A1

(Oh boy, you saved two whole bytes! Iım sure your mother is
proud... or
is that two whole keypresses? Iıd type the whole word if I
were you;
practice makes perfect! If you really really meant to just
say u
for stylistic reasons... ... well, you suck.
Hyperreductionist English
is evil!)
Assuming that you really mean taking HP48 programs and
changing them
to work on the HP49:
Well, one plausable but imperfect way is to write a program
that converts
all PTRs or GO* to an HP48G? entry point to an HP49G one.
This is
imperfect because it assumes that the programmer only used
certain entry
points -- which might not be true; there are valid reasons to
only jump
to a location after the typical one, say to avoid an
unnecessary test or
register juggle. When dealing with ML code you have the
additional
====
I tried several times to enter lim or li to get lim in CAT.
I remember it worked some times?
I am using ALPHA ALPHA LS ALPHA and it works Ŝne to enter
small letters,
but not in CAT?
What I am doing wrong in CAT.
====
Try this
EQW
lim (1/X) for X=0
SIMP UNDO
watch the result!
Do SIMP UNDO again
====
if you plug the cable end to the calc with nothing else
connected to the
other
====
Did you put the connector the right way up? I.e. with the
HP-emblem on
====
Do you use a HP49G cable or another connector that Ŝts (100LX
for
example)?
Anyway, if you see the screen change color, itıs time to pull
the plug!

====
Count me in, I got a nice black screen before unplugging it
(in a new world record time :-)
The amount of force wasnıt that big, just slightly more that
the right way
up
This is exactly what I did, I was downloading a lot of
different proggies
and I got careless with the plugging...
Not that I was worried, because when I pugged in my
F1015-80002
several times (applying 12V to my 49) it sustained no damage.

2. BNCıs one hole, earth on rim. Never experienced any
problems with them
3. Cable TV, I have some strange receptions on some channels
but when I
called them they told me that you had to pay for
the Playboy channel.....
4. Flat cables, never manage to plug one in the wrong way
6. USB, wonıt plug in the wrong way
====
No, everything looks right-side-up here -- you must be
looking at me
wrong.
Power plugs here now have the Œone prong is a little larger
than the otherı
feature, and some of them have Œgroundı prongs -- both can
only be plugged
in one way. All the cables and ribbons inside my computer
have safeties
====
Iıve updated my ŝags document to the revision 1.1 at
hpcalc.org. The URL
is:
http://www.hpcalc.org/hp49/docs/misc/ŝags11.zip
Some english imprvements (thanx Dennis!) and ŝag 61 added
(thanks to

====
Having some issues with my 49G, seems at random startups it
does a
warmboot. I read a thread on that awhile ago and canıt
remember if anyone
came up with any good ideas. I donıt loose my memory but it
is really
annoying loosing all my settings every day or so.
It seems to have started to happen since I went to 1.19-6.

You are completely right! And this explain why I canıt Ŝnd
lot of
informations about this variable. Anyway, going here and
there in the
hidden
directory (thanx to all that replied me giving hints in this
direction or
in
any other one ^_^) I discovered the variable Alrams, which I
think must
refer to the scheduled alarms. I havenıt the time to play
around to
====

I think i hate RPL !
a friend of mine is using a ti 89 and i m running a 48g+
I wish i could access a variable as easily as he does...

I wish there was a real C cross compiler for saturn....or asm
was like
80x86.. i can t manage to understand the smallest Saturn
code, whereas
80x86

====
nothing easier than accessing variables in RPL
Actually, you can perfectly do exactly:
Phrase
Good Bye ŒPhraseı STO
Phrase
Or you could have done:
blabha % do stuff
ŒIhaventreadthemanualandispokemeaninglessmessageı STO
blahblah %do other stuff
Œxı STO
====
There is. hp48xgcc.
http://sourceforge.net/projects/hp48xgcc/
There is such an assembler. Iıve never seen it, but there
obviously must
be one, because xgcc internally uses a syntax like x86
assembler, i.e.
mov.a a,c for example.
IMHO, the Saturn mnemonics MASD, Jazz and HPTools etc use are
*much*
====
Well, the command is {STO}, not {sto}.
has the effect of your example when I ignore the
aforementioned error.
I understand neither this nor your example.
... what donıt you understand? See ca-on.hpcalc.org for
documentation
on UserRPL.

====
====
This is probably a really stupid question but I have been
unable to
Ŝnd the answer in HPıs users guide.
I have just switched over from a TI-89 to an HP 49G.
In order to enter polar values for a complex number with the
TI all I
had to do was enter (magnitude<angle) and to enter
rectangular values
====

Itıs a pity how HPs are being sold. But it began in 1996/97,
when
production
was moved to Indonesia. All those painted keys... Perhaps if
HP had
received
enough negative feedback at that time...
I was just trying to make JYA talk about them... ;-)
But the question is: what are they doing there? Hope theyıre
happy.
http://blues.squidly.org
Oh, perhaps our dreams should be directed to more interesting
machines:
the HP-67, HP-41, HP-71 and, of course, the HP48 can still
Ŝll many pages
of magazines (even new books are published from time to time).
====
way, that very unusual (and sometimes unexpected) features are
possible. But the process of discovering about such things is
very
difŜcult, when one has only the manuals that come with the
calculator. The whole work of documenting features and Ŝnding
bugs
that has been done here, should have been done by HP itself.
( Or do
they expect us to test and document their product? ) Let me
pretend
for a moment that I am someone who only wants to use the 49
as a tool
in his/her work and doesnıt care about hows or whys or
CAS-programing.
What would such a person say? Something like: I really enjoy
reading
the postings here, and also post myself, if I Ŝnd something
that is
worth mentioning. But, buddies at HP, I am not an HP employee
and so
====

