Update:  Thanks to those who offered the suggestions, particularly the=20
barrel shifter.  Things look MUCH better today.
The left justify routine now takes 45 cycles worst case (Denominator =3D 1)=
,=20
16 cycles best case.  Went back, took a clean
sheet of paper, started working thru it.  Not nearly as ugly as I had=20
thought.  Pretty much linear code, no loops, very
few branches.  The process also popped up a nearly free divide by zero=20
detector. (1 conditional branch.).  It does
a <8 bit shift to get byte justify, then a conditonal byte swap to get word=
=20
justify, then a conditional word swap to get an
INT32 justify.
With that I am on to the rest of the effort.  That outside counsel put me o=
n=20
the right track.  I still have to make better
friends with the DSP core.  Right now it's kind of like a first date.  Ther=
e=20
is a bunch of stuff down there that I may be
able to use in the future.  I need to understand that whole DSP core better=
..=20
(Rainy day project and winter is coming.)
Once again, thanks for the suggestions.=20

--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.