				STENSOR

							Stockholm, sept 1992
							LH@vana.physto.se

They work together and with Reduce(for GCD,INTegration). Their specialties are:
SHEEP -components of tensors
CLASSI -fication of metrics, Petrov types.
STENSOR-"SymbolicTENSORs"=having letter indices. General noncommuting algebra.
	Sumsubstitution eg for trigsimplification. Diskusage for BIG work.
Please request separate further information about the features in your
area of interest. The below only describes the technical availability.


		MACHINES/LISPS THAT RUN THIS SYSTEM
		===================================
MACHINE:  	Requiring what LISP:		Relative speed:	Comment:
VAX(VMS/U-ix)	PSL(Reduce) or Franz		Eg 750:1  780:1.2 (Divide
Orion (HLH ltd)	Cambridge Lisp			       1   	-with users!)
ATARI ST  	Cambridge Lisp			      1/9  	(All for you)
AMIGA(As Atari,but untested) Cambridge Lisp	      1/9  	(All for you)
Sun 3,4,Sparcs etc.	PSL (or Franz?)		 from  4 (+>10 for disk i/o!)
DEC/VAX workstations.	PSL			 from  4 (+>10 for disk i/o!)
Other   -"- from eg Apollo should work similarly.
IBM 3090=If Need.	PSL			   ca 50??
Desktops IBM 386, 486 and MacIntosh within 1992?

Possible on request:	Eg Apollo, Cray, or most machines with PSL(Reduce).
			Rather similar to Atari: Acorn Archimedes/NS32016.

DO YOU HAVE REDUCE?  This is the first question - whether you have some
	machine with the computer algebra system REDUCE on it. If so,
	it also has "Bare-PSL" which REDUCE runs on, and which also is
	the most common platform for us.

IF YES,HAVE REDUCE:	-then it's preferable I send STENSOR,
	if you want the most recent version.

IF NO,HAVEN'T REDUCE:
	a:Get it separate (Valuable;can eg work together with STENSOR)
	b:If you don't need/want Reduce, you can get both SHEEP../STENSOR
	*AND* Bare-PSL, from Malcolm MacCallum, QMW, London (mm@qmw.ac.uk)
	who has aquired licence to distribute Bare-PSL for this purpose.
	(He is also one of the authors of a book on SHEEP/CLASSI/STENSOR).

	To be complete, there are currently 3 suitable lisp-alternatives:
	Portable Standard Lisp, 'PSL'   (USD 200),
		or better get REDUCE for USD 500 (includes PSL)
	Cambridge (University) Lisp	(USD 2-300?)
	Franz Lisp (comes eg with Vaxyma), now free?
Soon? porting to: LeLisp(fast,costly), and
	CMU Common Lisp (free), which has window system(dropdown menus)

Of MACHINES, a SUN or DEC workstation (with Reduce) has the most flexible
	memory. Next to those, VAXes are most common; any OS.

When ordering, specify your:
	Machine + Operating system:
	Tape formate: (TAR or VMS-BACKUP)
	Tape density: (800, 1600, 6250 BPI. XABYTE 8mm cartridge)
	Lisp system: (Cambridge, PSL or Franz)
	Whether you have got REDUCE or not:

IBM 3090: REDUCE exists in two versions; one based on PSL, and
======== one on a too old SLISP. Thus REDUCE isn't enough;
	You must ensure	it is based on *PSL*.
	Eg the one in LYON (FRCPN11) has NOT PSL, and is thus not feasible.

ATARI/AMIGA:	I'd send DS/DD floppys, and also later correspondence and
===========	complementary upgrades could use the very convenient floppys.
		Four Mbyte allow up to ca 10.000 term expressions in RAM
		-usually enough(remember bucketing). 1Mb hold max 2000 terms.