He, he, whenever we try, he says DeŜnitely no new projects. I
guess, HP has adviced him friendly, not to talk about
anything. ;-)
Yes, I wonder if some new enhanced TI92+ will be made. The
same
hardware but much better software. I mean the built-in
software
programed more efŜciently and squeeze any bit of performance
out of
the machine. Plus all the wonderful things that guys like
Bhuvanesh
have made so far. (Plus RPN, but this is never going to
happen...)
Oh you mean *these* guys. Well, Iım on a mission for BORG,
errrhh, GOD
of course.
I also Ŝnd the 67 and the 41 still amazing. But I think the
main
point of my interest towards these machines is, how much can
be done
with so few things. Amazing!!!
====

Well, can you give a concrete example? I must admit that I
didnıt follow
the ng lately as much as Iıd like to. And my 49G is getting
pretty
lonely these days... :( But from the little time I can spend
with it my
impression is that as long as you stick with the standard
apps and the
user rpl side of things, you can take pretty much anything
thatıs on the
stack or in a variable, unpack it, stuff it into some other
appropriate
data structure and feed that to another routine. You might
have to write
some little programs (I mean user rpl! John H. Meyers, please
take
notice! ;-) to Ŝll the gaps, but at least itıs possible.
As you all know there are some very nice math tools for UNIX
systems.
All the parts are there, but to solve a substantial problem
youıll have
to run pipes to several different interpreters, feed them
data unparsed
to ASCII, parse the ASCII results, process them ..., wash,
rinse,
repeat. AFAIK noone has yet come up with anything that really
integrates
the whole lot that it takes to do some serious scientiŜc
programming: a
state of the art functional programming language,
comprehensive numeric
and symbolic math features, GUI helpers and a good
visualization
package. Maybe that will remain a dream forever. But the 49G
is
something like this (on a calculator scale), and thatıs what
I like
about it.
[...]
Thatıs what Urrozı books are about, I can only recommend
them. What I
still miss though is a comprehensive CAS tutorial, with
exercises.
Anytime Iım banging around in the EQW I Ŝnd myself trying to
remember

Is there a way to exchange data/programs between an emulator
and the real thing?
Is it possible to upload Ŝles from the HP PC Kit to an
emulator?
Ŝnd all the documentation I need.

Letıs say you have EMU48 as your emulator (if you donıt have
it, go to
www.hpcalc.org and search)
Start EMU48, go to File/Settings menu and put the appropriate
COM port
in Wire Ŝeld. Now you can connect your real HP with the
emulated one
by your PC cable.
Sadly true (well, not exactly: thereıs a lot, but it isnıt
well
====
----- Original Message -----
Well, I downloaded Emu48-1.25 and Emu48 SP 2.8. I tried many
times, but
everytime I got a No System message, where Iım stuck. I
copied also from a
Yorkem distribution the BMP image and the rom.e49 Ŝle; I also
tried to
rename this Ŝle to ROM.49G. Nothing. Every time the same
story, I even try
to load different KML script I downloaded, but the emulator
seems not to
see them.

version works quite in the same way.
Emu48 has a free frontend, the view is always a bitmap
controlled by a KML
(Keyboard Macro Language) script Ŝle. In the KML script Ŝle
other Ŝles
maybe included. Finally you need a ROM image of the emulated
calculator.
This ROM image _must_ have a special format, the Win32
package (not the
update archive) contain a program doing this. The ROM format
is equal in
the
Mac and Win32 (and WinCE) version.
All these Ŝle _must_ be in the same directory.
Pierre made a Ŝle
http://tardyp.free.fr/emu48fm/download/emu48Ŝles.sit
containing all the needed additional Ŝles inclusive the
converted ROM
images. If you want to use other skins, the ones written for
the Emu48
Windows version should work.
But please donıt ask me how the HP48 Port2 is supported, the
Win32 version
has a special Port2 Ŝle maker (command line tool in the
ofŜcial
distribution, GUI version on seperate download which will
replace the
command line one in the next ofŜcial distribution), which the
Mac
distribution hasnıt. The Win32 Ŝle data is a Ŝle with double
size of
wanted port2 RAM Ŝlled with 0.
The generated emulator state Ŝles arenıt compatible from the
Mac to the
Win32 version, because Mac save the data in big endian order,
where the
====

1) How do I paste an object on the stack? I tried but the
Paste menu
item is not on.
2) How do I keep the emulator on constantly since saving
batteries are not an issue here?
3) Can I make the screen brighter? compare it to this picture:
http://www.gnarlodious.com/Time/Time.html#middle
====
Dragınıdrop a Ŝle onto the emulator window.
I know there are patch Ŝles for removing the auto off, but i
canıt Ŝnd
any on hpcalc.org. Try searching Googleıs archive of
comp.sys.hp48.
If you edit the color settings in the KML Ŝle youıre using,
you can
make the screen any colour you like.

