWOOD" <gatewood%ctl.decnet@natc-fw.navy.mil>]

I am running AKCL-1.492 on a SGI Iris 4D320VGX.  It was compiled with
GCC-1.37 as modified by Michail Meissner (file: gcc-osf-1.9.2.13.tar on
foobar.colorado.edu).  I am trying to port some code written on a Lispm
which does lots of floating point computation on large arrays.  I have
tried to pre-allocate enough space to avoid garbage collection and I have
been partially successful at that but the program is still suffering from
very poor performance.  How can I find out how to speed up my programs in
kcl?  How can I be sure the floating point hardware is being used when
appropriate?  How can I pinpoint the source of the garbage collections?

Any help would be appreciated.

Pat Gatewood
gatewood%ctl.decnet@natc-fw.navy.mil
301/737-6081