Newsletter of the SR-52 Users Club
published at
9459 Taylorsville Road
Dayton, OH 45424
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
A Few Observations on the Occasion of our First Anniversary
While most of you seem to be pleased with 52-NOTES subject mat-
ter and technical level, some have expressed a sense of difficulty
in following a few of the more detailed discussions, and have suggested
that a more elementary approach would be better. My aim has been to
consolidate and present input material in such a way as to reach PPC
users of above average intelligence, inquisitiveness, and attention
span, but who may not have pursued formal education past highschool.
As I have already suggested to a few, if after careful perusal of
specific articles, having worked (actually run) all examples, and
carefully referred to cited earlier issues of 52-NOTES you still don't
understand something, write me, and I'll do my best to help. While
the college professors among you have contributed significantly to
52-NOTES content, so have others with considerably less formal educa-
tion. In order to make 52-NOTES as attractive to all concerned as I
can, and keep the size of each issue within pleasantly managable
bounds, I have pursued comprehensiveness with brevity as a general
approach.
I expect to continue to give priority to software inventions and
discoveries that have potentially broad application. Non-TI-approved
hardware modifications are apt to be risky, and will not often be
discussed in 52-NOTES. Even when mods appear to work, insidious
problems such as power supply overloads and potentially dangerous (to
hardware) instruction sequences may cause serious trouble. I intend
to continue sharing the best of my own discoveries and inventions
covering all the TI PPCs, so long as other members continue to do the
same. The ability to "hide" 59 programs on protected cards will
motivate some to try to market their best programs. I will publish
only the names of members who have such programs to sell; interested
members can write to the individuals concerned for further details.
But I hope that most will be motivated to share their best ideas openly.
For the 58 and 59, a particularly rich and rewarding field to
explore is the efficient use of CROM code. There are apt to be many
subroutines, or portions thereof, that can be cleverly used in appli-
pations other than those for which they were written. The sequence:
*Pgm nn SBR mmm will call a code sequence in Program nn beginning at
step mmm, and ending with the first INVSBR encountered (a feature not
noted in the owner's manual). Since every 58 or 59 user has the Master
Library module, clever use of its code can be taken advantage of by
all users. As special applications modules become available, they too
can be tapped for clever, efficient use by those who have them.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
The program that follows this article was designed to enhance the
ML-02 matrix program, facilitating inputs, labeling outputs, and it
makes use of the *Pgm nn SBR mmm sequence.
For those of you who would like to explore topics applicable to
all (or most) of the TI PPCs, there are still some outstanding invita-
tions to the membership: approaches to teaching computer programming
(V1N2p1), specific requests for help (V1N2p4), printer techniques
(V1N3p1), rigorous diagnostic programs (V1N4p4), systematic exploration
of pseudo behavior (V1N5p3), comprehensive determination of all factors
affecting program execution time (V1N7p3), reviews of periodicals of
interest to PPC users (V2N2p4), application program generation input
(V2N2p5), random number generators (V2N3p2), and program challenges to
HP users (V2N4p2).
I'll close this commentary by repeating what I said some months
ago: "... it is with pleasant anticipation that I look forward to
getting inputs from members who have so far been among the silent
majority, as well as to getting more gems from the productive minor-
ity."
- - - - - - - - -
TI-58/59/PC-100A Program: Enhanced ML-02 Program Ed
User Instructions:
1. Key n, press A, see 8 displayed (address of first storage register);
n and N printed.
2. Key ith element, with columnwise catenation, press R/S; ei printed,
address of next register displayed; repeat for i=1,2,...n2; correct
entries by direct STO (display-cued); determinant calculated and
printed following input of en2.
3. For simultaneous equations: If Det ≠ 0, press B, "B" printed; key
bi, press R/S; bi printed; repeat for i=1,2,...n; repeat step 3
for new constant vector; xi calculated and printed (with labels)
following input of bn.
4. For inverse, press C; inverse printed and labeled by column.
Note: For TI-59, max n=8; for TI-58, max n=3, and change all references
to Reg 89 to Reg 29.