Just create a new KML Ŝle with the appropriate screen size,
most are in
2x mode. For an example of a small image, see the KML Ŝles i
made at

That would be up to the creator of the bitmap used in the KML
Ŝle. You
===

Ŝnd
http://groups.google.com/groups?group:comp.sys.hp48&hl=de&
selm=3a6cc698%40ne

like many people here I donıt have read the manual. :) It was
quite easier
for me to Ŝnd the patch position in the HP49 ROM after
detecting how the
auto off work in the HP48 ROM. So this works in all Emu48
state Ŝles
without creating the TOFF variable. I donıt also take care
about the ROM
CRC, because it has been corrupted by the =makebeep patch
before.
====
whatıs the syntax for the display area argument for this
command?
Page H-8 users guide isnıt to much help...
- . . .
___
====
Itıs a real (a ZINT is automagically converted to a real and
accepted).
You can add up the following three values:
1. Header
2. Stack
4. Menu line
So e.g. 7 is all three areas, while 5 is only header and menu
line. 0
====
thanx a lot!
====
--
Colin Croft
====
====
====
====
This is a newbie question:
Is there a way to make the values in the Ŝelds of an INFORM
form
to appear in small font? All the remaining text is already in
font8, my default.
=============================================================
Paolo Cavallo I am a teacher at heart, and
====
Al d216 Tue, 13 Nov 2001 14:52:01 +0100, Paolo Cavallo
Yes, there is. But you have to learn sysRPL and play with new
Inform
Box Generator (command IfMain2).
If you use UserRPL or sysRPLıs old engine there is no easy
way (or, at
====
Yes, I was thinking about that but diff.eqs with initial
conditions are
Ŝne
for LAP but all my investigations with integrals/INTVX/RITCH
and LAP never gives me a solution it just hangs thinking for
ever :(
Then Iım trying to Ŝnd a way to solve it by hand only to
imagine a way for tha HP ;)
====
Iıve written a documentation which explain what are the
interrupts, how to deturn it and how to write a new manager.

====
what do you think of this keyboard lay-out:
http://www.4p-online.com/4p/dat300_47_keyboard.html
i would like to make one very similar (the enter key double
size), but with HP48
keys (not membrane). I would like to run programs like
Mathematica or Maple

====
Sorry to bother you all again, please believe Iım really
trying...
How can I put my libraries in the APPS menu? Is there any
easy (i.e.,
User-RPL) way to do it?

====
You need to create a so called $EXTPRG it must contain the
name of a
variable in your source directory letıs say Œextprgı
It (extprg) is run on different moments in the 49ıs operations
For now I shall discuss only the APPS application.
The input for Œextprgı always has a BINT on level 1 of the
stack
and if that is a BINT0 then you can be sure that you have the
APPS
choose list on level 2
letıs look at an example of an extprg which adds the name My
Lib to
the APPS menu.
(it has to be in SysRPL)
::
DUP
#0= % only do this if in APPS
case
::
SWAP
DUPLENCOMP % get size of APPS list, leave list on stack
#1+ % add one
UNCOERCE % make a real
MY LIB % this will appear in APPS
&$ % merge the two strings
Π% put following program on stack
::
% 1000. InitMenu% % this will initiate menu number 1000
(change it
to your lib id)
;
====
Yes, there is. And no familiarity with SysRPL or the
involved variable $EXTPRG is assumed. Simply get the
small library APPSman from hpcalc.org. In the latest
version, the use of the customization-Ŝle APAR has
been generalized. You can infact do whatever you want
====
just to my mind. This lib has only one command
(IFB as far as I remember) but has no $EXTRPRG.
====
1. In RPN or algebraic mode?
2. The function you are surching for in algebraic mode is
LIMIT(f(X),X=0+0)
====
linking all your Ŝles together. A .LR Ŝle will be created.
Make sure that all the entries are resolved, and that you are
using the
right address (e.g. the one for the HP49)
As thereıs no problem with your code here, it should work Ŝne
====
BTW you can shorten this to:
GOSBVL =PopASavptr
====
worked out why it did not work.
When I make the call to MINI_DISP, C [A] must contain the
number of
characters in the string. The problem with the code was when I
divided the number by 2
The 5th nibble of C [A] was not equal to zero. for example,
after the
division
C [A] = 8000B (No wonder it did not like this)
To Ŝx the problem it was just a matter of setting C [W] to
zero
CODE
GOSBVL =SAVPTR * SAVE RPL Pointers
A=DAT1 A * Pointer to String
D1=A * Point D1 to Address
D1=D1+ 5 * Skip Prologue
C=0 W * Reset C
C=DAT1 A * Read Length
C=C-CON A,5 * Remove Length Ŝeld from size
CSRB A * Divide by two for number of char
D1=D1+ 5 * Skip length Ŝeld
GOVLNG =MINI_DISP * Display String
GOSBVL =GETPTRLOOP * Reset RPL Pointers
ENDCODE
I have since rewritten the code to include the supported entry
=GetStrLen (Which eliminates most of the code)
Input: D1=$ (Address of string)
Output: C.A = length, D1 = body (start of characters)
Code now looks like.
ASSEMBLE
NIBASC HPHP49-C
RPL
::
$ How are you going?
CODE
GOSBVL =PopASavptr * Pop string from Stack & Save Pointers
====
In this case the ŒAı will be recognized as comment.
You could have used CSRB.F A,
which operates on a selected Ŝeld of C.
or even shorter:
CODE
GOSBVL =PopASavptr
C=A A
GOSBVL =GetStrLenC
GOSBVL =MINI_DISP
GOVLNG =GETPTRLOOP
ENDCODE
The above code saves one nibble;-)
Or shorter again, if you donıt want to drop the string from
the stack:
====
Sorry. My mistake. I guess Iıve been working too much with
the MASD syntax.
The CSRB.A instruction using the HP syntax is CSRB.f A
===
In general as I remember (itıs been awhile since I took one)
the large
highly programable (IE you can use it to store pages of
notes, formuals
etc.) calculators and IR xfer calcs are not allowed on most
standardized
tests. SAT, ACT etc.
Basically the only thing they want you to use the calculator
for is for
adding, multiplying, dividing.
ACT lists the Ti 89/92, HP 40/49 as ŝat out not allowed due
to their
capability to handle algebraic expressions (simpliŜcation
etc.). Multiply
polynomials, Factoring is also a no no for any calc on the
ACT. Since from
what I know the 38/39 will do those things as well Iıd say
they would not
be allowed to either. However Iıd think just about all the
TIıs would be
to becuase Iım pretty sure they can multiply polynomials and
factor. SATıs

