Newsletter of the SR-52 Users Club
published at
9459 Taylorsville Road
Dayton, OH 45424
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Book Review: User Survival Guide for TI-58/59 Master Library, by Fred
Fish (606), $10.25 postpaid to members, 119 pp, copyright 1978
Fred has put together a comprehensive package designed for the
serious user of the ML programs, who wants to know how they work, how
to interface them with his own programs, and how to take shortcuts. He
brings to the user's attention conditional caveats, and pitfalls to
avoid, and details specific program calling sequences to optimize reg-
ister usage, execution speed, and required user code. This work appears
to be technically sound, includes listings, flowcharts, and even a few
cartoons, is well-proofed, and I am pleased to give it my endorsement.
Include your Club membership number with your order, and ask Fred about
quantity discounts.
A Review of the TI-57 Vis-a-Vis its Contemporary PPCs
The 57 has a few unique features which fall into both + and - cate-
gories when compared with other PPCs. This review is intended as an aid
to 57 users as they either start from scratch, or attempt to translate
other PPC programs, and to acquaint non-57 users with this somewhat dif-
ferent machine.
It is interesting that the low-end TI PPC's unique features are
common to some of the high-end HP's architecture (67 or 97): 1) The
numerals 0-9 are used as labels, 2) All composite instructions are merged
3) The step following a conditional can be any instruction, and 4) During
execution, the display flickers intelligibly (which the 67/97 can be
made to do under certain circumstances). Perhaps the most critical limi-
tation is the comparatively large amount of register sharing. While the
number of data registers (8) puts the 57 into the same league as the 56
or HP-25, shared use of some of them is excessive: 1) The last 2 (of a
max of 4) pending operations go into Reg 5 and 6, 2) Reg 7 is also the T
register, and 3) The Σ+ uses all the data registers except Reg 6. How-
ever, the latter results from a positive feature: The Σ+ function pro-
cesses 2 variables at a time. Another important limitation is the 8-
place display (with 3 guard digits). While this may be acceptable for
most applications, the user should bear in mind that actual accuracy
may in some cases be lower than expected (1 INV lnx produces e to only 9
places). Execution speed of the 57 is generally slower than that of
the other TI PPCs, especially for the trig and exponential functions. A
realistic comparison of the 50 merged steps with the 56's 100 unmerged
ones will be deferred until a few representative challenging test cases
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
surface. However, as a starter, when I attempted translation of Dix
Fulton's V2N10p1 program, it overflowed by about 20 steps.
A few programming tips come to mind: 1) Instead of inputting via
a S1 R/S S2 R/S ... programmed sequence, write S1 S2 ... and execute
with the SST key; 2) Take advantage of the anything-goes step following
a conditional, where such functions as R/S, rtn, Pause, ±, etc may be
better than GTO n; and 3) If you run out of user-defined labels, SBR
2nd nn from the keyboard (but not under program control) will start
execution at step nn. 57 users are invited to comment and share their
tips, inventions, and discoveries.
Fractured Digits: More Control (58/59)
Roger Gentry (398) has discovered some new ways to produce frac-
tured digits, one of which leads to a capability to display a few sel-
ected characters with or without specified numerals. For example, with
a 59 at turn-on or a 58 at turn-on and a 459.0 partition, key: GTO 441
LRN List 0 0 0 0 0 LRN Pgm 11 SBR 444 R/S LRN, and see 2 ° symbols
looking like a pair of eyes. The machine at this point appears to be
in LRN mode, expecting a 2-digit register address. Pressing any non-
numeral except LRN writes code in the usual way; pressing a numeral
"closes" the eyes, producing 2 - symbols in addition to writing a
number code. Press LRN BST BST and see the keycode for the numeral
you just keyed, or key another numeral than BST, and see the merged
code for both numerals. Other (but not all) permutations of fractured
digits in 3 consecutive display positions can be produced similarly,
following rules based on this discovery of Roger's: The current step
number in the user memory IAR which prevails when the sequence: Pgm 12
SBR 999 R/R LRN is keyed manually, determines what characters are
formed at display positions 8, 9, and 10 (using the 0-11 full display
convention). Roger observed that when a step is converted into 3 such
codes by dividing it by 8, interpretation of the quotient, and remain-
der in integer form a V1N2p5 character code produces the expected
characters. In the above example, the call to Pgm 11 is made at user
memory step 447. Division by 8 produces 55 remainder 7 which are the
codes for ° ° blank. But there are other requirements to be met to
assure the desired display. If we label the octet containing the
"launch" step (in this example step 447): AB, CD, ... OP (as in V1N1p5)
the following rules appear to be necessary to follow to produce just
the 2 eyes: A must be zero, B can be anything, C must be non-zero, and
OP can be anything. Note that MN and OP are set to 72 and 42 respec-
tively, following execution. Non-zero EF, GH, and IJ can produce up
to 6 numerals preceding the eyes; non-zero A and C show up to the right
of the open and closed eyes, respectively; fix n, n: 1-8 positions a
decimal point in accordance with n; launch at other than the last step
in the octet produces different results ... there seems to be no end
to the variations. Members are invited to explore further, and to
report useful (or what may appear to be potentially useful) creations.