Program Listing:
000: *Lbl D x:t 1 SUM 89 RCL 89 INV *x=t *5' + 2 = STO 89 *Lbl *5'
018: *Op 04 INVSBR *Lbl E R/S *Op 06 STO*Ind 01 *Lbl *1' *Op 21 RCL 01
034: R/S STO*Ind 01 *Prt RCL 01 INV *x=t *1' *Adv *Adv INVSBR *Lbl A
048: STO 07 STO 00 x2 + 7 = x:t 31 *Op 04 RCL 07 *Op 06 *Adv 13 *Op 04
070: 8 STO 01 E *Op 00 161737 *Op 03 *Op 05 *Pgm 02 C *Adv *Adv R/S
092: *Lbl *B 1 *Pgm 02 D RCL 07 STO 06 *Op 00 14 *Op 02 *Op 05 CLR
111: *Lbl *7' R/S *Pgm 02 SBR 355 *Dsz 6 *7' CLR *Pgm 02 E *Pgm 02 *A'
129: 4402 STO 89 *Op 04 RCL 07 STO 06 *Lbl *6' RCL*Ind 01 *0p 06
147: *Op 21 4408 D *Dsz 6 *6' *Adv *Adv *Adv CLR R/S *Lbl C CLR *Pgm
166: 02 *B' *Op 00 243142 *Op 03 *Op 05 *Adv 1502 STO 89 *0p 04 1
190: *Lbl *2' *Pgm 02 *C' RCL 07 STO 06 *Lbl *3' *Op 21 *Op 24
205: RCL*Ind 01 *Op 06 *Op 36 *Lbl *4' *Pgm 02 SBR 860 *Dsz 6 *4'
221: *Adv 1508 D 1 + RCL 03 = *Dsz 0 *2' *Adv *Adv *Adv CLR R/S
- - - - - - - -
52-NOTES V2N7p2
0x Definitions
G E Wilkins (25) notes that the various PPCs produce varying
results upon execution of 0x with the yx function, as the following
table shows:
Machine 00 0+x 0-x
SR-52 1 0 0
SR-56 1 0 0
TI-58/59 1 0 error
HP-25 error error error
HP-55 error 0 error
HP-65 error error error
HP-67 error 0 error
Perhaps this is not too surprising, since there appears to be no clear
cut mathematical approach to defining 0x. G E suggests that one can
argue that since log yx = xlogy, and log 0 is mathematically undefined,
then 0x should also be undefined. Or on the other hand, one can
examine the limit of yx as x and y approach zero, and find that 00
ought to be 1, and 0x (x ≠ 0) ought to be zero. In any case, G E is
looking for an efficient SR-52 routine that treats yx such that an
error is produced only when y is negative and x is not an integer, or
when y is zero. He wants a correct result with correct sign when y
is a negative real, with x any integer or 0, or y is a positive real
with x any real, all in less than the 65 steps it has taken him.
Members are invited to help out, and/or shed more light on the 0x
definition problem.
Hyperbolic Functions Shortcut
Roy Grubb (483) passes along the following approach to producing
hyperbolic trig functions, which he came across in a Sinclair Electron-
ics manual for its small PPC: *LBL E (((INV lnx + 1) 1/x X 2 - 1)
+/- INV tan X 2) *rtn. Using this basic routine, the desired functions
are produced by: sinh: E tan; cosh; E cos 1/x; tanh; E sin; sech:
E cos; cosech: E tan 1/x; coth: B sin 1/x. Inputs should be in radians;
any angular mode will do.
More on Reg 60, 61, ...69 Behavior (52)
In attempting to find a pointer-filling shortcut, Phil Sturmfels
(49) discovered a means of sensing which values held in the pending
arithmetic stack are "non-normalized". (Following a convention adopted
by the HP-65 Users Club, I will use the term "normalized" to apply to
numbers composed only of binary-converted-to-decimal (BCD) bytes;
"non-normalized" to numbers with one or more of their 16 bytes represent-
ing hexadecimal (base sixteen) numbers.) The only way to move the
contents of Reg 60 to a program register for detailed examination is
through the display register, which always normalizes a non-normalized
number (see V2N1p4). Register arithmetic on a non-normalized value
in Reg 60 also appears to produce normalized results. However, as
Phil discovered, a non-normalized number in Reg 60 is not affected by
its use for indirect addressing, and its pointer behavior can identify
it as non-normalized. For example, key: 1 + 2 X 3 yx in RUN mode.
Find that Reg 60, 61, and 62 appear to contain 1, 2, and 3 respectively.
Now use these registers as pointers, and find that only the 2 behaves
as a 2; the "1" and the "3" behave as 0, and it appears that they
are non-normalized. It may be that there are some non-normalized num-
bors that point normally, and this phenomenon probably deserves further
exploration.