site seems to be a bit more ŝexible but itıs hard to say how
they will
interpret those more general rules. Unlike ACT they donıt
speciŜcally
list any calcs that are not allowed just provide guidelines
similar but not

quite as picky as ACTıs.
And of course as with any set of standardized rules it
probably varies some from test to test. I know ACT lists the
HP48G as being okay if you cover
the IR port, but I know when I tried to take my ACT with my
48GX I was ŝat

====
I just started studying
EE and it depends highly on the Œprofı which calc is allowed.
I could
try to use a HP 32S in the exam (which can solve for example
an
quadratic equation, canıt it?), but HP calcs are well known
by the
profs (they used them too!), so it could be a risk (i donıt
dare) to
use one.

====
Which means that the following complaints are reproducible ;-)
Take the HP41 in dimm light and compare its legibility of
keys to the
HP49. Youıll see what I mean.
But I need. I am not very good in memory games. ;-)
Then you should try mine. After 5 minutes you will be the man
with the
smoking Ŝngers. ;-)
Or try an HP48xx to see what I mean. (Or even better an HP41
:-) )
The different people you already mentioned above. ;-)
Ahaaa! Thatıs how it is meant to be used! Glad to here that
because
until now I always hanged mine on a string that I have glued
on the

I`m just starting to think about a major mode for Emacs (not
1.07 but
====
I started a major mode for *emacs. It does syntax hilite,
indentation
and delimiter matching but no interpretation. I was working
on hooks
to the GNU tools when I stopped working on it. Perhaps it
would be
easier to do the interpretor in Perl and provide it as a hook
for the
never ported to the 49 rom but maybe it is doable?

Very interesting, is this for both sys and user rpl? Iıve
been looking
====
Making it work for USER rpl is actually pretty trivial. I
think I have
it now where it does anyway. It is a simple matter of how you
deŜne
-Al
--
-Al Arduengo
------------
Windows 3.1: The best $89 solitaire game you can buy.





--
====
====
Does anyone know of a computer software editor
to use to write programs for the HP48 on a computer
to later be downloaded to the HP48 ?
The absolute least would be one that you can enter
the hex code equivalent of hp code, but preferably one
that recognizes hp commands and automatically enters
====
Vous trouverez sur mon site web
(http://perso.atsat.com/pigallio) les
conversions que jıavais faites specialement pour la HP-Party
de ce week-end
: HP-Worms v1.1 pour HP39/40 et Trajectoires demo pour HP49
et HP39/40
A+
How about this?
LOBAT from Joe Hornıs Goodies Disk #10:
%%HP: T(3)A(D)F(.);
::
PTR 4544 @ AlertStatus (for GX only; use PTR 4546 for SX)
#2/ @ move System Batts bit to lsb
ONE #AND @ mask out the Port Batts bits
UNCOERCE @ convert bint to real number
;
If you donıt have Jazz or donıt want to compile it, youıll
have to download

it
for yourself at hpcalc.org
====
This will only tell you whether the batteries are good, not
how much
====
True. I seem to recall something about sampling this bit
several times in
====
There are some stuff that i donıt know how to do with the 49g
First why te calculator cannot handle inequalities with abs
value in
solvex.
Secondly how to solve this kind of equations with solvex:
log,base3(log,base4(12*log,base4(x)))=0
or log,base2(x)+log,base4(x)+log,base8(x)=0
if possible in step bvy step mode
====
You can try using the following equality:
====
====
anyone know the best low power ram to get for a 48G+? looking
to upgrade to
1.2meg total 128 in slot 1 merged and maybe 512 or 1 meg in
slot 2. thanks
====
And if you practice a bit more the 68K, you would know that
itıs possible

to
perform a multiply instruction on the 68K faster than the
multiply opcode..
If I remember correctly (and I havenıt checked for the past 8
years), MUL