Label Search Methods: A Comparison of TI/HP Machines
John Hirsch (763) notes that while considerable attention has been
given to the pros and cons of AOS vs RPN, label addressing comparisons
have received little attention. John points out that the TI machines
have absolute and label addressing, contrasting with the newer HP
machines' relative and label addressing.
52-NOTES V3N3p2
Perhaps a definition of terms at this point would help to avoid
confusion. In past issues of 52-NOTES I have used the term relative
to describe addressing within relocatable code, and absolute for un-
relocatable code. John uses the term relative with the HP machines to
mean an addressing method whereby an address is calculated by adding/
subtracting an integer value to/from the address of the currently
executed instruction. In other words, the program counter (or what I
prefer to call the IAR (V2N4p5)) is either incremented or decremented
by a specified amount. It is in this same way that the word relative
is used to describe general purpose computer addressing, and since it
is quite possible that a new TI PPC (or personal microcomputer) might
have such addressing, henceforth I will use the term relative addressing
with this meaning, and as John suggests: label addressing for what I
used to call relative addressing.
HP-67/97 relative addressing is backwards only (a negative value
stored in the I register is added to the contents of the IAR), so for
forward transfers, the user must count the number of steps back through
000, 224, 223, ... to the desired step, and then provide for storage
of the negative of this count in the I register. The primary advantage
of this scheme is that it shares the same speed enhancement and memory-
region independence of the TI machines' absolute addressing without the
unrelocatability restriction. Another important difference is the label
search mechanism: The HP machines start a search downward from the
instruction starting the search. This makes if difficult to optimally
locate subroutines called throughout a main program, but makes it
possible to use the same label more than once, and expedites transfers
within code located low in memory.
John made some label-search-execution time comparisons among the
SR-52, SR-56, TI-57, TI-59, and HP-29c machines, and reached the follow-
ing conclusions:
1) There is really not much difference in label search time when
differences in total program steps are considered, although the TI-57
came out best.
2) The HP disadvantages offset the advantages.
and suggests for 3rd generation machines a key which gives the
user the option of specifying label search to be forward, backward, or
from top to bottom.
Other members are invited to comment, and contribute further to
this topic.
Routines:
Fibonacci Number Generator Revisited (V3N1p3): Upon applying his
routine to large numbers, Joel found that correct answers could be
extended a bit by calculating the (1 + √5)÷2 constant (known to mathe-
maticians as Tau or Phi, or popularly as the Golden ratio as
(1 + (1 ÷ √.2 ))÷2. I will refer to (1 + √5)÷2 as t1, and (1 + (1÷√.2))
÷2 as t2, and Tau in general as t. Incidently, some interesting things
about t are due to the relationship: t2=t+1, which may also be written:
t=(t-1)-1. It turns out that t2 is a better approximation than t1 by
one place, because of machine rounding and truncation. As calculated,
t1=1.618033988749, t2=1.618033988750, but since t=1.61803398874989...
t1 is good to only 12 places, while t2 is good to 13.
52-NOTES V3N3p3
Efficient I/O Handling For Engineering Problems (58/59): Clyde
Durbin (618) found that the nature of some of his engineering problems
suggested some useful tricks/shortcuts, which should also be helpful
in related applications. One situation which arises often concerns
the user's desire to specify printing and tagging of a few outputs out
of many possible ones, following initial processing. The sequence:
LA S0 R*0 H8 GTO*0 when called by nA, where n is a register address,
will cause code to be executed and tagged results printed, both speci-
fied by the contents of Reg n, provided Reg n is initialized with:
sss.00tttttttt where sss is the step where specified processing is to
begin, and tttttttt the print code for the desired tag. The step sss
sequence ends with ... Op6 R/S. Clyde has also found that in cases
where 2 or more input parameters are in mutually exclusive numerical
ranges, they can all be entered with a single user defined key. The
program first compares an input with prestored boundary values, and
branches accordingly.