52-NOTES V2N7p3
Pseudo 73 Limitations (52)
Larry Mayhew (145) points out that p73 "... cannot be trusted as
the equivalent of rset without the flag modifications, because under
certain important conditions it will simply be ignored during program
execution. Example: Beginning at step 000, write: HLT INV *ifflg 1 *1'
*LBL *1' 12 p73 45 HLT. In RUN mode key *rset RUN RUN and see 12 as
expected. Now key *rset *stflg 1 RUN RUN and see 1245, which shows
that the p73 executed as a no op. In general, if any conditional test
has been made without a transfer occurring, and if after that test
p73 occurs and there has been no intervening STO, RCL, SUM, EXC, PROD,
CLR or fix, the p73 will be ignored."
Step 223 Odd Behavior (52)
Larry also notes that "If step 223 contains part of a numerical
address (program or register) that is left incomplete at 223, and if
the program runs through step 223, and a digit is then keyed manually,
a 'semi-crash' will occur: Every register is cleared, but flags remain
unchanged." Or, if a conditional instruction is positioned at step
223, and executed with RUN n RUN where n is a digit, the machine goes
into a read state. Larry has also gotten a step 224 to appear in the
display in a sort of through-the-lookingglass Alice-In-Wonderlandish
hocus pocus by way of a run-down battery. Write him for details.
Information Referrals
Largely to save time, I have adopted a policy of suggesting via
52-NOTES that members contact a contributor directly, without my having
obtained prior approval from him in cases where I consider topics
inappropriate for 52-NOTES coverage. I don't believe this has caused
contributors to be overwhelmed with solicitations, but it always helps
to include at least a SASE with each request. Members not wishing
such announcements to be made should so indicate when contributing
material.
More on Don Ellis' Publications (V2N3p3)
B K Patterson, Jr (272) has examined some of Don's statistics
programs and reports an error in one of them. Members using Don's
publications may wish to contact B K and/or Don for details. But send
B K some change (rather than a SASE) since US stamps won't be of much
use to him in Canada.
More on Sum-of-the-Digits (V2N5p6)
I've received a number of 13-digit sum routines now, most of which
won't handle all numbers, which makes them hard to compare. So here is
a new routine from Howard Cook that appears to handle all reals, but
requires 56 steps (Karl Hoppe (507) has one with 61 steps). Shorten
it if you can, but "better" routines must be able to handle all reals.
*LBL E INV *fix *ifpos +/- +/- *LBL +/- STO 01 EE div EE 00 = INV *Prod
01 *fix 0 *LBL *LBL CLR RCL 01 *ifzro *EXC - .5 = EE SUM 99 INV SUM 01
10 *PROD 01 GTO *LBL *EXC *EXC 99 *rtn. Comparable routines for the
56/57/58/59 ought be be shorter.
Decapower Zero Suppression Following Negative Tests (52)
Jared Weinberger (221) notes that any decapower zero is suppressed
following an unmet test using a defined label for its transfer instruc-
tion. The suppressed-zero display can only be seen when a test is
executed manually or SST'd, since a HLT or rtn restores the suppressed
zero.
52-NOTES V2N7p4
SR-56 Program Exchange Update (V1N6p6)
Dave Johnston's 1 July 77 Catalog lists 86 programs covering
math, statistics, physics, games, finance, operations research, and
misc. Dave plans to get a TI-57 and will broaden his exchange service
to include 57 programs. And since TI does not plan to provide a
program exchange service for the 58 either, there will be a need to
be filled for that machine too. Write to Dave if you plan to buy
either the 57 or 58. If response looks like more than he wants to
handle, we'll look for additional help. Those members with 58s who
join TI's PPX-59 and wish to contribute programs to it may send me
copies of programs on TI forms with sample problems, all checked out
on the 58, but scaled to the 59, and I will put them on mag cards
(until such time as there are too many such requests). Include blank
mag cards and a large SASE.
Machine Coverage in 52-Notes
Since the nature of contributed material influences how I priori-
tize newsletter topics, if you want to see more coverage of a par-
ticular machine, write about it; clever routines, inventions, and
discoveries are what I look for most, and appear to be what most of
you prefer too.
Data Entry Sensing
Izzy Nelken (576) asks if there is a way during SR-52 program
execution to sense whether data have been entered from the keyboard.