is
170 cycles isnıt it ?
====
It doesnıt really matter that the Saturn is able to perform
operation on a
nibble.
You barely use this feature. What needs to be compared is on
a common usage
basis, on which case the 68K is faster, no doubt about that.
The proof is with the TI89, this calculator is overall faster
than the
HP49,
despite the fact that the TI programmer have probably spent
far less time
has we did to make the software efŜcient. Dirty code will be
acceptable on
====
====
SASM.DOC states on page 3 ff.:
The Saturn CPU is proprietary CPU optimized for high-accuracy
BCD math
and
low power consumtion. [...]
Thereıs more interesting information about the Saturn in this
doc and
others.
====
I believe that, however the Saturn was originally designed
before the
HP-71B project started, and I donıt think they speciŜcally
had Forth
in mind. If they did, they would have designed in the extra
instructions
that were later added in the processor for the HP-18C and
HP-28C.
The Saturn was designed to be a more powerful processor than
the one
used in the HP-41 and HP-1x (Voyager) family, and to be
better-suited
as a general-purpose processor. It was their Ŝrst Von Neuman
architecture for a calculator processor, in that it has a
single address
space for instructions and data. This is necessary, but not
sufŜcient,
to make it a suitable processor for FORTH or RPL.
The main design criteria appear to be optimization for BCD
arithmetic,
and a low pin-count bus. The earlier calculator processors
were bit-serial
instead of nibble serial. This allowed a fairly low pin count
for the
HP-41 expansion modules, but limited the performance. A four
bit bus
was a good compromise at the time, still allowing for a low
pin count
expansion connector, but increasing the performance
substantially.
By the time of the HP-48, the pin count of the expansion bus
became nearly
irrelevant for several reasons:
1) Improvements in packaging technology for plug-in modules -
high pin
count Ŝne pitch connectors
====
I donıt agree. You can often speed up an algorithm by Ŝguring
out how
to exploit this capability. But you donıt even have to put
much
thought into it to get a speed up. For example if you have a
quantity
that you know is limited to a certain number of nibbles. It
takes a
lot less time to use Saturnıs capabilities and perform the
operations
on just the nibbles you need.
Faster if you mean slightly faster and sometimes slower. Most
basic operations on the 68K ( bitwise, arithmetic, etc. )
arenıt that
much faster than on the Saturn and some are slower. Some
common
usage examples for the Saturn might be a 4-bit, 8-bit, 12-bit,
20-bit and 64-bit add. For these the clock cycle counts are
3.5, 4,
4.5, 4.75 and 11 respectively. For the 68K these might be an
8-bit,
16-bit and 32-bit add with respective clock cycle counts of
4, 4, and
8. Which CPU is faster based on these numbers is a matter of
debate.
Also, a lot of the cycles in the above mentioned instructions
arenıt
really attributable to the Saturn core but to its ancient
external
data bus which it uses to communicate with the other parts of
the
Yorke. The core is ready for an 8-bit bus and in fact ( I
think ) it
uses one internally ( as cycle counts would indicate ).
And this owes nothing to the fact that the TI-89 has 3 times
the clock
speed?
You know, there is a C compiler for the 68K and modern C
compilers can
====
Maybe that comment about the Saturn being faster was a little
brash (
getting a little hot headed from my Saturn evangelism ;) but
the fact
still remains that most of the equivalent 68K instructions
arenıt
====
On the current 68K a 32-bit multiply instruction takes 70
cycles max.
I seriously doubt it would take less in software. The
multiply is
microcoded and as such doesnıt need to deal with instruction
fetch/decode so if the software version and the hardware
version are
using the same method ( Which they arenıt. You have many more
capabilities with direct access to the ALU and registers ) the
hardware version will automatically be faster. If youıre
using the
shift-add algorithm for the software version then it will *at
least*
take a maximum of something like 500 cycles owing to the fact
that a
====
====
Is it 64 bit BCD ? If so, that would actually translate to
about 53 bits of

M68K.
M68K does not support 64 bits of data except in
multiplication and division

commands.
I donıt think nibble ŝexibility is such an important factor.
It seems very

exotic
====
But then you should consider, that M68K has probably many
more exotic,
rarely used commands not available on Saturn, that can be
exploited to
speed up the program.
Well, TI92 has also 3 times the number of keys on the
keyboard to be
scanned
and 3 times more pixels on the screen to be fed with
information :-)
No matter how efŜcient is the compiler, code is slow, if
programmers
====
====
Because without wanting to be too much arrogant, Iım one with
of the the
strongest experience on writing software for the Saturn CPU.
Iıve had a
long
experience with the 68K too.
Back in the 80s when I had my Atari ST (M68K at 8Mhz), I was
able to make
things that you can just dream about with the HP48.
If the HP49 is able to compete in speed with the TI89, it has
little to do
with its processor but due to the algorithm used and how itıs
been
implemented.
performance than the Saturn.
The Saturn is extremely poor in the following area:
-addressing mode
-have very little registers (4 work registers, 5 storage)
(you canıt even
work between many registers)
-only two address registers (8 for the
-8 levels return stack (6 only can really be used), you can
access only the
Ŝrst level of the stack (just this poor return stack is a
reason to get
rid
of it)
-no maskable interrupts, and in fact can generate only one
interrupt common
for all events
-no protected mode
-Creating a debugger is a pain in the a***, you have to
emulate everything
In all these areas, the 68K is performing really well. Also,
a 68K is about
20 times cheaper than the Saturn ($11 in big quantity)
One of the advantage of the Saturn is its ability to work on
a Ŝeld, but

as
I said, these instructions are slow, and even buggy (C=C+2.WP
will not work
====
====
I see HP49G as a temporary improvement of the Œ48, on our way
to the real
deal. When HP has stopped this, there will nnot anytime soon
be the real
deal. The Œ49 and the Saturn is a dead end - I will not
develop software
for a dead end. I see Mr. Parisse has seen the light? I have
too.
Iım not willing to shell out any money for the Œ49 production
details - the
hardware is ancient, and the only thing holding it up is the
software. The
software is at itıs limits now, we canıt squeeze much more
power out of it.
Yes, the machine is customizable (I know thatıs not a
word...), but itıs
not
enough when we spend too much time doing trivial
calculations. We need
speed
improvement in the magnitude of 50 or 100 times. It can
easily be done, we
just need another platform.
Any improvement to the Œ49 performance-wise, will include a
different
platform. That, or NEC will sell the Saturn transistor
layout, and it could
be implemented in an FPGA. The latter solution will be
inefŜcient and
expensive, the Ŝrst will render all previous Œ49 software
useless. I opt
for a new platform now - preferably one with a C++ compiler.
HP have tried. The Œ49 will never bear itself. What we need
is a mass
market
TI killer. Whatever we think of the TI89/92 line of calcs,
they are the
biz.
The key is inexpensiveness, and marketing. HP could do the
latter, but
didnıt. The Saturn really prevented the former. Performance
comes in an
entirely different package.
HP does not share my ideas. The last years of HP progress has
lead me