Copying the Display into the T Register (56,57,58,59): There are
times when it is useful to effectively store the display in the T regi-
ster without changing the display. Jared Weinberger (221) suggests:
... (CP + x:t)... . Does anyone have a shorter way? (...CP + x:t =
... doesn't count!, and for the 57, STO 7 does the trick).
Digit Reversing (V2N11p6): Reinhold Patzer (689) has trimmed
Jared's by 3 steps and one register with: LA (CE ÷ 10 Prd 1 - INV Int
SUM1) INV x=t A (Exc1 X 10) rtn, and notes that both routines will
handle up to 13-digit numbers.
Tips and Miscellany
A New Hidden Facet of the Old Machines (52,56): Ken Whipple (849)
discovered that his SR-52 dimly displays a numeral representing the
number of digits to the right of a displayed decimal point in the other-
wise unused decapower LSD position for a real in fix format. Ten digits
to the right (the maximum) even shows up as an A (as if the machine
knew it should count in the hexadecimal number base)! I confirm Ken's
findings, and add that the SR-56 counts the same as follows: ' ° -
blank 5 6 " blank ' A. So here's one way to get fractured digits on
the 56. The new machines do not seem to respond in this way, but per-
haps there are related frontiers waiting to be explored via darkened
room and magnifying glass.
Extra Memory for the TI-58: Bob Edelen (100) reports that he has
successfully added 2 TMC 0598 chips (order from TI per V2N5p3 for the
TMC 0599, omitting Riggs' name) to his 58, effectively converting it
into a 59 without mag card read/write. Bob has a supply of these
chips, and is prepared to install them for 58 owners. Write Bob at his
business address: 5440 Roslyn #285 Denver, CO 80216.
Data Entry Sensing (V2N7p5): Reinhold Patzer (689) brought to my
attention the best approach I've seen yet for program-sensing of data
entered from the keyboard. The method is due to Heinrich Schnepf (376)
which he describes (in German) in V2N2p35 of his Display newsletter,
and is simply to follow a programmed data-entry R/S with a decimal
point and a zero test. It appears to work for all reals so long as the
display is zero or hard at the R/S, and is left soft following a datum
entry; a keyed zero is treated as no entry. This trick works because
the decimal point zeros a hard display, but does not alter a soft one.
52-NOTES V3N3p4
A New R-P Convention (57,58,59): J R Merrill (693) notes that
the new machines return positive angles in the 0-270° range, and negative
to -90°, unlike the older TI and all the HP machines, which return ±
180°, for rectangular to polar conversions. Which is better? J R
prefers the latter, but there may be applications favoring the former.
Page V-31 of the 58/59 owner's manual notes the 4th quadrant negative
angle convention, but doesn't say why it was chosen. The 57 manual
doesn't appear to address the matter of returned angle signs at all.
Fast Constants: When deciding whether to create or store a con-
stant, if execution speed is critical, note that creating a digit string
with more than 3 elements using program code takes longer to execute
than a RCL.
Runaway Mag Card Drive Motor (59): John Hirsch (736) reports
occasions when following card drive, the motor can only be stopped by
turning the machine off. The owner's manual (p VII-8) states that key-
ing R/S should stop a running motor, but implies that it may be "normal"
to have to turn the machine off in some cases. Anyone else experiencing
John's problem, let me know.
More on Pgm mm R/S (V3N1p3): John has found that there are indeed
applications for sequences of the form: Pgm mm R/S in user programs.
After a CROM routine in Pgm mm has been called by a user program, the
effect of Pgm mm R/S is to resume CROM execution at the step following
the last rtn encountered, just as a plain R/S does during keyboard inter-
action with a CROM program. A CROM pointer maintains this restart
address through intermediate user-code execution, but not following a
manual or programmed call to another specified Pgm label or step number.
For example, write: LA 3 Pgm 2 A 1 Pgm 2 B R/S LB Pgm 2 R/S R/S GTO B.
Press A, see 1; key a datum, press R/S, repeat for say 3 data. Now go
elsewhere in user memory, write a small routine (that doesn't clobber
the ML-02 registers, but may use all 6 subroutine levels), run it, then
key the next datum for input to ML-2, and press B; key a few more data
with R/Ss and find that all data were stored sequentially as expected
in Reg 8, 9, ... .