There is no built-in function in the TI machines comparable to the
HP-67 Flag 3 (which is automatically set and post-test cleared when
data are entered during a program halt or pause), so it appears that
the display would need to be examined, assuming that keyed data would
fall in a given numerical range outside of the possible display values
produced by the program. If anyone has a better idea, Izzy would be
pleased to hear from you. But it's probably more practicable to
design programs to always expect some data to be keyed at specific
break points. A keyed zero or one provides easily tested data for
a two-way branch: a handy way to respond yes or no to printed ques-
tions (see the TI-59/PC-100A program in V2N6p3).
Membership List Corrections
44: Box 233; 97: 18 Carty Ave Ft Nonmouth NJ 07703; 307: Collins
Radio Group MS 424-101 Dallas, TX 75207; 368: 2631 Navarre Ave #209
Oregon OH 43616; 376: Buchenweg 24 D-5000 Koeln 40 Germany; 518: Math
Dept SUNY Geneseo, NY 14454.
Assessment of the 58/59 Master Library CROM Programs
Members with appropriate expertise can contribute significantly
by assessing/analysing specific ML programs. All can be down-loaded
into 59 user memory; all but ML02 and ML19 into a 58's. If you have
a 58 but not a 59, and want listings of these two programs, send me
a SASE. A cursory look at ML-02 (the longest CROM program) shows a
repeated use of in-line code sequences, which may be justified by
increased execution speed. But it also appears that more data regis-
ters than necessary are used. I invite more analysis of this and
other ML programs, and will air pertinent discussion via 52-NOTES.
While we can't change the CROM cede, we can use it more effectively
if we understand its structure and limitations.
52-NOTES V2N7p5
58/59 Tips
Short Form Addressing: Although the omission of leading zeros
when keying register or step addresses can save manual keystrokes, it
is easy to forget that a non-digit keystreke must follow. So it can
be worth the extra manual steps to avoid address mistakes that can
be time consuming to find; the number of required program steps is
the same either way.
ML-01 Print Routine Use: The last sentence in the user instruc-
tions should read "... except that the program must not be called."
If a program other than ML-1 is accessed, automatic printing won't
occur. Also, only manually keyed user defined keys (A-E') will ini-
tiate the automatic prints (calls to other labels, absolute addresses,
and/or R/S sequences will not). This routine does make use of indirect
program accessing, and is worth examining (via Op 9 dowmload) to see
how it works.
Taking Advantage of Control Operations: From time to time, a
glance at page V-27 of the owner's manual will remind you of all the
things the 40 special control operations can do. Keep in mind the use
of Ops 20-39 in place of programmed 1 SUM n or 1 INV SUM n (where 0 <
n < 10); they save steps, don't alter the display, and run faster.
INV and Ind Combinations: There are a lot, and some can be com-
plicated. Page V-68 sorts them out, showing proper sequences, and
which Ind combinations are merged. One misprint that caught my eye
is in the 19th entry: replace Dsz with ifflg. Remember that in all
cases, Ind follows the associated direct instruction (unlike the 52 or
56). Note that for both flag and Dsz commands both the flag number or
Dsz register and transfer address can be indirectly specified. A
double indirect sequence such as: *Dsz *Ind 25 *Ind 26 must be com-
pletely rewritten if either indirect address is to be changed, other-
wise the paired digits would not be properly merged. Incidently, *Dsz
*Ind XX will dsz the contents of any addressable data register within
the current partition as specified by the contents of Reg XX (not just
Reg 0-9 as is the case for direct Dsz). For direct Dsz it is important
to remember to key only a sigle-digit register address. For example,
*Dsz 0 4 keyed in LRN node is interpreted as *Dsz 0 004, which means
decrement Reg 0 and go to step 004 if the contents of Reg 0 is not zero.
The Nop Instruction: Contrary to what the manual says on page
V-51, there are quite a few code sequences which an interposed Nop will
alter, notably between a Command and an address, and between merged
commands.
Getting Merged Code into the Last Partitioned Step: The mechanism
by which some instructions are merged won't work at the last parti-
tioned step. This appears to be because the first keystroke(s) of
some merged instructions increment(s) the program pointer, and when
this occurs at the last step, there is the automatic switch to RUN
mode. The most likely such merged instruction to be at the last step
is INVSBR (code 92), and here are two ways to get it there: 1) starting
at the next to the last step, key RCL 92, then overwrite the RCL with
the intended next-to-last instruction, or 2) if the last step is less
than 959 (479 for the 58) temporarily partition to a larger program
field, write INVSBR at the desired step, then repartition as desired.
52-NOTES V2N7p6 (end)