to
believe that I will never work with Hewlett Packard. Not as
their politics
are now at least. I do not owe the current HP anything. I owe
the Corwallis
team, but they are not anymore. I owe this community, and we
get fewer and
far between. Nowadays, much of the grunt of comp.sys.hp48 is
no more -
Mika,
BP, JYA, CdB, Christian Meland, Gerald and many many others
are gone.
No, I do not see that. FPGA is expensive. Shrinking the die
is not feasible
in any small scale scenario. We have a die fab and bonding
facility at my
work place for prototyping; The least expensive production we
can do of,
than US$ 20000. The quality will be mil.spec, but even lesser
quality will
not save us much - it will be hard to get the price down a
factor of 10.
Who
will pay US$ 100 for a 300 MHz Saturn? We still need to get
new LCD drivers
for a larger screen, and a modiŜcation of the 4-bit databus
wil be
expensive and very vary hard to do. Another processor will be
much cheaper
and faster to implement.
I sincerely disagree here. Even the TI89/92ıs are too slow in
my opinion. I
believe the performance of the TIs could be improved maybe a
factor of 5 in
most areas - maybe more. That would be a start, but also take
huge
ressources for a new OS to be made.
Yes, but still limited by the 20-bit address space.
Mildly put.
To sell large quantities we must deŜne minimal mathematical
background

as
very basic knowledge.
That would be excellent.
Hehe :-)
I also frown a bit at the very limited user interface on
current PDAs, so I
do not see that as a solution either. The PDAs have the
advantage of modern
high-speed processors (new ones at least have) and large
screens.
I agree here too. The problem is that it is expensive to
leave out this
market, but one will save money not trying to do it all in
one calc. I then
vote for increasing minimal mathematical background to at
least more
than
====
Oh, thank you very much. Iım not gone yet mind you - there is
just a very
big possibility that Iıll be pursuing new ventures. It may
even be on this
group, but Iım just telling you that in my eyes the Œ49 is
dead as disco
now.
This is a huge task, and Iım not sure Iıll ever Ŝnd time for
it. The
personal docs of Mr. Parisse would probably be of more help
;-)
Hehe, if you want the knowledge I have of the Œ49 (is that
something to aim
for?), you just need a few things:
http://www.hpcalc.org/hp48/docs/programming/stut_pdf.zip
http://www.hpcalc.org/hp49/docs/programming/hp49sysrpl_pdf.zip
http://www.hpcalc.org/hp48/pc/programming/hptools-3.0.6-win32
.zip
http://www.hpcalc.org/hp49/docs/programming/masddocs.zip
http://www.hpcalc.org/hp49/programming/misc/nosy40.zip
http://zon.astro.uva.nl/~dominik/hpcalc/emacs/emacs10701.zip
Add to that a healthy dose of Œ48 for reference, and youıre
on your way ;-)
My loyalty is not with the ACO solely either. Iım just being
realistic.
I considered the Œ49 to be a taylored (from the Œ48)
temporary object in
the
big memory of calculator history. We needed to implement a
cooperation
(ACO)
that supported inheritance, so that we could instantiate a
new and better
object of it - this would be our goal. Now a Ŝnal GC has
occured, and our
unreferenced baby is gone. Madness is eminent, as you see ;-)
(If Iıll be gone)
====
which shows that the average webserver downtimes are much
(around 8 to
100 most used German WWW servers over a month (IIRC...).
I think the old rule four eyes see more than two (German
saying - I
donıt know the English version) also applies to software. And
while
closed-source software can only be seen by the eyes of the few
programmer(s) hired to code it, an open-source code can be
seen by
kernel, a LOT of people.
Who made any money in the past? Nobody except the ACO and the
Corvallis
team made any money, not to say a life out of the HP
software. And who
====
that
When you maintain software, you need version control and
rigorous structure
and ŝow planning - not to speak about interface documents.
That is much
easier with few people than with many. This is something that
I do have
alot
of experience with - without control you end up doing the
same work over
and
over again.
I would. If we need to produce the calcs ourselves, Iım not
giving them
away
for free. The work we have done in the past is little
compared to the work
HP has done. If we need to take it all up, we need funding.
I just object about the fact that I would make HP money by
maintaining
====
Canıt resist to put in my 2 :
library outside HP eg) and 68k (built my own system & adapted
CP/M 68k
and OS/9 to it).
The CPU/OS design limits you to certain ways to implement
algorithms,
so IMO it has a lot to do w/ the CPU.
Not necessarily when it comes to mathematics. Also, how do
you use a
68k correctly ?? (I doubt that the CSEıs at TI are
exceptionally dumb).
True but most can be emulated very easy w/o giving away too
many cycles. Remember the more complicated an addressing
mode on the 68k, the more cycles it uses. Also given the OS
environments, you donıt really need that many modes.
True but remember the 4 main registers are fully supported
64bit in
size. Working w/ 64bit on the 68k is cumbersome and slow --
especially
when it comes to BCD operations.
The hardware stack is actually an advantage on the Saturn --
having
a software stack would have slowed the RPL system
signiŜcantly down.
I would like more entries though.
Not on the 68k either.
Thatıs why in-circuit emulators where used for debugging. IMO
thatıs only
a nuisiance, so what ?
At the time the Saturn was developt, the 68k wasnıt an option
for various
reasons: cost, low connection count HW design, low power, BCD
capabilities.
Now, I donıt agree -- the poor instruction set makes it quite
hard to

master
the CPU.
Huh, howıs that ? If you donıt want to work w/ the Saturn you
simply donıt.