Creating Large Mantissas (57,58,59): For all 3 machines, creating
a number in scientific notation whose mantissa is larger than 8 places
can be done, but requires a few steps. For the 57, one way is to store
the first 8 digits in a convenient register, SUM up to 3 more (appro-
priately scaled), then Prd 1 EE dd, where dd is the desired decapower.
For the 58/59, the SUM step can be skipped for mantissas less than 11,
or included for mantissas of 11 to 13 places.
Overflow and Underflow (57,58,59): The V2N1p3,4 discussion applies
to the new machines, for the most part, but with a few exceptions. For
all 3, display rounding of large numbers to 8 mantissa places corres-
ponds to 10 places for the 52 and 56. When the 57 transfers any condi-
tionally overflowed number from a data register to the display, it is
converted to the upper-limit overflow number: 9.9999999999D99. So
operations performed on a conditionally overflowed number should be
via register arithmetic.
Membership Address Changes: 128: Back to original address; 452:
3812 Kendate Dr Gautier, MS 39553; 553: 11 Plumeria St APRA Hts FPO
SF 96630; 664: 100 Barton St #10 London, Ont N6A 1N2 Canada.
52-NOTES V3N3p5
Absolute Code Execution Speed (52,56,58,59): Roger Gentry (398)
notes that absolute transfers near the top of memory are faster than
those near the bottom (high addresses). The effect is the same as for
label transfers, but is not so pronounced. Short Dsz loops placed
within a 1-register octet of steps run a bit faster than when straddling
2 or more registers.
HIR Operator Modifications (58/59): Rodger discovered that under
certain circumstances HIR operations in the middle of pending arith-
metic can change the originally designated arithmetic operators. For
example, n - H11 H1 = produces 2n, and n ÷ h11 H1 = produces n2. This
behavior suggests that numbers pushed into the HIRs via pending arith-
metic may not be so "clean" as once thought (V2N9p1).
More on CROM HIRs (V3N2p6): The Leisure Library program which
T S Cox meant to refer to is Craps (13), which contains a "real" HIR
at step 240. I say real, not "intended", because my definition for
the latter is poor in this example: The only way this HIR operation
could do anything is via a SBR 240 call, which would add the display
value to the contents of HIR 2; SBR CLR R/S would add zero to HIR 2,
and I don't see any other way for the program to even get to step 240.
Also, there doesn't appear to be any use made of the contents of HIR 2.
I make this assessment based on a listing kindly provided by T S (I'm
still waiting for my free copy of the Leisure Library).
Mag Card Tips: Jerry Johnston (493) suggests folding a strip of
paper around a mag card before inserting it into one of the TI wallets,
to facilitate extraction, and recommends Pelican #17 black ink for
permanent card marking.
More on the Hardware Interrupt (V3N2p5,6): Roy Chardon (515)
adds an improvement to make the interrupt handler module-independent:
Make the CROM call to Pgm 1 SBR 021, since every CROM appears to have
a rtn at that step. Roy also warns that the manual RST does, of course,
reset all flags and clear the subroutine-return register. The R/S at
step 000 of the V3N2p6 sequence serves no purpose, and should be dele-
ted or replaced by GTO N or GTO nnn, where label N or step nnn head an
interrupt processing sequence, if such is not to start at step 000.
CROM Pause: The statement concerning step 353 of Pgm 19 in
V2N8p6 is incorrect. That step is a rtn, not a pause. Code 66 in
CROM, either intended or unintended executes as pause.
DC Adapter: Mack Maloney (246) notes that TI's 12 volt DC to
120 volts AC convertor (model DC 9105) was designed for use with the
52,58, and 59 (but would presumably work with other PPCs), and is
available at the Washington DC TI Exchange Center.
Quick Manual Flag Tester (58/59): Nolan Tobias (853) notes that
the V1N4p6 method for the 52 can be shortened to Ifflgn N, where N is
an undefined label.
The R/S' Function (58): J R Merrill (693) notes that the 58's
equivalent of the 59's Write function appears to execute as Int, and
Jared Weinberger finds that when used as a label, R/S' behaves like R/S
(V2N11p4). Roger Gentry (398) has found that a prevailing fix n condi-
tion, n: 1-8, followed by R/S' multiplies the integer part of the display
by 10n.
CLR/CLR' Difference (58/59): Joel Pitcairn (514) wonders if
anyone has found an application where CLR and CLR' execute differently.
52-NOTES V3N3p6 (end)