COST:	SHEEP/CLASSI is hitherto free (except for tape, ca $50) but the
====	STENSOR code is so large that a symbolic contribution of $200 extra
is welcome. I could email an invoice in tex-form, or send a signed hardcopy,
to help your administration. The total payment corresponding to US Dollar 250
could be sent:
>>	as check by mail,

>>	or to the bank account nbr: (11 60) 241 081 521
	-at the office: SHB (Svenska Handelsbanken)
			Box 1057
			S-171 21    SOLNA,	S W E D E N

	Checks cost ca USD 7-8 for both sender and receiver.
	One can avoid this by going to a (main?) postoffice,
	and use the postal currency (transfer) account service,
	with a postal cheque (giro) form, sending to my:

>> Personal postal cheque account,	(postgiro) number: 433 20 98-5

The accounts are associated with my home address (phone +46 8 837114):
			Lars  Hornfeldt
			Vastra vagen 11B, 6tr
			S-171 46    SOLNA,	S W E D E N

Notify me of transfer,so I can send off documentation and latest version.

Sincerely Yours,
			Lars  Hornfeldt
			Institute of Theoretical Physics
			University of Stockholm
			Vanadisvagen 9
			S-113 46   STOCKHOLM,	S W E D E N


	    NOVELTIES IN THE LATEST RELEASE of STENSOR (2.31)
	    =================================================
-Are very briefly listed below:

1: New GAMMA ALGEBRA algorithms:
 a:Takes on VAX the trace of 22 gammas, while Reduce HEP stalls for 14 gammas.
 b:Combinations of several gamma_5 -traces may be compressed orders of magni-
   tude compared to other systems. This is due to STENSORs unique exploita-
   tion of the combinatorial SCHOUTEN IDENTITY for Levi-Civita symbols.
 c:It has always allowed mixing of gamma- with other user introduced non-
   commutative objects. This seems impossible in Reduce,Form/Schoonship etc.
 d:Prepare fast NUMERICAL EVALUATION (for FORTRAN) by subexpression-
	recognition and factoring. Savings, over eg Reduce-factoring, are
   important since evaluations are repeated a very large number of times.

2: Explicit contractions (generated by the TensorCOMPiler) will run over
   different ranges depending on the space (indextype) of the dummy used
   in the formula. This makes eg the Ashtekar package much more compact.
   Already before TCOMP declared the proper storage arrays. Also
   indicial manipulations and splitting already have utilized indextypes.

3: Runs also on SUN 3, SUN 4, Sparc and DEC workstations (+probably Apollo,..
   and others with REDUCE),beside as before VAX(Unix/VMS), ATARI,AMIGA, ORION.
   Versions for IBM 386/486 and MacIntosh are expected this year.

4: Index sequences: Each 'index' starting with two dots indicates any number
   of indices in substitutions, eg to turn: <A i ..><B ... i> into <C .. ...>

5:     COMMUTATORs can be inserted with [] eg exactly as: [a,b], and
   ANTICOMMUTATORs can be inserted with {} eg exactly as: {a,b*c}.
   Objects (a,b,c..) may of course also have indices, then surrounded by <>.
   You can nest (anti)commutators by nesting [] and {} in each other.
   It proves eg the Jacobi identity, a task one can feed in
   straightforwardly as: [[a,b],c]+...	Type SEE COMMUT

Also: FACTOROUT-command for partial factorization in a way suited for tensors,
TRACESUB-command to step through how complicated sets of substitutions act,
Edit substitutions, TRUNCATE powers, Save_or_compare with expression on disk.


	NEW APPLICATION FILES and FEATURES in eg the FOLLOWING AREAS:
	=============================================================

GRAPHS: Grow and canonize graphs, eg random growth, aids-simulation.

QUANTUM GROUPS and QUANTUM-PLANES: This fresh area again shows the ease
	by which noncommutative algebras can be declared.
	Such expressions are then also used as entries in matrices.

DECOMPOSITION of SU(n) in QCD/Color:
	By combining [] and {} one gets three Jacobi identities for the
	generators T. These translate to corresponding sumrelations for the
	3:rd-rank structure constants f and d. These are effectively exploited
	by the sumsubstitutor, giving considerable compression of expressions.

Similar payoff in reduction is given for expressions containing levi-civita
symbols as said in 1b above (again, not exploited by any other system).
For this, and more on eg graphs and Seeley-de Witt expansion/heat kernels
(1c) cfr a paper presented at the HEP-conference -92 in l'Agelonde, France.