;)
Comparing the 68k w/ the 1LT8 (or higher) doesnıt lead to
anything, the 68k
is a general purpose CPU designed to run compiled high level
languages
(especially C) whereas the 1LT8 is designed to power
calculators w/ very
little additional HW (!) and run interpreters.
I was always amazed what you can do w/ the Saturn running at
2MHz in
comparison to 68k and 86k systems just by having the software
env. linked
to the HW (reminds me to those FORTH CPUıs -- they had a real
astonishing
performance).
Bye,
Detlef
--
====
To qoute you: not really. The 71B was the 1st device using
the Saturn
architecture and came out in 1984. It didnıt use RPL at all,
the 1st RPL
machine came out 1986 (18C).
See also http://www.hpmuseum.org/rpl.htm,
http://www.hpmuseum.org/saturn.htm,
RPL: A Mathematical Control Language, by W. C. Wickes,
published in
====
Actually they are :-)
====
Well, but in that cas you have to use many registers that
many times you
donıt have.
The OS environment is limited due to the limited CPU, I guess
other choices
would have been made with a different CPU
Yes, the Saturn is really good at working in BCD, especially
compare to the
68K which can only work in 8 bits mode.
However, the choice of using BCD is due to the Saturn CPU
itself. Thereıs

no
obligation of working in BCD, and I believe that choosing BCD
format on the
TI89 was a mistake , because of the CPU they are using.
As a matter of fact, neither ALG48 nor Erable v3 on the HP48
used BCD
representation for their long reals. We chose to use BCD in
the HP49
because
the display of the numbers with internal BCD coding was
faster. However,
calculations done on the HP49 is slower than on the HP48 with
ALG48.
To summarize, yes there is a good BCD mode with the Saturn
(despite the
fact
that you can only add/substract 1 to numbers) but the choice
of BCD would
have been different with another CPU.
The way you work with the software stack on the 68K is just
as with a
hardware stack, I donıt see what difference it makes.
My mistake, I should have written: supervisor mode.
everything
Well, itıs very important to be able to develop easily
software isnıt it ?
But more importantly, the ability to generate an interruption
after each
instruction gives the capability of multi-tasking which the
Saturn can only
dream of. Of course you can emulate totally such a feature,
but with a huge
cost on speed.
quite easily to 68K machine (uClinux).
about
capabilities.
That is true, but at present, if you compare the embedded
version of the
68K
and the current Saturn CPU (which was the origin of this
thread), then I
believe the 68K is a better choice has itıs a more ŝexible CPU
master
I guess it depends on how you see it: thereıs less
instruction to learn and
donıthave to worry with addressing mode
68k
Well, Iıve always been amazed with what Apple did with their
Apple 2 (1Mhz

8
bits CPU) and their Apple 2GS (as good if not better than all
the Amiga and
Atari and PC at the time) with the 65C816 running at 2Mhz
(they even had a
====
====
====
====
The HP49G user manuals can be found here:
http://www.hp.com/calculators/techsupport/graphing/49g_
userguide/
====
====
1. Screen at last 480 x 160 with 16 levels of grey (much like
Psion
5-series).
2. Fold-out conŜguration with QWERTY keyboard (much like
Psion 5
series).
3. Proper digital, stereo sound (with a headphone socket so
the thing
can double up as an OGG player).
recognition (experimental, of course).
5. A good CPU to deal with all this (e.g. TM 3120, the one
used in
web-pads).
6. An Li+ or NiMH battery and charger, with option to power
from the
mains (like a mobile phone) and work in the background in a
low-power
mode.
7. USB, RS232 (some of us are still in the dark ages), and IR
ports
(some of us still get kicks out of the old calculator/remote
control
thing).
8. Two types of RAM: system RAM (e.g. 8Mb SDRAM) and backing
store
(somewhat slower, but requiring less power, around 32 Mb or
so). The
operating systemıs job would be to effectively cache between
these two
stores, moving [RPL] objects in and out as needed.
which runs an incredibly more sophisticated version of RPL.
Iım thinking
multitasking here and a professional CAS. The works - top
quality 3D
graphics, a new problem-solving approach that doesnıt revolve
around
notebooks, plus better object-orientation (e.g. the ability
to deŜne
your own classes).
10. TCP/IP networking and connectivity.
====
I donıt like such a wide display - it should keep the VGA
aspect ratio. I
like 320x240 most.
No QWERTY as standard.
Why OGG? MP3 can be VBR too. Whatever - what Iıll be making
will not be a
stereo, but a data device. It wonıt run WinCE either.
Why both QWERTY and touch screen? You have commented on the
resolution too

-
do you just throw something in the Ŝre, or did you think
Ŝrst? No touch
screen on my device.
Thatıs nice, but very expensive. I know of a processor that
at equal
performance cost a third and uses half the power. What about
that?
If you had looked at the technologies, youıd say lithium ion
- no other
technology.
Which IR? IrDA? Have you read what have been discussed prior
to your post?
What would be slower than SDRAM?
====
What do you want, a calculator or a PDA? Start to add
countless
functions to a device...
interesting operating system in Earth, and whenever they see
any sort
Same thing. Why not a PDA?
Calculators are good because they are designed for one
purpose. When

you design something as a general purpose device, you have to
make
some design decisions that will not be good for each and all
of the
intended purposes of the device.
Current calculators have two advantages: very low power and a
user
interface especially designed for that, for calculating. It
is much
faster for me to grab my HP from the desk and do some
operations,
than moving the mouse, going to the menu, selecting the
calculator
====
--
Thierry Morissette
Z8cI7.807$Bs1.188939@news000.worldonline.dk...
too -
OK. But, since the keyboard will, I hope, be at least as
versatile to
reassign as the 49Gıs, _please_ make the machine so that
keyboard templates
can be designed and used on it (like on the good old 41).
Once again: why not HP-IL?
I would be happy with, letıs say 8 Mb RAM and a port for
Flash ROM, any
size
available.
Expansion ports, yes!  la 41, which had four of them!

====
templates
The machine could have a PS/2 connector as an option. The
keyboard will of
course be user deŜnable.
post?
Too proprietary, but it wouldnıt be impossible to deliver it
as an I/O
option -. remember, the I/O capabilities of this device are
user deŜnable
as well (even after purchase).
size
The base conŜguration may be less than 32 MB SRAM then? The
OS will pretty
====
====
====
That would be something, and after reading everyoneıs ideas a
couple
things kept sticking in my skull (which is amazing as thick
as it is).
First, a lot of those ideas I donıt think would be easily
incorperatable into a small package.....a serial/paralell
port in
something that will Ŝt in a pocket, etc. I mean some $3000
labtops
donıt have all those features :) If you could build it youıd
need to
kill anyone that saw it and have a self destruct button on it
in case
you were compromised.
I want my calculator to be a calculator, not a PDA, not a
computer,
IMO from the stuff Iıve seen thatıs a combo of devices you
almost
always compromise on both devices and in some cases itıs more
$ than
buying both. If I want a PDA Iıll buy a PDA if I want a
palmtop Iıll
buy that, if I want a calculator iıll do that, if I really
need all 3
Iıll carry a labtop.
Lion batteries, nice yes, but extremely expensive compared to
other
rechargeable options, also you need more complicated internal
recharging circuits. IE yes itıs the most technical options
itıs also
by far the most expensive to impliment.
Cost is the big thing that stands out as a problem for me. I
mean how
much market is there for a $1000 calculator? If HP is selling
the 49G
for $179 and you add a ton of memory, a penitum, color
display, ion
batteries, Ŝrewire, etc. etc. where does it all end? I mean
at some
porint itıs going to be cheaper/better to either go with a
palmtop
computer and derive/48/49 emulator or a full labtop, for the
same $
that you donıt have to squint to see more than 2ı away.
Hereıs what *I*ıd like to see improved upon say the base of
the 48GX.
faster transfers, maybe USB
a 256meg ram at most, I mean hell my computer has less than
that.
possibly a small PDA style processor, I think they are using
low mhz
pentiums now
Iıd like to see about 8 meg of storeage stock, with expansion
cards
a better quality, larger, backlit display
Iıd like a folding design, say what you want about folding
being
weaker, but Iıve owned a lot of folding devices and havenıt
had any of
them fail yet. IMO the weakness of some folding devices is
simply
poor implimentation. That way you can have one side with much
more
screen and possibly a few buttons and a qwerty style keyboard
on the
other. Iıd like to see a touch screen for graph interfacing
etc.
Batteries, K.I.S.S AAAıs maybe something where you can use a
lion
battery or a adapter with AAAs in it. But it is going to
really ramp
up the cost if you use lion. Espesically if you donıt have a
normal
battery option, then you need a spare battery or two.
Color screen would be nice but I donıt know if youıd need
more than 16
or 256 colors at most for a calculator. Maybe borrow one from
TI
(gasp) since they seem to have by far the nicest calculator
screens as
far as contrast around, though no color.
If itıs going to be a power hog and as long as weıre going
off the
deep end how about a solar panel to recharge the ion battery
as you
use it when there is adequate light. Or maybe use the kinetic
power
of it being josseled around while you walk to recharge it
like the
====
What do you think a laptop display, HDD and processor cost?
Look at
http://www.junipersys.com/allegro_pc.htm for an idea to what
I mean. The
price is USD 5000.
You havenıt seen this.
This wonıt.
A laptop has many disadvantages.
For USD 0.25 worth of circuitry, yes. LiOn batteries are all
advantages and
no disadvantages. The added expense is neglible.
IE? Whatıs that?
Nobody said USD 1000, did they? Iıve said USD 150-250.
The manufacturing cost is maybe one fourth of that - what if
Iım satisŜed
with a 10% overhead instead of 3-400%?
Three years from now maybe?
I would very much like to see a comparable laptop with a USD
250 price tag.
Where does this differ from the specs mentioned?
Arenıt you adding to the price tag with these features? I can
promise you
this: NO folding design, NO QWERTY and NO touchscreen. Too
much PDA/laptop,
and too little sense.
More so if you implement an adapter that takes both.
Spare batteries wonıt be an option when you have 300 hours of
battery life.
You also want rechargables.
Why color? you donıt need one for a calc - you have grayscale
anyway.
LOL! You have not seen many LCDs, have you? I havenıt seen
many 320x240
LCDs
from TI either.
No.
Solar panels are very expensive.
Kinetic watches suck big time. I know, Iıve had a couple.
Talk about
mechanics that break.
Would be an option, if the smallest possible nuclear
powerplant didnıt
====
You seem to take my post as some personal attack on the
device your
building/thinking of building. I was simply making comments
on many
of the items people have asked for in this thread for the
ideal
your machine since you had no post listing them in this
thread.
I did go back to a previous thread and read your ideas for
designing a
new calc and they sound great. I was responding to the ideas
posted
in this thread though, not the guidelines of your
design/ideas. I
wish you the best of luck in your design. With HP slipping it
may be
the best hope for the next super calculator.
Todd